[SOLVED] Why is my Gpu not using 100%

[Reezyz]

CPU: I7-9700 (non k)
GPU:RTX 2070 super
Ram:1x16gb 2666mhz
PSU:650 W

My gpu is most of the time at 70-90% usage in games and i've tried a lot of things like doing a clean install of Windows, deleting drivers and updating them, etc. None of these "fixes" worked. There's literally no way im getting cpu bottlenecked and my settings on Nvidia are on maximum performance. This issue is really annoying. I play games at 1080p and my monitor is 165hz G-Sync compatible.
Any tips that would fix my problem would be appreciated.

 

[drivinfast247]

Ya'll realize you are all looking at it from entirely the wrong angle, even backwards.

Usage is Not how much of the gpu is used but how much resources it uses. Very important distinction.

Resources are things like speed, bandwidth, vram, cores, voltages, whatever. You are using 100% of the gpu, the gpu doesn't need to use 100% of its available resources to do so.

Think of it like hammering a nail into drywall. YOU will use 100% of the muscles in your hand to hols the hammer. YOU will use 100% of your muscles in your arm to swing the hammer. You will NOT need to use 100% of the strength in your arm to do so. You'll only use a small fraction of strength, even if you are using every single muscle.

Same with a gpu, it works to the best of its ability, working at 100%, but doesn't need to use the full extent of that ability to do so.

Those other ppl with 98% usage are actually in a worse position, their gpus are working harder, using more, just to get the same results as you get using 90%. It means they only have 2% gpu resources left before getting bottlenecked. So if they have a giant explosion, lots of physX and vectors and particles, and they hit 100% usage, they just ran out of strength, and fps goes in the toilet.

If data running through the vram only uses 6Gb,and your card has 8Gb of vram, that's 75% usage. Doesn't mean the data gets pushed through any slower, it's still getting pushed through at 100% of the vrams clock speeds and timings, it just doesn't need to use the full 8Gb of vram to do so. Someone at 100% usage is using all 8Gb of vram, so what happens when they need 8.1Gb, the data is slowed down a cycle, fps plummets due to the wait. You usage could spike could go upto 7.9Gb of vram, 99% usage, and still not get capped.

You are FAR better off never coming anywhere near 100% usage, it means there are still available resources the gpu Can use when it has a need to, depending on the scene, and doesn't take a hit the same as a gpu that is resource capped and has nothing left to give.

You basically just said that people are better off with a 3080-3090 for 1080p as it will never even use close to it's full capacity.

 

[Reezyz]

Ya'll realize you are all looking at it from entirely the wrong angle, even backwards.

Usage is Not how much of the gpu is used but how much resources it uses. Very important distinction.

Resources are things like speed, bandwidth, vram, cores, voltages, whatever. You are using 100% of the gpu, the gpu doesn't need to use 100% of its available resources to do so.

Think of it like hammering a nail into drywall. YOU will use 100% of the muscles in your hand to hols the hammer. YOU will use 100% of your muscles in your arm to swing the hammer. You will NOT need to use 100% of the strength in your arm to do so. You'll only use a small fraction of strength, even if you are using every single muscle.

Same with a gpu, it works to the best of its ability, working at 100%, but doesn't need to use the full extent of that ability to do so.

Those other ppl with 98% usage are actually in a worse position, their gpus are working harder, using more, just to get the same results as you get using 90%. It means they only have 2% gpu resources left before getting bottlenecked. So if they have a giant explosion, lots of physX and vectors and particles, and they hit 100% usage, they just ran out of strength, and fps goes in the toilet.

If data running through the vram only uses 6Gb,and your card has 8Gb of vram, that's 75% usage. Doesn't mean the data gets pushed through any slower, it's still getting pushed through at 100% of the vrams clock speeds and timings, it just doesn't need to use the full 8Gb of vram to do so. Someone at 100% usage is using all 8Gb of vram, so what happens when they need 8.1Gb, the data is slowed down a cycle, fps plummets due to the wait. You usage could spike could go upto 7.9Gb of vram, 99% usage, and still not get capped.

You are FAR better off never coming anywhere near 100% usage, it means there are still available resources the gpu Can use when it has a need to, depending on the scene, and doesn't take a hit the same as a gpu that is resource capped and has nothing left to give.

Punch something as hard and fast as you are physically able to. Give your utmost, 100% power in that swing, that punch. =20% usage. Why so low? Because you are not using 100% of the strength in your LEGS to punch something, you are not using 100% of the strength in your other arm, only the One ARM. You might use your legs and arm to stand, to balance, but that's not the same as using all the combined power your legs and other arm have.

so you are telling me that i dont have to change anything in my build? And btw i get what u mean but people using 100% of their gpu get more fps

 

[Karadjgne]

You basically just said that people are better off with a 3080-3090 for 1080p as it will never even use close to it's full capacity.

No, not even close. I was getting 50-80% on my gtx970, and even when moving upto my current 2070Super I'm still getting the same rough results from the same games, since I'm also using a stronger cpu which provides higher fps, so a higher workload comparatively.

1% usage or 99% usage is the exact same thing, there's no difference to the output to the screen. The difference is when the frame changes and demands more resources are used. If usage is lower, there's more room for expansion. If you are at 98% usage, an intense frame or frames can and will put that to 100% and then there's nothing left. No more resources. You are capped. So if you are at 98% and the bomb goes off and creates a giant fireball and gpu usage hits 100%, you'll get stutters, frame drops, freezes. You just ran out of cores, vram etc they can do no more, accept no more.

It's no different than trying to run gta5 on a 4690k vs a 4790k, same cpu except the i5 will see 100% usage and lower fps, but the additional threads on the i7 will lower usage and you don't get bottlenecks as a result.

Op was saying he was disgruntled because no matter what he did to the pc he was getting 90% gpu usage, whereas others were hitting 98%. He assumed that the others were doing better because of higher usage. I'm saying that's backwards, He was in a better position as he still has 10% available resources whereas the others were almost bottlenecked and most likely did have patches of 100% usage and were bottlenecking.

Usage isn't using a % of the clock speeds or memory speeds or timings, they run at 100%, usage is the amount of cores required to complete the frame, the amount of memory required to hold the frame etc. Just like you will use every muscle to swing the hammer, but don't need to use the entire strength of all those muscles to do it. You use all 5 fingers on one hand to type with, but you don't need to jam the keys down as hard as you can, just takes a light touch. Same motions, same fingers, same results, different usage in the amount of finger strength applied.

Cpu sends a frame to the gpu. Gpu takes that data and creates a wire frame first, placing every object provided at the addresses provided using the dimensions provided. When everything is placed, the gpu fills in colors, overlays affects from lighting, shadows, rtx, whatever. All that takes a certain amount of time. Usage of 1% or 99% is irrelevant, it is whatever the gpu needed to use to render that frame. Doesn't affect the amount of time it took at all.

The difference between using a 3080 and op's 2070S is in the core clocks, memory clocks, IPC etc. A 3080 takes less time to render the frame than a 2070S does, so can complete more frames per second, higher possible fps. Usage doesn't affect that.

If a gpu has to use 1000 cores and 1Gb of vram to complete a frame, that might be 60% usage for a 2070Super and only 40% usage for a 3080, but that usage doesn't matter at all. What matters is the faster clocks, faster vram, higher IPC of the 3080 dumping out that frame 2x as fast as the 2070Super, so doubling fps.

 

[RTX 2080]

so you are telling me that i dont have to change anything in my build? And btw i get what u mean but people using 100% of their gpu get more fps

There are lots of little bits and pieces of this conversation that can be (and have been) debated endlessly, but one thing is for certain: AMD CPUs are very sensitive to a) RAM speed & b) RAM bandwidth.

By having 1 stick of RAM, you are choking your CPU's ability to pre-render frames, keeping it from sending the GPU the information it needs quickly enough to run your game at higher frame rates. Thus, your 1 stick of RAM is leading to lower frame rates. If you use 2 identical sticks of RAM (preferably with a speed of at least 3200 MHz) I guarentee you that based on personal experience it will allow your CPU to run better, leading to higher GPU utilization and ultimately better frame rates.