Question Low GPU Usage In Game RTX 3060 Ti After Upgrading GPU ONLY

[ldefinis]

I've run out of options so I'm coming here as a last resort. I know there are lots of posts about this but none of them have worked for me and none of them really match my situation. I upgraded my GPU from a Asus 1070Ti turbo to the 3060 Ti. Everything else in my build stayed the same. My 1070ti was getting higher frames by 30-40 or more. Some games I cant even run because the frames drop below 10. I can see my GPU usage is stuck at 30% in task manager and it shows 0% load in NZXT CAM.

All power settings are set to performance mode.

The strange part is that when I run a benchmark the GPU usage is as expected 90% or more. I was using the Heaven Benchmark.

Games I was playing:
Destiny 2 low settings across the board getting 70-80 frames
Cyberpunk is unplayable on the rtx 3060ti but I was getting above 60 frames on my 1070ti

Build:
i7-8700
MSI Z390 gaming edge
600W Be Quiet power supply
Zotac RTX3060 Ti
32GB G.SKILL RAM

 

[rgd1101]

reinstall gpu driver?
turn off RT?

 

[ProfSeaGoat]

Change your windows power options to balanced, make sure in NVCP that you set the card to optimal power.

 

[ldefinis]

reinstall gpu driver?
turn off RT?

Like the GPU drives are freshly installed as it is when I upgraded. I don't even think Destiny 2 has RT but even still that's not much of a long term solution. I bought an RTX card for a reason..

 

[ldefinis]

Sounds like you have a FPS cap limit somewhere.

Can that be something outside of the game? there is no in-game cap, I specifically changed the settings but also the settings have not changed from my previous GPU.

 

[ProfSeaGoat]

Can that be something outside of the game? there is no in-game cap, I specifically changed the settings but also the settings have not changed from my previous GPU.

I re-read your post and 70-80 FPS sounds like there is no cap on thinking it over again.
I have suggested since that you may need to change your windows power plan to balanced and see if NVCP has your GPU at "Optimal power" in 3D settings menu.

 

[ldefinis]

I re-read your post and 70-80 FPS sounds like there is no cap on thinking it over again.
I have suggested since that you may need to change your windows power plan to balanced and see if NVCP has your GPU at "Optimal power" in 3D settings menu.

Is this the Power management mode? should I change it to normal instead of Max Performance?

Also, I changed the frame rate cap to ON and set it to 1000FPS and this has actually allowed 40% usage of GPU and around 110 frames which is good but this also tells me that I'm missing out on a lot of performance.

 

[ProfSeaGoat]

Is this the Power management mode? should I change it to normal instead of Max Performance?

Also, I changed the frame rate cap to ON and set it to 1000FPS and this has actually allowed 40% usage of GPU and around 110 frames which is good but this also tells me that I'm missing out on a lot of performance.

I did this exact thing with Crysis 3 on my RTX 3070 with Ryzen CPU, 1000 FPS cap in NVCP with Latency on Ultra and went from 72 FPS to over 200.

Seems reliable to do this in some games.

The i7 8700 is not that slow... but maybe it is? Bios settings maybe?

The 8700 is comparable to the 2600X which would bottleneck a 3060 Ti so perhaps the CPU is too slow?

Nvidia drivers have larger overhead with RTX series cards so the CPU could be seeing a way tougher load.

 

[ldefinis]

I did this exact thing with Crysis 3 on my RTX 3070 with Ryzen CPU, 1000 FPS cap in NVCP with Latency on Ultra and went from 72 FPS to over 200.

Seems reliable to do this in some games.

The i7 8700 is not that slow... but maybe it is? Bios settings maybe?

The 8700 is comparable to the 2600X which would bottleneck a 3060 Ti so perhaps the CPU is too slow?

Nvidia drivers have larger overhead with RTX series cards so the CPU could be seeing a way tougher load.

It would just be strange that I got higher frames on my older GPU with he same CPU and NOW its the bottleneck. I went back to Blanaced mode as suggested and I'm getting the same 30-40% GPU usage and 110 frames on average. My GPU doesn't even go above 57C while gaming. I'm still missing something here.

 

[ProfSeaGoat]

It would just be strange that I got higher frames on my older GPU with he same CPU and NOW its the bottleneck. I went back to Blanaced mode as suggested and I'm getting the same 30-40% GPU usage and 110 frames on average. My GPU doesn't even go above 57C while gaming. I'm still missing something here.

Driver overhead as mentioned above.

 

[hotaru.hino]

Nvidia drivers have larger overhead with RTX series cards so the CPU could be seeing a way tougher load.

This is only a problem if you're using a low-end CPU and people have pointed out the games they tested were in DX12, which AMD tends to have a lead in.

 

[ProfSeaGoat]

This is only a problem if you're using a low-end CPU and people have pointed out the games they tested were in DX12, which AMD tends to have a lead in.

People pointed out that they tested with DX12, yet their up to date benchmarks are almost all new games so obviously it would be DX12 included.
It shows that this is no evidence against my original thoughts.

It would help tremendously if someone did try out all API's.

EDIT, I am downloading Destiny 2 right now to see how the game runs on my own system, we can calculate the general increase in FPS to see if it is reasonably mathematically accurately ahead of this persons i7 8700 at which point the problem will be solved as a CPU bottleneck.
If it is proportionately way faster then there is something else that is the problem.
If my system performs similarly then it is likely a game issue.

 

[hotaru.hino]

It shows that this is no evidence against my original thoughts.

Well, I'd hardly call an i7-8700 or an R7 2700X "low end" these days.

And digging around regarding this, someone did some testing on AnandTech's forums (though they didn't really give much specifics). Between a RX 6900 and RTX 3090, they performed basically the same relative to 100% until it gets to an i7-4770K or Ryzen 5 1400.

So I don't think a 3060 Ti's going to be hurting for performance due to driver overhead. As long as you have spare CPU cycles, driver overhead is not going to affect performance that much.

 

[ProfSeaGoat]

Well, I'd hardly call an i7-8700 or an R7 2700X "low end" these days.

And digging around regarding this, someone did some testing on AnandTech's forums (though they didn't really give much specifics). Between a RX 6900 and RTX 3090, they performed basically the same relative to 100% until it gets to an i7-4770K or Ryzen 5 1400.

So I don't think a 3060 Ti's going to be hurting for performance due to driver overhead. As long as you have spare CPU cycles, driver overhead is not going to affect performance that much.

In CPU limited titles which Destiny 2 is, the 2600x is barely any better than a 4790k.

They are both almost exactly the same in IPC. I used to own a 2600X that would bottleneck my 5700-XT by a good amount.

 

[ldefinis]

People pointed out that they tested with DX12, yet their up to date benchmarks are almost all new games so obviously it would be DX12 included.
It shows that this is no evidence against my original thoughts.

It would help tremendously if someone did try out all API's.

EDIT, I am downloading Destiny 2 right now to see how the game runs on my own system, we can calculate the general increase in FPS to see if it is reasonably mathematically accurately ahead of this persons i7 8700 at which point the problem will be solved as a CPU bottleneck.
If it is proportionately way faster then there is something else that is the problem.
If my system performs similarly then it is likely a game issue.

I am very curious to hear the results and thanks in advance for the help.

 

[ProfSeaGoat]

I am very curious to hear the results and thanks in advance for the help.

No problem, I will get back to you tomorrow as this is a big download and it is rather late where I live.

 

[ProfSeaGoat]

I am not sure about GPU usage, I assume it is hitting my GPU's limits but I have double your FPS. I tested at low maxed and medium and was still comfortably way ahead of your setup even though the 3070 is not much faster than the 3060 Ti.

I have a video of it uploading, I used OBS which takes around 10 FPS off with the settings I use as I dislike Geforce experience. I was averaging around 170 FPS at 2560x1440.

I am going to work shortly so will post the video when I am home later.

 

[ldefinis]

I am not sure about GPU usage, I assume it is hitting my GPU's limits but I have double your FPS. I tested at low maxed and medium and was still comfortably way ahead of your setup even though the 3070 is not much faster than the 3060 Ti.

I have a video of it uploading, I used OBS which takes around 10 FPS off with the settings I use as I dislike Geforce experience. I was averaging around 170 FPS at 2560x1440.

I am going to work shortly so will post the video when I am home later.

Thanks for confirming that my card is definitely underperforming. Does the video show temps, power settings, and CPU usage?

 

[Karadjgne]

I hate the term 'bottleneck' because it's barely understood, often misused and more often used by those who want to sound like they kniw what they are talking about.

A bottleneck is a component that slows down the flow of data. Since the cpu is the source of data, it cannot be a bottleneck. It is what it is. A gpu can most definitely be a bottleneck, as can a monitor, software, OS.

If a cpu is setting 100fps limit, and gpu is only rendering 60fps, that's a bottleneck, gpu bound. If the cpu is setting 100fps and the gpu can hit that 100fps and has room to exceed that, the cpu is not a bottleneck, the software is, the gpu isn't under performing, it's performing at 100% ability, just not 100% capacity. Cpu bound.

Most titles today are cpu bound at 1080p, the cpu is hitting its limits and that's all you get, regardless of what the gpu is capable of.

Usage is also highly misunderstood. It's not the amount of the processor used, but the time in idle between use. All processors, cpu or gpu, are normally at 100% ability. It'll put out all the fps it can. If that only requires 30% actual working time per cycle to do, then that's all it requires. 100% usage means the processor has zero time for anything else in the cycle. So if you are at 100% just walking through town and a tank blows out the wall next to you, the fps will go in the toilet because now the cpu has to also deal with vector analysis, motion on massive amounts of tiny objects, lighting affects and other things and has zero time to do so, so takes much longer to prep a frame.

Op. Did you use DDU through safe mode to uninstall the old drivers before installing new ones? Have you updated bios and chipset drivers to the last released? Many older bios don't recognise RTX for what it is, they'll recognise the card but don't have a clue how to deal with it.

 

[ldefinis]

I hate the term 'bottleneck' because it's barely understood, often misused and more often used by those who want to sound like they kniw what they are talking about.

A bottleneck is a component that slows down the flow of data. Since the cpu is the source of data, it cannot be a bottleneck. It is what it is. A gpu can most definitely be a bottleneck, as can a monitor, software, OS.

If a cpu is setting 100fps limit, and gpu is only rendering 60fps, that's a bottleneck, gpu bound. If the cpu is setting 100fps and the gpu can hit that 100fps and has room to exceed that, the cpu is not a bottleneck, the software is, the gpu isn't under performing, it's performing at 100% ability, just not 100% capacity. Cpu bound.

Most titles today are cpu bound at 1080p, the cpu is hitting its limits and that's all you get, regardless of what the gpu is capable of.

Usage is also highly misunderstood. It's not the amount of the processor used, but the time in idle between use. All processors, cpu or gpu, are normally at 100% ability. It'll put out all the fps it can. If that only requires 30% actual working time per cycle to do, then that's all it requires. 100% usage means the processor has zero time for anything else in the cycle. So if you are at 100% just walking through town and a tank blows out the wall next to you, the fps will go in the toilet because now the cpu has to also deal with vector analysis, motion on massive amounts of tiny objects, lighting affects and other things and has zero time to do so, so takes much longer to prep a frame.

Op. Did you use DDU through safe mode to uninstall the old drivers before installing new ones? Have you updated bios and chipset drivers to the last released? Many older bios don't recognise RTX for what it is, they'll recognise the card but don't have a clue how to deal with it.

Hey first off thanks a lot of the explanation and for taking the time to respond.

I actually did not use DDU through safe mode to uninstall the old drives before installing new ones. This was my first time upgrading. I was under the impression that going from Nvidia GPU to another Nvidia GPU would be plug and play for the most part. What steps would you suggest at this point?

What CPU you would recommend I get that is significantly better than my current one to maximize my GPU.

Also, update on some potentially key information. I'm running windows 11 and my power supply is 3ish years old (if that matters).

Am I able to post a screenshot of my temps/settings? I'm pretty new here.

 

[ProfSeaGoat]

I am unsure of what the other person is even talking about. Your 3060Ti is not anywhere close to being almost 2x slower than my 3070.

The FPS difference is all CPU which means the CPU you use is bottlenecking your GPU.

Here is the video. I have HDR enabled so things look overly bright.

 

[ldefinis]

I am not sure about GPU usage, I assume it is hitting my GPU's limits but I have double your FPS. I tested at low maxed and medium and was still comfortably way ahead of your setup even though the 3070 is not much faster than the 3060 Ti.

I have a video of it uploading, I used OBS which takes around 10 FPS off with the settings I use as I dislike Geforce experience. I was averaging around 170 FPS at 2560x1440.

I am going to work shortly so will post the video when I am home later.

I realized an important detail. Is your FPS while in orbit or in a PVP match? The numbers I were giving you are in a PVP match mid-game.

 

[ProfSeaGoat]

I realized an important detail. Is your FPS while in orbit or in a PVP match? The numbers I were giving you are in a PVP match mid-game.

I can do a PVP match when I am home.
Off my break now.

 

[ldefinis]

I can do a PVP match when I am home.
Off my break now.

Regardless I ran the same exact mission you were doing in your video and I'm sitting around 110 frames average on lowest settings. The gap in frames still doesn't make sense to me.
How much better is your CPU?

 

[Karadjgne]

I'm sitting around 110 frames average on lowest settings

Change that to High. If the gpu is whats at fault, your fps will tank hard with the additional workload. If fps really doesn't change much, then you are cpu bound for some reason. If that's the case, you'll need to think in a totally different direction.

But first get the bios and drivers straight.