mersenneforum.org  

Go Back   mersenneforum.org > Fun Stuff > Lounge

Reply
 
Thread Tools
Old 2020-10-29, 19:26   #34
Runtime Error
 
Sep 2017
USA

3·61 Posts
Default

What load % is your CPU showing? I'm suspicious because 29C is very cool. I have a machine with a corsair iCUE H150i that is 40C under the wavefront FFT load. However, it quickly gets up to 70C on small FFTs.

Regarding your errors, my money is that your RAM is having issues. Are you running your XMP profile? You might consider increasing voltage slightly in bios. Edit: what is your RAM speed?

Last fiddled with by Runtime Error on 2020-10-29 at 19:28 Reason: 2 separate points here my bad
Runtime Error is offline   Reply With Quote
Old 2020-10-29, 20:04   #35
mathematizer
 
Oct 2020

1000002 Posts
Default

Quote:
Originally Posted by Runtime Error View Post
What load % is your CPU showing? I'm suspicious because 29C is very cool. I have a machine with a corsair iCUE H150i that is 40C under the wavefront FFT load. However, it quickly gets up to 70C on small FFTs.

Regarding your errors, my money is that your RAM is having issues. Are you running your XMP profile? You might consider increasing voltage slightly in bios. Edit: what is your RAM speed?
Where would I determine my CPU load? Task Manager (see attached image?)

I'm using 2x8GB 3000Mhz Corsair Vengeance. I have another two sticks available that I could try swapping in that had come from a friend - I swapped them out for new sticks of the same type (same make/model/speed) in the process of troubleshooting my build, and haven't tried putting them back in. The new ones that I had purchased, that are in it now, match the mobo specs exactly, down to the version number. The original ones match all but the version number.

I rebooted my machine, checked in the BIOS and saw that the XMP profile was enabled.
Attached Thumbnails
Click image for larger version

Name:	task manager.png
Views:	16
Size:	11.3 KB
ID:	23670  

Last fiddled with by mathematizer on 2020-10-29 at 20:13
mathematizer is offline   Reply With Quote
Old 2020-10-29, 20:36   #36
VBCurtis
 
VBCurtis's Avatar
 
"Curtis"
Feb 2005
Riverside, CA

106258 Posts
Default

XMP = telling the motherboard to go faster than "stock" to run the memory at whatever speed the memory claims it can run at.

If you disable that XMP option, your memory speed will slow down to stock. That will cost you a little speed on Prime95, but if your errors disappear you'll know it's a memory issue with XMP being a little too fast.

Usually XMP works well, but it's worth it to test your rig without to see if stability improves.
VBCurtis is online now   Reply With Quote
Old 2020-10-29, 20:40   #37
mathematizer
 
Oct 2020

3210 Posts
Default

Quote:
Originally Posted by VBCurtis View Post
XMP = telling the motherboard to go faster than "stock" to run the memory at whatever speed the memory claims it can run at.

If you disable that XMP option, your memory speed will slow down to stock. That will cost you a little speed on Prime95, but if your errors disappear you'll know it's a memory issue with XMP being a little too fast.

Usually XMP works well, but it's worth it to test your rig without to see if stability improves.
Thanks for the advice. I've turned XMP off and will let it run for a bit and see if it posts more errors.
mathematizer is offline   Reply With Quote
Old 2020-10-29, 20:42   #38
Runtime Error
 
Sep 2017
USA

3·61 Posts
Default

Quote:
Originally Posted by mathematizer View Post
Where would I determine my CPU load? Task Manager (see attached image?)

I'm using 2x8GB 3000Mhz Corsair Vengeance. I have another two sticks available that I could try swapping in that had come from a friend - I swapped them out for new sticks of the same type (same make/model/speed) in the process of troubleshooting my build, and haven't tried putting them back in. The new ones that I had purchased, that are in it now, match the mobo specs exactly, down to the version number. The original ones match all but the version number.

I rebooted my machine, checked in the BIOS and saw that the XMP profile was enabled.
Interesting. You do have a very good cooler, but I am still surprised that your i5-9600k isn't throwing off more heat.

If you are still seeing frequent errors, and 1 a day is definitely frequent, you might try disabling the XMP profile for a bit. This will severely gimp your processor (no pun intended), but if it runs for a while error-free, then you know that your CPU is probably fine. Then you can try tinkering with the voltage in bios, very gradually increasing it. You said both tried sets of RAM are the same, but they both might have the same voltage issue.

Edit: Looks like VBCurtis beat me to it. Good luck!!!

Last fiddled with by Runtime Error on 2020-10-29 at 20:43
Runtime Error is offline   Reply With Quote
Old 2020-10-29, 21:42   #39
mathematizer
 
Oct 2020

25 Posts
Default

I’m trying to decide how long to let it run with XMP disabled. I could try to let it finish the current PRP test, and maybe one more? That’ll likely be a few weeks.

I’ll need to see if the gimped processor interferes with other applications - I may not be able to let it go as long as I’d like.

Would you re-enable XMP before tinkering with the voltage, and how would you quantify a “gradual increase” of the voltage. This is all new territory for me, but I’m willing to put a little time into learning if it can help resolve those errors, isolate problems, and eliminate hardware errors.
mathematizer is offline   Reply With Quote
Old 2020-10-29, 22:11   #40
Runtime Error
 
Sep 2017
USA

3×61 Posts
Default

If you were previously seeing an error (say) every few hours, I'd let it run for at least a day error-free with XMP disabled. Other folks might recommend longer but I'm not the most patient person. That will give you some confidence that the error is indeed coming from the RAM.

For stepping up the voltage, yes enable XMP first, and then I might go up by 0.05v at a time. Then wait until you see errors.
Runtime Error is offline   Reply With Quote
Old 2020-10-29, 23:25   #41
VBCurtis
 
VBCurtis's Avatar
 
"Curtis"
Feb 2005
Riverside, CA

10001100101012 Posts
Default

If you got a error every so often, and now run 4-6x longer than that period without an error, you have your answer that XMP is correlated to the problem. If you were getting one error a day, I'd let it run 4-5 days; some changes only make errors less frequent, rather than fully curing the problem.

You won't notice a change in computer speed or responsiveness from slower memory- 99% of the world runs memory at stock speeds, after all. But Prime95 is so well tuned for speed that it uses every bit of memory bandwidth available, and usually the CPU has to wait for more data from memory while running Prime95. That's why XMP helps specifically Prime95 work so much, but doesn't matter for other things much at all- "other things" pretty much never fill all of the memory access bandwidth, so they're not slowed down by slower memory.

You can quantify this effect, if you wish- run P95 with just one worker one thread. Memory speed won't affect the time reported per iteration.
Then, go to 2 workers or 2 threads (doesn't really matter which); XMP-off and XMP-on timings should be about the same. Finally, go to your full production # of workers & threads, and the XMP-on version will have lower timings.
If I'm wrong, and it doesn't, then memory is not restricting P95 from full speed and you can run without XMP forever- or you have a setup problem with P95, such as only running 1 worker / 1 thread all the time. That would also explain your low temperature reading!
VBCurtis is online now   Reply With Quote
Old 2020-10-29, 23:57   #42
mathematizer
 
Oct 2020

25 Posts
Default

Thanks for the explanation!

Since errors seem to occur once every 2-3 days, with an average of maybe 10 per PRP, I’ll let it run for a bit longer before making adjustments.

Before tweaking the voltage, perhaps I should try swapping the RAM with the other two sticks I have, to see if that produces errors (with/without XMP) - I haven’t had them in since I swapped my CPU and got windows installed.

I’d be curious to try all 4 sticks together, but if I understand it correctly, it’s sounding like the consensus is that there may be a problem with the RAM itself or how that type interacts with XMP. Guess we will know more in a few days. Appreciate all the help.

Last fiddled with by mathematizer on 2020-10-29 at 23:58
mathematizer is offline   Reply With Quote
Old 2020-10-30, 00:48   #43
mathematizer
 
Oct 2020

25 Posts
Default

Hmm. Just got a system freeze when I tried to tell prime95 to stop. No hardware errors yet.
mathematizer is offline   Reply With Quote
Old 2020-10-30, 01:21   #44
Prime95
P90 years forever!
 
Prime95's Avatar
 
Aug 2002
Yeehaw, FL

52×172 Posts
Default

Quote:
Originally Posted by mathematizer View Post
Before tweaking the voltage, perhaps I should try swapping the RAM with the other two sticks I have, to see if that produces errors (with/without XMP) - I haven’t had them in since I swapped my CPU and got windows installed.

I’d be curious to try all 4 sticks together, but if I understand it correctly, it’s sounding like the consensus is that there may be a problem with the RAM itself or how that type interacts with XMP. Guess we will know more in a few days. Appreciate all the help.
Quote:
Originally Posted by mathematizer View Post
Hmm. Just got a system freeze when I tried to tell prime95 to stop. No hardware errors yet.
Sorry, you are having such difficulty getting the system stable. Prime95 can be pretty brutal on hardware running "near the edge".

You will definitely want to try running with all 4 sticks of RAM at some point. Several have reported improved timings with 4 sticks of RAM.

It is not at all uncommon for RAM to not run at its rated XMP specs. You can either return it if under warranty or usually an extra 0.05V solves the problem. This is generally considered quite safe.

I had a Haswell-era CPU that would crash stopping prime95. I theorized the chip "overshot" in its response to the sudden drop in required power. The problem went away when I disabled C states, which tells the CPU to run at full power even at idle.
Prime95 is online now   Reply With Quote
Reply

Thread Tools


Similar Threads
Thread Thread Starter Forum Replies Last Post
Greetings from a newbie piforbreakfast Lounge 53 2020-10-28 16:41
Newbie :) ednaemars Information & Answers 2 2018-04-24 07:24
I'm a Newbie davieddy Information & Answers 17 2010-11-22 00:47
newbie needs help 111393 Hardware 9 2004-03-28 05:01
Newbie ThePotato Lounge 7 2003-03-02 15:24

All times are UTC. The time now is 05:57.

Fri Dec 4 05:57:12 UTC 2020 up 1 day, 2:08, 0 users, load averages: 1.25, 1.12, 1.08

Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2020, Jelsoft Enterprises Ltd.

This forum has received and complied with 0 (zero) government requests for information.

Permission is granted to copy, distribute and/or modify this document under the terms of the GNU Free Documentation License, Version 1.2 or any later version published by the Free Software Foundation.
A copy of the license is included in the FAQ.