What does TechPowerUp conisder as a "max load"?

[Kajiger09]

When they are testing CPU coolers they have idle, typical load, and max load (running aida64). What would playing a game maxed out at 1440p, a few tabs in chrome, and a word document constitute? Typical load?

 

[TheoneandonlyMrK]

Can't be that, because a higher spec pc might not be loaded adequately, that's typically why people use stress test software, I just Fold ,that's a true full load test (including endurance stability test) imho, though you do choose how much load obv.

 

[Kursah]

Welcome to TPU @Kajiger09 !!

They explain their CPU load tests in the reviews, here's an example: https://www.techpowerup.com/reviews/CRYORIG/H7_Universal/6.html

But they leave a very generalized statement beyond that, and I would really expect no less because there isn't much for a static set of what would be typical CPU load on our systems. We can assume though! And if we know what we run on our systems regularly, odds are that is going to be the best typical CPU load indicator. At least that's my mindset on the situation. Here's a quote from the review I linked above:

The idle test will consist of the CPU sitting idle at the desktop for 15 minutes. This will allow for a stable temperature reading that will be recorded at the end of those 15 minutes.

Wprime's and AIDA64's CPU test represent typical multi-threaded loads. Both offer consistent results, with one being a benchmarking application and the other a stability test. Both are run for 15 minutes before the peak reading during the test is recorded and taken as the result. This test lets enthusiasts know what temperatures they can expect to see with games and applications. Wprime is set to eight threads while AIDA64 is configured to stress the CPU, FPU, cache, and system memory.

AIDA64 offers maximum heat generation when set to stress just the FPU in the stability test, which will really push the CPU. This test represents extreme loads much like LinX, Prime95, and other extreme stress tests many users are familiar with.

So typical load for the CPU would likely be close to what you're stating depending on the game and if it requires many resources or thread utilization from the CPU.

Both Wprime and AIDA64's CPU test are not the heaviest stress tests. My CPU averages more around the 20-35% per-thread range unless I play Star Citizen or Ashes of the Singularity or Elite Dangerous, etc, etc. As they state just below the bolded statement above, the tempurature results from that testing method is what users can expect to see with games and applications. It is going to vary based on system. software, games, ambient temperature and a number of things so I don't think there is a set finite rule for what is going to be an average load on a TPU test system compared to your system or my system. Though maybe there could be a different set of tests, but then different games have different loads affected by different resolutions, settings, drivers, patches, etc. It gets messy fast and wouldn't be worth the hassle IMHO.

Best thing you can do is either duplicate their typical load tests, or compare to what your system's typical load is...which could be lighter or heavier or even the same as what TPU reviewers are using for testing. That's how I treat it, since their values are a mere guideline from a specific result in consistent testing environments I tend to run a little bit lower average load than they test it seems. I know this probably doesn't clear anything up as-far-as a TPU reviewer explanation, but its my take on it. Hope it helps!

 

[EarthDog]

I don't know how they define it (they don't tell you in the article?!!!!)... but I would call your load 'typical'.

I mean Chrome and a word document don't use much (any?) CPU after its up so that is almost not worth mentioning... I suppose there can be active things on the webpage but... its nothing. The CPU is being used for a game and so long as its a quad core or greater (quad with HT +) won't be stressed 100%.



EDIT: That's what I get for not hitting post, LOL

 

[crazyeyesreaper]

Typical load in Wprime is much like playing a game with a few tabs open in Chrome etc. Its not a perfect test but its close enough.

Typical load = the temps you are likely see day in day out when doing 3D gaming / Video Editing / Photoshop /Twitch streaming a game etc. Those temps are what you can expect under typical tasks. The FPU load for Max Load = unrealistic will never happen loads it stresses the coolers in a way typical software won't. However for those that overclock and then stress test for stability the Max Load is what they will run into. This means that if a cooler keeps the CPU from thermal throttling at Max load it will handle daily tasks without issue.

@Kursah did a great job explaining it.

But suffice to say testing games / Chrome tabs etc is not scientific it is not easily produced as results can change to drastically depending on version of browser / addons / adblockers the game selected the settings used.

Wprime / Aida64 the same version is used on all coolers on that test bench. Its reliable easy to produce it gives typical max and extreme max that users can see and understand. In fact the load level and how that load is applied is very different. However both are selected because they give a similar result. Wprime is very integer focused. While the typical Aida64 load test not only stresses CPU it stresses memory controller/ system memory / floating point units etc. It runs a large variety of tests and usually comes close to the Wprime temp. Basically one gives an easy to produce result the other runs the gamut of what programs require. Both average out about the same.

Example If i used Chrome + Total War Atilla, depending on the camera location in the game the load can change etc. So unless its perfectly the same every time the variables result in changes that cannot be controlled. Then of course is the fact that most benchmarks do not run long enough to test a CPU load during gaming. so requires baby sitting a game constantly to monitor that benchmark is run in succession.

Simply put. Wprime / Aida64 = different workloads similar results that are on par with most demanding tasks an end user will put their system through. While Aida64 stability test FPU only load = worst case situation that should never come about. That said unlike most reviews that offer Idle / Load. I actually give a larger data set that covers more situations.

 

[MrGenius]

IntelBurnTest for max load. Typical load is too subjective to be definable. IMO it's anything shy of max load and more than idle. IOW between 1% and 99% CPU usage.

Seems like such a pointless question to ask. As in why do you want to know? What good does it do you to have a definition of "typical load"? If your system is configured to handle max load with stability, it will be stable with any "typical" load and at idle.

 

[cadaveca]

IntelBurnTest for max load. Typical load is too subjective to be definable. IMO it's anything shy of max load and more than idle. IOW between 1% and 99% CPU usage.

Seems like such a pointless question to ask. As in why do you want to know? What good does it do you to have a definition of "typical load"? If your system is configured to handle max load with stability, it will be stable with any "typical" load and at idle.

Because he's asking about what our cooler reviewer is doing to generate his test results, not what an average user does, it isn't that pointless? SMH. Makes sense that a user would ask such a question so that they might be able to draw some comparison to their needs or what they are looking to buy.

 

[D007]

Max load to me is 100% or dang close...

 

[EarthDog]

Max load to me is 100% or dang close...

That doesn't mean much... I can make my PC hit your definition of max load in 5 different ways with 5 different max temperatures and power use. See post #7.

 

[MrGenius]

Because he's asking about what our cooler reviewer is doing to generate his test results, not what an average user does, it isn't that pointless? SMH. Makes sense that a user would ask such a question so that they might be able to draw some comparison to their needs or what they are looking to buy.

I still don't get it. You always want the best cooler you can get for your money. Basing that on some hypothetical and/or indefinable load is to me...IMO...pointless. And I could say worse about it. But I'm playing nice today. Max load temps are the only thing you can use to determine the actual effectiveness of a given cooler with a given CPU. Without factoring in case design, case air flow, and other highly relevant factors(as EarthDog is getting at). By that I mean even max load temps are subjective to a certain extent too. But they mean the most as far as giving a good idea of cooler effectiveness. Under a given set of repeatable circumstances.

I don't know what else to say. To me...Idle temps mean nothing...typical load temps mean nothing. Max load temps mean everything. Why bother with anything else. I don't care about it.

 

[crazyeyesreaper]

I still don't get it. You always want the best cooler you can get for your money. Basing that on some hypothetical and/or indefinable load is to me...IMO...pointless. And I could say worse about it. But I'm playing nice today. Max load temps are the only thing you can use to determine the actual effectiveness of a given cooler with a given CPU. Without factoring in case design, case air flow, and other highly relevant factors(as EarthDog is getting at). By that I mean even max load temps are subjective to a certain extent too. But they mean the most as far as giving a good idea of cooler effectiveness. Under a given set of repeatable circumstances.

I don't know what else to say. To me...Idle temps mean nothing...typical load temps mean nothing. Max load temp means everything. Why bother with anything else. I don't care about it.

Tell that to people that were shitting bricks when i wanted to drop Idle temps about a year or so ago.

 

[MrGenius]

And I'm not saying you're doing your job incorrectly. I just don't look at most of it for meaningful information.

 

[EarthDog]

I still don't get it. You always want the best cooler you can get for your money. Basing that on some hypothetical and/or indefinable load is to me...IMO...pointless. And I could say worse about it. But I'm playing nice today. Max load temps are the only thing you can use to determine the actual effectiveness of a given cooler with a given CPU. Without factoring in case design, case air flow, and other highly relevant factors(as EarthDog is getting at). By that I mean even max load temps are subjective to a certain extent too. But they mean the most as far as giving a good idea of cooler effectiveness. Under a given set of repeatable circumstances.

I don't know what else to say. To me...Idle temps mean nothing...typical load temps mean nothing. Max load temp means everything. Why bother with anything else. I don't care about it.

That is why reviewers that know what they are doing (they do at TPU), test with the same tests and same loads so its comparable. Of course its a "YMMV" with case design/airflow and other relevant factors (which I was not getting at) as far as temps go. In other words, yeah, no system will hit the same exact temp his will for a WIDE variety of reasons.

Understand that REGARDLESS of case and airflow etc, that the difference between the coolers will be the same if the case air hitting the cooler is 22C or 30C. For example, If cooler A cools 4C better than cooler B at the same exact heat load (which again is why they test with the same loads...) The actual temperature will be different, but the temperature between them will still be 4C as was tested. Heat load is heat load is heatload. There is a 1:1 on ambient to actual temps. So if your ambient goes up 4C, so will your temps.

About the only thing I agree with, more or less mind you, is that idle temperatures are pretty useless. I mean, most CPU sensors arent even accurate at ambient temps. The closer they get to TJmax, the more accurate they are. Near any cooler can keep a CPU idle. T

he fact that people are braying about keeping that results speaks volumes. Sometimes a little education goes a long way for the readers.

 

[cadaveca]

And I'm not saying you're doing your job incorrectly. I just don't look at most of it for meaningful information.

good for you. Everyone has different opinions, yours isn't any better than anyone else's, so trying to force your opinion only makes you look like an ass.

 

[MrGenius]

If stating my opinion is considered forceful. Then all I can do is laugh.

 

[Caring1]

When they are testing CPU coolers they have idle, typical load, and max load (running aida64). What would playing a game maxed out at 1440p, a few tabs in chrome, and a word document constitute? Typical load?

Max load to me is utilising the cpu at around 100%

 

[crazyeyesreaper]

Aida64 stress test (CPU / FPU / Memory etc) uses 100% CPU
Aida64 stress Test (FPU) uses 100% CPU,

obviously very drastic temp differences.

Example
Silentium PC Fortis 3 Stock / OC

Wprime 59 / 74
Aida64 Stress 60 / 75
Aida64 FPU 65 / 84

CPU Usage in all 3 = 100%
Type of workload has a huge effect on Temps.

Wprime isn't using the FPU on the CPU, thus lower temps, Aida64 Stressing all aspects of the CPU same situation its doing a more general CPU usage situation, tends to be a tiny bit warmer since FPU gets used. Full on FPU load... yeah maximum heat output.