CPU Overclocking - is it possible?

Message boards : Number crunching : CPU Overclocking - is it possible?
Message board moderation

To post messages, you must log in.

Previous · 1 · 2

AuthorMessage
Profile petri33
Volunteer tester

Send message
Joined: 6 Jun 02
Posts: 1668
Credit: 623,086,772
RAC: 156
Finland
Message 1745121 - Posted: 26 Nov 2015, 23:33:49 UTC - in response to Message 1744948.  
Last modified: 26 Nov 2015, 23:48:54 UTC

I can't get Boinc to a reliable state with ANY kind of OC going on, even the slightest bit. Is there something specific I need to tweak for Boinc or is it just a bad idea to OC for it? I have an otherwise fairly stable overclock, passed multiple stress tests for 6-8 hours each and the temps are also pretty comfortable. Running Boinc though, will result to random BSODs, from 5 minutes in to 3 hours in, I can't really replicate it. Anyone has had any better luck with overclocking and crunching?

PS: It's a 5820k, pushed to 4.5ghz. Boinc wasn't stable at even 3.9ghz, which is just 0.1 more than the default turbo speeds of this cpu.


Hi,

the Seti CPU applications, both Multibeam and Astropulse, are higly optimized. They utilize the CPU to the maximum and there is a lot of heat generated. Most applications sit idle and have time for the heat to dissipate. Thats why your OC is unstable with Seti.

You sure? I always thought the stock apps are not very optimized, so they can run on almost any CPU.
Anyway, I don't see any reasons, why the CPU would pass a stress test for hours but immediatly chokes on Seti. That's really strange. There are/have been ppl running Seti on oc'd CPUs. Mark and his famous machine for example. Salad Fingers, I would try Lunatic apps instead of stock.


I have a feeling that the lunatics CPU optimizations were adopted to be the official CPU versions a couple of years ago. And it is also possible that the same thing happened to the Cuda x41zc. I'm not sure.

I do not stress my CPU any more. It is an old 3930K and runs nowadays @3.9GHz. It used to run 4.3 24/7/365. 6 cores Seti and the remaining 6 ht cores serving GPU and me. I still have it running 6 CPU AP/MB and 6 free for other tasks. It seems to be doing quite OK.
To overcome Heisenbergs:
"You can't always get what you want / but if you try sometimes you just might find / you get what you need." -- Rolling Stones
ID: 1745121 · Report as offensive
Profile William
Volunteer tester
Avatar

Send message
Joined: 14 Feb 13
Posts: 2037
Credit: 17,689,662
RAC: 0
Message 1745212 - Posted: 27 Nov 2015, 10:39:18 UTC - in response to Message 1745121.  

I have a feeling that the lunatics CPU optimizations were adopted to be the official CPU versions a couple of years ago. And it is also possible that the same thing happened to the Cuda x41zc. I'm not sure.

Nope, the CPU versions are not from us, and neither are the android ones. We have,however, been supplying the GPU versions for several years now. That includes x41zc.

_some_ optimisations made it into the stock CPU code over the years, but stock only supports SSE and AVX SIMD levels and that makes a hell lot of difference if you have something that supports say SSE2.
A person who won't read has no advantage over one who can't read. (Mark Twain)
ID: 1745212 · Report as offensive
Profile petri33
Volunteer tester

Send message
Joined: 6 Jun 02
Posts: 1668
Credit: 623,086,772
RAC: 156
Finland
Message 1745218 - Posted: 27 Nov 2015, 11:12:16 UTC - in response to Message 1745212.  

I have a feeling that the lunatics CPU optimizations were adopted to be the official CPU versions a couple of years ago. And it is also possible that the same thing happened to the Cuda x41zc. I'm not sure.

Nope, the CPU versions are not from us, and neither are the android ones. We have,however, been supplying the GPU versions for several years now. That includes x41zc.

_some_ optimisations made it into the stock CPU code over the years, but stock only supports SSE and AVX SIMD levels and that makes a hell lot of difference if you have something that supports say SSE2.


Thanks for clarification.
To overcome Heisenbergs:
"You can't always get what you want / but if you try sometimes you just might find / you get what you need." -- Rolling Stones
ID: 1745218 · Report as offensive
Claggy
Volunteer tester

Send message
Joined: 5 Jul 99
Posts: 4654
Credit: 47,537,079
RAC: 4
United Kingdom
Message 1745270 - Posted: 27 Nov 2015, 17:51:47 UTC - in response to Message 1745212.  
Last modified: 27 Nov 2015, 17:52:42 UTC

I have a feeling that the lunatics CPU optimizations were adopted to be the official CPU versions a couple of years ago. And it is also possible that the same thing happened to the Cuda x41zc. I'm not sure.

Nope, the CPU versions are not from us, and neither are the android ones. We have,however, been supplying the GPU versions for several years now. That includes x41zc.

_some_ optimisations made it into the stock CPU code over the years, but stock only supports SSE and AVX SIMD levels and that makes a hell lot of difference if you have something that supports say SSE2.

The Function choices that are available to the Stock 7.01 Linux x64 app, there are some SSE2 and SSE3 optimisations, But only for ChirpData:

setiathome_v7 7.00 Revision: 1772 g++ (GCC) 4.4.6 20110731 (Red Hat 4.4.6-3)
libboinc: BOINC 7.1.0

Work Unit Info:
...............
WU true angle range is : 0.008955
Optimal function choices:
--------------------------------------------------------
name timing error
--------------------------------------------------------
v_BaseLineSmooth (no other)

v_GetPowerSpectrum 0.000312 0.00000 test
v_vGetPowerSpectrum 0.000262 0.00000 test
v_vGetPowerSpectrum2 0.000279 0.00000 test
v_vGetPowerSpectrumUnrolled 0.000246 0.00000 test
v_vGetPowerSpectrumUnrolled2 0.000370 0.00000 test
v_avxGetPowerSpectrum faulted
v_vGetPowerSpectrumUnrolled 0.000246 0.00000 choice

v_ChirpData 0.018884 0.00000 test
fpu_ChirpData 0.033732 0.00000 test
fpu_opt_ChirpData 0.021543 0.00000 test
v_vChirpData_x86_64 0.123544 0.03953 test
sse1_ChirpData_ak 0.017488 0.00000 test
sse1_ChirpData_ak8e 0.014017 0.00000 test
sse1_ChirpData_ak8h 0.014662 0.00000 test
sse2_ChirpData_ak 0.019267 0.00000 test
sse2_ChirpData_ak8 0.012947 0.00000 test
sse3_ChirpData_ak 0.019511 0.00000 test
sse3_ChirpData_ak8 0.010893 0.00000 test
avx_ChirpData_a faulted
avx_ChirpData_b faulted
avx_ChirpData_c faulted
avx_ChirpData_d faulted
sse3_ChirpData_ak8 0.010893 0.00000 choice

v_Transpose 0.043648 0.00000 test
v_Transpose2 0.023649 0.00000 test
v_Transpose4 0.012405 0.00000 test
v_Transpose8 0.019241 0.00000 test
v_pfTranspose2 0.025504 0.00000 test
v_pfTranspose4 0.012925 0.00000 test
v_pfTranspose8 0.020227 0.00000 test
v_vTranspose4 0.013009 0.00000 test
v_vTranspose4np 0.011504 0.00000 test
v_vTranspose4ntw 0.010808 0.00000 test
v_vTranspose4x8ntw 0.011445 0.00000 test
v_vTranspose4x16ntw 0.005536 0.00000 test
v_vpfTranspose8x4ntw 0.010703 0.00000 test
v_avxTranspose4x8ntw faulted
v_avxTranspose4x16ntw faulted
v_avxTranspose8x4ntw faulted
v_avxTranspose8x8ntw_a faulted
v_avxTranspose8x8ntw_b faulted
v_vTranspose4x16ntw 0.005536 0.00000 choice

FPU opt folding 0.003694 0.00000 test
ben SSE folding 0.003866 0.00000 test
AK SSE folding 0.003208 0.00000 test
BH SSE folding 0.003255 0.00000 test
JS AVX_a folding faulted
JS AVX_c folding faulted
AK SSE folding 0.003208 0.00000 choice

Test duration 10.30 seconds


Claggy
ID: 1745270 · Report as offensive
Profile Dr Grey

Send message
Joined: 27 May 99
Posts: 154
Credit: 104,147,344
RAC: 21
United Kingdom
Message 1745319 - Posted: 27 Nov 2015, 22:46:33 UTC

Just to check in with my 5820K experience with Boinc. While it seems to have been stable for a while at 4.3 GHz, over the last few days I've had a series of reboots - about once a day. So, I've dropped it down to 4.2 GHz and it's been steady for the last 48 Hrs.
I'm thinking I could add a bit more voltage or turn the fans up a bit to keep it cooler for stability at the higher speed but I don't need the extra heat or noise in the room. I'm looking at the temperature now and its sitting at 67 C being cooled by an H100i.
Maybe as the temperature drops outside over the next months I'll push the voltage up a bit instead of turning on the heating and run it a little hotter.
ID: 1745319 · Report as offensive
kittyman Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 9 Jul 00
Posts: 51468
Credit: 1,018,363,574
RAC: 1,004
United States
Message 1746789 - Posted: 4 Dec 2015, 9:16:09 UTC - in response to Message 1744969.  
Last modified: 4 Dec 2015, 9:18:48 UTC


Speaking of Kitties you could do what Mark Sattler does. Use Freon based cooler and freeze your ass off

Sorry mark. Your great in my book.

Michael Miles

Thank you very much, Michael for that kind comment......

The frozen experiment ended a while back.
Very hard to maintain, and almost certainly resulting in mobo or cpu demise...LOL. I lost a couple of both to the freezer. Condensation control is near impossible at those temperatures. Tube after tube of silicon grease in the cpu socket and all around it.....
But I was able to do it for long enough to show that at least it COULD be done.

All air cooled now, and the GPUs have taken over, so CPU ocing does not have near as much impact anyway.
The compressor setup is now in my basement. Maybe if one day there is some kind of Seti museum, it should go on display there....

It was a very interesting bit of Seti and computing history, however.
"Freedom is just Chaos, with better lighting." Alan Dean Foster

ID: 1746789 · Report as offensive
J. Mileski
Volunteer tester
Avatar

Send message
Joined: 9 Jun 02
Posts: 632
Credit: 172,116,532
RAC: 572
United States
Message 1746946 - Posted: 5 Dec 2015, 0:40:38 UTC - in response to Message 1746789.  


Speaking of Kitties you could do what Mark Sattler does. Use Freon based cooler and freeze your ass off

Sorry mark. Your great in my book.

Michael Miles

Thank you very much, Michael for that kind comment......

The frozen experiment ended a while back.
Very hard to maintain, and almost certainly resulting in mobo or cpu demise...LOL. I lost a couple of both to the freezer. Condensation control is near impossible at those temperatures. Tube after tube of silicon grease in the cpu socket and all around it.....
But I was able to do it for long enough to show that at least it COULD be done.

All air cooled now, and the GPUs have taken over, so CPU ocing does not have near as much impact anyway.
The compressor setup is now in my basement. Maybe if one day there is some kind of Seti museum, it should go on display there....

It was a very interesting bit of Seti and computing history, however.


Modern phase change cooling
ID: 1746946 · Report as offensive
kittyman Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 9 Jul 00
Posts: 51468
Credit: 1,018,363,574
RAC: 1,004
United States
Message 1746960 - Posted: 5 Dec 2015, 1:23:12 UTC

Modern, hell....LOL.

Looks very much like the Danfoss compressor I used on the Frozen Penny.....
It took MONTHS of trial and error for the kid I was working with to get the freon blend right.....
Too much of one freon got the temps down alright, but the compressor locked up when sucking liquid freon when temps got too low to evaporate the fluid.
Another blend just did not cool very well. Several month later...he hit the right mix. He never did tell me exactly what the mix was. He was college bound, and long ago left the freezing scene.

It had to achieve the ultimate low temp, but also had to avoid slugging the compressor if the rig went down and temps crashed to minus 40.........

I trashed two stock fridge units before I had the custom one built.
It still resides in my basement.....and ya never know.......
Some day, somewhere, I may fire it up again just for amusement.
\
I am rather low on funds these days, but if anybody wishes to fund the experiment again, PM me with about 3 grand and I shall happily rebuild the Frozen One with a current build and see what gives.
I can get many more tubes of silicon grease from the auto parts store.
I am very curious, but I suspect that the current gen processors won't get the boost that the old high powered toasters did. \\

I'd need access to a experimental CPU with unlocked multi as well.

If anybody wants to toss it at me, I will build it.

Meow.
"Freedom is just Chaos, with better lighting." Alan Dean Foster

ID: 1746960 · Report as offensive
Previous · 1 · 2

Message boards : Number crunching : CPU Overclocking - is it possible?


 
©2024 University of California
 
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.