Brand New Radeon 6750 XT with very poor performance compared to global averages

[RobHydell]

I was hoping someone could help me figure out why my brand new card is performing so poorly in comparison to global averages (based on Passmark scores). I understand I would lose a little performance due to the card being PCIE 4.0 (and the motherboard being 3.0) but not a loss of about 50% off the average). All drivers have been updated and I just cleanly reinstalled my OS. Something else I should add I thought possibly this was a CPU Bottleneck issue as I do see the CPU Utilization in the high 90%'s while the GPU is in the high 80%'s during Starfield gameplay. The only thing is during the passmark testing (all below) I never once saw the CPU go above GPU utilization (would this indicate no CPU bottlenecking)?

1) Do you see any glaring issues that may lead me to understand why the scores are so low?
2) Any suggestions I should try to get the performance up?


[PASSMARK SCORES]
TEST | My Scores | Passmark Global Averages
3D Graphics Mark | 15,130 | 21,145
DirectX 9 | 176FPS | 260FPS
DirectX10 | 122 | 133FPS
DirectX11 | 175 | 184FPS
DirectX12 | 46FPS | 77FPS
GPU Compute | 7329 | 11065


[SYSTEM INFORMATION]
OS Name Microsoft Windows 11 Pro
Version 10.0.22621 Build 22621 (Windows 11 22H2)
Motherboard MS-7917 (MSI Z97 Gaming 5) (PCIE 3.0x16)
System Type x64-based PC
Processor Intel(R) Core(TM) i7-4790K CPU @ 4.00GHz, 4001 Mhz, 4 Core(s), 8 Logical Processor(s)
Graphics AMD Radeon RX 6750 XT 12GB Vram (PCIE 4.0) (Driver version 23.9.2 = Latest)
Drive Western Digital 960GB WD Green SN350 NVMe WDS960G2G0C
Power Supply Antec EarthWatts EA-650 Platinum (650W)
BIOS Version/Date American Megatrends Inc. V1.13, 2/16/2016
Installed Physical Memory (RAM) 16.0 GB (G.Skill F3-2133C10D-16GAB)
Total Virtual Memory 63.9 GB
Available Virtual Memory 50.2 GB

[RobHydell]

Correction, after playing Starfield for a while on Ultra settings my CPU Utilization rarely ever goes over my GPU Utilization numbers. Not sure if this makes a difference.

[David (PassMark)]

CPU percentage load figures are deceptive when you have hyperthreading on. The first 4 CPU cores can do a lot more work than the additional 4 hyperthreaded cores. In some cases the hyperthreading cores decrease performance. So on a 4/8 Core CPU, anything above 50% load basically means the CPU is fully loaded.

This is a 9 year old CPU with a 9 year old motherboard with slow DDR3-2133 RAM. Probably isn't reasonable to expect modern day performance from it.

Also what monitor resolution are you running? You might be getting penalties on the DX12 test for not having 4K support.

[RobHydell]

Thanks for the quick response (amazing)! I appreciate your taking the time.

I am running a resolution of 1920 x 1080 both in windows and in Starfield
I agree with the old CPU, Motherboard (PCIE 3.0), RAM (DDR3) (and additionally a M.2 drive using only half the lanes) being an issue.
I am not sure if this tells us anything more but In order to test the above I swapped cards with my sons computer (See details below) to see what type of results I would get. Oddly the results were about the same on both sides even though his computer is slightly newer. Could this mean an issue with the card or same answer both machines are just too old to take advantage of the new card? I don't expect to get full use out of this card until I upgrade the full rig, I just want to make suer there might not be a config change I should be doing or slight chance it's the card itself.

[Son's Computer]
OS Name Microsoft Windows 11 Pro
Version 10.0.22621 Build 22621 (Windows 11 22H2)
CPU=AMD Ryzen 5 2600
Video Card=GeForce GTX 1070 Ti
Motherboard=ASUS PRIME B450M-A
RAM=2x8GB DDR4 Corsair Vengeance CMK16GX4M2B3200C16
Storage=Samsung SSD 850 EVO

Radeon 6750 XT running on the above computer = Passmark 3D = 14888 (Versus 15,130 on my computer)
NVIDIA 1070 Ti running on my computer = Passmark 3D = 13424 (Versus 14562 on his computer)


Again thank you in advance for any help, suggestions, or comments.

[David (PassMark)]

DX12 test was designed to run at 4K. So not having 4K will hurt DX12 result, and thus the overall result as well.

Yes, I also would have expected slightly better results from the Ryzen 5 2600. But looking at the single thread chart, the i7-4790K is actually the faster CPU, despite being older
The i7-4790K was a great CPU when it was released. And from a CPU point of view the 3D tests are largely single thread. Not having PCIe4.0 hurts a bit as well.

But maybe there is some tweaking that can be done to get slightly better performance from it.

[RobHydell]

Great information to digest thank you very much. I feel much better that the poor results are due to my CPU and other computer components as a whole (and not something wrong with the GPU). Since it is only 4 days old I wanted to make sure while I could still return it. It sounds like the best option is to keep the 6750 XT (as opposed to buying an earlier generation high end card like a 2080 ti) with my current rig (games run pretty good for me) and then I should definitely get scores closer to the average once I can afford to upgrade my CPU, RAM, Monitor....well everything but the case.

Let me know if you disagree with this, otherwise thank you so much for the assistance. You have set my mind at ease and were a great help.

[David (PassMark)]

Yes, but also check cooling/temperatures & the speed of the PCIe slot is running at.

[RobHydell]

Not sure about what is a normal fan speed but I have been keeping an eye on GPU Temperature. It's generally at about 70C while playing Starfield on High or Ultra. While running the passmark tests it was at about 60C (see image)

As far as the PCIe / Bus GPUZ says the card is 4.0 x 16 (true) and it's currently running at 3.0 x 16 (also true due to the old motherboard). (see also image for full GPU-Z results.