mersenneforum.org  

Go Back   mersenneforum.org > Great Internet Mersenne Prime Search > Software

Reply
 
Thread Tools
Old 2019-02-05, 22:18   #265
Mysticial
 
Mysticial's Avatar
 
Sep 2016

14B16 Posts
Default

Quote:
Originally Posted by GP2 View Post
But run-of-the-mill hardware errors are caught by Gerbicz error checking.

This is the only system we know of that reports erroneous final PRP residues to Primenet.
I'd suggest at least 2 things before shipping anything:
  • Give George remote access. (as mentioned earlier)
  • See if someone can replicate it on a different system by tweaking it unstable.

Last fiddled with by Mysticial on 2019-02-05 at 22:19
Mysticial is offline   Reply With Quote
Old 2019-02-05, 22:23   #266
Mark Rose
 
Mark Rose's Avatar
 
"/X\(‘-‘)/X\"
Jan 2013

55618 Posts
Default

Quote:
Originally Posted by simon389 View Post
The 9800X doing a LL Doublecheck with AVX512 seems to be oscillating between 4096 and 4115 Mhz. I would be sad to have to underclock my CPU to get it stable at AVX512, which sort of defeats the purpose of the boost it gives to my iterations per second. Although maybe the voltage simply needs to be bumped up.
The AVX512 implementation should be doing fewer operations (less than half of AVX2), so it should require reading and writing fewer operands and results to RAM. RAM bandwidth is usually the bottleneck and almost certainly is on your system using AVX512 at 4 GHz, unless you have some insanely fast RAM. So the reduced RAM bandwidth requirements of AVX512 are a win, even if you have to run at a slightly slower core clock.
Mark Rose is offline   Reply With Quote
Old 2019-02-05, 22:23   #267
simon389
 
Aug 2013

3·29 Posts
Default

Quote:
Originally Posted by Mysticial View Post
4.1 GHz all-core AVX512 is way too high for any chip that isn't overclocked. I can't tell for sure, but it looks like the BIOS is not applying any AVX512 offset.

What's the all-core frequency for non-AVX? Is it the same?

If it's the same, then it confirms my suspicion that the BIOS isn't doing the offsets. I looked up your motherboard and the option does exist. So you might have to find it. When you do, I recommend setting the AVX offset to -3 and the AVX512 offset to -5. That will get you closer to the "true" stock settings.

If my suspicions are true, your chip is already running overclocked for AVX(2) and AVX512. It just hasn't been crashing for AVX2.

Yes, increasing the voltages will also work. But of course that counts as overclocking. Normally, I don't try to overclock until I get the system stable at stock.

These chips are going to be very hard to cool with an air cooler. If you do plan to overclock it, you're gonna need at least a 280/360 AIO or full custom water.

EDIT:

Also... before you do anything. How old is the BIOS? Your chip is the refresh line, not the original run. If your BIOS is older than late last year, it might not be properly updated for your chip.
Without AVX512 it is also running at 4.1Ghz. So you're correct, the BIOS is not lowering the clock speed while running AVX512. Also, my BIOS is the most recent version version 1.09 released 12/07/2018 and the notes for the firmware state that it has been updated for the refresh Skylake-X CPUs, which is what I'm using.

Here are screen grabs for the BIOS. No AVX offset settings that I can see:

Main screen: https://imgur.com/IG1geel

CPU config screen: https://imgur.com/DT0Nil5

CPU info screen: https://imgur.com/ILT4bVx

So unless the Uncore Voltage Offset setting actually means AVX512, I'm not sure an AVX or AVX512 offset setting exists in the latest 121-SX-E296-KR - EVGA X299 Micro ATX 2 BIOS.

Edit: As you can see I increased the VCore to 1.15 to see if I can get a stable AIDA64 test.

Last fiddled with by simon389 on 2019-02-05 at 22:24
simon389 is offline   Reply With Quote
Old 2019-02-05, 22:30   #268
Mysticial
 
Mysticial's Avatar
 
Sep 2016

331 Posts
Default

Quote:
Originally Posted by simon389 View Post
Without AVX512 it is also running at 4.1Ghz. So you're correct, the BIOS is not lowering the clock speed while running AVX512. Also, my BIOS is the most recent version version 1.09 released 12/07/2018 and the notes for the firmware state that it has been updated for the refresh Skylake-X CPUs, which is what I'm using.

Here are screen grabs for the BIOS. No AVX offset settings that I can see:

Main screen: https://imgur.com/IG1geel

CPU config screen: https://imgur.com/DT0Nil5

CPU info screen: https://imgur.com/ILT4bVx

So unless the Uncore Voltage Offset setting actually means AVX512, I'm not sure an AVX or AVX512 offset setting exists in the latest 121-SX-E296-KR - EVGA X299 Micro ATX 2 BIOS.

Edit: As you can see I increased the VCore to 1.15 to see if I can get a stable AIDA64 test.
According to this: https://www.anandtech.com/show/12126...board-review/3

Quote:
Also like the FTW K, when switching to manual mode, either Per Core or Ratio Limit, it automatically sets a negative AVX and AVX2 offset of 3. Whereas other boards allow the user to set this manually, this is pre-applied and cannot be swtiched off. The result of this setting if it isn't manually adjusted is slower performance on AVX based tasks.
Ugh... So there really is no offset control... You're kinda out of luck.

Really EVGA? How can you omit such an important feature?
Mysticial is offline   Reply With Quote
Old 2019-02-05, 22:34   #269
chalsall
If I May
 
chalsall's Avatar
 
"Chris Halsall"
Sep 2002
Barbados

19×499 Posts
Default

Quote:
Originally Posted by simon389 View Post
Edit: As you can see I increased the VCore to 1.15 to see if I can get a stable AIDA64 test.
But this suggests that Mysticial is correct. It's really tricky to get your CPU to perform to spec.
chalsall is online now   Reply With Quote
Old 2019-02-05, 22:40   #270
simon389
 
Aug 2013

3·29 Posts
Default

Quote:
Originally Posted by Mysticial View Post
According to this: https://www.anandtech.com/show/12126...board-review/3



Ugh... So there really is no offset control... You're kinda out of luck.

Really EVGA? How can you omit such an important feature?
I think AVX512 should maybe only be an *option* in the next release of Prime95, since some people (like me) will suddenly not have stable default systems.
simon389 is offline   Reply With Quote
Old 2019-02-05, 22:46   #271
chalsall
If I May
 
chalsall's Avatar
 
"Chris Halsall"
Sep 2002
Barbados

19×499 Posts
Default

Quote:
Originally Posted by simon389 View Post
I think AVX512 should maybe only be an *option* in the next release of Prime95, since some people (like me) will suddenly not have stable default systems.
I would argue differently...

If I've bought kit which claims it can do things, it should be stable while doing said things, no matter how hard it is worked.
chalsall is online now   Reply With Quote
Old 2019-02-05, 22:47   #272
Mysticial
 
Mysticial's Avatar
 
Sep 2016

331 Posts
Default

Quote:
Originally Posted by simon389 View Post
I think AVX512 should maybe only be an *option* in the next release of Prime95, since some people (like me) will suddenly not have stable default systems.
TBH, I'd do the opposite. Use as much AVX512 as possible to expose the mobo vendors so they can fix their sh1t.

One of the main reasons the whole Skylake X platform got into this situation is because of the lack of software that used AVX512. So they didn't have anything to validate their systems before shipping.

I'd also consider filing a bug report on EVGA saying that stock system fails under Prime95. No system should be failing at stock under any workload, period. And call them out on the offset problem. lol
Mysticial is offline   Reply With Quote
Old 2019-02-05, 23:02   #273
simon389
 
Aug 2013

3×29 Posts
Default

Progress! If I change the CPU multiplier setting from Auto to Manual - Per Core the option for AVX2 offset and AVX3 offset appear at the bottom of the screen, each with a default setting of 3. And when I go to save the BIOS it tells me the "AVX2 Offset" setting is being switched from 0 to 3. So AVX2 and AVX3 offset does exist, but it's zero by default, and if you want to edit it you have to change the CPU multiplier setting to manual. So now I have the system running at 3.9Ghz in AIDA64 (with AVX512 setting). Hopefully that's stable, and I can see if it's also stable at 4.0Ghz. I just wish this stuff worked out of the box...

Quote:
Originally Posted by Mysticial View Post
TBH, I'd do the opposite. Use as much AVX512 as possible to expose the mobo vendors so they can fix their sh1t.

One of the main reasons the whole Skylake X platform got into this situation is because of the lack of software that used AVX512. So they didn't have anything to validate their systems before shipping.

I'd also consider filing a bug report on EVGA saying that stock system fails under Prime95. No system should be failing at stock under any workload, period. And call them out on the offset problem. lol
Well I might have an argument if Prime95 29.5b9 was an official release. But as of right now I'd be complaining about a beta version. So maybe I could say that AIDA64 stress test is failing at stock.

Last fiddled with by simon389 on 2019-02-05 at 23:21
simon389 is offline   Reply With Quote
Old 2019-02-05, 23:12   #274
chalsall
If I May
 
chalsall's Avatar
 
"Chris Halsall"
Sep 2002
Barbados

19×499 Posts
Default

Quote:
Originally Posted by simon389 View Post
So maybe I could say that AIDA64 stress test is failing at stock.
OK. Let's talk about that.

Do you have an understanding about why that is happening?
chalsall is online now   Reply With Quote
Old 2019-02-05, 23:56   #275
GP2
 
GP2's Avatar
 
Sep 2003

32×7×41 Posts
Default

Quote:
Originally Posted by simon389 View Post
I think AVX512 should maybe only be an *option* in the next release of Prime95, since some people (like me) will suddenly not have stable default systems.
In your local.txt file you can set CpuSupportsAVX512F=0
GP2 is offline   Reply With Quote
Reply

Thread Tools


All times are UTC. The time now is 02:06.

Fri Mar 5 02:06:08 UTC 2021 up 91 days, 22:17, 0 users, load averages: 1.21, 1.40, 1.59

Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2021, 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.