XMP unstable after installing 3080

[Kamov23]

Hello!

 

So it's taken me a while to narrow down this issue but I would appreciate any help!

 

Specs:

OS: Win 10 Pro 20H2 19042.928

MB: X570 Aorus Master (BIOS F32)

CPU: AMD 3700X

Ram: G Skill 3600Mhz CL16 (4x8GB)

PSU: Corsair HX 750 Watts (Platinum)

GPU: EVGA 3080 XC3 Black Ultra

Storage: Samsung 1 Tb 970 Evo M.2

 

Originally built this system with everything (minus the 3080) in November of 2019 with a GTX 1060 while I waited for the 3000 series. It's been great and completely stable. December of 2020, managed to get a 3080. Since then, some games have been crashing on me. No error message, just a CTD. Reliability monitor would either say "APPCRASH" for whichever game I was playing or occasionally would throw a "Hardware error" message. At first I though it was the GPU since that's the only thing I replaced, but after stress testing it...it would always pass. Lately I tried turning off the XMP profile....and what do you know, the crashes are all gone.

 

I've run benchmarks and stress tests on every component. Memtest86 ran for 7 hours with no issue on my XMP profile. OCCT ran on the CPU, GPU, ram and PSU with no issues. So individually my components are fine (maybe?). It's just when I enable XMP with the 3080 and play demanding games that something doesn't work (Also, some games are completely fine, the issue is pretty hit or miss). So for now I'm running my ram at 2133 Mhz, but would appreciate some help in getting it back up to spec .

 

Cheers!

[Plermpel]

You could try to taise the DRAM and SoC voltage a little bit. Maybe the current spikes of the GPU lead to unstable voltages.

Did you maybe update your BIOSfor ResBAR?

[Guest]

Try XMP enabled, but manually set dram frequency to 3200mt/s. It should be a bit more stable with 4x8gb installed.

[Kamov23]

25 minutes ago, Plermpel said:

You could try to taise the DRAM and SoC voltage a little bit. Maybe the current spikes of the GPU lead to unstable voltages.

Did you maybe update your BIOSfor ResBAR?

I will try! What is "a little bit"? I've never done anything manually with ram speed, timings or voltage before so this is pretty new to me.

I've had this issue since December of last year (been trying to narrow it down lol). Since then I've updated my drivers, BIOS, and GPU firmware multiple times to the latest available with the issue still persisting. Currently have ReBar enabled.

 

12 minutes ago, ShrimpBrime said:

Try XMP enabled, but manually set dram frequency to 3200mt/s. It should be a bit more stable with 4x8gb installed.

Thanks! I will try that too!

 

[Plermpel]

What is you SoC Voltage, should be 1.1 for 3600Mhz

Memory is usually 1.35, try to raise Ram to 1.36 if that doesn't help raise SoC to 1.11

 

DRAM calculator is also always a good way to get desired RAM speeds.

 

You can also tryto "manually" set your XMP 

 

Infinity Fabric to 1800

Memory 3600

 

And the first 5 timings your RAM was advertised with.

 

Set the SoC and DRAM voltage and leave the rest on Auto.

 

 

Ram can be frustrating on Ryzen. For example my Patriot Viper doesn't run 3600Mhz CL16 on my B550 Pro4 but it does run on my X570 Strix. My Ballistix sticks do work full speed on the B550 but not on the X570. Both single rank, almost same timings but Micron vs. Hynx die.

 

Be prepared to reset your BIOSabout 200x and invest a few hours and you might even get above stock speed.

 

 

 

 

 

[Mister Woof]

One thing that's possible is that the 3080 dumps a LOT of waste heat into your case and can drastically increase temperatures of your other components.

 

Your otherwise xmp stable RAM at 45-50c could possibly become unstable at 60c (or whatever, made up numbers) after being heat soaked by your 3080 at 340+ watts.

 

Samsung modules in my understanding will be much more sensitive to heat compared to say, micron. If the other suggestions here don't work, perhaps try adding some positive airflow over your RAM.

[Kamov23]

So an update on this:

 

I tried lowering the speed down to 3200Mhz. Although it did seem to improve things, the issue did crop up again.

 

Raising SOC voltage and Memory voltage seemed to have fixed it for now though (even at 3600mhz). My only question is that although I set the ram to run at 1.36V, hwinfo reports that it's running between 1.38 and 1.39V. I know that's safe, but why the discrepancy?

 

For those wondering, the ram chips are made by Hynix (didn't know until I looked into it). And it runs at around 55 - 58C under load.

[Mister Woof]

20 minutes ago, Kamov23 said:

So an update on this:

 

I tried lowering the speed down to 3200Mhz. Although it did seem to improve things, the issue did crop up again.

 

Raising SOC voltage and Memory voltage seemed to have fixed it for now though (even at 3600mhz). My only question is that although I set the ram to run at 1.36V, hwinfo reports that it's running between 1.38 and 1.39V. I know that's safe, but why the discrepancy?

 

For those wondering, the ram chips are made by Hynix (didn't know until I looked into it). And it runs at around 55 - 58C under load.

I think it's mainly the Samsung chips that are most sensitive to temperature. IIRC they start getting funky after 50-60c or so.

[thezp]

On 5/12/2021 at 7:52 PM, Kamov23 said:

Hello!

 

So it's taken me a while to narrow down this issue but I would appreciate any help!

 

Specs:

OS: Win 10 Pro 20H2 19042.928

MB: X570 Aorus Master (BIOS F32)

CPU: AMD 3700X

Ram: G Skill 3600Mhz CL16 (4x8GB)

PSU: Corsair HX 750 Watts (Platinum)

GPU: EVGA 3080 XC3 Black Ultra

Storage: Samsung 1 Tb 970 Evo M.2

 

Originally built this system with everything (minus the 3080) in November of 2019 with a GTX 1060 while I waited for the 3000 series. It's been great and completely stable. December of 2020, managed to get a 3080. Since then, some games have been crashing on me. No error message, just a CTD. Reliability monitor would either say "APPCRASH" for whichever game I was playing or occasionally would throw a "Hardware error" message. At first I though it was the GPU since that's the only thing I replaced, but after stress testing it...it would always pass. Lately I tried turning off the XMP profile....and what do you know, the crashes are all gone.

 

I've run benchmarks and stress tests on every component. Memtest86 ran for 7 hours with no issue on my XMP profile. OCCT ran on the CPU, GPU, ram and PSU with no issues. So individually my components are fine (maybe?). It's just when I enable XMP with the 3080 and play demanding games that something doesn't work (Also, some games are completely fine, the issue is pretty hit or miss). So for now I'm running my ram at 2133 Mhz, but would appreciate some help in getting it back up to spec .

 

Cheers!

I am having a similar issue with my RAM after installing a 3080 on a Ryzen system. I am also using Gskill memory, 2x8GB sticks, 3600Mhz CL16. What SOC voltage are you running at now?