-
Competitor rules
Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.
3090FE low GPU usage?
- Thread starter RipGroove
- Start date
Must be game related as it's 99% usage in Unigine Heaven with a benchmark score of 6313 (250fps) (High Preset at 3440x1440)
That particular game (BFV) must not be loading CPU/GPU up enough then?
I always thought GPU would run at 99% usage all the time no matter what?
That particular game (BFV) must not be loading CPU/GPU up enough then?
I always thought GPU would run at 99% usage all the time no matter what?
Some games are FPS or CPU thread limited. So if the game has some code that is slow/clock locked to the compute/ai/etc, then the FPS can hit a hard limit.
Especially true in some driving games where framerate was locked as it's physics calculations are locked to framerate. Also true for some other games (IIRC Fallout 3 had this).
Without knowing the games code though, hard to accurately speculate.
Especially true in some driving games where framerate was locked as it's physics calculations are locked to framerate. Also true for some other games (IIRC Fallout 3 had this).
Without knowing the games code though, hard to accurately speculate.
Interesting, my old 1080ti and 7700k would have the GPU pinned at 99% in BFV so I just assumed the new hardware would be the same.
Hi ish, 3440x1440 Ultrawide
99% of what?
Percentage is a measure of use, not unit. 99% at what fps, and what resolution? 99% of what compute ability vs the 3090? It could be the game engine not understanding how to use a 3090 (mistaking 24gb vram for 2gb vram due to poor coding not allowing 2 digits of ram utilisation and just reading "2.4" in error, but rarely does that happen nowadays, more a 1990s/2000s game engine problem).
A Nissan Micra doing 99% performance might be it going 80mph, a Porsche doing 80mph is only like 49% performance. So if the road is limited to 80mph, you'd get the same result as a 3090 and the game being limited to 160fps.
"Fair enough, I just seem to remember all my GPUs always sitting at 99%, maybe I've just been hammering them all along lol."
Then you were not CPU limited, but GPU limited. Now you are CPU (or game engine) limited. You mistook 99% to *be* the unit measurement. It's not. It's telling you the GPU is working full speed, but not what that speed is.
PS, providing you are getting the FPS results though. If your getting 50% gpu use and only 30fps, then something is wrong. That's when you look for power failures, bottlenecks or driver/game bugs.
99% useage, as in GPU useage.99% of what?
Percentage is a measure of use, not unit. 99% at what fps, and what resolution? 99% of what compute ability vs the 3090? It could be the game engine not understanding how to use a 3090 (mistaking 24gb vram for 2gb vram due to poor coding not allowing 2 digits of ram utilisation and just reading "2.4" in error, but rarely does that happen nowadays, more a 1990s/2000s game engine problem).
A Nissan Micra doing 99% performance might be it going 80mph, a Porsche doing 80mph is only like 49% performance. So if the road is limited to 80mph, you'd get the same result as a 3090 and the game being limited to 160fps.
"Fair enough, I just seem to remember all my GPUs always sitting at 99%, maybe I've just been hammering them all along lol."
Then you were not CPU limited, but GPU limited. Now you are CPU (or game engine) limited. You mistook 99% to *be* the unit measurement. It's not. It's telling you the GPU is working full speed, but not what that speed is.
PS, providing you are getting the FPS results though. If your getting 50% gpu use and only 30fps, then something is wrong. That's when you look for power failures, bottlenecks or driver/game bugs.
Associate
- Joined
- 1 Oct 2009
- Posts
- 1,033
- Location
- Norwich, UK
In most game engines the GPU can only create a new frame if the state of the world has been updated, otherwise any new frame would be identical to the previous one and thus be a waste of time. This means the CPU needs to re-calculate all game states for the next "tick" as fast as the GPU can draw frames, if the CPU is too slow to do this then the GPU has periods of being idle while waiting to start the next frame and this shows up as less than 99% GPU usage. In this case you'd be CPU bound/limited.
The main problem that confuses this scenario in modern games is that they're notoriously bad for using CPUs efficiently. Modern CPUs have multiples cores and % utilization of the entire CPU when given as just 1 number is referring to average use across all cores. So if you have a 8 core CPU and a game engine only uses 4 then the game will never exceed 50% usage even in perfect conditions. Typically the way most multi-core game engines work is they have a primary/master thread and then many worker threads which each run on a different CPU cores, and the primary thread delegates work to the other threads, but doing this efficiently is extremely hard and so games can become CPU limited long before the CPU actually runs out of utilization. I think this is typically because once that master thread hits 100% utilization on its own core it can limit the delegation of work to other threads. It's common to see a game CPU limited with 1 core at 100% which is the master thread, while the total CPU usage is only at 60-70% because the other threads are not being fully loaded.
The better metric is the GPU utilization. The GPU its much simpler to measure performance because the workload it deals with is very well optimized for multi-core in the first place, as moden GPUs have 1000's of cores. So you can say with a high degree of reliability that if you're using 99% GPU then you're GPU limited, and if you're using 70% GPU then you're CPU limited. But if you show 70% CPU usage you have no real idea what that means for the game, it's tempting to think the game must be GPU limited because there's overhead left on the CPU but it's just not necessarily true, the game engine itself can be CPU limited while only reporting say 70% usage.
The easiest way to remain not confused by this is just trust the GPU usage, that's way more reliable in most cases.
If using vsync or a hard fps cap then yes then pretty much normal, if not and it’s off then it should be at 99% or there abouts even with an RTX 3090.
Id find it hard to believe that the 11900k couldn’t keep up with it especially at 3440x1440. Especially to the point of only using 40-60%.
Make sure you’re using normal DX11, DX12
On BFV is usually just trouble and never has ran right.
Last edited:
This is what I thought, I've played some Cyberpunk and the GPU usage is pinned at 99% as expected, with CPU at 30-40%.
So my initial problem is probably related to BFV then, maybe some weird combination of settings including DX11/12. I'll see if I can dig out the settings I used with my old CPU/GPU and apply those to see what happens.
I've set everything up exactly as I had it before with my old CPU/GPU as below:
NVCP
G-Sync - Enable for windowed and full screen
Low latency mode - off
Max frame rate - off
Power - Prefer Max
Preferrered refresh rate - Highest available
Threaded optimization - Auto
V-Sync - Off
Battlefield V
Screen mode - Borderless
DX12 - Enabled
DXR - Off
Framerate limiter - 200 (That's the max the game will allow)
Future frame rendering - Off
V-Sync - Off
Graphics - Ultra
RTSS
Framerate limit - Off
So with these exact settings with both my old CPU/GPU and new CPU/GPU the results are below:
BFV (7700k/1080ti)
BFV (11900k/3090)
Something does not seem right about the performance of the new CPU/GPU, the FPS fluctuates by a large amount, the GPU useage is very low low and fluctuates and the whole experience is nowehere near as smooth as it should be. Could this be symptoms of the PSU not having enough power?
NVCP
G-Sync - Enable for windowed and full screen
Low latency mode - off
Max frame rate - off
Power - Prefer Max
Preferrered refresh rate - Highest available
Threaded optimization - Auto
V-Sync - Off
Battlefield V
Screen mode - Borderless
DX12 - Enabled
DXR - Off
Framerate limiter - 200 (That's the max the game will allow)
Future frame rendering - Off
V-Sync - Off
Graphics - Ultra
RTSS
Framerate limit - Off
So with these exact settings with both my old CPU/GPU and new CPU/GPU the results are below:
BFV (7700k/1080ti)
- FPS - 90-100 (it would never go over 100 as if there was a hard cap somewhere)
- CPU Usage - Don't acually recall this
- GPU Useage - 99% always
- Smoothness - Always super smooth
BFV (11900k/3090)
- FPS - 90-160 (seems to fluctuate wildly)
- CPU Usage - 40-70%
- GPU Useage - 70% average
- Smoothness - fairly stuttery at times
Something does not seem right about the performance of the new CPU/GPU, the FPS fluctuates by a large amount, the GPU useage is very low low and fluctuates and the whole experience is nowehere near as smooth as it should be. Could this be symptoms of the PSU not having enough power?
use higher resolution scaling options
or use DSR
you're cpu bottlenecked near 120-160 fps. this is pretty normal and common with nvidia gpus. they eat %20-30 more cpu so your cpu bound max ceiling fps will always be lower than AMD gpus due to nvidia's hack
conisider urself lucky if that cpu can hit 100-110 fps in bf 2042 lmao. that game will be designed to push 120 fps with zen 4+ddr5 in mind
oh disable dx12
or use DSR
you're cpu bottlenecked near 120-160 fps. this is pretty normal and common with nvidia gpus. they eat %20-30 more cpu so your cpu bound max ceiling fps will always be lower than AMD gpus due to nvidia's hack
conisider urself lucky if that cpu can hit 100-110 fps in bf 2042 lmao. that game will be designed to push 120 fps with zen 4+ddr5 in mind
oh disable dx12
You think I'm not taxing the new hardware enough? That did also cross my mind. I did try ray tracing but it just made the game jerky and the FPS went to 80. Can't use DLSS on BFV because it's DLSS v1 which does not support Ultrawide resolutions. I could try upping the Res scaling in game to see what that does.
well, battlefield 5 runs at 1080p 60 fps with a ps4
its not anormal for it to underutilize 3090
if you want %99 gpu util you have to push higher settings
but %99 gpu util will introduce additional input lag
i would suggest putting a frame limiter of 120 with rivatuner
that would both save you from %99 gpu util input lag and provide you smooth consistent frametimes
its not anormal for it to underutilize 3090
if you want %99 gpu util you have to push higher settings
but %99 gpu util will introduce additional input lag
i would suggest putting a frame limiter of 120 with rivatuner
that would both save you from %99 gpu util input lag and provide you smooth consistent frametimes
Point: Whoosh.
If I measure 99% on my ruler, 99% of what? The ruler yes. But what distance is the ruler. What processing power is the GPU, and what is it processing? It's a deep and difficult question, but the right one. Until you realise what 99% *is* you'll be stuck wondering why the new 3090 is only at 40%.