Question GPU Benchmarking Fine, Game Performance bad

[Gazza160]

Performace in game is not comparable to benchmarking

- Re Seated all components
- Swapped out CPU and RAM (No Luck)
- Cleaned out System
- Re Installed Windows three times
- Cleared CMOS
- Updated BIOS
- Re Installed Drivers
- Cleaned Dust
- Re Applied Thermal Paste
- Control Panel 3D Settings
- Power Plans
- Game Bar, etc Disabled

Before someone tells me to lower game/visual settings. I know 100% that is not the issue due to the fact that prior to the issues occuring I was running most games I play at a solid 120fps+ on high/very high settings, 80-100fps for FiveM/GTAV.

My specs are as followed:
MB: Asus Prime B450M-A II
CPU: Ryzen 5 5600X
GPU: Zotac Gaming RTX 3060 12GB LHR
RAM: 32GB Corsair Vengeance LPX 3200Mhz
PSU: Gigabyte 80+ GOLD 750W, Fully Modular
STORAGE: 2TB TeamMP M.2 SSD

I have your standard air cooler for my 5600X and 3x Case Fans.

I returned my GPU back to the vendor and they ran some tests on it for 5 hours and found nothing wrong, my ram and CPU are all fine and are running normally at standard temps I've been running at for the last 3 years even when I had my old parts installed. Everything in my current system is ~6 months old and my new PSU I just bought on Tuesday.

Download both these as HTML Files
HWINFO64: https://pastebin.com/xTpKnDGt
Unigine Heaven Benchmark: https://pastebin.com/U7Jc618z

 

[Lutfij]

Welcome to the forums, newcomer!

Updated BIOS
+
Asus Prime B450M-A II
What BIOS version are you on at this moment of time?

Re Installed Drivers
Use DDU to remove all GPU drivers(Intel, AMD and Nvidia) from your platform and then manually reinstall with the latest GPU driver sourced from Nvidia's support site in an elevated command, i.e, Right click installer>Run as Administrator.

Re Installed Windows three times
Where did you source the installer for the OS?

PSU: Gigabyte 80+ GOLD 750W, Fully Modular
Gigabyte is the brand of the PSU, while 750W is the advertised wattage for the unit. 80+ Gold is the advertised 80+ efficiency rating. What is the model of the unit?

RAM: 32GB Corsair Vengeance LPX 3200Mhz
Can you pass on a link?

 

[Gazza160]

Welcome to the forums, newcomer!

Updated BIOS
+
Asus Prime B450M-A II
What BIOS version are you on at this moment of time?

Re Installed Drivers
Use DDU to remove all GPU drivers(Intel, AMD and Nvidia) from your platform and then manually reinstall with the latest GPU driver sourced from Nvidia's support site in an elevated command, i.e, Right click installer>Run as Administrator.

Re Installed Windows three times
Where did you source the installer for the OS?

PSU: Gigabyte 80+ GOLD 750W, Fully Modular
Gigabyte is the brand of the PSU, while 750W is the advertised wattage for the unit. 80+ Gold is the advertised 80+ efficiency rating. What is the model of the unit?

RAM: 32GB Corsair Vengeance LPX 3200Mhz
Can you pass on a link?

Afternoon,


RAM:
https://www.scorptec.com.au/product/memory/ddr4-desktop-memory/88788-cmk32gx4m2e3200c16

Windows 11 23H2 (Latest Update): Official Microsoft Site

PSU MODEL: GP-UD750GM

BIOS: PRIME B450M-A II BIOS 4402

Additionally, I've utilized DDU 3 times while running safemode to uninstall ALL of my drivers.

 

[thestryker]

Do any of the games you're having issues with have built in benchmarks you can test to see if the problem is replicated there?

Have you checked to see what GPU and CPU utilization are looking like when having performance problems?

It's not particularly clear has it always been this way with your current components or did it randomly start?

 

[Gazza160]

Do any of the games you're having issues with have built in benchmarks you can test to see if the problem is replicated there?

Have you checked to see what GPU and CPU utilization are looking like when having performance problems?

It's not particularly clear has it always been this way with your current components or did it randomly start?

1. Yes the issue is replicated in gameplay/benchmarking on the few games that allow it.

- Ready Or Not normally I'd get 100-200fps while on medium/low settings.
- FiveM sadly doesnt have a benchmarking tool. However, without any graphical modifications installed, I'd normally get 120-140fps.

2. GPU and CPU utilization are roughly the same.

3. It just randomly started last week after booting up my PC on the Saturday.

Since then I've done everything I have stated in my original post. Including taking my individual parts back to the vendor to have them run stress tests which yeilded average/above average results for a 3060 12GB + 5600X

----- FiveM

https://imgur.com/5rnNzPt

View: https://i.imgur.com/5rnNzPt.jpeg
as you can see in the top left that is what my stats in FiveM have been like for the last week in every server I play on. Including my localhost

----- Ready or Not

https://imgur.com/nmADLxB

View: https://i.imgur.com/nmADLxB.jpeg
If you look in the top left my FPS fluctuates between 10-60fps with vsync turned on. 20-170fps with vsync off

 

[thestryker]

2. GPU and CPU utilization are roughly the same.

What do you mean by this? If the GPU utilization is low then that means it's being bottlenecked by something. Properly running GPU should be well over 80% unless it's an old game or super low settings in which case the CPU would seem to be the source of the issue.

While it's not a good benchmark userbenchmark can be helpful for figuring out which component isn't running right.

What CPU cooler are you using?

With low utilization I'm guessing temperatures aren't high, but it'd be worth checking to see what they look like.

 

[Gazza160]

What do you mean by this? If the GPU utilization is low then that means it's being bottlenecked by something. Properly running GPU should be well over 80% unless it's an old game or super low settings in which case the CPU would seem to be the source of the issue.

While it's not a good benchmark userbenchmark can be helpful for figuring out which component isn't running right.

What CPU cooler are you using?

With low utilization I'm guessing temperatures aren't high, but it'd be worth checking to see what they look like.

In games, monitoring with Ryzen Master and MSI Afterburner,

My CPU is usually at 50-60c
My GPU is usually at 50-65c

Utilization for both while under load is roughly 10-20% (Generally the same, if not a 2-5% difference)

For my CPU Cooler I have a

---

Cooler Master Hyper 212 Black Edition CPU Cooler R2.0

 

[thestryker]

So your GPU is basically not being used at all.

What do your CPU and GPU clockspeeds look like?

 

[Gazza160]

While it's not a good benchmark userbenchmark can be helpful for figuring out which component isn't running right.

I've been trying to run Userbenchmark however keep getting the server is full error. So right now not having any luck.

 

[Gazza160]

So your GPU is basically not being used at all.

What do your CPU and GPU clockspeeds look like?

https://imgur.com/fdesBMc

View: https://i.imgur.com/fdesBMc.png
| GPU Clock Speeds have been a stable 1777 which is the boost clock rated for my card.

https://imgur.com/7TmocIA

View: https://i.imgur.com/7TmocIA.png
CPU is also running at its max boost clock, base clock is 3.7Mhz while Boost is 4.7Mhz for the Ryzen 5 5600X

 

[Gazza160]

So your GPU is basically not being used at all.

What do your CPU and GPU clockspeeds look like?

Finally got UserBenchmark to work by running my VPN.

https://www.userbenchmark.com/UserRun/67220433

 

[thestryker]

| GPU Clock Speeds have been a stable 1777 which is the boost clock rated for my card.

If it's never doing anything other than going to 1777 that's not really a good thing. The boost clocks on nvidia cards don't really mean a whole lot and 1777 is basically what the minimum boost should be. The GPU power is super low so whatever it going on certainly seems to be related to the GPU exclusively. Saw your userbenchmark run which confirms the GPU is the problem area.

Use GPU-Z sensor polling to see what the power consumption looks like as that will show where the power is being drawn from. I believe your card should have a single 8 pin so you should see PCIe Slot Power and a listing for 8-pin power on the sensors tab.

 

[Gazza160]

If it's never doing anything other than going to 1777 that's not really a good thing. The boost clocks on nvidia cards don't really mean a whole lot and 1777 is basically what the minimum boost should be. The GPU power is super low so whatever it going on certainly seems to be related to the GPU exclusively. Saw your userbenchmark run which confirms the GPU is the problem area.

Use GPU-Z sensor polling to see what the power consumption looks like as that will show where the power is being drawn from. I believe your card should have a single 8 pin so you should see PCIe Slot Power and a listing for 8-pin power on the sensors tab.

https://imgur.com/VaWXLku

View: https://i.imgur.com/VaWXLku.png
| Here's what my GPU Z Sensors Tab is looking like with Ready Or Not running in the background.

 

[thestryker]

For Bus Interface does it show x16?

The video card is plugged directly into the primary slot of the motherboard no riser cable or anything correct?

Most games don't render when they're in the background, or are limited if they do, but that graph doesn't indicate anything really different happening as long as GPU-Z was open.

 

[Gazza160]

For Bus Interface does it show x16?

The video card is plugged directly into the primary slot of the motherboard no riser cable or anything correct?

Most games don't render when they're in the background, or are limited if they do, but that graph doesn't indicate anything really different happening as long as GPU-Z was open.

Directly plugged into the MB, no Cable Extensions. Like I said, I just bought my PSU on Tuesday so that is a fresh unit.

Bus Interface is showing PCIE x16 4.0 @ x16x 3.0

 

[Gazza160]

For Bus Interface does it show x16?

The video card is plugged directly into the primary slot of the motherboard no riser cable or anything correct?

Most games don't render when they're in the background, or are limited if they do, but that graph doesn't indicate anything really different happening as long as GPU-Z was open.

I noticed something strange. I launched GTA and realized that everything was fine so it seems like its just certain games, browsing, etc that I have the issue. But I get the issue in mostly every game besides GTA?

Power Draw

https://imgur.com/VJBpwh0

View: https://i.imgur.com/VJBpwh0.png
seems to be fine and whats expected for GTA here but even though its fine here, when launching FiveM. Which is GTA just different launcher due to RP servers I get the issue and same with Ready Or Not, FS22, ATS, 7D2D. So I am clueless

The above IMGUR link is in Regular GTA

This is in FiveM on a server I play on

https://imgur.com/m1HuOie

View: https://i.imgur.com/m1HuOie.png

 

[thestryker]

I noticed something strange. I launched GTA and realized that everything was fine so it seems like its just certain games, browsing, etc that I have the issue. But I get the issue in mostly every game besides GTA?

Power Draw

https://imgur.com/VJBpwh0

View: https://i.imgur.com/VJBpwh0.png
seems to be fine and whats expected for GTA here but even though its fine here, when launching FiveM. Which is GTA just different launcher due to RP servers I get the issue and same with Ready Or Not, FS22, ATS, 7D2D. So I am clueless

The above IMGUR link is in Regular GTA

This is in FiveM on a server I play on

https://imgur.com/m1HuOie

View: https://i.imgur.com/m1HuOie.png

You're still having problems with it barely using your GPU even though it happens to be better than in other games. Your system is acting like it's heavily CPU bottlenecked and there's a hint of it in UBM (the normal section of the benchmark) even though the scores aren't terrible. You can run the CPU-Z benchmark test and your CPU should be in the 4700-4800 range in multi and over 600 single.

While the SSD performance seems off on UBM that's a poor benchmark for it from what I've seen so it'd be worth checking CrystalDiskMark to see if there are any obvious problems.

There's certainly something stopping your video card from performing like it should and it's just a matter of figuring out what that could be.

 

[Gazza160]

You're still having problems with it barely using your GPU even though it happens to be better than in other games. Your system is acting like it's heavily CPU bottlenecked and there's a hint of it in UBM (the normal section of the benchmark) even though the scores aren't terrible. You can run the CPU-Z benchmark test and your CPU should be in the 4700-4800 range in multi and over 600 single.

While the SSD performance seems off on UBM that's a poor benchmark for it from what I've seen so it'd be worth checking CrystalDiskMark to see if there are any obvious problems.

There's certainly something stopping your video card from performing like it should and it's just a matter of figuring out what that could be.

This is what CPU Z ran for the benchmark

https://imgur.com/Wm5NbMf

View: https://i.imgur.com/Wm5NbMf.png


This is what the SSD ran

https://imgur.com/VnltcT9

View: https://i.imgur.com/VnltcT9.png

 

[thestryker]

This is what CPU Z ran for the benchmark

https://imgur.com/Wm5NbMf

View: https://i.imgur.com/Wm5NbMf.png

So your single core is really low so I'd like to think we're onto something here. What's weird is that the multicore is pretty much right where it would be expected to be. Need to try to figure out what that is, and I'm not totally sure what will reflect it. HWInfo/HWMonitor will show clockspeeds so you could check those while running the CPU-Z benchmark or Cinebench single core. Maybe that will expose what's going on, but this is just something I've never seen before.

This is what the SSD ran

https://imgur.com/VnltcT9

View: https://i.imgur.com/VnltcT9.png

SSD is a little slow on the sequentials, but nothing looks obviously amiss here at all.

 

[Gazza160]

So your single core is really low so I'd like to think we're onto something here. What's weird is that the multicore is pretty much right where it would be expected to be. Need to try to figure out what that is, and I'm not totally sure what will reflect it. HWInfo/HWMonitor will show clockspeeds so you could check those while running the CPU-Z benchmark or Cinebench single core. Maybe that will expose what's going on, but this is just something I've never seen before.

SSD is a little slow on the sequentials, but nothing looks obviously amiss here at all.

I now believe that it may be my CPU thats the issue here, I ran the test again with HWINFO open and this is what I got

https://imgur.com/wcDKres

View: https://i.imgur.com/wcDKres.png
but then also had this while running the test

https://imgur.com/Hrfr78R

View: https://i.imgur.com/Hrfr78R.png


I ran the CPUZ Benchmark again and got this

https://imgur.com/46W5O4Y

View: https://i.imgur.com/46W5O4Y.png

 

[thestryker]

Yeah definitely something very wrong there with the CPU. I don't have any direct experience with AMD CPUs new enough to be worthwhile to have much insight as to what could be causing it. I don't think it has anything to do with cooling or temperature though because the multi core would be bad too if it was and single might not be impacted much. It's possible that this may have something to do with PBO settings or the curve optimizer, but I'm not sure how you'd go about resetting everything to default settings.

On my machine I can't use the sensors from HWInfo while running CPU-Z's benchmark so I use HWMonitor instead. Basically what I see while running it is maximum clocks, but 100% usage across all cores during multi, 100% usage on a single core during single.

 

[Gazza160]

Yeah definitely something very wrong there with the CPU. I don't have any direct experience with AMD CPUs new enough to be worthwhile to have much insight as to what could be causing it. I don't think it has anything to do with cooling or temperature though because the multi core would be bad too if it was and single might not be impacted much. It's possible that this may have something to do with PBO settings or the curve optimizer, but I'm not sure how you'd go about resetting everything to default settings.

On my machine I can't use the sensors from HWInfo while running CPU-Z's benchmark so I use HWMonitor instead. Basically what I see while running it is maximum clocks, but 100% usage across all cores during multi, 100% usage on a single core during single.

I reset my clocks back to default and still getting the same. So might just be the CPU.

Will have to get a new one next week due to not having the box for this one.

I'll keep this thread open until I replace my CPU and know it was 100% that