Question Why is my graphics card never at 99%? It's driving me crazy

[Seeryus]

First of all I'll give the specs, as I'm not overly familiar with what components might clash and it could be a simple fix.
GPU - 3060Ti
CPU - i5-9600k
RAM - Corsair 3600mhz (2x8gb)
Motherboard - Z390 Gaming
SSD - 512GB PCS PCIe M.2
HDD - 1TB
So basically whenever I try to play games on a 1080p 144hz monitor, I encounter very minor stutters as well as FPS far lower than what I should be getting from a 3060TI (CoD Mw2 Warzone gives roughly 70-80fps). This is consistent with every game, I assumed it was a bottleneck (CoD MW2's utilization is usually around 60% CPU and 70-80% GPU usage). Is it as simple as just needed a CPU upgrade? I though the i5-9600k would be perfectly fine, although a little behind the power curve, to pair with a 3060TI? Note that if I crank the resolution to 8K, only then do I see 99% GPU utilization and under 25% CPU. Thanks for any help or suggestions.

 

[Lutfij]

Welcome to the forums, newcomer!

What GPU were you on prior to migrating to the RTX3060Ti? I say prior since the rest of your platform specs would indicate that you had a discrete GPU about 2 generations older before the RTX3060Ti.

Z390 Gaming
What is the make and model of your motherboard? Is it one off of Gigabyte's portfolio? What BIOS version are you currently on for said motherboard?

 

[geofelt]

You do not want a graphics card to be running at 99%
That would indicate that the card was a performance limiter.
Ditto for the cpu.
cpu and gpu usage will alternate, and not necessarily overlap.

9600K, as good as it is, may be limiting your performance.
Games will depend on the performance of the single master thread.
To that end, the single thread performance of your cpu becomes very important.
Run the cpu-z bench and look at the single thread rating.
You should see about 566:
https://valid.x86.fr/bench/kn2lme/1
Many submissions will have overclocked their 9700K so do not panic if your results are not as good.

If you play multiplayer games with many participants, then the number of threads becomes important.

 

[Nighthawk117]

I encounter very minor stutters as well as FPS far lower than what I should be getting from a 3060TI (CoD Mw2 Warzone gives roughly 70-80fps)

Stutters are usually CPU related, Warzone is a very demanding game for a 6 core/6 thread CPU.

Can I ask what CPU usage your getting in game?

 

[Seeryus]

Welcome to the forums, newcomer!

What GPU were you on prior to migrating to the RTX3060Ti? I say prior since the rest of your platform specs would indicate that you had a discrete GPU about 2 generations older before the RTX3060Ti.

Z390 Gaming
What is the make and model of your motherboard? Is it one off of Gigabyte's portfolio? What BIOS version are you currently on for said motherboard?

Thanks!

I was on a 1660TI prior to the 3060TI
The motherboard is a TUF Z390-PLus Gaming
Bio was recently updated to 3006

 

[Seeryus]

You do not want a graphics card to be running at 99%
That would indicate that the card was a performance limiter.
Ditto for the cpu.
cpu and gpu usage will alternate, and not necessarily overlap.

9600K, as good as it is, may be limiting your performance.
Games will depend on the performance of the single master thread.
To that end, the single thread performance of your cpu becomes very important.
Run the cpu-z bench and look at the single thread rating.
You should see about 566:
https://valid.x86.fr/bench/kn2lme/1
Many submissions will have overclocked their 9700K so do not panic if your results are not as good.

If you play multiplayer games with many participants, then the number of threads becomes important.

Ah right, I see. I always thought 99% GPU and relatively low CPU was what you wanted, going off all the benchmark tests on Youtube.

I've just ran the bench test and got a score of 480, is that rather low?

 

[Seeryus]

Stutters are usually CPU related, Warzone is a very demanding game for a 6 core/6 thread CPU.

Can I ask what CPU usage your getting in game?

The CPU usage use to be over 90%, but through various tips on the web I've got it down to around 60-70%.

 

[Zerk2012]

To add to what Nighthawk said the game is also poorly optimized.

 

[Phaaze88]

I've got it down to around 60-70%.

What does that mean for each core though?

 

[helper800]

Ah right, I see. I always thought 99% GPU and relatively low CPU was what you wanted, going off all the benchmark tests on Youtube.

I've just ran the bench test and got a score of 480, is that rather low?

That is very low. What are your CPU temps and clocks when running the test? You should be getting around 515 stock. Close all programs that are running in the background and start the test again with the aforementioned stats being watched.

 

[Roland Of Gilead]

Hmmmm. I disagree with some of the points here.

IMO, you do want your GPU to be running at 98/99% This means it's working as hard as it can to output the pre-rendered frames the CPU is supplying it.

If the GPU usage was lower, lets say 70-80%, it's hardly breaking a sweat. Or, is artificially limited by in game frame rate settings, or frame rate cap, or NVCP settings, or monitor res/hz,

CPU usage is going to high for your CPU. Although it's pretty decent, it hasn't aged well, and is limited for host processing. Although 6c/6t, this can hold back your GPU in some CPU demanding games. If it's maxed out all the time, invariably it will then cause stuttering if the GPU is demanding more grunt.

 

[Roland Of Gilead]

Note that if I crank the resolution to 8K, only then do I see 99% GPU utilization and under 25% CPU

This is exactly what you would expect to see, as the load switches from the CPU to the GPU. GPU utilisation will be up, and CPU usage will normalise.

 

[Phaaze88]

Hmmmm. I disagree with some of the points here.

IMO, you do want your GPU to be running at 98/99% This means it's working as hard as it can to output the pre-rendered frames the CPU is supplying it.

If the GPU usage was lower, lets say 70-80%, it's hardly breaking a sweat. Or, is artificially limited by in game frame rate settings, or frame rate cap, or NVCP settings, or monitor res/hz,

CPU usage is going to high for your CPU. Although it's pretty decent, it hasn't aged well, and is limited for host processing. Although 6c/6t, this can hold back your GPU in some CPU demanding games. If it's maxed out all the time, invariably it will then cause stuttering if the GPU is demanding more grunt.

When some gpu heavy effect does come along, like explosions or other, if the gpu has leeway, it can take the hit and keep pace with the cpu.
If it's already sitting at max and the same effects come along, the gpu takes the hit, but 'slows down'.

Put another way:
Cpu sending 100 frames per second to gpu, gpu can send that same rate to your screen throughout if there's enough overhead. You consistently see 100fps in game. [Ideal.]
Cpu sending 100 frames to gpu, gpu dropping to 80 frames per second at times, because it's 'too busy', i.e., 99% or such. You see 100fps, with occasional dips to 80fps. [Less desirable.]
The gpu's 'turn' is always after the cpu's(plus ram).

Maxing out, great for benchmarks/stress tests. Not so much for the actual games.

 

[CelicaGT]

When some gpu heavy effect does come along, like explosions or other, if the gpu has leeway, it can take the hit and keep pace with the cpu.
If it's already sitting at max and the same effects come along, the gpu takes the hit, but 'slows down'.

Put another way:
Cpu sending 100 frames per second to gpu, gpu can send that same rate to your screen throughout if there's enough overhead. You consistently see 100fps in game. [Ideal.]
Cpu sending 100 frames to gpu, gpu dropping to 80 frames per second at times, because it's 'too busy', i.e., 99% or such. You see 100fps, with occasional dips to 80fps. [Less desirable.]
The gpu's 'turn' is always after the cpu's(plus ram).

Maxing out, great for benchmarks/stress tests. Not so much for the actual games.

This is the way. In my opinion there's a lot of malarky surrounding "bottlenecks" and component utilization. Add in variable refresh and things really get fuzzy as most people don't even know how it works or more importantly how it is MEANT to work.

When designing a system I start with the display. Resolution and max refresh rate dictate the rest of the components. If it's a 1440p panel with a 120Hz refresh then I choose components that will support that at about 50-60% average GPU uptime, hopefully same for CPU uptime PER LEVERAGED CORE. These are variables that will change with the game and with time as new game engines require more resources, so, for the latter my chosen system will give me several years of smooth, quiet performance before I need to consider upgrades.

Note that this is how I do it, everyone else can do as they wish but for most users that fact remains....100% GPU uptime is a BAD thing unless you are playing a competitive shooter with a fully unlocked (no Vsync or variable refresh) framerate. It means you bought too little GPU/CPU, or you shot for the moon display wise and fell short. 100% uptime means extra heat and power use, extra noise, wear and tear on fans/pumps, extra dirt on heat sinks due to all the extra airflow needed.

 

[Roland Of Gilead]

If it's already sitting at max and the same effects come along, the gpu takes the hit, but 'slows down'.

This might be the case for Vram usage (as in maxing vram usage), but GPU's (cores) by design are preferred to run at 98/99%. The cores are designed to do this and for sustained gaming sessions. Sure, for a light game you might expect low GPU/CPU usage. For other games, usage may fluctuate during gameplay, between 70-100% this is also to be expected. But for more demanding games yes, you can expect your GPU cores to run at max usage.

And yes, as already pointed out there are many variables that can effect usage.

A google search for 'should GPU usage be at 99% reveals a lot of reference here. They all say the same thing. 99% usage is normal for demanding gaming.

Here's just one example of many: Is 100% GPU Usage Bad? (Read THIS First) | Streamers Playbook

 

[Phaaze88]

A google search for 'should GPU usage be at 99' reveals a lot of reference here. They all say the same thing. 99% usage is normal for demanding gaming.

Here's just one example of many: Is 100% GPU Usage Bad? (Read THIS First) | Streamers Playbook

Gonna' have to agree to disagree on this one.
The same crowd would likely say a cpu AIO/CLC should always be top mounted, and not second guess anything.

 

[helper800]

Gonna' have to agree to disagree on this one.
The same crowd would likely say a cpu AIO/CLC should always be top mounted, and not second guess anything.

There is nothing wrong with top mounting an AIO... The problems come with the confusion behind how to front mount to reduce noise and increase pump life. You can bottom mount an AIO, just dont expect it to last longer than a year or two.

The argument about 70-99% gpu usage being a good or bad thing in either situation is moot. If your CPU is powerful enough to keep the GPU fed at 99%, all that means is Your FPS is being maximized. If you are getting 70-80% GPU usage that means you are leaving room on the table for more FPS in that moment but are not getting it for various reasons. Being at 70-80% is arguably worse because this means that instead of going from 120 down to 100 fps when more effects come onto the screen you just get stuck at 100fps all the time. I would rather be at 120fps part of the time with dips to 100fps than be constantly stuck at 100 or less because there will also be dips from there because the rest of the platform cannot keep up with the GPU.

Another thing is that when a GPU is waiting on the CPU to draw frames you typically see more stuttering issues than if it was the reverse and an increased variability in frame-times.

 

[Roland Of Gilead]

There is nothing wrong with top mounting an AIO... The problems come with the confusion behind how to from mount to reduce noise and increase pump life. You can bottom mount an AIO, just dont expect it to last longer than a year or two.

The argument about 70-99% gpu usage being a good or bad thing in either situation is moot. If your CPU is powerful enough to keep the GPU fed at 99%, all that means is Your FPS is being maximized. If you are getting 70-80% GPU usage that means you are leaving room on the table for more FPS in that moment but are not getting it for various reasons. Being at 70-80% is arguably worse because this means that instead of going from 120 down to 100 fps when more effects come onto the screen you just get stuck at 100fps all the time. I would rather be at 120fps part of the time with dips to 100fps than be constantly stuck at 100 or less because there will also be dips from there because the rest of the platform cannot keep up with the GPU.

Another thing is that when a GPU is waiting on the CPU to draw frames you typically see more stuttering issues than if it was the reverse and an increased variability in frame-times.

Exactly!

 

[Roland Of Gilead]

Gonna' have to agree to disagree on this one.
The same crowd would likely say a cpu AIO/CLC should always be top mounted, and not second guess anything.

There are many example of the same explanation for 99% usage and whether it's right or wrong. They all say the same thing. If you know otherwise, I'd love to see.

Edit: Sorry, that wasn't mean to be some kind of throw down challenge! Rather, I'd just be interested to see where you're coming from.

We're not talking about one AIO or another and where it's mounted, only the singular purpose (mostly) for what a GPU does. As mentioned, they are designed to run at 98/99/100% if not limited by artificial means, or CPU not powerful enough to provide data as quickly as the GPU can process them, or resolution bound etc.

Anywho, I guess you are right about 1 thing. We can agree to disagree

 

[Phaaze88]

There is nothing wrong with top mounting an AIO... The problems come with the confusion behind how to front mount to reduce noise and increase pump life. You can bottom mount an AIO, just dont expect it to last longer than a year or two.

The argument about 70-99% gpu usage being a good or bad thing in either situation is moot. If your CPU is powerful enough to keep the GPU fed at 99%, all that means is Your FPS is being maximized. If you are getting 70-80% GPU usage that means you are leaving room on the table for more FPS in that moment but are not getting it for various reasons. Being at 70-80% is arguably worse because this means that instead of going from 120 down to 100 fps when more effects come onto the screen you just get stuck at 100fps all the time. I would rather be at 120fps part of the time with dips to 100fps than be constantly stuck at 100 or less because there will also be dips from there because the rest of the platform cannot keep up with the GPU.

Another thing is that when a GPU is waiting on the CPU to draw frames you typically see more stuttering issues than if it was the reverse and an increased variability in frame-times.

Never said there was, but some take it as an absolute - don't think or try to do things for themselves, and I don't like that.

There are many example of the same explanation for 99% usage and whether it's right or wrong. They all say the same thing. If you know otherwise, I'd love to see.

Edit: Sorry, that wasn't mean to be some kind of throw down challenge! Rather, I'd just be interested to see where you're coming from.

We're not talking about one AIO or another and where it's mounted, only the singular purpose (mostly) for what a GPU does. As mentioned, they are designed to run at 98/99/100% if not limited by artificial means, or CPU not powerful enough to provide data as quickly as the GPU can process them, or resolution bound etc.

Anywho, I guess you are right about 1 thing. We can agree to disagree

I'm not saying maxed gpu usage is bad - never did. Bad =! less desirable/ideal.
But I'm just serving to take the thread off topic, so that's why I'd chosen to agree to disagree, back off, and only watch the thread.

 

[Roland Of Gilead]

I'm not saying maxed gpu usage is bad - never did. Bad =! less desirable/ideal.
But I'm just serving to take the thread off topic, so that's why I'd chosen to agree to disagree, back off, and only watch the thread.

Agreed.