For comparison purposes. Review published on March 23rd, 2023.
In the screenshot above, the difference between the old and new measurements is the chipset drivers.
The single threaded result with the latest drivers is 26% higher than OP’s latest run.
Multi is 6% higher.
The raw clock rate of DDR memory is half the rated speed, but the doubling of the data processing per clock cycle effectively doubles the speed. So a 3000MHz raw clock speed would make sense for a 6000MHz rated stick. I’ve also seen tech channels starting to refer to the rated speeds as megatransfers per second, which I guess is the more technically correct phrasing.
What is throwing me is your screenshot shows 64 GB memory, but your initial post said you had 32 GB (2x16). Did I miss a change somewhere in the discussion?
Gotcha. I would still try swapping sticks around and only running one at a time just to rule that out, because that is very similar to @Baracus250 problem, and he got acceptable performance with only one 32 GB stick when testing.
@kido007dz did I understand the comment to you on the thread about 7800X3d performance, is he saying we should expect to get frame limited long before the cpu reaches 100% utilization? Does that mean the 50-70% utilization I’m seeing is to be expected?
If this is the comment about LaGuardia, there was mention of an overall CPU utilization. So if you have a 16 cores CPU you can have 30% overall CPU utilization, and at the same time the main thread, the core that MSFS uses most can approach 100%, in which case you are CPU limited.
You can also be GPU limited, in some cases.
In some cases, one can be thermal limited: if cooling is not efficient, CPU can throttle due to high temperatures.
That’s why when troubleshooting, watching both RTSS data and the MSFS dev tool helps to have an understanding of the situation.
Good board, congratulations. If was on my shortlist as well.
For my virtualization setup, very peculiar needs, it will be either the Asrock Taichi or the Asus Proart
testing in VR with Ultra this evening, Varjo Aero at 39ppd. dx11 DLSS DLAA
If I turn down the TLOD and OLD I can see it is gpu bound, but when they are at 200 it seems like it is going back and forth from gpu bound to limited by main thread.
the gpu looks like it is working hard
the cpu doesn’t look like it is working hard and frame times are long
HAGS off Game Mode off, Gamebar off, Core Isolation, turned off driver monitoring, everything else was already off.
stuttery and image quality not as good as it was.
The above snapshot was used by @joeaudacious a few posts up and I just noticed something funny. Looking at the upper left hand side where it lists the L1, L2, and L3 cache amounts. L1 and L2 cache are correct for the 7950x3D, however, the L3 cache indicates only 64M and it should be 128M.
Perhaps another tool such as CPUID can be used to double check as perhaps HWiNFO64 has not been updated yet to support the 7950x3D.
My machine came with AMD chipset driver 5.02.19.2221 which is the latest one available from AMD website. I’ve re-installed it to make sure they installed all the needed things for cache parking. However, I notice that version Asus makes available on the support page for this motherboard is 5.01.03.005. Is there any point in trying this slightly older version?