Can anyone explain this "none" PerfCap Reason?

[raink]

I searched this reason and got nothing. The fps and power draw did dropped, but the frequency got higher. I have no clue what's going on.....

 

[blobster21]

It may sound totally dumb but let me ask it anyway : isn't it because the screenshot was taken in beetween two perfcap peaks, so it was not performance capped when you took the screenshot, hence the "none" ?

 

[raink]

It may sound totally dumb but let me ask it anyway : isn't it because the screenshot was taken in beetween two perfcap peaks, so it was not performance capped when you took the screenshot, hence the "none" ?

Hi, thanks for your reply.
The screenshot was taken when I was running a furmark and the fps dropped from 100+ to 30. And when the GPU is on idle, the perfcap reason should shows "idle".

 

[hat]

What CPU and graphics card do you have? Perfcap none means there is nothing limiting the performance of the card, but fps drops accompanied by lower usage can be explained by CPU bottleneck.

 

[raink]

What CPU and graphics card do you have? Perfcap none means there is nothing limiting the performance of the card, but fps drops accompanied by lower usage can be explained by CPU bottleneck.

6700HQ and 980M. I didn't thought about that. Thanks, I will try to redo the test.

 

[W1zzard]

And when the GPU is on idle, the perfcap reason should shows "idle".

I think that's for GeForce 20 Series, GeForce 10 Series was "VRel" when not limited, and GeForce 9 is "None". Not 100% sure I remember correctly

 

[hat]

6700HQ and 980M. I didn't thought about that. Thanks, I will try to redo the test.

I'm not too sure how those chips jive together, but one way to see if you're GPU bottlenecked is to increase the resolution and go back and play the same game again and see if it looks any different. If you're already playing at max resolution for this game, you can head over to the Nvidia control panel, and enable DSR to render the game at up to 4x your native resolution. If you are CPU bottlenecked right now, that means your GPU can't work to its full potential because it has to wait on the CPU. If you enable DSR, you'll flip it around. CPU should be waiting on GPU instead, in this case. This doesn't mean your game will run better, in fact you'll certainly have less FPS than before... but if GPU-Z readout changes to show the GPU pegged all the time, you'll know perfcap "none" means GPU is waiting on CPU.

 

[raink]

I'm not too sure how those chips jive together, but one way to see if you're GPU bottlenecked is to increase the resolution and go back and play the same game again and see if it looks any different. If you're already playing at max resolution for this game, you can head over to the Nvidia control panel, and enable DSR to render the game at up to 4x your native resolution. If you are CPU bottlenecked right now, that means your GPU can't work to its full potential because it has to wait on the CPU. If you enable DSR, you'll flip it around. CPU should be waiting on GPU instead, in this case. This doesn't mean your game will run better, in fact you'll certainly have less FPS than before... but if GPU-Z readout changes to show the GPU pegged all the time, you'll know perfcap "none" means GPU is waiting on CPU.

Thanks! So I tried to increase the MSAA all the way up to 16X to keep GPU busy, and the result is finally normal.

So I guess you are right, it's CPU bottlenecked.

But there is a wired thing... I tried redo the thermal paste and ignored the Video memory chips and some small things around the CPU that was supposed to be something about power supply. Then the system stuttered while CPU and GPU both under 80 degrees and PerfCap Reason shows None. Is it possible for Video memory chips to overheat?

 

[hat]

VRAM can overheat and cause issues, sure. I'm not sure if stuttering is a common symptom, but it's definitely not good. You said you "ignored" the memory chips, what exactly does that mean? Usually, at least on desktop graphics cards, VRAM makes contact with the heatsink by way of thermal pads. Sometimes you can reuse them, sometimes you can't. If they're in poor condition, once you break that bond, they have to be replaced.

I have similar concerns about "small things around the CPU..." those are probably VRMs (voltage regulation modules) which are responsible for regulating the power that goes to the CPU. Like VRAM, these generally don't have temperature sensors, but they can get quite toasty and bad things can happen if they get too hot. In my experience with desktops, sometimes these have heatsinks, sometimes they don't... but they all, at a minimum, do rely on some amount of case airflow (even if they don't have heatsinks) to keep the heat under control.

Once you get that under control, instead of artificially making your GPU work harder by inflating the resolution or AA settings, take a look at Throttlestop. You may be able to overclock your CPU a little bit with it. Side note: don't rely on Furmark to find CPU bottlenecks. It's designed to torture your GPU and not much else. Saying you're CPU bottlenecked in Furmark is like saying you're GPU bottlenecked in Prime95. Keep GPU-Z open and possibly logging while you're playing a game, or at least running an all-around benchmark like 3DMARK or something.

Actually, on that note, it's possible your CPU may be throttling to some extent under heavy load, like gaming. As a laptop chip, it's going to have more stringent limits for power and temperature than its desktop counterparts. This is one of the reasons why Throttlestop exists. To monitor CPU frequency, you might be able to use HWiNFO64. I believe it supports logging as well, so you can play a game, go back and check the clockspeed it was running at, at any given time.

 

[raink]

VRAM can overheat and cause issues, sure. I'm not sure if stuttering is a common symptom, but it's definitely not good. You said you "ignored" the memory chips, what exactly does that mean? Usually, at least on desktop graphics cards, VRAM makes contact with the heatsink by way of thermal pads. Sometimes you can reuse them, sometimes you can't. If they're in poor condition, once you break that bond, they have to be replaced.

I have similar concerns about "small things around the CPU..." those are probably VRMs (voltage regulation modules) which are responsible for regulating the power that goes to the CPU. Like VRAM, these generally don't have temperature sensors, but they can get quite toasty and bad things can happen if they get too hot. In my experience with desktops, sometimes these have heatsinks, sometimes they don't... but they all, at a minimum, do rely on some amount of case airflow (even if they don't have heatsinks) to keep the heat under control.

Once you get that under control, instead of artificially making your GPU work harder by inflating the resolution or AA settings, take a look at Throttlestop. You may be able to overclock your CPU a little bit with it. Side note: don't rely on Furmark to find CPU bottlenecks. It's designed to torture your GPU and not much else. Saying you're CPU bottlenecked in Furmark is like saying you're GPU bottlenecked in Prime95. Keep GPU-Z open and possibly logging while you're playing a game, or at least running an all-around benchmark like 3DMARK or something.

Actually, on that note, it's possible your CPU may be throttling to some extent under heavy load, like gaming. As a laptop chip, it's going to have more stringent limits for power and temperature than its desktop counterparts. This is one of the reasons why Throttlestop exists. To monitor CPU frequency, you might be able to use HWiNFO64. I believe it supports logging as well, so you can play a game, go back and check the clockspeed it was running at, at any given time.

I tried replace the thermal pads on VRAM and VRM. But the pads are little bit thicker, which may causing a bad contact for heatsink and GPU&CPU. So I removed them for a little test. And I put them on after that test.

Thanks for your advice! I will try these methods.

 

[hat]

With no thermal pads, contact is going to be... bad. I wouldn't be surprised if that stuff is overheating. You don't want thermal pads that are too think either... it's just as bad to use thermal pads that are too thick. You need the correct size thermal pads.