- last edited on by
ROGBot
Hi everyone
I currently find myself slowly losing my sanity (lol) trying to figure out what the heck is going on with my system.
Around a month or so ago I overclocked my 8700k successfully to 5Ghz with an adaptive voltage of 1.350V (in BIOS) and LLC5. My cache ratio topped out at 46. I ran AIDA64 FPU stress test for an hour, Realbench stress test for an hour and prime95 26.6 for an hour to make sure it was stable. I had no issues for a while. Then a new BIOS came out for the Maximus X Code (1602), so I updated from the previous BIOS and ran Realbench for 30mins and AIDA64 for 30mins just to make sure there weren't any obvious stability issues with this BIOS version. No problems.
Last week there was a huge sale on PC parts on eBay and I found a great deal on another 16GB of the same RAM I already had (Trident Z RGB 3200Mhz C16) so I bought that. Before anyone mentions it, yes I know now that you aren't supposed to combine RAM kits even if they are identical. I found the article written by Raja about that just the other day. Despite this it seemed that I got lucky. I seemingly had all 32GB RAM running in XMP 3200mhz without any issues. I ran HCI memtest pro for over 12 hours overnight and got 0 errors. I ran memtest86 overnight and got no errors. I ran the system memory stress test in AIDA64 twice 20 mins each without any issues. I ran 30 mins of Realbench with 32GB ram selected and passed with flying colours. All sounds good right?
Now the part where I am stumped. Today one of my programs locked up and when I clicked restart my system blue screened. This was the report from WhoCrashed afterward:
Googling those results brought up some threads suggesting display driver issues and to use DDU and reinstall the latest driver. I did that and even rolled back the application that locked up (Corsair iCUE) to the previous release just incase it was at fault. I then decided to re-check system stability by running AIDA64 FPU test again. 25mins in BSOD. WTF. I backed off the cache ratio to 45 and repeated, immediate BSOD. I then backed off the cache ratio to 44. BSOD after 20 mins. 43, BSOD after 25mins. I thought to myself "ok, maybe the RAM is causing this" and removed the two sticks of RAM I recently added and restored my overclock settings to where they were originally. another BSOD after 35mins of AIDA.
Now I am really unsure of what's going on, or how I am to determine exactly what is the cause. I thought if there was a problem with the RAM kits, it would definitely show after lots of testing in HCI memtest pro and memtest86. And now my overclock is no longer stable whether I have one kit or two kits installed.
Is it safe to say that my RAM kits are okay? I could sell this 16GB kit since I got a good deal on it and fork out for a matched 32GB kit but I don't want to do that without being sure that the RAM is a problem. I have my doubts now that I am failing AIDA64 with the original 16GB in there.
If anyone has advice to help me solve this it would be greatly appreciated!!
I currently find myself slowly losing my sanity (lol) trying to figure out what the heck is going on with my system.
Around a month or so ago I overclocked my 8700k successfully to 5Ghz with an adaptive voltage of 1.350V (in BIOS) and LLC5. My cache ratio topped out at 46. I ran AIDA64 FPU stress test for an hour, Realbench stress test for an hour and prime95 26.6 for an hour to make sure it was stable. I had no issues for a while. Then a new BIOS came out for the Maximus X Code (1602), so I updated from the previous BIOS and ran Realbench for 30mins and AIDA64 for 30mins just to make sure there weren't any obvious stability issues with this BIOS version. No problems.
Last week there was a huge sale on PC parts on eBay and I found a great deal on another 16GB of the same RAM I already had (Trident Z RGB 3200Mhz C16) so I bought that. Before anyone mentions it, yes I know now that you aren't supposed to combine RAM kits even if they are identical. I found the article written by Raja about that just the other day. Despite this it seemed that I got lucky. I seemingly had all 32GB RAM running in XMP 3200mhz without any issues. I ran HCI memtest pro for over 12 hours overnight and got 0 errors. I ran memtest86 overnight and got no errors. I ran the system memory stress test in AIDA64 twice 20 mins each without any issues. I ran 30 mins of Realbench with 32GB ram selected and passed with flying colours. All sounds good right?
Now the part where I am stumped. Today one of my programs locked up and when I clicked restart my system blue screened. This was the report from WhoCrashed afterward:
Googling those results brought up some threads suggesting display driver issues and to use DDU and reinstall the latest driver. I did that and even rolled back the application that locked up (Corsair iCUE) to the previous release just incase it was at fault. I then decided to re-check system stability by running AIDA64 FPU test again. 25mins in BSOD. WTF. I backed off the cache ratio to 45 and repeated, immediate BSOD. I then backed off the cache ratio to 44. BSOD after 20 mins. 43, BSOD after 25mins. I thought to myself "ok, maybe the RAM is causing this" and removed the two sticks of RAM I recently added and restored my overclock settings to where they were originally. another BSOD after 35mins of AIDA.
Now I am really unsure of what's going on, or how I am to determine exactly what is the cause. I thought if there was a problem with the RAM kits, it would definitely show after lots of testing in HCI memtest pro and memtest86. And now my overclock is no longer stable whether I have one kit or two kits installed.
Is it safe to say that my RAM kits are okay? I could sell this 16GB kit since I got a good deal on it and fork out for a matched 32GB kit but I don't want to do that without being sure that the RAM is a problem. I have my doubts now that I am failing AIDA64 with the original 16GB in there.
If anyone has advice to help me solve this it would be greatly appreciated!!
Alright I've had some time to play around with this some more. I've carried out the following:
- Another AIDA64 FPU stress test with the same settings as the last one (5Ghz@1.350V LLC5 Cache x43 and just the original 16GB RAM in XMP) only this time it gave me a BSOD after 35 mins despite passing an hour without issues last time. No idea why this seems to happen. Perhaps I'm on the bleeding edge of stability or something..
- Stuck the other two RAM modules back in (2x16GB kits total), reset to defaults and ran an hour of AIDA64 FPU stress test. Pass.
- Manually dialed in the RAM settings instead of using XMP. 3200Mhz 1.35V 16-18-18-38 2N (I know you mentioned 1N but these kits are tested at 2N so I thought that was the right thing to do) and VCCIO and System Agent at 1.15V. Although the actual VCCIO and System Agent voltage read by the BIOS and AIDA64 is around 1.17V for some reason. Ran an hour of AIDA64 FPU stress test. Pass. No issues with the RAM then it would seem. Lucky.
- Dialed in my CPU overclock same as before. 5Ghz with an adaptive Vcore of 1.350V in BIOS, LLC5 and left the cache on auto this time (which gives me x44). Ran AIDA64 FPU stress test - BSOD again. Must be an issue with the CPU overclock.
- Bumped up the max CPU voltage in the BIOS to 1.365V. Under full load software reads 1.34V. I considered bumping the LLC to level 6 instead but I'm a little nervous to use anything higher than 5 as I understand there can be overshoot that you don't see in software, and I don't have an oscilloscope or any of the right tools to measure that sort of thing, So I figured just adding a little extra Vcore would be a safer option. This seems to have worked (for now at least) as it passed an hour in AIDA64 and an hour in Realbench with 32GB ram selected.
Hopefully it stays stable this time. I can only imagine the reason for being stable previously and unstable now is either a fluke or something about this newer BIOS version is requiring slightly more voltage to be stable. It's hard to be sure.
Thanks again for your help Nate.
- Another AIDA64 FPU stress test with the same settings as the last one (5Ghz@1.350V LLC5 Cache x43 and just the original 16GB RAM in XMP) only this time it gave me a BSOD after 35 mins despite passing an hour without issues last time. No idea why this seems to happen. Perhaps I'm on the bleeding edge of stability or something..
- Stuck the other two RAM modules back in (2x16GB kits total), reset to defaults and ran an hour of AIDA64 FPU stress test. Pass.
- Manually dialed in the RAM settings instead of using XMP. 3200Mhz 1.35V 16-18-18-38 2N (I know you mentioned 1N but these kits are tested at 2N so I thought that was the right thing to do) and VCCIO and System Agent at 1.15V. Although the actual VCCIO and System Agent voltage read by the BIOS and AIDA64 is around 1.17V for some reason. Ran an hour of AIDA64 FPU stress test. Pass. No issues with the RAM then it would seem. Lucky.
- Dialed in my CPU overclock same as before. 5Ghz with an adaptive Vcore of 1.350V in BIOS, LLC5 and left the cache on auto this time (which gives me x44). Ran AIDA64 FPU stress test - BSOD again. Must be an issue with the CPU overclock.
- Bumped up the max CPU voltage in the BIOS to 1.365V. Under full load software reads 1.34V. I considered bumping the LLC to level 6 instead but I'm a little nervous to use anything higher than 5 as I understand there can be overshoot that you don't see in software, and I don't have an oscilloscope or any of the right tools to measure that sort of thing, So I figured just adding a little extra Vcore would be a safer option. This seems to have worked (for now at least) as it passed an hour in AIDA64 and an hour in Realbench with 32GB ram selected.
Hopefully it stays stable this time. I can only imagine the reason for being stable previously and unstable now is either a fluke or something about this newer BIOS version is requiring slightly more voltage to be stable. It's hard to be sure.
Thanks again for your help Nate.
I tried 1602 on my X Apex and couldn't run my RAM at the same timings as 1503 and earlier bios's so I flashed back to 1503
