My RX 6700 XT results are similar. But our CPU's make a difference, as your R7 5700X is approx 9-10% better performer than my R9 3900X, and it affects the results to a similar degree.
What are the two Menu drop down menus (Ablage & Bearbeiten) provided for? I can't select anything from either menu, Ablage has no child entries. All the child entries for Bearbeiten are greyed out.
Suggestions:
Adding an option to run the test manually, or to run a second test might be useful.
Also being able to run the test with just the CPU or the DGPU selected as the OpenCL device. The log tab shows the app uses both the CPU and DGPU, deselecting one or the other might give a better indication of the OpenCL result from either device.
Hi,
thanks for your comments.
1. The OpenCL API is only used to measure PCIe & VRAM Speed! CPU is there for sure involved because ists in the que of the transferway CPU-PCIe-VRAM. CPU speed for that two values doesn't matter.
2. Metal API is used for measure GFLOPS. Its GPU only .. but as always CPU Speed/Bus Speed also does matter a bit.
I will add the functin to run the GFLOPS again manually. The PCIe& Vram Speed already measures 3 times & use AVG values at startup time of the app. VRAM & PCIe Speed normally would not change with further runs (after the 3 runs at startup).
Here some M3 Max result.
All Mx CPUs have main benefit of very fast CPU-GPU Transferspeeds - much faster than our PCIe*4 Systems.
One reason why they are very good for FCP / Videocut Apps even the GFLOPS, CPU Compute doesnt reach RX 6800++ speed.
This M3 Max is only on the cpu compute level around the RX 6700XT BUT also has at least double fast CPU<>GPU transferspeeds.
Updated GPU Bench Nbody to V2.5:
- added function to run nbody Gflops again (Button)
- Nbody now runs longer - 30 Sec and also Result will be the average GFlops
- added Info about autor of orig. Nbody Metal , James Price - click on the the link to shown his github website viewn within the app
- scaled PCIe Speed bars better - M3 Max .. are too fast
PCIE Transferspeed is a bit low for an fullspeed PCIe4 x16 Lane system. Can you run the app again if same low 12 GB/s?
OS X system info shows also basic PCie Info abour the gpu card: AMD Radeon RX 5600 XT:
Chipsatz-Modell: AMD Radeon RX 5600 XT
Typ: GPU
Bus: PCIe
PCIe-Lane-Breite: x16
Is the card in the main PCIe Slot (near CPU)? Perhaps also more than 1 NVME cost some Lanes for PCIe Speed.
You can also check MB Bios Settings about PCIe settings.
Here some other user (german hackintosh forum) with also RX 6900XT:
VRAM & PCIe Speeds below normal , much higher than your 127 GB/s VRAM and 13 GB/s PCIE with RX 6900XT!!!
Nice dGPU benchmark App, not seen this one before.
Results on my main desk top system (See White Knight 2.0 in my build sig for full specs)
My GPU is ROG-STRIX-LC-RX6800XT (Liquid Cooled)
AMD Radeon RX 6800 XT:
Chipset Model: AMD Radeon RX 6800 XT
Type: GPU
Bus: PCIe
Slot: PCIEX16
PCIe Lane Width: x16
VRAM (Total): 16 GB
Vendor: AMD (0x1002)
Device ID: 0x73bf
Revision ID: 0x00c1
ROM Revision: 115-D412BS0-101
Metal Support: Metal 3
As my CPU is 10th Gen the dGPU is running on PCIe3 x 16
This site uses cookies to help personalise content, tailor your experience and to keep you logged in if you register.
By continuing to use this site, you are consenting to our use of cookies.