• Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.

RTX 3070 - PerfCap reason PWR?

M

matixitam

New Member
Joined
Dec 20, 2020
Messages
4 (0.00/day)
Hello everyone, I am kind of a newbie when it comes to overclocking so I am hoping someone can help me with my problem.

Yesterday i wanted to overclock my GAINWARD RTX 3070 PHOENIX via MSI Afterburner. I opened the GPU-Z sensors and started the MSI Kombustor stress test, so that I could proceed with increasing core and memory clock.

While I was still running stock clock settings I noticed that the PerfCapReason was PWR, as seen on the bottom picture.

Picture 1.PNG



I then proceeded to increase the power limit in afterburner to 113% hoping that would fix the issue, but the PerfCap reason was still stuck at PWR, as seen on the next picture.

Picture 2.PNG


I then stress tested the GPU via AIDA64. This time the PerfCapReason changed to VRel, since the GPU Voltage reached the threshold 1.08 V (next picture), whereas before when stressing with Kombustor the voltage was much lower.

Picture 3.PNG



My question is if there is something wrong with my GPU, or am I not understanding something correctly?

Thanks!
 
If you look at the board power draw between the 1st and 2nd pictures, you can see that your card was indeed consuming more power. Even if you didn't manually increase the core or memory frequency, the card will still boost given the additional power headroom. You can see that in the pictures as well, your FPS increased.

The third picture can be explained simply: different workloads stress your GPU differently, some requiring more voltage. The same applies to CPUs where, for example, when running AVX either the clock is reduced or additional voltage needs to be applied. It just so happens that you are hitting a voltage ceiling in that benchmark instead of power.

Nothing out of the normal here.
 
Thank you for the response kind sir.

I check out JayzTwoCents video regarding this matter. He is running an ASUS STRIX RTX 3070 and testing it in Unigine Heaven bench. During this test, he is constantly hitting VRel limit, as shown in this timestamp (
).

I tried the same bench on my GPU, but i am constantly hitting Pwr limit, even at 113% power limit setting in afterburner. My clocks are fluctuating a lot as well as a consequence, while Jays card has a very stable clock since he is VRel limited.

Is there any point of overclocking my card then? It is my understanding that Incresing clocks in afterburner increases power consumption, therefore there is no performance gain since i am already bound by power in stock clocks.
 
Do you have dual BIOS? Couldn't find the specs on the card's website, perhaps you're in "quiet" BIOS.
 
Use something else other than Kombustor, it is a power virus.
 
The problem with kombustor is that it puts so much of a load on the VRM that they heat up, throttle(or smash into pwr limit), then the core voltage drops and then the clocks will drop as defined by the frequency/voltage curve boost plots out for your gpu.(you should be able to see this in afterburner I think) So even if you are stable in kombuster, since the gpu throttled the voltage down your still not testing the card for oc stability. This is why you can be stable in kombutster and then crash ingame.
 
I dont think is anything wrong with the GPU.
If I'm not mistaken PerfCapReason: Pwr means that your card is using all the power headroom its been given. Meaning temp and voltage is not stopping it from reaching top perf.

You can use UnigineHeaven 4.0 in a loop (default), which can been run windowed to test your GPU and/or OCing realtime as you run it.
 
you're using a testing program that maxes out the power usage of the GPU, and that's exactly whats happening.

go test with a game, you should see better results.
 
This is what I'm talking about

Unigine_Heaven_4.0_001.png
 
Ampere cards run in their power limit all of the time, this is normal and the way NVIDIA designed their card.

Raising the power limit does work, it will give you higher FPS, but the limit is still not high enough
 
MSI AfterBurner 4.6.2 is too old for the RTX 30x0 cards, use the beta version below or search for it on Guru3d forums.

MSI AfterBurner 4.6.3 Beta 4
Download link below

Microsoft OneDrive

1drv.ms 1drv.ms

1608537795965.png
 
Zach_01 said:
I dont think is anything wrong with the GPU.
If I'm not mistaken PerfCapReason: Pwr means that your card is using all the power headroom its been given. Meaning temp and voltage is not stopping it from reaching top perf.

You can use UnigineHeaven 4.0 in a loop (default), which can been run windowed to test your GPU and/or OCing realtime as you run it.
Click to expand...

My Zotac GTX 1660 Super throttles a lot, even in Heaven 4.0, IIRC. Down from 1.9 Ghz to 1.8 Ghz repeatedly. Heck, even Halo MCC makes it throttle, too! With the same reason. (IIRC)

It's like it thinks Unigine and Halo MCC are power viruses, while it possibly keeps it at 1.9 Ghz in DirectX9-based Halo Custom Edition. (That's actually Halo version 1x)

W1zzard said:
Ampere cards run in their power limit all of the time, this is normal and the way NVIDIA designed their card.

Raising the power limit does work, it will give you higher FPS, but the limit is still not high enough
Click to expand...

Sounds just like my Turing, my Zotac GTX 1660 Super, see above as an example! That's why I have been OC'ing the VRAM only.
 
Last edited:
RJARRRPCGP said:
My Zotac GTX 1660 Super throttles a lot, even in Heaven 4.0, IIRC. Down from 1.9 Ghz to 1.8 Ghz repeatedly. Heck, even Halo MCC makes it throttle, too! With the same reason. (IIRC)

It's like it thinks Unigine and Halo MCC are power viruses, while it possibly keeps it at 1.9 Ghz in DirectX9-based Halo Custom Edition. (That's actually Halo version 1x)
Click to expand...
For maintaining max boost all the time depends on a lot of things, and in most circumstances it’s not happening. Today’s GPUs (and CPUs also) are highly dynamic devices. Different games and/or benchmarks will load very differently a GPU with different voltages and boost clocks.
Its normal. Ideally a GPU is not thinking anything and doesn’t know what’s it’s running. Its tracking its own parameters (temp, voltage, current, power... etc) and act accordingly.
 
Zach_01 said:
For maintaining max boost all the time depends on a lot of things, and in most circumstances it’s not happening. Today’s GPUs (and CPUs also) are highly dynamic devices. Different games and/or benchmarks will load very differently a GPU with different voltages and boost clocks.
Its normal. Ideally a GPU is not thinking anything and doesn’t know what’s it’s running. Its tracking its own parameters (temp, voltage, current, power... etc) and act accordingly.
Click to expand...
For me, I exceed TDP setting, even in what are considered the usual-type of stuff! Especially at the default 125W setting.
Looks like the VBIOS will only let me go to 130W and it looks like that's not enough, LOL.
 
matixitam said:
Is there any point of overclocking my card then? It is my understanding that Incresing clocks in afterburner increases power consumption, therefore there is no performance gain since i am already bound by power in stock clocks.
Click to expand...

You'll still see some extra frames by increasing the power limit. If you are fine with the extra power consumption I'd say keep it elevated. That said I'm afraid going further than that is not going to yield much results. As Wizzard pointed out, it's by Nvidia design. It's a shame really as your card has plenty of thermal headroom.
 
Last edited:
Hello guys, thanks for all the replies. I did as @Zach_01 and @Mussels suggested, i tried running Heaven 4.0. Perfcap reason is now cycling between Pwr and VRel. Seems all is good on this kind of load.
I will overclock the card with AIDA64 stress test, since is maxes out the GPU voltage so that I wont have crashes mid game. I will not be using kombustor in the future. Thanks for all the advices.
1608976111356.png
 
You must log in or register to reply here.
Back
Top