x570 Aorus Ultra, Ryzen 3700x, F4-3600C16D-32GTZNC

[mass85]

I am having problems with following setup (brand new PC):
- Gigabyte x570 Aorus Ultra - latest BIOS (F10a)
- AMD Ryzen 3700x
- G. SKill Trident Z Neo (AMD) DDR4 32GB (2x16GB) 3600MHz CL16 1.35V XMP 2.0 - F4-3600C16D-32GTZNC, installed in the slots recommended in Gigabyte manual (A2, B2)

When I run RAM at default frequency it seems fine. However I bought this memory in order to work at its max performance - X.M.P. profile 3600 16-19-19-39-58-1.35V. So when I enable this profile and leave the rest of BIOS settings at default values, I get unstable operation under WIndows 10 (with latest updates). System frequently freezes for a second or two, once it also rebooted and cleared CMOS.

I run Windows memory diagnostics in standard mode with 2 runs and it did not detect any errors. However running it in extended mode hang at 21% of the very first run - this repeated twice. Is my memory broken?



I did a bit more testing and it turns out that system is also unstable when using single memory bank in slot A2. Checked both banks with the same results - 3D Mark test ends with reboot, extended Windows memory diagnostics hangs at 21%. When XMP is disabled, 3D Mark completes.

I've run MemTest86 and it did not detect any errors. Does it mean RAM is OK and I should look for root cause elsewhere? At some point I thought this issue was related to Core Performance Boost, because MemTest86 did not boost cores and there were no crashes, errors. But computer restarts also in 3d Mark, when I disable CPB and enable XMP. What should be my next steps in debugging?

[GSKILL TECH]

With XMP profile enabled, manually set DRAM Frequency to DDR4-3200 to see if it can work better.

[mass85]

With XMP profile enabled and DDR4-3200 it no longer crashes in Cinebench and 3d Mark, however I don't see performance gain compared to XMP disabled (perhaps due to timings).

EDIT: actually I did not see performance gain in Cinebench, but in 3d Mark, DaVinci Resolve benchmark and Far Cry Primal I can see improvement by a few percent, in Far Cry even above 10%. However it does not make me happy, as I bought this RAM to work at 3600MHz CL16.

[emissary42]

Even though the modules are tested for it, maybe your CPU & MB just can't handle the full DDR4-3600 at XMP settings. With that being said, try increasing the memory frequency step by step to DDR4-3466/3533 while keeping the XMP atctive and recheck for performance and stability.

[mass85]

I have checked both DDR4-3466 and 3533 with positive results. I think I've read somewhere that increasing RAM voltage may improve its stability when overclocking, so maybe I should try that with DDR-3600?
emissary42, are you suggesting that my pieces of hardware are performing below advertised levels and I should request a replacement?

EDIT: Actually in case of DDR4-3533 running a CPU stress test in CPU-Z causes computer restart after an hour. DDR4-3466 restarts after 5hours of that test.

[TechyMike]

I understand that you have the latest BIOS but that doesn't mean Gigabyte has a problem that they are working on for a fix. I'm guessing that you have tried running just one stick at a time to verify it isn't a defect in the memory?

[mass85]

Yes, I mentioned in the first post that I reproduced the problem with each single stick. I'm going to run more MemTest86, maybe one attempt isn't enough.

EDIT: Unfortunately MemTest86 is now unable to finish, it keeps hanging in the middle before it detects any errors, just like Windows extended memory diagnostics.

[mass85]

I decided to return this PC (except from RAM and SSDs) and buy a new set with just one change - high end PSU. It arrived with a motherboard upgraded to latest version F11 and so far it has been working very well in XMP at DDR-3600. Not sure what helped, different piece of hardware, new PSU or the new BIOS (which is supposed to have better memory compatibility according to release notes).