Hello,
i just got a new msi rtx3090 (no OC, just stock), in a dell 7920 dual proc, 64gb ram, so a quite good workstation.
i ran the lastest benchmark and have very low results compared to other rtx3090 benchmarks
in cuda mode (gpu only), i get around 1250, when others reach 2000
in rtx, i get around 2450, while others have easily 3000
i installed tha very latest nvidia drivers.
Do you guys have any idea what might cause this ?
edit : i checked also with msi afterburn, gpu is at 100%
Thanks!
Have a buddy with an EVGA FTW3 Ultra that is also seeing Vray 5 Cuda scores lower than what is posted - 1554 (by changing the Windows GPU scheduler). His RTX score is 2348.
So, are the benchmark values at this point due to severe overclocks or are there some significant driver differences or …
We did note a 15% to 20% increase in performance by using the hardware GPU scheduler in Windows 10. You need a recent Windows build and driver for this to be available.
Also, I noticed the scaling numbers are no longer as linear as they were in V-ray Next’s CUDA benchmark.
Edit: We did some slight overclocking and got scores around 1700 and 2500 which seem to be more inline. I noticed the scaling on the multiGPU setups is more like 2500, so I wonder if those 3000 numbers are just aggressive overclocking or a particular vendor’s board that is set more aggressively.
yeah, well strange still i 'm in the verry bottom of all the rtx3090 benhmarks. I’m more like the 3080…
could there be a bottleneck somewhere?
The top scores you are looking at, are achieved with some OC, for example:
V-Ray 5 Benchmark | Chaos, and
V-Ray 5 Benchmark | Chaos.
The average scores for RTX 3090, although, are 1415 (CUDA), and 2462 (RTX), currently, so your scores seem normal to me (we’ll make average scores visible on our website soon).
haaa cheaters ! 
didn’t see immediately the small OC comments, as i’m not used to do them. I like quite and cool workstations
Well, still strange behaviors while testing. I got another msi rtx3090 on another dell7920 workstation. There are exactly the same configs (cpu/proc/sdd/rtx), and i got from 2150 to 2500 on the rtx, and 1225 to 1475 on the cuda.
i found a 10 % performance increase on my 3090 FE when installing the game ready drivers ( 460.89 ) instead of the studio drivers ( 457.30 )…have on RTX = 2534 vs 2378 and on Cuda = 1479 vs 1380
By activating hardware accelerated gpu scheduling i got a big performance increase on the Cuda score= 1928 | RTX score stayed almost the same = 2548
Are there any disadvantages when having this option in Win10 activated at all ?
Anyone did a moderat OC to the new RTX 3090 FE ?