Freeze and Reboot Issue - AMD 5900x - 4 x 8GB Corsair 3600MHz RAM

SpyderTracks

SpyderTracks

We love you Ukraine
ChrisCooney said:
Is this because of the increments for AMD Processors? I read that on another forum when I was doing research around this issue.

If this does stabilize, I'd be more than happy to help you guys investigate. Gotta be a community player, right :)
Click to expand...
AMD processors have unusual DOCP (XMP) step ups, it's not like you'd expect like 3000, 3100, 3200, it's more random figures.

Because of this, I'm wondering if 3600Mhz isn't configured properly on some boards, and the manufacturer missed it or something random because they were concentrating on the 3533 or whatever that naturally came before it.

I have no idea really, but I think quite possibly, just a bit lower (or maybe even higher) than 3600MHz may well run stable with 4 DIMMS.... that's my hope anyway.
 
D

Deleted member 17413

Guest
Wasnt the "perfect" spot 3733MHz before it starts having diminishing returns?
 
D

Deleted member 17413

Guest
SpyderTracks said:
That’s quite a hefty overclock though on 3600mhz, it’s definitely worth a try but you’d have to strike it lucky with the silicon gods for the RAM to accept that much of an OC
Click to expand...
Talking just theoretically, if you took 4000MHz RAM and underclocked it to that figure, would that be the best way to achieve it?
Talking in terms of process, rather than pricepoint.

LOL sorry, this should probably be in a different thread! I'm diverting...ignore me.
 
SpyderTracks

SpyderTracks

We love you Ukraine
sibun1 said:
Talking just theoretically, if you took 4000MHz RAM and underclocked it to that figure, would that be the best way to achieve it?
Talking in terms of process, rather than pricepoint.

LOL sorry, this should probably be in a different thread! I'm diverting...ignore me.
Click to expand...
Yeah, that would be much easier.

Just quickly, with RAM chips, the higher rated speeds is simply because it's better quality silicon and can handle a higher "overclock" and tighter voltage control.

Same with CPU's really, you can have 5 of the same CPU, and they'll all overclock to different maximums, due to silicon lottery. But you can get "binned" chips where they've been manually tested to be certain of reaching a certain speed. obviously you pay a premium for higher binned silicon.
 
S

scarpa1

Enthusiast
I currently have my 4x8gb 3600mhz ram running at 3600mhz on a X570- F MB

I have had issues but all mine have been on restarts of the PC or an example last night.

I bought a game and had to download it via Origin. Up on restarting the machine it did 3 beeps, turned off, came back on, 3 beeps, same again, then rebooted but ram was back at 2133mhz.
Ive got it up to 3600mhz again and its running fine, it had been running with no issues for 2 months after some problems early on, PC came with the ram running at 2133mhz.

It has never just crashed when I'm playing games, turning on after PC has been off or anything, never got any errors or random restarts. All my issues came after I restarted the PC after being prompted to.

Running at 3533mhz however has never caused an issue, I ran it at this for a month after the first few issues until I decided to give 3600 ago again.
I do need to select 3533 before 3600 though, it never boots up properly, always resets to 2133 if I go from 3400 to 3600mhz. When it has gone wrong I go from 3200, to 3400, 3533, 3600 but it generally works better by going to 3533 then go into EZ mode and de select DOCP in here, reselect it again and the 3600 works, choosing 3600 in the tweaker part in generally doesn't like it.

Hope that sort of makes sense, I generally have no idea what I'm taking about here, haha.
 
SpyderTracks

SpyderTracks

We love you Ukraine
scarpa1 said:
I currently have my 4x8gb 3600mhz ram running at 3600mhz on a X570- F MB

I have had issues but all mine have been on restarts of the PC or an example last night.

I bought a game and had to download it via Origin. Up on restarting the machine it did 3 beeps, turned off, came back on, 3 beeps, same again, then rebooted but ram was back at 2133mhz.
Ive got it up to 3600mhz again and its running fine, it had been running with no issues for 2 months after some problems early on, PC came with the ram running at 2133mhz.

It has never just crashed when I'm playing games, turning on after PC has been off or anything, never got any errors or random restarts. All my issues came after I restarted the PC after being prompted to.

Running at 3533mhz however has never caused an issue, I ran it at this for a month after the first few issues until I decided to give 3600 ago again.
I do need to select 3533 before 3600 though, it never boots up properly, always resets to 2133 if I go from 3400 to 3600mhz. When it has gone wrong I go from 3200, to 3400, 3533, 3600 but it generally works better by going to 3533 then go into EZ mode and de select DOCP in here, reselect it again and the 3600 works, choosing 3600 in the tweaker part in generally doesn't like it.

Hope that sort of makes sense, I generally have no idea what I'm taking about here, haha.
Click to expand...
That makes a lot of sense, very helpful, thank you (y)

I think it's definitely worth trying 4 DIMMS at 3533 and see if it's stable. The difference between that and 3600 wouldn't be noticeable anyway so if you can get it stable at 3533, then you're laughing.
 
ubuysa

ubuysa

The BSOD Doctor
@ChrisCooney has kindly made his logs available to me, I've also analyzed the kernel dump posted yesterday.

In the System log I can see the (now familiar) WHEA error logs refering to a processor cache error. An example is this one, just before a bug check (BSOD)...

Clipboard01.jpg

There is nothing of note in the Application log related to this issue. In the kernel dump the stop code is WHEA_UNCORRECTABLE_ERROR which is of course a hardware error. The hardware error record field contains a reason of DCACHEL1_DRD_ERR on processor 8 bank 0 and this seems to indicate that there was a problem restoring data deferred to the L1 Cache when the instruction was needed (in other words the instruction pipeline failed). This could be a CPU error and I know that at least one other user has RMA'd his build for this problem and PCS replaced the CPU - seemingly successfully.

Although it appears from this kernel dump (and several other dumps from other users with this issue) that the problem is a CPU L1 cache that doesn't explain why it can be apparently mitigated by either reducing the number of RAM sticks and/or down-clocking the RAM slightly.....
 
C

ChrisCooney

Silver Level Poster
ubuysa said:
@ChrisCooney has kindly made his logs available to me, I've also analyzed the kernel dump posted yesterday.

In the System log I can see the (now familiar) WHEA error logs refering to a processor cache error. An example is this one, just before a bug check (BSOD)...

View attachment 23003

There is nothing of note in the Application log related to this issue. In the kernel dump the stop code is WHEA_UNCORRECTABLE_ERROR which is of course a hardware error. The hardware error record field contains a reason of DCACHEL1_DRD_ERR on processor 8 bank 0 and this seems to indicate that there was a problem restoring data deferred to the L1 Cache when the instruction was needed (in other words the instruction pipeline failed). This could be a CPU error and I know that at least one other user has RMA'd his build for this problem and PCS replaced the CPU - seemingly successfully.

Although it appears from this kernel dump (and several other dumps from other users with this issue) that the problem is a CPU L1 cache that doesn't explain why it can be apparently mitigated by either reducing the number of RAM sticks and/or down-clocking the RAM slightly.....
Click to expand...
I suppose I should say that we don't know if it has been mitigated. PC has been crash free for 36 hours though, which is a good sign, but don't want to get ahead of myself.

It can't be a coincidence that error logs stop coming through when we drop down to 2 RAM sticks. That would be so strange!
 
jamiephillips909

jamiephillips909

Godlike
SpyderTracks said:
That makes a lot of sense, very helpful, thank you (y)

I think it's definitely worth trying 4 DIMMS at 3533 and see if it's stable. The difference between that and 3600 wouldn't be noticeable anyway so if you can get it stable at 3533, then you're laughing.
Click to expand...
i think now it has been crash free after our talk about removing 2 sticks think we can hazard a guess that the system is now stable with 2 sticks so may be worth giving spyders method a try to see if we can get it back up to 4 sticks running at 3533 if its not can easily drop back down to 2 sticks
 
C

ChrisCooney

Silver Level Poster
jamiephillips909 said:
i think now it has been crash free after our talk about removing 2 sticks think we can hazard a guess that the system is now stable with 2 sticks so may be worth giving spyders method a try to see if we can get it back up to 4 sticks running at 3533 if its not can easily drop back down to 2 sticks
Click to expand...
I know I'm being a bit difficult but I'm going to hold off until tomorrow evening. That will make it 72 hours (or close enough) and eliminate any doubt. Will update tomorrow evening and, all being well, I'll reinstall the 2 sticks and change the frequency in the BIOS to 3533 rather than 3600 and monitor.
 
ubuysa

ubuysa

The BSOD Doctor
ChrisCooney said:
I know I'm being a bit difficult but I'm going to hold off until tomorrow evening. That will make it 72 hours (or close enough) and eliminate any doubt. Will update tomorrow evening and, all being well, I'll reinstall the 2 sticks and change the frequency in the BIOS to 3533 rather than 3600 and monitor.
Click to expand...
I think you're being wise. You need a solid baseline before making changes. This is a good plan in my book. :)
 
C

ChrisCooney

Silver Level Poster
Evening everyone! No crashes in 72 hours of constant uptime, and regular use. I think we can consider this stable!!

I won't have time this evening to mess about with RAM and frequency settings but I'll be sure to do it at some point this week and get back to you. Big thanks to everyone for helping me triage and solve.

Do you recommend I now get in touch with PCS and explain that I'll be needing 2x16GB RAM rather than 4x8GB?
 
Last edited:
ubuysa

ubuysa

The BSOD Doctor
ChrisCooney said:
Evening everyone! No crashes in 72 hours of constant uptime, and regular use. I think we can consider this stable!!

I won't have time this evening to mess about with RAM and frequency settings but I'll be sure to do it at some point this week and get back to you. Big thanks to everyone for helping me triage and solve.

Do you recommend I now get in touch with PCS and explain that I'll be needing 2x16GB RAM rather than 4x8GB?
Click to expand...
That's good to hear. :)

Phoning PCS is always the most effective way to contact them, even though the wait is terrible.
 
C

ChrisCooney

Silver Level Poster
ubuysa said:
That's good to hear. :)

Phoning PCS is always the most effective way to contact them, even though the wait is terrible.
Click to expand...
Awesome. I'll wait until I'm in a particularly boring meeting and call them I think!

In the mean time, I'll have a play with the RAM and let you guys know. Are there any other experiments you'd like me to run? I really want to help contribute back into the forum a little, since you guys were all so friendly and helpful!
 
ubuysa

ubuysa

The BSOD Doctor
ChrisCooney said:
Awesome. I'll wait until I'm in a particularly boring meeting and call them I think!

In the mean time, I'll have a play with the RAM and let you guys know. Are there any other experiments you'd like me to run? I really want to help contribute back into the forum a little, since you guys were all so friendly and helpful!
Click to expand...
If you're brave enough to experiment a little then yes please, we'd all be interested to know whether you can keep it stable at higher RAM frequencies. If/when it BSODs copy the file C:\Windows\Memory.dmp out to somewhere safe so it's not overwritten by the next BSOD. Upload any and all kernel dumps you get, they all help to give a bigger picture. Don't bother uploading minidumps, they don't help with this particular problem. :)
 
C

ChrisCooney

Silver Level Poster
ubuysa said:
If you're brave enough to experiment a little then yes please, we'd all be interested to know whether you can keep it stable at higher RAM frequencies. If/when it BSODs copy the file C:\Windows\Memory.dmp out to somewhere safe so it's not overwritten by the next BSOD. Upload any and all kernel dumps you get, they all help to give a bigger picture. Don't bother uploading minidumps, they don't help with this particular problem. :)
Click to expand...
I’ll get right on it this weekend ☺️ - I’ll begin with four x 3533 and see if that’s stable. If not, full kernel dumps will be made available on this thread 👍🏼👍🏼
 
You must log in or register to reply here.
Top