Jump to content
Corsair Community

Corsair One Pro A200 acceptance test?


Recommended Posts

I ordered a loaded A200 (5950x, 3080 Ti, 64GB, and 2TB SSD) and FedEx is shipping it to me. After reading a few threads on these forums, I believe that I need to do a stress load acceptance test in the first 30 days. Can some of you more experienced "gamers" please suggest some realistic tests for this purpose? AMD R9 5950x is a desktop grade CPU and thermo throttle is an expected behavior for this type of CPUs. 

First, I am NOT a gamer. So I need some help in identifying reasonable tests for gaming PCs. I plan to use this new PC for machine learning software prototyping and advanced photo processing. My current desktop is a Mac Pro 6,1 (aka Trashcan and has a Xeon server-grade CPU) which has a similar heat dissipation/fan design as Corsair One. I really like its quietness. But Apple does not use Nvidia GPUs. So I have not used it for Deep Learning/Neural Network model development. Since my MP 6,1 is aging, I thought that I might try a Windows machine with Nvidia GPU. 

Note that this new PC will serve as a machine learning model prototyping system, but not full scale model development. In other words, I may run models that stress the system (CPU and GPU) for a day, but not for weeks. I have been using Macs for the last decade, but have been doing PC tech support for family and extended family members. I am a ML/AI software developer and hold a few EE degrees. So I can follow instructions if some of you can provide test setup instructions (e.g., what benchmark software to use to stress the system? what to look for? etc.)

Link to comment
Share on other sites

if you do not utterly push both CPU and GPU at the same time, you shouldn't throttle.

Throttling is expected in laptops, sometimes, but on desktop it should never happen. the 5950x is 105w TDP which should be no problem for the AIO.

Maybe try some Cinebench R23 tests for the CPU. there's a dropdown list to select the test duration, just set it to 30 min, to really heat-soak the AIO, and see if you get any thermal throttling. CPU alone, it shouldn't happen.

For the GPU, since you're not going to game, maybe try the Blender benchmark, and do some GPU rendering. That will be the biggest heat load in the case at 350W.

Combining both tests may be excessive, i do not know how much what you intend to do will load both components, but either way, the PC should safely throttle and not crash

 

You should be able to use iCue's dashboard to monitor components temperatures. Ideally the GPU should stay under the typical temperature above which it reduces frequency, around 83°C. Also watch the NVME drive temperature. Since there's limited airflow it could get quite hot. they usually throttle read/write speed over 75°C.

Now, it's a small form factor case, so it will always be at a disadvantage when it comes to cooling, so, it will be up to you to decide if for your use case it's acceptable.

Link to comment
Share on other sites

Thanks. Does iCue also show CPU and GPU clock speeds? Additionally, is there a single core stress test so that one can verify turbo boost? From what I have been reading, AMD R9's boost is only with single core. I want to verify Corsair One does not thermo throttle the CPU and GPU under various situations. 

FYI, there are 3 kinds of CPUs for general computing: laptop, desktop, and server. The server grade CPUs usually have better heat dissipation than desktop and laptop ones. My 2013 Mac Pro has a 10-core 20-thread 135W server-grade CPU (Intel Xeon). This MP is about 1/3 in size as compared to Corsair One. The MP has only 1 fan to dissipate the heat from the Xeon CPU and 2 AMD graphic cards. The system can run full CPU loads for days and stay under 80 degrees. I don't expect Corsair One to achieve similar performance level. Rather, I want to understand how gaming desktops are usually benchmarked.

Link to comment
Share on other sites

I don't remember exactly what iCUE displays, i have stopped using it a couple of years ago.. maybe soneone else will answer that 🙂

You can use HWinfo64 to get all the hardware readings you want though, just remember not to let it running in background when you're done as it can interfere with iCUE

Cinebench has single core test too.

As for the thermal dissipation, the desktop CPUs are not really worse than servers, but we get crap coolers by default, and the voltages and frequencies are higher than servers. 🙂 with good quality coolers, they never throttle at factory settings.

As for gaming, usually it's 3Dmark for a pure performance score, and running the tests in loop to test cooling. Other realistic load tests could be the Unigine ones like Valley or Superposition.

For extreme thermal load, some use Furmark to completely cap the GPU, but it's not exactly a realistic load for gaming.

For CPU in the same category that would be Prime95, and i have no doubt that one would make the Corsair One throttle.

 

 

Link to comment
Share on other sites

Thanks for the info about various benchmark/stress test. Since my A200 is advertised as a workstation grade desktop, I'll make sure it can sustain the workload before the 30-day return period ends.

How do I enter the BIOS menu to verify the configuration? My unit has a 350w 3080Ti GPU and a 105w 5950x CPU. Is there a way to check if Corsair has limited the power to these devices in order to fit in a small case?

 

Link to comment
Share on other sites

  • 2 weeks later...
Posted (edited)

I use HWMonitor to analyze the system performance. I found the GPU performance is 'utilization limited', as shown below. What does that mean?

image.thumb.png.abf7fff174731d8cd17d6d1712c5724b.png

Edited by vett93
Link to comment
Share on other sites

that means the GPU is not pushed to the point of needing to boost more.

That's pretty much its idle/light task state

Apparently it's running a memory intensive thingie, but the GPU itself is not stressed much, so, no need to push it.

Link to comment
Share on other sites

Thanks, LeDoyen. Does that mean the GPU RAM size of 12GB is too small for my computation? 

I also see 'reliable voltage limit' jumps to 1 from time to time. Does that mean the power supply is not stable enough?

Link to comment
Share on other sites

3 hours ago, vett93 said:

Thanks, LeDoyen. Does that mean the GPU RAM size of 12GB is too small for my computation? 

eeeeeeh it's hard to say ^^ but, since the bus interface (PCIE traffic) is extremely low, i don't think the GPU is struggling with data, swapping constantly between vram and SSD.

One thing is, i am not sure upgrading to a 3090 would be a good idea in the One anyway.. the vram on the backside gets so hot it wouldn't play nice with the NVME drive right next to it.

Seems fine as it is since there's no huge traffic on the interface.

 

As for reliable voltage limit, that's not the PSU, it's the GPU core voltage that can become unstable when boosted or overclocked too much.

When the GPU runs hard, the GPU boost algorithm will bump on and off on the first hard limit it encounters, be it vcore stability, power limit, temperature etc..

those options on HWmonitor shows you why the GPU won't boost past the frequency it shows.

Link to comment
Share on other sites

23 hours ago, LeDoyen said:

eeeeeeh it's hard to say ^^ but, since the bus interface (PCIE traffic) is extremely low, i don't think the GPU is struggling with data, swapping constantly between vram and SSD.

One thing is, i am not sure upgrading to a 3090 would be a good idea in the One anyway.. the vram on the backside gets so hot it wouldn't play nice with the NVME drive right next to it.

Seems fine as it is since there's no huge traffic on the interface.

 

As for reliable voltage limit, that's not the PSU, it's the GPU core voltage that can become unstable when boosted or overclocked too much.

When the GPU runs hard, the GPU boost algorithm will bump on and off on the first hard limit it encounters, be it vcore stability, power limit, temperature etc..

those options on HWmonitor shows you why the GPU won't boost past the frequency it shows.

Thanks again for the info. I changed the computation software and now it could drive the GPU to 80% utilization, which is very good. Judging from the low utilization of the bus interface, there is not a lot of traffic between vram and system ram. So the 12GB vram size is probably not a bottleneck, I think...

Link to comment
Share on other sites

Update: I believe the fan control software is still buggy. This is not a deal breaker for me. But I hope Corsair can fix it soon.

My Corsair One Pro's fan can spin at 2,000rpm. When multi-cores and GPU are running harder, the fan also gradually increases the speed. This is nice. However, for some of my work, the software runs in single thread and runs for a long time, and without GPU processing. As a result, the CPU runs at 91C for a long time (like hours) and I don't see the fan speeding up. 

This is not a good idea for consumer products! I set the min fan speed at 975rpm. When CPU runs at 90C, I would hope the fan could speed up a bit.

 

Link to comment
Share on other sites

it depends what temperature the water is at.

If it barely warms up, speeding up the fan won't do anything. it cools the water, not the CPU directly 😛

In single thread, the CPU is supposed to boost higher, so, it's probably the cause for that higher temperature.

As long as the water temperature is low, there is no need to speed up the fan. But what you can try is to speed up the pump when running those single thread tasks.

It could shave a few degrees off, if iCUE also slowed the pump down because water is cold.

Link to comment
Share on other sites

For my Machine Learning work (as opposed to Deep Learning work), it is CPU 100% for hours or days, but not much GPU or disk I/O. The CPU stays at 80C. The top fan stays at 1,200rpm. The CPU coolant stays at 48C, for room temperature at 22C. I think the clock speeds, as shown below, look reasonable too. 

image.thumb.png.96c2a74ac801b71d76c09181b1991f0e.png

Link to comment
Share on other sites

It looks like I have found another weak point of the system: the M.2 SSD. Without any R/W to it, it reached 69C while I was doing CPU or GPU stress test. The top fan was at 1,200rpm during these tests.  I should probably open up the computer to take a look.

Link to comment
Share on other sites

  • 2 weeks later...

Update: So I opened the case and looked inside my Corsair One Pro. The M.2 SSD is on the front of the motherboard and it has the stock ASRock heatsink already. The heatsink looks a bit wimpy to me. If you are interested in what it looks like, you may see it on ASRock's website: ASRock > B550 Phantom Gaming-ITX/ax

I am not sure upgrading the heatsink would do much. It gets the hot air from the CPU watercooler radiator. So when the CPU is hot, the SSD gets the heat from its radiator.  It has a great performance when it is cool though:

image.png.69531cb03840fa2e1ea3f71efa48f6ff.png

I am at the end of the 30-day return period. I think I will keep it. Then I will replace the M.2 SSD in a few years.

Link to comment
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
 Share

×
×
  • Create New...