Yes. All three runs, Standard, Deep and Extended, the plain-words verdict, the curve, the decay figures and the before/after comparison are free. A report unlock pays for the cause and the fix plan.

Do I need to install anything or make an account?

No. The test runs in the browser tab with Web Workers and WebGL or WebGPU. Nothing is uploaded unless you ask for an emailed copy of the report.

How long does it take?

The standard run is exactly 3 minutes across 4 stages. The Deep run is exactly 9 minutes across 8: your screen, two and a half minutes of CPU load to reach thermal saturation, memory, the GPU pass looped three times, CPU and GPU together, soft shadows, path tracing, and a hot re-test. Extended is the Deep run with 10 minutes of CPU load, 16.5 minutes in all. Each is the same length on every machine.

Which processor should I pick?

The one in the machine. The browser cannot read its name, so the report is labelled with your choice from a catalogue of about 1,600 Intel, AMD, Apple and Qualcomm parts. On Windows, Ctrl+Shift+Esc → Performance → CPU shows it; on macOS, About This Mac.

Does it work on a phone or tablet?

No. The test needs a desktop-class browser with Web Workers and a GPU backend, and a touch-first device is turned away at the gate rather than given a misleading number.

What should I check first on a refurbished laptop?

In this order: that it holds its performance under sustained load, the battery cycle count, the drive's SMART hours, and then the physical items you can see. The first two are the expensive ones to get wrong and the ones a listing is least likely to describe honestly.

The seller said it was tested. Is that enough?

It depends entirely on what was tested. Most refurbishment testing confirms that a machine boots, that the ports work and that the drive is wiped. Almost none of it holds the processor at full load for three minutes, which is the only way tired cooling shows up.

It slows down but the listing said grade A. What now?

Grades describe cosmetics, not thermals, and an immaculate chassis says nothing about the paste under it. A measured curve is a concrete thing to put in front of a seller, and most reputable refurbishers would rather service it than take a return.

Everyone says it is my internet. How do I tell?

By the timing. Network problems are erratic: they come and go, and they do not care how long you have been on the call. A thermal limit is punctual, arriving at roughly the same point in every long call and never in a short one. If it is always the second half, suspect the machine.

Why does it only happen on long calls?

Because heat takes minutes to build. A ten-minute stand-up never gets there. An hour-long review does, and by then the machine has quietly stepped down to a level where real-time encoding no longer fits.

Does closing my camera help?

Often, and that is itself diagnostic. If turning the camera off recovers the call, you were short of processing headroom rather than bandwidth, which points back at the machine.

Is it safe to stress test a machine that already crashes?

A browser workload is lighter than a dedicated stress tool and stops when the tab does, so it is a gentler place to start than a power-virus utility. Even so, a machine that freezes from heat should be looked at physically before it is loaded repeatedly.

How do I tell throttling from a real fault?

Throttling is graceful: the machine gets slower and keeps working, and the curve slopes. A fault is abrupt: the machine stops, reboots or shows a driver error, and there is often no slope at all before it.

It only crashes in games, not here. Does that rule out hardware?

No. Games load the processor, the card, the memory and the power supply together, and a browser test is lighter on all four. A clean result here narrows the field; it does not clear the machine.

Why is my computer fast at first and then slow?

The processor boosted, then hit a limit. If the limit was a power budget it stepped down once and held; if it was heat it slid gradually; if another program took the cores it became erratic. A sustained-load test shows which, because each has a different shape.

Is it thermal throttling?

A slide is consistent with it. A single clean step is consistent with a power limit rather than heat, which cooling does not change; comparing on mains in performance mode, or a sensor tool, confirms which. The report names the shape and states how confident it is given the noise in the measurement.

What should I try first?

For a slide: clean the fans and check airflow. For a step: check the power plan and the vendor's performance mode. For erratic: open the task list and look for what is using the processor. Then run again; the report compares the two runs.

Can a browser test measure laptop temperature?

No, and any web page that shows a temperature is guessing. This test measures throughput over time; heat shows up as decay with a sliding shape. For the temperature itself use a monitoring tool with sensor access.

My laptop is hot but the test says it held. Is that possible?

Yes. A hot chassis with a curve that holds means the cooler is carrying the load; the heat is being moved, not accumulated. Warm is not the same as throttling.

Does a cooling pad fix overheating?

It helps a sliding curve, because a slide is consistent with the cooler being the limit, and more airflow raises its ceiling. It does nothing for a stepped curve, which is consistent with a power limit. Run before and after; the report compares them.

How do I test if my CPU is throttling?

Run a sustained all-core load and watch throughput over time rather than a single score. If it keeps falling after its boost ends and it settles, that is consistent with throttling; a short boost and one step down early in the run is normal and shown separately as the boost step. The shape of the fall points toward a power limit or toward heat, and a sensor tool confirms which. This test does exactly that in a browser tab.

What is the difference between power-limit and thermal throttling?

A power limit is a wattage or current ceiling set in firmware: the chip drops to it in one step once its boost window closes. Thermal throttling is the chip stepping down progressively as its temperature rises: a slide. The fixes are different, which is why the report names the kind.

Under about 8% is ordinary settling. Around 10% is a mild loss after the boost and is common on laptops. A short boost and one step down early in the run is not part of the figure. Above 15% with a sliding shape is consistent with cooling being the limit, and a sensor tool confirms which. The report puts the figure next to the shape so the number is read correctly.

Why does my FPS drop after 10 minutes?

One of the two chips reached a limit and stepped down. The test runs each under sustained load and shows which lost performance and how: a slide is consistent with heat, a step with a power limit, noise with another program. The fix depends on which chip and which shape.

Is it my CPU or my GPU?

The report gives both a decay figure. The one that fell is the one to look at; if the CPU held and the GPU decayed, the card's cooling is the place to start, and the other way round.

Can I fix FPS drops without new hardware?

Often. A sliding curve responds to cleaning, paste and airflow. A stepped curve responds to a power limit or a performance mode. An erratic curve responds to closing whatever else is running. The before/after comparison shows whether the change worked.

Why is my laptop so slow on battery?

Most laptops switch to a lower sustained power limit on battery to protect the cell and extend runtime. The performance loss is the size of that limit. Run the test once plugged in and once on battery; the report compares the two and states the difference.

Can I make my laptop as fast on battery as plugged in?

Usually not fully. The vendor's power tool can raise the battery limit at the cost of runtime and heat. The comparison shows how much a setting change recovered.

Does the test drain the battery?

A standard run uses a few percent on most laptops; the report states how much. A deep run uses proportionally more.

Why is my new PC slower than it should be?

Three causes cover most cases: a motherboard power limit lower than the chip can use, a memory kit running at its fallback speed, or a cooler that is not seated. Each shows differently in the test: a step, low bandwidth, or a slide.

How do I know if my RAM is running at the right speed?

The memory stage reports bandwidth from one thread and from all threads. Enable the kit's rated profile in the BIOS and run again; the report compares the two runs and the bandwidth figure moves if the profile took effect.

Should I be worried about a single step down?

No. A step is consistent with a power limit, which on a new build is usually the board's default. Raising it is a BIOS setting. A slide is the one to act on.

How do I know if my PC is throttling?

Run a sustained load and watch throughput over time. If it keeps falling after its boost ends and it settles, that is consistent with throttling, and a sensor tool confirms it. A short boost and one step down early in the run is normal. This test does that in a browser tab and states the answer in words on the first screen of the report.

Not by itself. A single step to a power limit is the design of almost every laptop. A slide that loses a fifth of the machine's performance is consistent with cooling that needs attention. The report distinguishes the two.

Do I need to install anything?

No. The test runs in the browser with nothing installed and nothing uploaded unless you ask for an emailed copy of the report.

Why does my FPS drop after a few minutes of playing?

Almost always one of two things: the processor or graphics card reached a power or temperature limit and stepped its clock down, or something else is now sharing the cores. A sustained-load test separates the two: a limit shows as a curve with a shape, contention shows as noise. The report names which.

Can a browser test measure gaming performance?

It measures the thing that decides gaming performance over time, which is how much of its peak a machine holds when hot. It cannot measure a specific game's frame rate; use the game's own counter for that and this test for why the counter fell.

Should I run on battery or plugged in?

Plugged in, unless you play on battery. On battery most laptops apply a lower power limit, and the report flags a run on battery so the two are never compared as equals.

Can I run this on every machine that comes through?

Yes. There is nothing to install, nothing to uninstall afterwards, and no account, so it does not add a step to the wipe-and-reimage routine. Open the browser, run it, and the machine you hand over is exactly as you imaged it.

Does a good result actually help me sell?

It changes what the listing is arguing about. “Runs great” is a claim; a report a buyer can open on their own phone and read for themselves is evidence. It is most valuable on the machines that are genuinely good, which are the ones currently being priced as though they might not be.

What should I do about a machine that fails?

Look at the shape before you write it off. A steady slide is usually thermal paste and dust, which is cheap to fix and turns a marginal unit into a good one. A sawtooth or a cliff is a cooling fault, and those are worth diagnosing before you spend money on parts.

Will a buyer trust a test I ran myself?

More than they trust an adjective you wrote yourself. It is a report generated by a third party, in plain words, that they can open without installing anything. It does not make you unimpeachable; it moves the conversation from what you say to what the machine did.

What if the result is bad?

Then you have found out before your buyer did. A disclosed flaw costs you some of the price; an undisclosed one costs you the sale, the return, and the rating.

Running the test is free, both lengths. A report unlock is optional and is about getting the cause and the fix plan; the curve and the headline figures you would want for a listing do not need it.

Why not just use the tools I already have?

Keep them. This one has a specific property they do not: the customer can open the result themselves, on their own phone, and read it in plain words. It is not a better measurement than a sensor log. It is a more explainable one.

Can I run it on a machine I have not imaged yet?

Yes, that is the point of it running in a browser. Nothing is installed and nothing is left behind, so an intake test does not change the state of a machine you may need to hand back untouched.

Does it work on a machine that will not hold a charge?

Run it plugged in. On battery most machines apply a lower power limit and the report flags that, so a battery run and a mains run are never compared as if they were the same test.

How do I use this across hundreds of machines without an agent?

Most teams do it as part of something a person is already doing at the desk: a support visit, an OS upgrade, a hardware refresh survey. It costs the technician three minutes and no install, which is usually the difference between a check that happens and one that does not.

Our machines are all the same age. Why would they differ?

Because ageing is environmental, not chronological. Dust, ambient temperature, how a desk is positioned, whether a vent has been against a partition for two years. Identical purchase orders produce very different machines by year three, and the spread is the useful signal.

Is a low score a reason to replace a machine?

Not on its own, and this is the mistake worth avoiding. A machine that slides badly is often a machine that needs an hour of cleaning, which is far cheaper than a replacement. Look at the shape first: a slide is usually serviceable, a hard limit usually is not.

Is my laptop faulty, or is it just hot here?

Run the same test twice: once at the coolest hour you can manage, once in the heat of the afternoon. A machine that holds up in the morning and slides in the afternoon is not faulty, it is thermally limited by the room. A machine that slides in both is worth opening up.

Questions

Every question, answered once.

221 questions the site answers, each linked to the page that answers it at length.

The short answer to most of them: SystemCheck is a free stress test that runs in the browser, holds the processor and graphics card under a fixed load for a fixed time, and tells you whether performance held up across the run. It cannot read temperatures or clock speeds, so it reports what throughput did and which explanations fit, never a cause it did not see. Start with the benchmark test, read how each stage works on the methodology page, and see what is free and what is paid on the pricing page.

The test

Is SystemCheck free?: Yes. All three runs, Standard, Deep and Extended, the plain-words verdict, the curve, the decay figures and the before/after comparison are free. A report unlock pays for the cause and the fix plan. Pricing →
Do I need to install anything or make an account?: No. The test runs in the browser tab with Web Workers and WebGL or WebGPU. Nothing is uploaded unless you ask for an emailed copy of the report. Privacy →
How long does it take?: The standard run is exactly 3 minutes across 4 stages. The Deep run is exactly 9 minutes across 8: your screen, two and a half minutes of CPU load to reach thermal saturation, memory, the GPU pass looped three times, CPU and GPU together, soft shadows, path tracing, and a hot re-test. Extended is the Deep run with 10 minutes of CPU load, 16.5 minutes in all. Each is the same length on every machine. Methodology →
Which processor should I pick?: The one in the machine. The browser cannot read its name, so the report is labelled with your choice from a catalogue of about 1,600 Intel, AMD, Apple and Qualcomm parts. On Windows, Ctrl+Shift+Esc → Performance → CPU shows it; on macOS, About This Mac. Run the test →
Does it work on a phone or tablet?: No. The test needs a desktop-class browser with Web Workers and a GPU backend, and a touch-first device is turned away at the gate rather than given a misleading number. Disclaimer →

Symptoms

What should I check first on a refurbished laptop?: In this order: that it holds its performance under sustained load, the battery cycle count, the drive's SMART hours, and then the physical items you can see. The first two are the expensive ones to get wrong and the ones a listing is least likely to describe honestly. Checking a refurbished laptop →
The seller said it was tested. Is that enough?: It depends entirely on what was tested. Most refurbishment testing confirms that a machine boots, that the ports work and that the drive is wiped. Almost none of it holds the processor at full load for three minutes, which is the only way tired cooling shows up. Checking a refurbished laptop →
It slows down but the listing said grade A. What now?: Grades describe cosmetics, not thermals, and an immaculate chassis says nothing about the paste under it. A measured curve is a concrete thing to put in front of a seller, and most reputable refurbishers would rather service it than take a return. Checking a refurbished laptop →
Everyone says it is my internet. How do I tell?: By the timing. Network problems are erratic: they come and go, and they do not care how long you have been on the call. A thermal limit is punctual, arriving at roughly the same point in every long call and never in a short one. If it is always the second half, suspect the machine. Video calls stutter →
Why does it only happen on long calls?: Because heat takes minutes to build. A ten-minute stand-up never gets there. An hour-long review does, and by then the machine has quietly stepped down to a level where real-time encoding no longer fits. Video calls stutter →
Does closing my camera help?: Often, and that is itself diagnostic. If turning the camera off recovers the call, you were short of processing headroom rather than bandwidth, which points back at the machine. Video calls stutter →
Is it safe to stress test a machine that already crashes?: A browser workload is lighter than a dedicated stress tool and stops when the tab does, so it is a gentler place to start than a power-virus utility. Even so, a machine that freezes from heat should be looked at physically before it is loaded repeatedly. It freezes under load →
How do I tell throttling from a real fault?: Throttling is graceful: the machine gets slower and keeps working, and the curve slopes. A fault is abrupt: the machine stops, reboots or shows a driver error, and there is often no slope at all before it. It freezes under load →
It only crashes in games, not here. Does that rule out hardware?: No. Games load the processor, the card, the memory and the power supply together, and a browser test is lighter on all four. A clean result here narrows the field; it does not clear the machine. It freezes under load →
Why is my computer fast at first and then slow?: The processor boosted, then hit a limit. If the limit was a power budget it stepped down once and held; if it was heat it slid gradually; if another program took the cores it became erratic. A sustained-load test shows which, because each has a different shape. PC slows down after a few minutes →
Is it thermal throttling?: A slide is consistent with it. A single clean step is consistent with a power limit rather than heat, which cooling does not change; comparing on mains in performance mode, or a sensor tool, confirms which. The report names the shape and states how confident it is given the noise in the measurement. PC slows down after a few minutes →
What should I try first?: For a slide: clean the fans and check airflow. For a step: check the power plan and the vendor's performance mode. For erratic: open the task list and look for what is using the processor. Then run again; the report compares the two runs. PC slows down after a few minutes →
Can a browser test measure laptop temperature?: No, and any web page that shows a temperature is guessing. This test measures throughput over time; heat shows up as decay with a sliding shape. For the temperature itself use a monitoring tool with sensor access. Laptop overheating →
My laptop is hot but the test says it held. Is that possible?: Yes. A hot chassis with a curve that holds means the cooler is carrying the load; the heat is being moved, not accumulated. Warm is not the same as throttling. Laptop overheating →
Does a cooling pad fix overheating?: It helps a sliding curve, because a slide is consistent with the cooler being the limit, and more airflow raises its ceiling. It does nothing for a stepped curve, which is consistent with a power limit. Run before and after; the report compares them. Laptop overheating →
How do I test if my CPU is throttling?: Run a sustained all-core load and watch throughput over time rather than a single score. If it keeps falling after its boost ends and it settles, that is consistent with throttling; a short boost and one step down early in the run is normal and shown separately as the boost step. The shape of the fall points toward a power limit or toward heat, and a sensor tool confirms which. This test does exactly that in a browser tab. CPU throttling test →
What is the difference between power-limit and thermal throttling?: A power limit is a wattage or current ceiling set in firmware: the chip drops to it in one step once its boost window closes. Thermal throttling is the chip stepping down progressively as its temperature rises: a slide. The fixes are different, which is why the report names the kind. CPU throttling test →
Is 10% decay bad?: Under about 8% is ordinary settling. Around 10% is a mild loss after the boost and is common on laptops. A short boost and one step down early in the run is not part of the figure. Above 15% with a sliding shape is consistent with cooling being the limit, and a sensor tool confirms which. The report puts the figure next to the shape so the number is read correctly. CPU throttling test →
Why does my FPS drop after 10 minutes?: One of the two chips reached a limit and stepped down. The test runs each under sustained load and shows which lost performance and how: a slide is consistent with heat, a step with a power limit, noise with another program. The fix depends on which chip and which shape. FPS drops over time →
Is it my CPU or my GPU?: The report gives both a decay figure. The one that fell is the one to look at; if the CPU held and the GPU decayed, the card's cooling is the place to start, and the other way round. FPS drops over time →
Can I fix FPS drops without new hardware?: Often. A sliding curve responds to cleaning, paste and airflow. A stepped curve responds to a power limit or a performance mode. An erratic curve responds to closing whatever else is running. The before/after comparison shows whether the change worked. FPS drops over time →
Why is my laptop so slow on battery?: Most laptops switch to a lower sustained power limit on battery to protect the cell and extend runtime. The performance loss is the size of that limit. Run the test once plugged in and once on battery; the report compares the two and states the difference. Laptop slow on battery →
Can I make my laptop as fast on battery as plugged in?: Usually not fully. The vendor's power tool can raise the battery limit at the cost of runtime and heat. The comparison shows how much a setting change recovered. Laptop slow on battery →
Does the test drain the battery?: A standard run uses a few percent on most laptops; the report states how much. A deep run uses proportionally more. Laptop slow on battery →
Why is my new PC slower than it should be?: Three causes cover most cases: a motherboard power limit lower than the chip can use, a memory kit running at its fallback speed, or a cooler that is not seated. Each shows differently in the test: a step, low bandwidth, or a slide. New PC underperforming →
How do I know if my RAM is running at the right speed?: The memory stage reports bandwidth from one thread and from all threads. Enable the kit's rated profile in the BIOS and run again; the report compares the two runs and the bandwidth figure moves if the profile took effect. New PC underperforming →
Should I be worried about a single step down?: No. A step is consistent with a power limit, which on a new build is usually the board's default. Raising it is a BIOS setting. A slide is the one to act on. New PC underperforming →
How do I know if my PC is throttling?: Run a sustained load and watch throughput over time. If it keeps falling after its boost ends and it settles, that is consistent with throttling, and a sensor tool confirms it. A short boost and one step down early in the run is normal. This test does that in a browser tab and states the answer in words on the first screen of the report. Is my PC throttling? →
Is throttling bad?: Not by itself. A single step to a power limit is the design of almost every laptop. A slide that loses a fifth of the machine's performance is consistent with cooling that needs attention. The report distinguishes the two. Is my PC throttling? →
Do I need to install anything?: No. The test runs in the browser with nothing installed and nothing uploaded unless you ask for an emailed copy of the report. Is my PC throttling? →

For you

Why does my FPS drop after a few minutes of playing?: Almost always one of two things: the processor or graphics card reached a power or temperature limit and stepped its clock down, or something else is now sharing the cores. A sustained-load test separates the two: a limit shows as a curve with a shape, contention shows as noise. The report names which. Gamers →
Can a browser test measure gaming performance?: It measures the thing that decides gaming performance over time, which is how much of its peak a machine holds when hot. It cannot measure a specific game's frame rate; use the game's own counter for that and this test for why the counter fell. Gamers →
Should I run on battery or plugged in?: Plugged in, unless you play on battery. On battery most laptops apply a lower power limit, and the report flags a run on battery so the two are never compared as equals. Gamers →
Can I run this on every machine that comes through?: Yes. There is nothing to install, nothing to uninstall afterwards, and no account, so it does not add a step to the wipe-and-reimage routine. Open the browser, run it, and the machine you hand over is exactly as you imaged it. Refurbishers and resellers →
Does a good result actually help me sell?: It changes what the listing is arguing about. “Runs great” is a claim; a report a buyer can open on their own phone and read for themselves is evidence. It is most valuable on the machines that are genuinely good, which are the ones currently being priced as though they might not be. Refurbishers and resellers →
What should I do about a machine that fails?: Look at the shape before you write it off. A steady slide is usually thermal paste and dust, which is cheap to fix and turns a marginal unit into a good one. A sawtooth or a cliff is a cooling fault, and those are worth diagnosing before you spend money on parts. Refurbishers and resellers →
Will a buyer trust a test I ran myself?: More than they trust an adjective you wrote yourself. It is a report generated by a third party, in plain words, that they can open without installing anything. It does not make you unimpeachable; it moves the conversation from what you say to what the machine did. Selling a laptop privately →
What if the result is bad?: Then you have found out before your buyer did. A disclosed flaw costs you some of the price; an undisclosed one costs you the sale, the return, and the rating. Selling a laptop privately →
Is it free?: Running the test is free, both lengths. A report unlock is optional and is about getting the cause and the fix plan; the curve and the headline figures you would want for a listing do not need it. Selling a laptop privately →
Why not just use the tools I already have?: Keep them. This one has a specific property they do not: the customer can open the result themselves, on their own phone, and read it in plain words. It is not a better measurement than a sensor log. It is a more explainable one. Repair shops →
Can I run it on a machine I have not imaged yet?: Yes, that is the point of it running in a browser. Nothing is installed and nothing is left behind, so an intake test does not change the state of a machine you may need to hand back untouched. Repair shops →
Does it work on a machine that will not hold a charge?: Run it plugged in. On battery most machines apply a lower power limit and the report flags that, so a battery run and a mains run are never compared as if they were the same test. Repair shops →
How do I use this across hundreds of machines without an agent?: Most teams do it as part of something a person is already doing at the desk: a support visit, an OS upgrade, a hardware refresh survey. It costs the technician three minutes and no install, which is usually the difference between a check that happens and one that does not. Fleets and contact centres →
Our machines are all the same age. Why would they differ?: Because ageing is environmental, not chronological. Dust, ambient temperature, how a desk is positioned, whether a vent has been against a partition for two years. Identical purchase orders produce very different machines by year three, and the spread is the useful signal. Fleets and contact centres →
Is a low score a reason to replace a machine?: Not on its own, and this is the mistake worth avoiding. A machine that slides badly is often a machine that needs an hour of cleaning, which is far cheaper than a replacement. Look at the shape first: a slide is usually serviceable, a hard limit usually is not. Fleets and contact centres →
Is my laptop faulty, or is it just hot here?: Run the same test twice: once at the coolest hour you can manage, once in the heat of the afternoon. A machine that holds up in the morning and slides in the afternoon is not faulty, it is thermally limited by the room. A machine that slides in both is worth opening up. Hot rooms and hot countries →
Does a cooling pad actually help?: Sometimes, and this will tell you whether it did on your machine rather than in general. Run it once without, once with, and compare the settled level. That is a real answer for your laptop instead of an argument in a forum thread. Hot rooms and hot countries →
Should I just use it on battery to keep it cooler?: No. On battery most machines drop their power limit, so you get a cooler machine that is slower for a different reason, and the report flags battery runs precisely so the two are never confused. Hot rooms and hot countries →
Can I run this on a university-managed laptop?: Usually yes, because there is nothing to install. It runs in the browser tab, which is why it works on machines where you are not an administrator and cannot approve software. Students →
Is a second-hand laptop a bad idea?: Not at all, and the test exists partly to make second-hand buying safer. A used machine that holds its performance is often better value than a new budget one that does not. Test before the return window closes. Students →
The result looks bad. Do I need a new laptop?: Check the shape first. A steady slide is very often dust and old thermal paste on a machine that has had three years of student life, and cleaning it costs almost nothing compared with replacing it. Students →
Can I run this on a locked-down work laptop?: Usually, because there is nothing to install and nothing to approve. It runs in a browser tab. Company policy is still company policy, but a web page does not need administrator rights. Remote and hybrid workers →
How do I make IT take this seriously?: Send the report link rather than a description. “Slow in the afternoons” is a feeling; a curve showing the machine holding barely half of its opening throughput after three minutes is a measurement, and service desks are built to act on measurements. Remote and hybrid workers →
Is it my laptop or the video call software?: Run the test with everything closed. If the machine slides on its own, the machine is the limit and the software is only the first thing to notice. If it is flat, look at the software and the network. Remote and hybrid workers →
Is my laptop throttling?: Run the test with the laptop plugged in and quiet. If throughput keeps falling, by more than about 15%, after its boost ends and it settles, and the curve slides rather than steps, that is consistent with the cooling limiting it. A short boost and a clean single step down early in the run is normal and consistent with a power limit; the report shows it as the boost step and does not count it as decay. A sensor tool confirms which. Laptop owners →
Why is my laptop slower on battery?: Most laptops switch to a lower sustained power limit on battery to protect the cell. The report flags a run on battery; run once plugged in and once not, and the before/after comparison shows the size of the gap on your machine. Laptop owners →
Does a laptop cooling pad help?: It helps when the curve is a slide, because a slide is consistent with the cooler being the limit. It does nothing for a step, which is consistent with a wattage limit set in firmware. The shape tells you whether to spend the money, and a run before and after confirms it. Laptop owners →
Why does my export slow down partway through?: The machine reached a sustained limit, thermal or power, and settled at a lower level than it started at. The test's curve shows when that happened and which kind of limit the shape is consistent with; a sensor tool confirms which. If the curve is noisy rather than shaped, something else is competing for the cores during the export. Video and photo editors →
Is memory bandwidth important for editing?: Yes, for anything that streams large frames or raw files. The memory stage reports bandwidth from one thread and from all threads, and the latency to memory and to cache, so a slow or single-channel memory configuration shows up. Video and photo editors →
Can this test my GPU encoder?: No. The GPU stages load the raster pipeline and a path tracer, which tells you whether the card holds its performance under sustained load. Encoder throughput needs the editing application's own benchmark. Video and photo editors →
Why is my build slower than the benchmark suggests?: Build times reflect sustained all-core throughput and memory latency, not the single-core boost most benchmarks headline. The test reports both, and the curve shows whether the machine holds its all-core level for the length of a build or steps down after the boost window. Developers →
Can I use this to compare two machines?: Yes, on the measured figures: sustained throughput, decay, memory bandwidth and latency, frame time. Run each machine twice so the report can state the spread; a difference inside the spread is noise. Developers →
Does Docker or a VM affect the result?: Anything using the cores during the run lowers throughput and raises noise, and the report flags starved workers when it can see it. Quit what you can before a baseline, then run again with the usual load to measure the cost of it. Developers →
Does it need admin rights or an install?: No. It runs in the browser tab with Web Workers and WebGL or WebGPU. A strict content security policy can block workers, in which case the report says it fell back to a single-core measurement rather than reporting a wrong number. IT and support teams →
Can users share the result with us?: The report prints to PDF, and where the deployment has email configured the user can send themselves a link. The link opens the same report, with the measurements and the diagnosis. IT and support teams →
Will a security agent affect the result?: If it uses the cores during the run, yes: throughput falls and noise rises, and the report flags starved workers when the effect is large. Comparing against a known-clean machine of the same model isolates the agent's cost. IT and support teams →
How do I test a used laptop before buying?: Open systemcheck.live/benchmark-test in the seller's browser, plug the laptop in, close other windows, and run the standard test. Read the first screen: is the cooling good, and what is the machine good for. A slide or a large decay figure is consistent with cooling that needs work. Buying a used PC →
Can I run it without installing anything on their machine?: Yes. It runs entirely in the browser tab, stores nothing on the machine unless you ask for a saved run, and leaves nothing behind when the tab closes. Buying a used PC →
What decay figure is acceptable on a used machine?: Under about 8% is ordinary settling for a healthy machine. Between 8 and 15% is a mild loss, common on laptops and on desktops with a modest cooler. The figure starts after the boost, so a short burst and one step down early in the run do not count against it. Above 15% with a sliding curve, the cooling is the limit and the machine will need attention. Buying a used PC →
How do I check a new PC build is cooling properly?: Run the deep test. A healthy build holds within a few percent of its peak for three minutes. A curve that slides is consistent with a cooler or airflow problem; a curve that steps once is consistent with a power limit, which may be the board's default rather than a fault. New PC builds →
How can I tell if my RAM is running at the right speed?: The memory stage reports bandwidth from one thread and from all threads. A kit running at its fallback speed shows markedly lower bandwidth than the same kit at its rated profile; compare against the figure after enabling the profile in the BIOS. New PC builds →
Should I save the first run?: Yes. The browser keeps the last five reports, and any later run on the same processor is compared against the previous one, curve over curve. The first run on a new build is the baseline every change is measured against. New PC builds →

Guides

Can SystemCheck tell me my CPU temperature?: No. Browsers expose no temperature API, so no browser-based tool can read your thermal sensors. SystemCheck infers thermal behaviour from throughput decay over a sustained load. For actual degrees Celsius, use HWiNFO64, HWMonitor, or your vendor's utility. How to diagnose and fix thermal throttling →
Is thermal throttling damaging my CPU?: No. Throttling is the protection mechanism that prevents damage. Modern chips are designed to run at their temperature limit indefinitely. Sustained high temperatures accelerate the ageing of the thermal interface material and nearby components, but the chip itself is not being harmed by hitting the limit its firmware enforces. How to diagnose and fix thermal throttling →
What decay percentage is normal?: SystemCheck has no population figure for normal yet. It reads decay after the boost against lines it chose, pending calibration: under 8% is treated as ordinary settling, 8-15% is mentioned, 15% and above is called a sustained loss, and 30% and above caps the grade at C. The methodology page explains where each line comes from. How to diagnose and fix thermal throttling →
My decay is high but the machine feels fast. Should I care?: Only if your workloads are long. Decay is invisible in short bursts: browsing, launching applications, compiling a small project. It becomes the dominant factor in video export, long compiles, extended gaming sessions, and anything else that keeps the chip loaded for minutes at a time. How to diagnose and fix thermal throttling →
How many times should I run a benchmark?: Three, with a couple of minutes of idle between runs so the machine starts from a comparable thermal state. If the three results fall inside a 3% band, the middle one represents your machine. If they are spread wider than that, the environment is unstable and the results should not be used for comparison. How to read a benchmark result without fooling yourself →
Why is my score lower than someone with the same CPU?: Cooling, power limits, background load, browser, and GPU backend all move results, and identical part numbers do not imply identical sustained behaviour. A laptop and a desktop with the same CPU model can differ by 40% or more in sustained throughput because their power budgets are different. How to read a benchmark result without fooling yourself →
Should I use peak or sustained throughput to compare machines?: Sustained, unless your actual workloads are all short bursts. Peak measures the best few seconds under ideal thermal conditions; sustained measures the equilibrium the machine settles into, which is where any workload longer than about ten seconds spends its time. How to read a benchmark result without fooling yourself →
Is having a bottleneck bad?: No. Every system is limited by something; a bottleneck simply names which component sets the ceiling. It only becomes a problem when the imbalance is large enough that one component is substantially idle while the other is saturated, or when the limiting component is not the one you paid to prioritise. How to tell whether your CPU or GPU is the bottleneck →
Can a browser benchmark detect my GPU utilisation?: No. There is no web API for GPU utilisation, VRAM capacity, or power draw. SystemCheck derives its bottleneck indication by measuring CPU and GPU throughput separately under loads designed to saturate each and comparing them, not by reading utilisation counters. For real-time utilisation, use MSI Afterburner or your vendor's overlay. How to tell whether your CPU or GPU is the bottleneck →
Why does my bottleneck change between games?: Because it is a property of the workload, not of the machine. A strategy game simulating thousands of units is CPU-heavy; the same PC running a graphically dense shooter at 4K is GPU-heavy. Resolution alone shifts the balance point, which is why the resolution test works. How to tell whether your CPU or GPU is the bottleneck →
Is undervolting risky?: It carries no risk of hardware damage, because you are reducing voltage rather than increasing it. The risk is instability: too aggressive an offset causes crashes or errors under load. Move in small steps, validate each one under a sustained load, and back off one step from the first instability you see. How to improve sustained performance, not just peak scores →
Why would lowering a power limit make my machine faster?: Because frequency versus power is non-linear at the top of the curve, and the last few hundred megahertz can cost a third of total package power. A chip pushed there heats fast, throttles, and settles below where a capped chip would have held steadily. You lose a little peak and gain sustained throughput. How to improve sustained performance, not just peak scores →
How much sustained performance can I gain without buying anything?: It depends entirely on how the machine was set up, but airflow, fan curves, power plans, and background contention together account for the majority of real-world sustained shortfalls. Machines that have been in service for a few years without maintenance are where the largest free gains live. How to improve sustained performance, not just peak scores →
Can a browser benchmark read my CPU temperature?: No. There is no thermal sensor API on the web platform, in any browser, with any permission. SystemCheck infers thermal behaviour from throughput decay under sustained load and says so explicitly. Any browser tool that displays a temperature in degrees is estimating or fabricating it. How accurate is a browser benchmark, really? →
Why do I get different scores in Chrome and Firefox?: Different JavaScript engines make different JIT decisions, and the browsers may select different GPU backends and driver paths. Double-digit percentage differences on the same hardware are normal. Cross-browser comparison is not a hardware comparison. Hold the browser constant when comparing machines. How accurate is a browser benchmark, really? →
Is a browser benchmark accurate enough to be useful?: For relative measurement on one machine, yes. Sustained throughput, decay, and consistency are measured directly and repeat reliably under controlled conditions. For absolute hardware capability or cross-machine ranking, treat the result as directional and use a native benchmark when precision matters. How accurate is a browser benchmark, really? →
Why not just build a native app instead?: Native tools already exist and are better at absolute measurement, and SystemCheck points to them where they are the right answer. What a browser adds is that it requires no download, no elevated privileges, and no trust decision about running an unfamiliar binary, which for many people is the difference between measuring and guessing. How accurate is a browser benchmark, really? →
Are peak benchmark scores useless?: No, they accurately describe short, bursty workloads, which is a large share of interactive computing. They become misleading when used to predict long workloads, because the machine leaves the boost window and never returns to it during the job. Why peak benchmarks mislead →
How long does a benchmark need to run to be meaningful?: Long enough for the system to reach thermal and power equilibrium, which on most machines takes between thirty seconds and two minutes of genuinely saturating load. A measurement that ends before equilibrium is measuring the boost policy, not the sustained capability. Why peak benchmarks mislead →
Will browsers ever expose temperature sensors?: It is unlikely. Sensor access is a fingerprinting surface and a side-channel risk, and the platform's consistent direction has been to reduce rather than expand hardware-identifying APIs. Assume it will not happen and plan measurement around behaviour instead. What browser benchmarks can and cannot measure →
Can a web page tell how much VRAM my GPU has?: No. WebGPU exposes adapter limits, which are specification-defined ceilings a device must meet, not the physical memory installed on the card. A page cannot distinguish an 8GB card from a 24GB one on that basis. What browser benchmarks can and cannot measure →
If a browser cannot read temperature, how does SystemCheck report thermal behaviour?: By measuring throughput continuously under sustained load and reporting how far it decays from its early peak. Decay is a behavioural consequence of thermal and power limiting. It tells you the machine slowed and by how much; it does not tell you what temperature it reached. What browser benchmarks can and cannot measure →
Does SystemCheck measure my temperature to calculate decay?: No. Decay is computed entirely from throughput over time. Browsers expose no thermal sensor API, so no web page can read temperature. Decay is a behavioural proxy for thermal and power limiting, and SystemCheck reports it as exactly that. Thermal decay, explained →
What decay percentage should I aim for?: Aim for high sustained throughput rather than for a low decay percentage. The report treats decay after the boost under 8% as ordinary settling and calls 15% and above a sustained loss, but these are lines SystemCheck chose, pending calibration, not typical values. A low ratio achieved by starting slower is not an improvement. Compare absolute sustained throughput before and after any change. Thermal decay, explained →
My decay is high only on battery. Why?: That is a power limit rather than a thermal one. Many laptops enforce a substantially lower sustained power budget on battery, and the signature is a fast drop to a hard plateau within seconds rather than a gradual slope. Cooling changes will not affect it; plugging in will. Thermal decay, explained →
Is lower latency or higher bandwidth more important?: It depends on the work. Large streaming jobs lean on bandwidth; work that follows pointers from one value to the next leans on latency. SystemCheck reports both and does not combine them into one memory figure for that reason. Memory latency vs bandwidth: what a random-walk test shows →
Why does the latency figure look higher than my memory kit's rating?: A rated latency describes the memory chips alone. The walk measures a whole round trip from a browser worker, including address translation and the sandbox's overhead, so it is not comparable with a rated figure. Compare it with other runs measured the same way. Memory latency vs bandwidth: what a random-walk test shows →
Can SystemCheck tell whether I am running in dual channel?: No. A browser cannot read the channel count. A low all-thread bandwidth that barely rises over one thread is worth checking in a native tool, but the browser figure alone does not prove how the memory is installed. Memory latency vs bandwidth: what a random-walk test shows →
Why only up to six threads?: Six workers at 48 MiB each already move far more data than a cache holds, and the cap keeps the stage short. A machine with fewer logical processors uses one worker per processor. Single-thread vs all-thread memory bandwidth →
My all-thread bandwidth changed between two runs. Is something wrong?: Not necessarily. Background work competes for the same memory. Each bandwidth figure carries an interval for the spread inside its run; compare two runs only when their conditions match, and repeat before reading anything into a small change. Single-thread vs all-thread memory bandwidth →
Will running again fill in a not-measured figure?: Often, if the reason was a hidden tab, a lost focus or a stage that stopped early. Keep the tab in front and the window focused for the whole run. If the reason was the browser itself, try a different current browser. What "not measured" means in your report →
Why does SystemCheck not show my CPU temperature?: Browsers expose no temperature sensor to web pages, so no browser tool can read it. SystemCheck measures throughput over time instead and says so. What "not measured" means in your report →
My two runs' intervals do not overlap. Did something change?: Possibly, but first check the conditions: power source, background apps, how warm the machine was and whether the tab stayed in front. The interval covers the spread inside a run, not the change between runs, so different conditions can move the figure outside it. Reading the confidence interval on every figure →
Why does the interval sit off-centre around the figure?: A bootstrap interval follows the shape of the data rather than assuming a symmetric spread, so it can extend further on one side. Reading the confidence interval on every figure →
Is a boost step a fault?: No. Running above the sustained level for a short time is ordinary behaviour for many laptops. SystemCheck shows the step so you can see it, and leaves it out of the score. Your boost is not your speed: what the first 10 seconds hide →
Can SystemCheck tell why the step happened?: Not from a browser. A web page sees no clock, power or temperature reading, so the report describes the shape and leaves the cause to a sensor tool. Your boost is not your speed: what the first 10 seconds hide →
Is a browser benchmark accurate enough to compare two runs?: On the laptop measured, the CPU sustained figure and the overall score repeated within a few percent. Compare runs under matching conditions, repeat each, and treat small differences as noise until a repeat confirms them. Ten runs on one laptop: how repeatable is a browser test? →
Why was the outlying run kept?: Because nothing measured explained it. Dropping a run because it looks wrong would hide exactly the kind of spread a reader needs to know about. Ten runs on one laptop: how repeatable is a browser test? →
Is WebAssembly always about five times faster than JavaScript?: No. The ratio here is for one kernel, on one laptop, in one browser. How much faster WebAssembly is depends on the code, and modern JavaScript engines run many workloads close to it. Why the WebAssembly kernel scores 5.4x the JavaScript one →
What if my browser blocks WebAssembly?: The workers fall back to the JavaScript kernel and the report says so. Its figures are read on the JavaScript kernel's own scale. Why the WebAssembly kernel scores 5.4x the JavaScript one →
Is 80C too hot for a laptop CPU under load?: It depends on that chip's published maximum. Many laptop processors are designed to run warm under sustained load. If the machine also holds its performance, the temperature is doing what the design allows. Is 80C safe for a CPU? What a browser can tell you →
Can SystemCheck show my CPU temperature?: No. Browsers expose no temperature sensor to web pages. SystemCheck measures throughput over time instead and says so in every report. Is 80C safe for a CPU? What a browser can tell you →
What should I do if performance drops while the machine is hot?: Check the vents for dust and blockage, use the machine on a hard surface, and compare a run before and after any change under the same conditions. Is 80C safe for a CPU? What a browser can tell you →
Is my Mac slower in Safari?: Not the machine. The browser admits fewer threads to the page, so the multi-thread figure covers fewer cores. Single-thread work is measured the same way. Mac in Safari: why only 8 threads show →
Should I test my Mac in Chrome instead?: If you want the multi-thread figure to use every logical processor, a browser that reports the full count will. Keep each comparison within one browser. Mac in Safari: why only 8 threads show →
Can a browser stress test damage my PC?: It runs ordinary computation inside the browser's limits, the same kind of load as heavy everyday work. Processors protect themselves by slowing down near their limits. Is a browser stress test as good as Prime95 or Cinebench? →
Can I compare my SystemCheck score with a Cinebench score?: No. They measure different work in different environments. Compare SystemCheck runs with other SystemCheck runs made under the same revision and conditions. Is a browser stress test as good as Prime95 or Cinebench? →
Do processors wear out and get slower?: Not in a way you would notice in normal use. A processor runs at the speed its firmware allows for the conditions. What changes over time is the load and the cooling around it. Why a PC slows down over time, and how to tell which kind →
Can SystemCheck tell me why my PC is slow?: It can tell you whether the machine holds its level under sustained load and what shape its curve has. It cannot read temperatures, clocks or startup programs from a browser, so it names what the shape is consistent with and leaves confirmation to a sensor tool. Why a PC slows down over time, and how to tell which kind →
Is 3 minutes enough to find a fault?: It is enough to see whether the machine holds its level under sustained load and completes every stage. Intermittent faults can need much longer, which is what the Deep run is for if the seller allows it. How to check a used laptop before you pay →
Should I test on battery?: Test plugged in first, so the result is not limited by battery power. If battery life matters to you, read the battery report as well; a stress test is not a battery capacity test. How to check a used laptop before you pay →
Can SystemCheck read my exact processor and graphics card?: No. Browsers report the logical processor count, which can be capped, and an approximate graphics renderer name that is increasingly masked. Your operating system's About page names the parts. PC specs and performance: what a test's numbers mean →
Is a higher boost clock always faster?: Only for as long as the machine can hold it. Over sustained work the cooling and power limits decide the level, which is why SystemCheck reports the settled level apart from the boost. PC specs and performance: what a test's numbers mean →
Is a 10 minute GPU stress test enough?: For a health check it is usually long enough to see whether the card settles and holds. For overclock stability it is not; test with your own games for longer. GPU stress test: how long is long enough? →
Can a GPU stress test damage my card?: A working card protects itself by lowering its clocks when it reaches its limits. A test only adds the load a demanding game does. Stop if you see artefacts, crashes or smell burning. GPU stress test: how long is long enough? →
Can a browser benchmark tell me my FPS in a game?: No. It runs different work from any game. It tells you whether the hardware holds its level and paces frames evenly, which is what decides whether the machine is the limit. How to benchmark a PC for gaming without installing →
Does a browser test use my graphics card?: Yes, through WebGL2, provided hardware acceleration is on in the browser's settings. On a laptop with two graphics chips, the browser may use the integrated one unless the system is told otherwise. How to benchmark a PC for gaming without installing →
Is higher FPS always smoother?: No. Higher FPS shortens the gap between frames, but uneven delivery still shows as stutter. Even pacing at a lower rate can look smoother than uneven pacing at a higher one. Frame pacing vs FPS: why smooth is not the same as fast →
Is frame pacing the same as input lag?: No. End-to-end input lag (from the device to the screen) is the delay between an action and its result on screen, which a web page cannot measure; a page sees only the browser's part of that chain. Pacing is whether frames arrive evenly once they are drawn. Frame pacing vs FPS: why smooth is not the same as fast →
Is a GPU bottleneck bad?: No. In a game, the GPU being the limit is the usual and expected case, because the processor is rarely asked to do as much per frame. It only matters if the frame rate is lower than you want. How to reduce a GPU bottleneck once you have measured it →
Will a faster processor fix a GPU bottleneck?: No. If the graphics card finishes last, a faster processor makes it wait longer. Lower the GPU's load or upgrade the GPU. How to reduce a GPU bottleneck once you have measured it →
Is CPU throttling bad for the processor?: No. Throttling is what keeps the chip inside its safe limits. Frequent heavy throttling means the cooling or the power setting is the ceiling on your sustained speed, which is a performance problem rather than a safety one. What CPU throttling means, read off one chart →
Does a falling line on SystemCheck prove throttling?: No. It shows the machine did less work later than earlier. Throttling is one cause consistent with that shape; a sensor log taken during the same run is what confirms it. What CPU throttling means, read off one chart →
Why is my GFLOP/s much lower than my GPU's specification?: A specification sheet's figure is a theoretical peak. This test measures one dependent workload through a browser, including the browser's overhead, so it is expected to be lower. It is most useful compared with another run of the same test. What the WebGPU compute test measures, and what it does not →
Does the WebGPU test change my SystemCheck score?: No. It runs on its own page, is not a stage of the Standard or Deep run, and contributes nothing to the score. What the WebGPU compute test measures, and what it does not →
Will the sustained write test wear out my SSD?: It writes up to 8 GiB once and deletes it. That is a small amount next to the total writes a drive is built for, but it is not nothing, so there is no reason to run it repeatedly. Testing an SSD's sustained write speed in a browser →
Why is the browser's figure lower than my drive's rating?: The rating is usually measured with a dedicated tool writing straight to the drive, often into its cache. The browser adds its own storage layer, so its figure is what a web app can get, which is a different thing. Testing an SSD's sustained write speed in a browser →
Is this the same as a mouse or monitor input lag test?: No. It measures the delay the computer adds inside the browser under load. The mouse's polling, USB and the monitor's processing are outside what a web page can time. Responsiveness test: input delay your PC adds under load →
Does the responsiveness test affect my score?: No. It runs on its own page, is not a stage of the Standard or Deep run, and contributes nothing to the score. Responsiveness test: input delay your PC adds under load →
How long should I wait between stress test runs?: Long enough for the machine to return to the state you want to measure. For a fair comparison, start each run from the same state: either from idle after the same break, or straight after the same load. Recovery test: how fast a machine comes back after load →
Does the recovery test change my score?: No. It runs on its own page, is not a stage of the Standard or Deep run, and contributes nothing to the score. Recovery test: how fast a machine comes back after load →
Will a SystemCheck report be accepted for a warranty claim?: That depends on the maker. The report is a checkable record of measured performance under stated conditions. Most warranty desks also run their own diagnostics, so treat it as supporting evidence. How to document laptop throttling for a warranty claim →
Is throttling a warranty fault?: Not by itself. Every modern processor throttles to stay inside its limits. It may be a fault when a machine falls far below what the same model does under the same conditions, for example from a failed fan or poor contact between the chip and its cooler. The maker judges that. How to document laptop throttling for a warranty claim →
Does repasting a CPU improve performance?: Only if heat was the limit. When old paste had dried or a fan was clogged, the chip may hold a higher level for longer afterwards. When a power limit or the maker's settings were the ceiling, cooler running will not raise the score. Before and after: measuring a repaste or a fan clean →
How soon after repasting should I test?: Some pastes are described by their makers as settling over the first heat cycles. A fair approach is to test once straight away and again after a few days of normal use, under the same conditions. Before and after: measuring a repaste or a fan clean →
If I pass SystemCheck's exam profile, will my exam work?: It means your camera, microphone, connection, screen and browser worked in this browser when you ran it. Run the exam provider's own check too, on the same machine and network you will use on the day. Online exam system check: what each check actually tests →
Why was a check left out on my device?: A check that cannot apply is skipped rather than failed. The screen check is not skipped: in a browser that cannot share its screen it records screen sharing as Not available and still judges the screen's size and the number of displays. Online exam system check: what each check actually tests →
What frame rate should a webcam have for video calls?: SystemCheck checks for at least 15 frames a second at 640 by 480, and for smooth delivery while you move. Many calling apps send less than the camera can capture when the connection is slow, so a good camera alone does not guarantee a smooth picture at the other end. The webcam, microphone and connection lines we use for calls →
Why does my webcam's frame rate drop in a dark room?: Many webcams lengthen each frame's exposure in low light to gather more of it, which lowers the frame rate. More light on your face is the usual fix. The webcam, microphone and connection lines we use for calls →
Why is my laptop slower on battery?: Most laptops apply a lower sustained power limit and a lower boost ceiling on battery, and the operating system's power mode can lower them further. A browser cannot read that policy, but it can measure how much work the machine does under each. Battery or plugged in: what changes in a laptop's run →
Should I run a benchmark on battery?: Only to answer a battery question, such as how a laptop performs away from the desk. For a figure to compare with later runs or with other machines, plug in and record the power mode. Battery or plugged in: what changes in a laptop's run →
Is thermal throttling the same as power limit throttling?: No. Thermal throttling lowers the clock to stay under a temperature limit. A power limit lowers it to stay under a power budget, often after a short boost allowance runs out, whatever the temperature. Both make a curve fall, which is why a curve alone cannot tell them apart for certain. Heat, a power limit or a busy machine: what a curve shows →
Can a browser test tell me which one my laptop has?: It can say which shapes the curve fits and whether another program was sharing the cores. It cannot read temperatures, power or the processor's limit reasons, so it names what a shape is consistent with and leaves the cause to a sensor log. Heat, a power limit or a busy machine: what a curve shows →
Is bufferbloat my provider's fault?: Usually not. The queue that fills is most often in the home router or modem, on the upload side. The provider's line speed is rarely the cause on its own, though a provider's own router may lack the setting that fixes it. Bufferbloat: why your ping rises when the line is busy →
Does a faster plan fix bufferbloat?: Not by itself. A faster line fills more slowly, but a deep queue still adds delay once it is full. Smart Queue Management keeps the queue short at any speed. Bufferbloat: why your ping rises when the line is busy →
What is a good jitter for gaming and calls?: As low as you can get it. Our warn line is 50 ms, and VoIP planning guides commonly use 30 ms for voice. Single-digit jitter on a cable is common. Jitter: what an uneven ping does to calls and games →
Can I have low ping and high jitter?: Yes. A line can average a low ping while individual round trips swing widely, and the swings are what freeze video and make games stutter. That is why the test reports jitter beside the median. Jitter: what an uneven ping does to calls and games →
Why is Bluetooth delay fine on YouTube but bad in games?: A video player knows the delay and holds the picture back to match. A game or a call is live, so nothing can be held back and the whole delay shows. Bluetooth headphone delay: measure it, then cut it →
Can a website tell which Bluetooth codec I am using?: No. The browser is not told which codec the link uses. The system's Bluetooth settings, its developer options or the earbuds' own app may show it. Bluetooth headphone delay: measure it, then cut it →
Can headphones play 5.1 surround?: Stereo headphones have two drivers. Virtual surround software can simulate direction on them, but a test that plays real channels will find two. Check 5.1 and 7.1 surround on Windows and macOS →
Why does the test say only two channels when my receiver is 5.1?: The system is giving the browser stereo. Set the output to 5.1 or 7.1 in the system's sound settings, check the receiver accepts multichannel PCM over HDMI, and reload. Check 5.1 and 7.1 surround on Windows and macOS →
Should hardware acceleration be on or off?: On, for nearly everyone. Off makes the processor draw everything, which is slower and uses more power. Turn it off only to test whether a graphics driver is causing a crash, then update the driver and turn it back on. Is hardware acceleration on? A check for each browser →
Does this check video decoding too?: No. It reads the drawing path. Whether the browser decodes AV1, HEVC, VP9 or H.264 on the graphics chip is a separate question, which the video decode test answers. Is hardware acceleration on? A check for each browser →
Is WebGPU faster than WebGL?: It can be, because it does more work per call and allows compute shaders, but the same chip draws both. The speed you see depends on the page and the browser's implementation more than on the name of the API. WebGPU vs WebGL: what each is and which one you have →
Why does WebGPU say no adapter found?: The browser has WebGPU but would not give an adapter: usually a blocked graphics driver, acceleration switched off, or a remote or virtual machine with no usable graphics chip. The browser's diagnostic page gives the reason. WebGPU vs WebGL: what each is and which one you have →
Do I need a wide-gamut screen?: Only for work or play made in a wide gamut: photo editing for print or for P3 devices, HDR video and some games. Most web pages and documents are sRGB, and look the same on either. sRGB, Display P3 and Rec. 2020: what a browser can tell →
Is a wide colour gamut the same as HDR?: No. Gamut is how saturated the colours can be; HDR is how bright and dark the picture can go. A screen can be wide gamut without HDR. The HDR test reads whether HDR is on. sRGB, Display P3 and Rec. 2020: what a browser can tell →
Why does my mouse measure a different DPI from its setting?: A sensor can deviate from its nominal setting, and the measured distance carries your own error. Measure several passes on a straight edge before reading a few percent as the sensor's. How to find your mouse's real DPI →
Does the test measure polling rate too?: No. Polling rate is how often the mouse reports, which the mouse test measures. DPI is how far each count is; the two are separate settings. How to find your mouse's real DPI →
Does the operating system's video player use the same decoder as the browser?: Not always. A native player can use a different decoder path, so a codec the browser decodes in software may play in hardware there, and the other way round. The test reports what the browser offers. How to check if your GPU decodes AV1 in hardware →
Why would AV1 decode in software on a new laptop?: Hardware acceleration may be off in the browser, the graphics driver may be blocked or old, or on Windows the AV1 Video Extension may be missing. The hardware acceleration test shows the first two. How to check if your GPU decodes AV1 in hardware →
Can a website read how much VRAM my graphics card has?: No. WebGPU and WebGL report limits such as the largest buffer, not installed memory. A page learns what it can use only by allocating, which this test does with your permission and a cap. How much VRAM can a browser use, and why not all of it →
Is it safe to allocate graphics memory from a browser?: The test stops at the cap you set and at the first error, and frees everything when it ends. The browser itself may reset the graphics device if memory runs out, which the result reports. How much VRAM can a browser use, and why not all of it →
Does system RAM count?: On an integrated chip or Apple silicon the graphics memory is system RAM, so yes, up to what the system lets the chip use. On a discrete card, runtimes can split a model between the card and RAM, which fits more and runs slower. Can my laptop run a local LLM? The memory arithmetic →
Why not estimate tokens per second from the bandwidth?: Because the runtime, the model's design, the context length and the browser all change it, and none of them can be known from this page. A number would be a guess. Can my laptop run a local LLM? The memory arithmetic →
What do I do if the drive fails partway?: Do not store anything on it you cannot lose. Format it and run the full check again; if it fails at the same point, its real capacity ends there. Failing flash can lose data too, so a repeat at a different point is also a reason to stop using it. How to spot a fake USB drive or SD card →
Why does the test need Chrome or Edge?: It writes to a folder you pick through the File System Access API, which Chrome and Edge offer on a computer and Firefox and Safari do not. How to spot a fake USB drive or SD card →
Can a website tell my battery's health or wear level?: No. The Battery Status API gives only the level and whether it is charging. Wear needs the design and full-charge capacity, which the operating system reports: on Windows, powercfg /batteryreport writes both. How long will my laptop battery last under load? →
Why does the test stop when I plug in?: A rate is only a drain rate while the battery is draining. Charging, even briefly, breaks the staircase the rate is worked out from, so the test asks you to unplug and start again. How long will my laptop battery last under load? →
Why does my controller read 250 Hz in the browser but more in other tools?: Because Chromium polls gamepads every 4 ms, so the Gamepad API cannot deliver updates faster than about 250 Hz. Use the WebHID reading, where the controller offers HID, for its own rate. What controller polling rate means, and what a browser sees →
Does overclocking a controller's polling rate help?: It shortens the average wait for a report by half the change in interval, which is a millisecond or two at most. Measure before and after on the same link; the spread can matter more than the median. What controller polling rate means, and what a browser sees →
Why does my 1080p webcam look soft in calls?: Call apps often send less than the camera's size, starting small and stepping up as the connection allows, and lower it again when the processor is short. The call section of the test shows what this computer would send and why. What resolution is my webcam really sending? →
Does the test record or upload video?: No. The call runs between two connections inside the page and nothing leaves it. What resolution is my webcam really sending? →
Why is my reaction time different on another site?: Each site times from a different point, draws differently, and runs on whatever screen and mouse you have. Two figures from different sites or setups are not the same measurement. What a reaction time test actually measures →
Can a browser measure end-to-end input latency?: No. From the screen changing to the click arriving needs a photodiode on the screen and a sensor on the button. A page can measure only the parts it times itself. What a reaction time test actually measures →
How does path tracing work in simple terms?: It follows light backwards. For each pixel a ray leaves the camera, bounces off surfaces in random directions, and stops when it finds a light. The light it gathered is one sample for that pixel. Many samples are averaged to make a clean image. How path tracing works, from one ray to a finished image →
Why is a path traced image noisy at first?: Each sample takes a random path, so a single sample may or may not find the light. Averaging more samples per pixel smooths the randomness out. Halving the noise takes four times as many samples, so the image improves quickly at first and slowly afterwards. How path tracing works, from one ray to a finished image →
What rules does path tracing follow?: Light is conserved, so a surface never reflects more than it receives. Each material has a scattering function that decides where a bounce is likely to go, and directions are sampled in proportion to it. Paths are ended at random with the survivors weighted up, so the average stays correct. How path tracing works, from one ray to a finished image →
Is path tracing the same as global illumination?: Global illumination is the effect: light bouncing between surfaces so that shadows soften and colours bleed. Path tracing is one way of computing it, and the most complete, because it follows every bounce rather than approximating some of them. How path tracing works, from one ray to a finished image →
Can I test my own graphics card with path tracing?: Yes. The SystemCheck run includes a path tracing stage that runs in the browser with nothing to install, and reports samples per second and how consistently they were delivered. Start the free test and the stage runs at the end. How path tracing works, from one ray to a finished image →
Is path tracing better than ray tracing?: It is more complete. Ray tracing in games usually means one or two ray traced effects on a normally rendered image. Path tracing computes all the lighting by following full light paths, so it looks more correct and costs far more. Better depends on whether your card can afford it. Path tracing vs ray tracing: what the difference actually is →
Do I need an RTX card for path tracing?: For path tracing modes in modern games, yes, in practice: they depend on dedicated ray tracing hardware plus a denoiser and upscaling. A browser path tracer is different, since it runs as a shader on any card with WebGL2, which is what lets the SystemCheck stage run on integrated graphics. Path tracing vs ray tracing: what the difference actually is →
Why is path tracing so much slower than ray tracing?: Because it replaces every lighting calculation with traced paths and needs many samples per pixel to converge, while hybrid ray tracing traces only a single bounce for one or two effects and leaves the rest to fast rasterization. Path tracing vs ray tracing: what the difference actually is →
Which is a better test of my graphics card?: Path tracing, for a sustained throughput test: the work per sample is identical on every card and there is no denoiser to hide a slow one. Hybrid ray tracing tests are closer to how a specific game behaves, but their result depends on the mix of effects in that scene. Path tracing vs ray tracing: what the difference actually is →

Against other tools

Is SystemCheck a replacement for UserBenchmark?: No. They answer different questions. UserBenchmark inventories a Windows system and places each part against a large database. SystemCheck measures how a system behaves across roughly three minutes of sustained CPU and GPU load, in a browser, on any desktop OS. If you want a component inventory, UserBenchmark is the right tool. SystemCheck vs UserBenchmark →
Why does SystemCheck not test storage?: Because a browser cannot measure it meaningfully. There is no web API for raw disk throughput, and anything we reported would be an artefact of the browser's storage layer rather than the drive. We would rather omit a number than publish a misleading one. Memory is different: workers streaming large arrays and a pointer chase through a shuffled one do measure bandwidth and round-trip latency, so the memory stage reports those, unscored. SystemCheck vs UserBenchmark →
Does SystemCheck need a download?: No. SystemCheck runs entirely in the browser tab and requires no account to run the core four-stage test. UserBenchmark is a program you download and run on the PC. SystemCheck vs UserBenchmark →
Can SystemCheck tell me my exact CPU or GPU model?: Not reliably. Browsers deliberately mask detailed hardware identifiers to limit fingerprinting, so what SystemCheck sees is an approximate renderer string. A native tool like UserBenchmark is much better at this. SystemCheck vs UserBenchmark →
Is SystemCheck as accurate as 3DMark?: No, and it does not claim to be. 3DMark runs natively against the graphics driver; SystemCheck runs through a browser's WebGL2 implementation, which adds a translation layer and makes results sensitive to browser version and background activity. SystemCheck results are directional. For a defensible figure, use 3DMark. SystemCheck vs 3DMark →
Can SystemCheck show my GPU temperature like 3DMark does?: No. The web platform exposes no API for temperature, power draw, fan speed or VRAM size, so no browser-based tool can report them. If you need to see the thermal curve, you need a native tool. SystemCheck vs 3DMark →
Why use SystemCheck at all if 3DMark exists?: Access. 3DMark requires a Windows machine, a multi-gigabyte install and, for most of its tests, a purchase. SystemCheck needs a browser tab and about three minutes, which makes it usable on machines and operating systems where 3DMark is not an option. SystemCheck vs 3DMark →
Does 3DMark also measure sustained performance?: Yes. It includes a stress test mode that loops a scene and reports a frame-rate stability percentage, and it charts hardware telemetry alongside it. Sustained measurement is not something SystemCheck does and 3DMark does not. SystemCheck's difference is that it does it without an install. SystemCheck vs 3DMark →
Is SystemCheck's CPU test the same as Cinebench's?: No. Cinebench renders a scene with a production render engine. SystemCheck runs parallel synthetic workloads in Web Workers, chosen because they behave consistently across browsers. Cinebench's workload is more representative of real rendering; SystemCheck's is more portable. SystemCheck vs Cinebench →
Which is better for testing a cooler?: Cinebench, if you can install it. Configured for an extended run it loads the CPU for far longer than SystemCheck's 60-second stage, and large air or liquid coolers can take many minutes to reach steady state. SystemCheck will show you early decay, but it will not show you a slow thermal saturation curve. SystemCheck vs Cinebench →
Can SystemCheck show me CPU temperatures during the run?: No. No browser can. There is no web API for temperature, power or fan speed. Cinebench does not report them either; for that you need a separate hardware monitoring utility. SystemCheck vs Cinebench →
Does SystemCheck test the GPU as well?: Yes. SystemCheck runs a GPU raster stage and a path-tracing stage in addition to the CPU stage. Cinebench is primarily a CPU benchmark, though recent versions include a GPU test built on Maxon's Redshift renderer. SystemCheck vs Cinebench →
Does SystemCheck work on a phone?: No. SystemCheck requires a desktop or laptop browser with WebGL2 and does not support mobile in v1. Geekbench has Android and iOS applications and is the right tool for a phone or tablet. SystemCheck vs Geekbench →
Why does SystemCheck report decay when Geekbench does not?: They are answering different questions. Geekbench's workloads are short by design, which makes them a clean measure of peak capability. SystemCheck holds a multi-core load for 60 seconds specifically to see how much of that peak survives, which is a different and narrower thing to measure. SystemCheck vs Geekbench →
Are SystemCheck and Geekbench scores comparable?: No. They use different workloads on different scales, and SystemCheck runs through a browser while Geekbench runs natively. Neither score converts into the other. SystemCheck vs Geekbench →
Which one should I trust for a purchase decision?: Geekbench, generally. Its database is large enough to place almost any device, and its native measurement is more precise. SystemCheck is more useful after the purchase, for checking that the machine in front of you behaves the way its class should under load. SystemCheck vs Geekbench →
Does SystemCheck test storage and memory like PassMark does?: No. A browser has no API for raw disk throughput or memory bandwidth, so any figure SystemCheck published would reflect the JavaScript engine rather than the hardware. If storage or memory is what you need measured, use PerformanceTest. SystemCheck vs PassMark PerformanceTest →
Is SystemCheck free?: The four-stage benchmark is free to run and needs no account. The detailed report is a one-time paid unlock. PassMark PerformanceTest's basic tests are free and do not expire, and its Advanced Test Suite is free for 30 days, by its Microsoft Store listing. SystemCheck vs PassMark PerformanceTest →
Which has the better comparison data?: PassMark, without question. Its baseline database has been accumulating for many years and its component charts are widely referenced. SystemCheck is new and its result history is correspondingly small. SystemCheck vs PassMark PerformanceTest →
Does PassMark measure sustained load?: PerformanceTest's tests are throughput measurements rather than soak tests. PassMark's separate BurnInTest product is the one aimed at sustained stability testing. SystemCheck reports decay across a sustained load as part of its standard result. SystemCheck vs PassMark PerformanceTest →

Find out why. Three minutes, in the tab you have open.

Run the test How we measure