View Full Version : why S-PI instabillity?
mag|c
05-15-2005, 01:33 PM
Hey
I'm having problems getting my naked venice 3000+ @ 2,7Ghz running spi 16m & 32m, but it primes and do memtest all day long, what can cause this?
I have tried more vcore, ldt, nb & vdimm voltage.
And it seems my memory only likes 2,6v, it's the only voltage is does a error free memtest, both with and without active cooling.
Using this hardware n' settings:
Naked A64 Venice 1.8 @ 2.8 (9x300) - vcore 1,47v - TR SLK-948u w. Enermax 92mm
DFI nF4 SLI-DR (Bios 510-3)
G.Skill 1GB LA (2,6v) - 2,5-4-6-3-7-14-2-2-1-2-AUTO-AUTO-ENA AUTO-0-AUTO-AUTO-NORMAL-AUTO-DIS-16x-7x-DIS
TAGAN 480w
Thanks
mag|c
Eversor
05-15-2005, 04:06 PM
Hey
I'm having problems getting my naked venice 3000+ @ 2,7Ghz running spi 16m & 32m, but it primes and do memtest all day long, what can cause this?
I have tried more vcore, ldt, nb & vdimm voltage.
And it seems my memory only likes 2,6v, it's the only voltage is does a error free memtest, both with and without active cooling.
Using this hardware n' settings:
Naked A64 Venice 1.8 @ 2.8 (9x300) - vcore 1,47v - TR SLK-948u w. Enermax 92mm
DFI nF4 SLI-DR (Bios 510-3)
G.Skill 1GB LA (2,6v) - 2,5-4-6-3-7-14-2-2-1-2-AUTO-AUTO-ENA AUTO-0-AUTO-AUTO-NORMAL-AUTO-DIS-16x-7x-DIS
TAGAN 480w
Thanks
mag|c
had the same problem. Going from cas 1.5 to cas 2 fixed it. Go figure, memtest was stable. Guess my winnie doesnt like.
halcyon
05-15-2005, 10:26 PM
"what can cause this?"
Sorry to be an ass, but running your parts out of spec is causing it.
However, with the modern A64 design with the dedicated memory controller, it is very hard, if not impossible to accurately pin-point whether the problem is in:
- CPU calculation
- CPU mem controller
- memory dimms
- general mobo/psu instability
However, if your system is not stable in S-PI, then it's not stable, even if it's stable in Prime95 and memtest.
Alsok, I'd guess that if you run your system long enough in Prime95/torture/blend (or small FFT) and Memtest (esp. test 5), then you'll encounter errors in those too.
An additional tool (esp. for heat generation) that you might want to test your rig for stability with is S&M:
http://www.testmem.nm.ru/snm_test.zip
mag|c
05-16-2005, 01:27 AM
An additional tool (esp. for heat generation) that you might want to test your rig for stability with is S&M:
http://www.testmem.nm.ru/snm_test.zip
According this program my mem is unstable, wierd memtest says different.
Anyone know the full stock specs, for G.Skill 512x2 LA?
halcyon
05-16-2005, 08:00 AM
Look, different programs test differently and finding instability is not a deterministic effort, it's stochastic.
Sometimes you run the test and it passes, sometimes it doesn't.
However, even a single failure means it's not FULLY stable.
Accidents don't happen all the time, just every now and then (unless the overlock is so outrageous that the computer crashes immediately).
Also, memtest does not test the cpu load (thermal output) nearly as well as S&M does.
So if S&M says it's unstable, trust it, it IS unstable.
Sorry, can't help you with G.Skill specs.
LexDiamonds
05-16-2005, 08:08 AM
Also, check your cpu utilization under Prime95. I have noticed on some occasions that a freshly downloaded prime95 doesnt always peg the cpu like it should. I have seen this on a variety of platforms... With anything less that %100 cpu usage, Prime is worthless.
Super Pi always seems to opperate like it should.
As others have mentioned, a 32Mb Super Pi is just a good start on the road to stability..until you have this, back down your OC.
:toast:
ben805
05-16-2005, 02:35 PM
try OCCT, I like it WAY better than Prime95, SuperPI, memtest or any other stability programs out there.
halcyon
05-16-2005, 11:30 PM
You may like it more, but it doesn't generate heat as much as S&M nor does it find memory errors nearly as fast as S&M.
If your time is precious and you only want to run one program, run S&M at full 100% cpu load and the memory test on top of that (both from S&M).
mag|c
05-17-2005, 08:14 AM
try OCCT, I like it WAY better than Prime95, SuperPI, memtest or any other stability programs out there.
its the half hour test in OCCT stable no problem.
CD 1986
05-17-2005, 08:33 AM
Have you tried loosening memory timings to 2.5-4-4-8 or something similar? Since you know it's a memory error i'd suggest using windows memtest to quickly test for memory errors, and the start loosening timings / adjusting mem voltage again to find your sweet-spot at 300mhz.
Reduce your CPU multiplier and put LDT / CPU voltages back to stock so that only the memory is being stressed. Then you can work on the memory till its stable and increase your CPU overclock again.
http://hcidesign.com/memtest/download.html
For my experience spi 16/32MB > dram or memory controller issues
In my opinion Memtest86 (bios vers.) is not a good tester for dram
I have a bad memory controller (CBBID) but I am rock solid (prime, spi, S&M, memtest, Gold Memory) at 300x8 with a good bandwidth
This is my dram configuration for GSkill (TCCD 440) :
2.8V
200
Enabled (1T)
2.5
4
8
4
8
14
2
2
2
2
4708
Auto
Enabled
Auto
0
Auto
Level2
8
5
256
Enabled
16x
7
Disabled
Just a start point, then you can improve settings ;)
Geebs
05-17-2005, 04:50 PM
Man halcyon,
You were right about it testing thermal generation better than Memtest or SuperPi or Prime95... S&M seems to reallly stress a system...
I just downloaded it and ran it for the first time... My temps werea good 3 degrees celcius hotter than any of the other stress programs I have tried...
But, I am proud to say, my system passed ALL the default tests... :woot:
Thanks for posting the link to S&M... a nice tool indeed
:toast:
Why don't you try upping your vDimm? It shows you that only have 2.6V for 300FSB...that's very impressive, but may also be holding you back. Kick it to 2.7V or 2.8V even.
[G.N.U.]Fragman
07-31-2005, 09:09 PM
just running folding@home 24/7 for over a week now.....and a full 32mb s-pi wasnt a problem...but occt wont run. about 3-20 min inside the test it crashes.......i think perhaps it has some flaws on some systems.....
[G.N.U.]Fragman
07-31-2005, 10:38 PM
not a problem...but dont like that program.....crashes randomly for me...perhaps some tweak in my windows xp it doesnt like....but folding should bed very stressfull....avi to mpg/vcd and so on also is quite good for testing stability