G.Skill Trident Z5 NEO RGB Series 32GB (2 x 16GB) DDR5 6000 CL30-38-38-96 1.35V F5-6000J3038F16G AMD Expo Issue w/ 7950X3D

[Keiko-chan]

Build: 7950X3D, 4090, 360mm AIO and 3 case fans, sabrent rocket 4 plus, asus ROG x670e crosshair hero mobo

Okay, I have expo one enabled and the PC boots and works perfectly fine, but I've had it do like a quick restart a couple times (no blue screen except like a flash of it one time). I ran Karhu and Prime95 both failed at the 45 minute mark. Prime95 says "Self-test 384K passed!
FATAL ERROR: Rounding was 0.4999953054, expected less than 0.4
Hardware failure detected running 392K FFT size, consult stress.txt file." (Nothing but links to forums in stress.txt and i read changing CPU SOC could help, but didnt want to mess with it without proper advice. Also mentioned thermals but mine seem fine, but should I change the preapplied thermal paste on AIO anyway to be safe? I hope I didnt press down to hard or screw it in to far when installing it.). Ive messed with the DRAM Voltage a bit lowering my .005 and raising it, but it made it fail quicker. So I decided to post here. Is this a common issue? Is anyone else having this issue with this RAM? Could it be something else? Or is that sufficient evidence to say its the RAM since all my temperatures are fine? Should I ignore it since it ran all my games fine and wait for another BIOS update currently running 1415 (newest version) or wait for more new chipset drivers(currently running latest on chipset drivers too ? The only two times it shut off was randomly watching YouTube and it literally started right back up in the same spot on the website when I turned back on (nothing closed). Id hate to have to downclock it, but what do you all think? Thanks in advance. TTY soon   

PS. Im also currently running it still on as well. Is this something I should change asap because it could damage my CPU or system? 

[Somerandomtechyboi]

Set vsoc to 1.3v

Set cl to 32

See if that might help or not

[RONOTHAN##]

43 minutes ago, Keiko-chan said:

Is this something I should change asap because it could damage my CPU or system? 

No, RAM instability won't damage the hardware itself. The most that will break is your Windows install, requiring you to reinstall Windows which usually isn't that big a deal. 

 

Do note that ASUS has seemed to have issues with their latest BIOS revisions for memory support. You might get 6000 to work, but you also might not. The issue as far as I can tell seems to be more with the 6000 ratio itself, so counterintuitively you might have better luck at 6200, though given you're this close to being stable I'd say messing with some of these voltages will help a bit more. 

46 minutes ago, Keiko-chan said:

i read changing CPU SOC could help, but didnt want to mess with it without proper advice

For memory frequency itself, higher = better, but for FCLK overclocking lower = better. This all depends on the part of the CPU that's actually experiencing instability. 

 

At 6000MT/s, the FCLK should default to 2000MHz which should be very easy to run for most chips, even at 1.3V. Lock the FCLK into place at 2000MHz, set the SOC voltage to 1.3V (the highest value recommended by AMD), and see if that helps. If that doesn't help, you might be having FCLK instability in which case lowering the SOC voltage down to 1.2V might in theory help. If none of those help, set it to 1.25V and mess with the other voltages instead. 

 

51 minutes ago, Keiko-chan said:

Ive messed with the DRAM Voltage a bit lowering my .005 and raising it, but it made it fail quicker.

Try going in larger steps, 0.005V for memory is within margin of error for the PMIC, it's effectively the same voltage. Go in increments of closer to 0.05V, or at minimum 0.01V and see if that helps. If you're worried about that, 6000 CL30 is a guaranteed Hynix memory bin, and Hynix memory chips are very voltage resistant, being sold in memory bins of over 1.5V. I'd still try to stay at or below 1.4V, partially because it shouldn't be necessary, partially because you're going to likely run into cooling issues at that voltage level, but those higher voltage levels are safe for the memory itself. 

 

Anyway, some other memory voltages you can try are VDDIO_MEM, VDDP, the VDDG voltages, and VDD Misc. Usually, these don't help or only really help with FCLK overclocking, but could in theory help raising these to ~1.1V for the VDDG/VDDP voltages (these sweet spot though, so don't assume higher = better), ~1.2V for the VDD Misc voltage, and anywhere between 1.2 and 1.45V for VDDIO_MEM. If you don't want to do this steps (I wouldn't blame you), lower the frequency down to 5800MT/s and call it a day, that should still work. 

[Keiko-chan]

4 hours ago, RONOTHAN## said:

No, RAM instability won't damage the hardware itself. The most that will break is your Windows install, requiring you to reinstall Windows which usually isn't that big a deal. 

 

Do note that ASUS has seemed to have issues with their latest BIOS revisions for memory support. You might get 6000 to work, but you also might not. The issue as far as I can tell seems to be more with the 6000 ratio itself, so counterintuitively you might have better luck at 6200, though given you're this close to being stable I'd say messing with some of these voltages will help a bit more. 

For memory frequency itself, higher = better, but for FCLK overclocking lower = better. This all depends on the part of the CPU that's actually experiencing instability. 

 

At 6000MT/s, the FCLK should default to 2000MHz which should be very easy to run for most chips, even at 1.3V. Lock the FCLK into place at 2000MHz, set the SOC voltage to 1.3V (the highest value recommended by AMD), and see if that helps. If that doesn't help, you might be having FCLK instability in which case lowering the SOC voltage down to 1.2V might in theory help. If none of those help, set it to 1.25V and mess with the other voltages instead. 

 

Try going in larger steps, 0.005V for memory is within margin of error for the PMIC, it's effectively the same voltage. Go in increments of closer to 0.05V, or at minimum 0.01V and see if that helps. If you're worried about that, 6000 CL30 is a guaranteed Hynix memory bin, and Hynix memory chips are very voltage resistant, being sold in memory bins of over 1.5V. I'd still try to stay at or below 1.4V, partially because it shouldn't be necessary, partially because you're going to likely run into cooling issues at that voltage level, but those higher voltage levels are safe for the memory itself. 

 

Anyway, some other memory voltages you can try are VDDIO_MEM, VDDP, the VDDG voltages, and VDD Misc. Usually, these don't help or only really help with FCLK overclocking, but could in theory help raising these to ~1.1V for the VDDG/VDDP voltages (these sweet spot though, so don't assume higher = better), ~1.2V for the VDD Misc voltage, and anywhere between 1.2 and 1.45V for VDDIO_MEM. If you don't want to do this steps (I wouldn't blame you), lower the frequency down to 5800MT/s and call it a day, that should still work. 

Thank you very much for all of this information. I was worried about messing with the SOC because I'd heard of X3D chips getting fried, and also, when I was putting on the AIO cold plate I couldn't get it to go down (didn't have screw loose enough) and had to press moderately hard to try to get the top latch over, and in doing so, the pre-applied thermal paste pressed against CPU. Then I had to lift it back off the CPU and reposition it a few times, so I thought I may need to reapply paste, but I think im fine on that front considering im at 60C gaming and underload and 40 C browsing. As low as 29(when ambient temps are 22C)-40C when idling. Finally, I ran Karhu and Windows memory diagnostic and got errors also so that pretty much guarantees it's the RAM and not the CPU, right?  I'm going to try all of your suggestions and hope it stabilizes. Currently, I'm slightly above average or average on all my benchmarks (Cinebench r23, TimeSpy, TimeSpy Extreme, AIDA64 Memory Bench), and not had any crashes in games, just the two strange ones while browsing YouTube. But I have my fingers crossed that you have solved this issue for me  I'll keep you posted when I adjust everything. 

 

4 hours ago, Somerandomtechyboi said:

Set vsoc to 1.3v

Set cl to 32

See if that might help or not

Thanks very much, i'll give it a try ASAP! 

[RONOTHAN##]

15 minutes ago, Keiko-chan said:

Finally, I ran Karhu and Windows memory diagnostic and got errors also so that pretty much guarantees it's the RAM and not the CPU, right?

It can be a bad memory controller which is part of the CPU, though dead/not fully seated RAM is more likely. Absolute first thing I'd try is a reseat, then I'd move on to some other stuff. 

[Keiko-chan]

25 minutes ago, RONOTHAN## said:

It can be a bad memory controller which is part of the CPU, though dead/not fully seated RAM is more likely. Absolute first thing I'd try is a reseat, then I'd move on to some other stuff. 

Okay, Ill all so switch from DIMM A2, B2. To A1 and B1  too, thanks for all your help again. Reseating now, doing the rest after. 

 

[RONOTHAN##]

8 hours ago, Keiko-chan said:

Okay, Ill all so switch from DIMM A2, B2. To A1 and B1  too, thanks for all your help again. Reseating now, doing the rest after. 

 

It probably won't work in A1 and B1 because of how memory systems are designed. You're able to try, but you'll almost certainly have to swap back to A2/B2 when it doesn't work. 

[Keiko-chan]

1 hour ago, RONOTHAN## said:

It probably won't work in A1 and B1 because of how memory systems are designed. You're able to try, but you'll almost certainly have to swap back to A2/B2 when it doesn't work. 

Okay, ill leave them as I made sure my RAM was seated well. I didnt reseat yet, i just now turned VDDCR_SOC Voltage to 1.3 and am about to run prime 95. But I also saw this warning before it crashed earlier today in event viewer 

: The embedded controller (EC) returned data when none was requested. The BIOS might be trying to access the EC without synchronizing with the operating system. This data will be ignored. No further action is necessary; however, you should check with your computer manufacturer for an upgraded BIOS.

[RONOTHAN##]

9 minutes ago, Keiko-chan said:

But I also saw this warning before it crashed earlier today in event viewer 

: The embedded controller (EC) returned data when none was requested. The BIOS might be trying to access the EC without synchronizing with the operating system. This data will be ignored. No further action is necessary; however, you should check with your computer manufacturer for an upgraded BIOS.

Random errors like that aren't out of the ordinary. If you start crashing in stress tests, then I'd be concerned, but just a random error in the event viewer isn't something I'd pay too much attention to. 

[Keiko-chan]

21 hours ago, RONOTHAN## said:

Random errors like that aren't out of the ordinary. If you start crashing in stress tests, then I'd be concerned, but just a random error in the event viewer isn't something I'd pay too much attention to. 

I just simply kept Expo I on and higered the VDDCR_SOC voltage to 1.3 and passed the two test I originally failed at 45 min. (Prime95 Small FFT AVX Enabled and Karhu both failed at 45 min). Ran Karhu for an hour and 10 with no issues and same Prime95 AVX Enabled for just over 2 hours and had no issues.) So im actually running Expo now, just had to higher SOC.  Thank you so so much, And I did have another question. In HWiNFO I saw a random split second spike that said 128C on CPU(Tctl/Tdie). Idk if this was error cause it went right back to normal. Could've been HWiNFO bugging. And on the prime95 test it never got above 79 and CCD1 and CCD2 never got above 74.8 and 67.1 C. Averaged about 63C during it. What do you think?  Also, although I do think it is now fixed ill run blend with AVX enabled tonight for about 6- 8 hours just to make sure? Or should I run Large? Your call :))

[RONOTHAN##]

54 minutes ago, Keiko-chan said:

In HWiNFO I saw a random split second spike that said 128C on CPU(Tctl/Tdie). Idk if this was error cause it went right back to normal. Could've been HWiNFO bugging.

Almost certainly this was just a bug, the thermal shutoff for those chips IIRC is ~110C. The system would've shutdown if that was real. 

 

55 minutes ago, Keiko-chan said:

And on the prime95 test it never got above 79 and CCD1 and CCD2 never got above 74.8 and 67.1 C. Averaged about 63C during it. What do you think?

Sounds about normal, the X3D chips are very power/current limited in those high power stress tests like Prime95. 

 

57 minutes ago, Keiko-chan said:

Also, although I do think it is now fixed ill run blend with AVX enabled tonight for about 6- 8 hours just to make sure? Or should I run Large?

Large is more likely to find memory instability, though doing blend isn't a bad idea either. My usual go to for memory controller stress testing on Ryzen is Linpack Xtreme for 40 loops 10GB dataset, usually if your IMC is unstable it'll crash pretty quickly so it might be a good idea to run that as well, but given you aren't crashing now you're probably stable. Y cruncher VST is also really good at stress testing IMC, though I'd use that more for testing FCLK. 

[Keiko-chan]

10 hours ago, RONOTHAN## said:

Almost certainly this was just a bug, the thermal shutoff for those chips IIRC is ~110C. The system would've shutdown if that was real. 

 

Sounds about normal, the X3D chips are very power/current limited in those high power stress tests like Prime95. 

 

Large is more likely to find memory instability, though doing blend isn't a bad idea either. My usual go to for memory controller stress testing on Ryzen is Linpack Xtreme for 40 loops 10GB dataset, usually if your IMC is unstable it'll crash pretty quickly so it might be a good idea to run that as well, but given you aren't crashing now you're probably stable. Y cruncher VST is also really good at stress testing IMC, though I'd use that more for testing FCLK. 

Thanks very much for all your help, ill run large later just to be extra sure. All still good so far. On Large FFTS, AVX enabled or disabled? 

Also, when I run cinebench R23. My CPU stays in the lower to upper 70's but my package hits 87-89C (Fans at Full speed). And I scored 35,067 MC and 1968 SC. I know this is normal on the 7950X3D from what I have read but with a 360mm AIO I feel like it should be a little bit cooler. Maybe I should wipe off the pre-applied thermal paste and repaste it since I had trouble getting the AIO on when I installed it, I touched it against cpu couldnt get it down to latch, removed it, then tried again (2-3 times) that could be making it a little warmer but surely wouldn't make a huge difference, but may as well try. What do you think? All stock settings (which mean PBO is on auto). So only changes ive made is Expo I enabled SOC @ 1.3V.

[RONOTHAN##]

3 minutes ago, Keiko-chan said:

On Large FFTS, AVX enabled or disabled? 

Doesn't matter much. AVX should in theory be more stressing, though in practice because of the way AVX 512 works on Ryzen 7000 it's actually a little lighter of a load than AVX2. 

 

4 minutes ago, Keiko-chan said:

I know this is normal on the 7950X3D from what I have read but with a 360mm AIO I feel like it should be a little bit cooler.

Problem is the limitation in this case is the water block itself, not the AIO capacity. The issue is getting the heat out of a 7950X3D, not radiator capacity. You can try a repaste if you really want to, but it shouldn't make a difference. Only real way to run them cooler is to use liquid metal and/or a block with thicker copper. 

[Keiko-chan]

26 minutes ago, RONOTHAN## said:

Problem is the limitation in this case is the water block itself, not the AIO capacity. The issue is getting the heat out of a 7950X3D, not radiator capacity. You can try a repaste if you really want to, but it shouldn't make a difference. Only real way to run them cooler is to use liquid metal and/or a block with thicker copper. 

Okay, I'll hold off on repasting for now, I was just afraid I may have accidentally screwed it unevenly or me removing it 2-3 times while trying to get it on may have affected it slightly (even though I had never even turned the computer on by that point), but everything sounds pretty much normal to you then? Oh by the way on top my PSU is a Thor Platinum II 1000W so its pulling in Air in to my Hyte Y60 Case from the back too. 

[RONOTHAN##]

6 minutes ago, Keiko-chan said:

but everything sounds pretty much normal to you then?

Yeah, everything you've described is about what I'd expect. 

[Keiko-chan]

31 minutes ago, RONOTHAN## said:

Yeah, everything you've described is about what I'd expect. 

Okay, thank you again for all you're help