• Guest, The rules for the P & N subforum have been updated to prohibit "ad hominem" or personal attacks against other posters. See the full details in the post "Politics and News Rules & Guidelines."

PrimeGrid Challenges 2021

Page 10 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

StefanR5R

Elite Member
Dec 10, 2016
4,174
4,679
136
Switching from PPS-sieve to llrSOB isn't straightforward: One is good on GPUs, the other tends to prefer CPUs. ;-)
 
  • Like
Reactions: Ken g6

Markfw

CPU Moderator, VC&G Moderator, Elite Member
Super Moderator
May 16, 2002
21,556
9,622
136
I spent about 2 days worth of work tightening the timings and adjusting the voltages for my 5950 on my Crosshair hero viii (wifi). The default settings w/ XMP enabled used way, way too much voltage on the memory controller, which was unstable at higher clockspeed. Dropping the SOC voltages allowed me to get to 3866 stable, and 4000 mostly stable. My guess is that 4000 would be stable if I had less than 64GB ram, but I don't have extra ram (or the patience) to buy a smaller kit to see if it changes the stability at higher clocks.
curious what the SOC and memory voltages that you set, and what they were. I might be able to up my memory speed before the next set (2 hours to go)

And what memory did you get ??
 

Justinus

Platinum Member
Oct 10, 2005
2,696
821
136
It's back in fast mode, running with an average effective clock of 4.1 GHz. 25% done in 3:11:53.

It's at a 4.4GHz clockspeed only running at 4.1 average effective clock, which leads me to believe I'm hitting memory limitations at this speed.

For reference, the stock run was only at 3.3 GHz and ran at an average effective clock of 3.2 GHz (97% of actual clock) and now at 4.4/4.1 it's only at 93% of actual clock.
 

Endgame124

Senior member
Feb 11, 2008
745
474
136
curious what the SOC and memory voltages that you set, and what they were. I might be able to up my memory speed before the next set (2 hours to go)

And what memory did you get ??
I used this memory (it was much cheaper when I bought it)

Auto voltage Settings:
VSOC: 1.08
CLDO VDDP: 1.0979
VDDG CCD: .9474
VDDG IOD: 1.0477
 

Markfw

CPU Moderator, VC&G Moderator, Elite Member
Super Moderator
May 16, 2002
21,556
9,622
136
I used this memory (it was much cheaper when I bought it)

Auto voltage Settings:
VSOC: 1.08
CLDO VDDP: 1.0979
VDDG CCD: .9474
VDDG IOD: 1.0477
and what did you lower the voltages to ??? and 1.35 for the memory ? or 1.4 or 1.45 ?

Thats exactly the memory I have ! I paid $269 for it.
 

Endgame124

Senior member
Feb 11, 2008
745
474
136
and what did you lower the voltages to ??? and 1.35 for the memory ? or 1.4 or 1.45 ?

Thats exactly the memory I have ! I paid $269 for it.
Sorry, I posted earlier than I meant to.

Final setting:
VSOC: 1.03
CLDO VDDP: .8975
VDDG CCD: .8975
VDDG IOD: .9975

IF, memclock & uclock 1:1

Memory 1.45, though I have a fan blowing directly on top of the dimms, so I wasn't overly concerned about heat. I never tried it at 1.4.

Its also 98% stable at 4000. It will run memtest for 4 hours without a detectable issue, but it never survives overnight. It feels like I'm one setting away from 4000, but I have no idea what that might be...
 

Justinus

Platinum Member
Oct 10, 2005
2,696
821
136
Sorry, I posted earlier than I meant to.

Final setting:
VSOC: 1.03
CLDO VDDP: .8975
VDDG CCD: .8975
VDDG IOD: .9975

IF, memclock & uclock 1:1

Memory 1.45, though I have a fan blowing directly on top of the dimms, so I wasn't overly concerned about heat. I never tried it at 1.4.

Its also 98% stable at 4000. It will run memtest for 4 hours without a detectable issue, but it never survives overnight. It feels like I'm one setting away from 4000, but I have no idea what that might be...
Are you sure your IF is stable at 2000? I can run memtest for a long time at 4000 without errors but in windows I get correctable whea errors out the wazoo. I'm stuck at 3800 in order to not get any whea errors.
 

Endgame124

Senior member
Feb 11, 2008
745
474
136
Are you sure your IF is stable at 2000? I can run memtest for a long time at 4000 without errors but in windows I get correctable whea errors out the wazoo. I'm stuck at 3800 in order to not get any whea errors.
Like I said, I can't get IF stable at 2000. It will, always whea and eventually random reboot over the course of a day.

It works great at 1866. 1900 won't even boot, but 1933 and 1966 are very similar to 2000. They work, but end up with whea over time.
 
  • Like
Reactions: Justinus

Justinus

Platinum Member
Oct 10, 2005
2,696
821
136
Like I said, I can't get IF stable at 2000. It will, always whea and eventually random reboot over the course of a day.

It works great at 1866. 1900 won't even boot, but 1933 and 1966 are very similar to 2000. They work, but end up with whea over time.
My hard limit is 1900. 1933, 1966, and 2000 all soft-whea like mad. I can definitely run 4000 ram memtest stable if I run async but the performance hit is not good.

In the other post you said you have 3866 ram stable - did you mean 3733 (2x1866)?
 

Endgame124

Senior member
Feb 11, 2008
745
474
136
My hard limit is 1900. 1933, 1966, and 2000 all soft-whea like mad. I can definitely run 4000 ram memtest stable if I run async but the performance hit is not good.
IF 1933 is better for me. It won't randomly reboot like 1966 and 2000 will over time, but it will still build up whea errors at a rate of about 3 per day. I tuned voltages for 2 days, timings for 2 days, and then spent weeks on and off trying to get 1933 to run without whea errors. Eventually I said that the extra 66mhz probably wasn't worth the effort I was putting into it, so I quit playing with it. The system is still a dramatic improvement over my old A10-7870K.
 
  • Like
Reactions: Justinus

Endgame124

Senior member
Feb 11, 2008
745
474
136
My hard limit is 1900. 1933, 1966, and 2000 all soft-whea like mad. I can definitely run 4000 ram memtest stable if I run async but the performance hit is not good.

In the other post you said you have 3866 ram stable - did you mean 3733 (2x1866)?
Sorry, 3733 is stable. I have 1933 / 3866 on the brain because i spent so much time on it. I was posting while on a conference call and apparently the brain doesn’t work so well with split attention.
 

Markfw

CPU Moderator, VC&G Moderator, Elite Member
Super Moderator
May 16, 2002
21,556
9,622
136
OK, BIOS was 1.4, now at 1.61. Memory is now at 1900/3800. All memory voltages are as close as I can get to endgame124's.

Now lets see how long it takes, and if its stable.

Its only been 2 minutes, and the ETA is already down to 18 hours and dropping. (from 23)

Based on my last problem, if it stays up for 30 minutes, it will probably run all day.

Edit:: Odd, now its running at 3.45 ghz, and I did not touch the CPU settings, just the memory. ETA (total time) is holding at 18 hours. still stable.
 
Last edited:

Justinus

Platinum Member
Oct 10, 2005
2,696
821
136
OK, BIOS was 1.4, now at 1.61. Memory is now at 1900/3800. All memory voltages are as close as I can get to endgame124's.

Now lets see how long it takes, and if its stable.

Its only been 2 minutes, and the ETA is already down to 18 hours and dropping. (from 23)

Based on my last problem, if it stays up for 30 minutes, it will probably run all day.

Edit:: Odd, now its running at 3.45 ghz, and I did not touch the CPU settings, just the memory. ETA (total time) is holding at 18 hours. still stable.
My guess is the CPU cores were spending more time idling waiting for memory/cache/etc. due to slower IF/memory speeds so they were drawing less power and were able to maintain a higher clockspeed. Now that they are working harder they are using more power and hitting the current limits at lower frequencies.

That's why seeing the effective clockspeed is key to understanding how fully utilized the cores are - I just don't know how you can see a metric like that, if possible, on Linux.
 

Markfw

CPU Moderator, VC&G Moderator, Elite Member
Super Moderator
May 16, 2002
21,556
9,622
136
My guess is the CPU cores were spending more time idling waiting for memory/cache/etc. due to slower IF/memory speeds so they were drawing less power and were able to maintain a higher clockspeed. Now that they are working harder they are using more power and hitting the current limits at lower frequencies.

That's why seeing the effective clockspeed is key to understanding how fully utilized the cores are - I just don't know how you can see a metric like that, if possible, on Linux.
Well, 18 hours is certainly better than 23 hours (my last batch) So I am happy now, thanks to you and endgame124.

BTW, the whole problem could have been that it was trying to run 1.22v for memory, even using A_XMP. But anyway, endgame124's voltages are now being used@1900/3800, so I am very happy.
 
  • Like
Reactions: Endgame124

Endgame124

Senior member
Feb 11, 2008
745
474
136
Well, 18 hours is certainly better than 23 hours (my last batch) So I am happy now, thanks to you and endgame124.

BTW, the whole problem could have been that it was trying to run 1.22v for memory, even using A_XMP. But anyway, endgame124's voltages are now being used@1900/3800, so I am very happy.
Glad I could assist, and just a touch jealous that you got 1900/3800.

also a touch jealous you got something out of the new egg shuffle... my big win was a RX 560...
 

Markfw

CPU Moderator, VC&G Moderator, Elite Member
Super Moderator
May 16, 2002
21,556
9,622
136
Glad I could assist, and just a touch jealous that you got 1900/3800.

also a touch jealous you got something out of the new egg shuffle... my big win was a RX 560...
well, I got a 3070, but not excited about it... I entered today for another 3070, an EVGA one, no word yet, so I probably lost.

Oh, and I have a 3080 on order (for 2 months) with shopblt. I should know by tomorrow if the current ETA of today is valid.
 

Endgame124

Senior member
Feb 11, 2008
745
474
136
well, I got a 3070, but not excited about it... I entered today for another 3070, an EVGA one, no word yet, so I probably lost.

Oh, and I have a 3080 on order (for 2 months) with shopblt. I should know by tomorrow if the current ETA of today is valid.
First one I’ve missed in a while. Got stuck in a 5 hour meeting today :(
 

Markfw

CPU Moderator, VC&G Moderator, Elite Member
Super Moderator
May 16, 2002
21,556
9,622
136
So, @StefanR5R , how is 18 hours on this project ?? It looks like that is the maximum for a well configured 5950x.
 

crashtech

Diamond Member
Jan 4, 2013
9,848
1,644
126
It's a bit distressing to me that memory doesn't seem to be "set it and forget it" anymore. I'm not a big fan of tinkering with memory settings!
 

Markfw

CPU Moderator, VC&G Moderator, Elite Member
Super Moderator
May 16, 2002
21,556
9,622
136
My current round is going to be completed in under 13 hours ;)
Well, before I get pissed, I will wait until these units finish, and see how the next does (like yours). But at least I know the memory is finally tweaked and working.
 

Justinus

Platinum Member
Oct 10, 2005
2,696
821
136
Well, before I get pissed, I will wait until these units finish, and see how the next does (like yours). But at least I know the memory is finally tweaked and working.
Oh definitely don't get pissed, this is with overclocking/allowing higher power limits. Runs pretty hot doing this even at only 4.4GHz on my water loop.
 

Markfw

CPU Moderator, VC&G Moderator, Elite Member
Super Moderator
May 16, 2002
21,556
9,622
136
I trashed my water setup, as it was noisy, leaked and sucked power like no other system I had. So now I have the best air cooling available, but not sure the temps. Does not show in linux under that software.
 

ASK THE COMMUNITY