Ugh..Prime95 and my overclock

[nealh]

I just have no clear understanding of why Prime95 and O/C seems to have changed

I flashed the bios on my dvd writer and winxp on my Neo2 plat with NF3 driver would not recognize it...so I had to switch to the ms ide drivers...reflash up my Benq 1620 to B7P9....then reinstall the NF3 drivers version 2.6...everything seems fine...

so I went ahead and reran Prime95 after playing some hl2....and it failed after like 2.5 hrs....WTF...this exact setup was running Prime95 earlier this week for 10hrs and I stopped it...never failed

so I reflashed my mobo bios figuring maybe when I flashed the dvdwriter something changed...reflashed same bios..1.36b..replaced all my setting

A64 3200@2550 cpu 10x255 4xhtt (this was ok before) 1.58v
ram 133 2.5-3-3-10-1t
ddr 2.7v
nv/ati speed up off
speed spectrum disabled
dynamic o/cing off
agp 1.5v(or 1.6v both times prime 95 ram for 10 hrs)
I have disabled cool n quiet, parallel, serial, IR, Nvidia lan, sata 1&2 disabled...I use sata 3

these exact settings ran for 10hr no prime95 failures twice...within last 5 days...

so I reran this am... come home and now I get 3.5hrs....

why would prime95..with same hardware settings, software, no chnages...start failing...prime 95 only failed with 1.41 modded bios at 2550....2550 was bulletproof...now it fails ...WTF

I just give up...this makes no sense to me at all...I guess something could have changed with Benq fw flash(which I did once before with 1.41modded bios and had similar dvdrw recognition issue) and now there is a theoretically hardware or software issue but this seem very strange

 

[Duvie]

That is weird...Mine has always stayed consitent...I have tested 3 times at my current speed for 8 hours and it has passed every time...I have done much addding and tweaking so that is why I reran it....

I dont use blend test...I test no ram in prime95...that is what memtest is for. I then run 8-12 hours of it and it usually gets 20 passes of extended test, cache on, etc.... I then run 100 passes of each test #5 and #6....

 

[nealh]

Duvie...which test are you running?

I am confused...I run the torture test default so to speak which also tests ram...in version 23.8 the tests are not broken down by number...so I do you get test 5&6..so you run extended test??

could you be a more specific on how you configure Prime95

BTW I am rerunning Prime95 at 2550 and lowered ram back to 2t and I am at 4hrs and counting

 

[Duvie]

I pm'd you this as well....

Test 5 and 6 was in reference to memtest86, no tprime....I was saying just run small FFT or large FFT, but no blended test. The blended test test ram (not very good, IMO and experience) and due to this does not build the stress and heat on the chip...that is why most do not use it....there will be times in blended where the cpu is under very little use and temps are just a hair above idle....

 

[nealh]

Thanks as always Duvie

so the blended test is no good..great I have been using this as my stress test..I have found it keeps a near 100% cpu load at all times except for occasion test change and temp at general at 51/52c(allow I have found my neo2 plat will report different temps at time 2-3C higher but if I reboot and rerun prime same test etc I get lower temp..general load temps are 51/52C occasional I will see 53/54C...I think this a difference with core center..literally I can reboot and rerun and it is lower weird)..I have no concern about temps...

I live in FL

anyway the blended test is running now nearly 5hrs....

 

[Duvie]

Also shitcan the corecenter....At least do not have it on while running apps...this is widely known as a possible source of instability....I use speedfan or everest.....much better though everest doe not read my +12v rail right......says like 2.x volts....

 

[nealh]

Hehe...I always have core center open when I use prime95..lol

I will get speedfan

 

[nealh]

For what it is worth..I reran Prime95 blended torture test last night at ram 133 2.5-3-3-10 2t at 2550 4x htt...I stopped this am at 15.5hrs....I do not get it

I did turn off core center as you suggested and will no longer use it

I wonder if this could have been causing my issues as it does "check" temps ...I am for grins going to try 2600 again with no core center since when I first set the system up I started prime95 and ran it for a couple hrs...reboot at 2600 and the P-1 test(not sure how this started ran for 3.5 hrs before I stopped it)...but I was never able to duplicate this


so I will retest and see if core center was the issue

 

[Tyrant222]

hey neal we pretty much got the same cpu (posted on xs as well)

hits that wall at 2550 at 1.575 v. Ive been just running mine at 2.6 @ 1.6v. Seems like MANY wk 44s hit this wall.

for speedfan, your cpu temp will be reported as Temp2.

id be overjoyed if you can break out of that 2550 barrier. will give me hope.

 

[nealh]

I came home and retried 2600 at 1.5 w/10% over vid....1.63v....died on Prime95 small fft within minutes...

I give up..I can like you run fine at 2600 (at 1.58v)...but I figure I will stick with my prime 95 stable speed..at least I know my o/c is fine

I am very happy with 15.5hrs of blended torture test in version 23.8..100% load

core center was not the issue..its the cpu

oh well someday I will get a sweet one...I hope, maybe

BTW what are the other reported temps from speedfan?..which one is system temp..is there one that is a videocard temp

 

[Duvie]

Actually the 3rd temp I believe is my aux temp. My 1st temp I though use to read right but now it is reading -22c...that aint right!!! Also take a look at everest home edition...

theonly thing I can say is I have yet to see 1 temp monitoring app be 100% compatabile with this board....I use about 3 -4 and ry to decipher the truth...they seem to be close on the importnat stuff....vcore, vdimm, +3.3v, +12v (somewhat close to one another), and cpu temp....they also all agree on my HDD temps

 

[Turin99]

Hey Duvie, I have a question for you if you don't mind:
How have you setup your voltage (1.57v) with the msi neo2 plat?
Did you set it 1.55 + x%/ or what combination did you use?
Thanks!

 

[Duvie]

1.5v + 8.3%...board undervolts...I answered in your thread!!!

 

[nealh]

In fact I found that even using the higher vcores seem to give the same voltage....ie 1.475 w/ 8.3 % over vid = 1.5v with 8.3% over vid = 1.55v with 8.3% over vid at least with 1.36b

 

[nealh]

Duvie..I know you run version 23.6 but in 23.8 with blended torture test ..I found cpu load was 100% and temps never seem to move off load temps of 52C...the only time it changed was if it changed tests for a second but you see this with any Prime95 setup...ie it runs test 1...for x time then goes to test2 x time etc....

I am sick of Prime95 now..I may go back and test small fft/large fft....but I fail comfortable(maybe I am wong) that after 15.5 hrs with no crashes that my system is stable

too bad 2600 fails....I may try 2565(9x285) again and see if it crashes again or 9 x 284...but Super Pi was no faster since I had to lower memory settings with my crappy ram...2550 seems as fast (10x255) as 2565(9x285)

one very interesting thing I have found with the A64 o/c when things are pushed to far...like 2700 I actually felt things seemed slower