Prime fails almost instantly, what to try? (Update: Success)

[SithSolo1]

Just OCed my 2500+ to 3200+ and I tried to run Prime95. This is what I get: Primefail.gif

Antec True380 psu
Shuttle AN35N-Ultra
2500+ week 43, locked, FSB at 200mhz stock voltage
2x256MB Mushkin Black Level 2 PC3500 running 200mhz @ 2-2-2-11 2.6v
Current idle temp of the CPU is 31C

Anything else you need to know?

Thanks,
J.

Update:
Swapped the Shuttle AN35N-Ultra for an Abit NF7-S v2.0
Set the proc as a 3200+ in the BIOS and booted windows, tried Prime and it failed after less than a min.
Went back into the BIOS and up'd the vdimm to 2.8
Now running Prime stable @ 2.2Ghz 1.61-1.63v(according to CPU-Z) and dual-channel memory 200mhz 2-2-2-11 2.8v.

For those who want the proc info here it is:

AXDA2500DKV4D
Z361557241169
AQYFA0343XPMW

 

[Dman877]

More voltage is the first step...

 

[fleflikr]

up the cpu voltage to 1.7 for starters

 

[SithSolo1]

Up'd the CPU to 1.7 and the memory to 2.7(max voltage)


The CPU intenstive ran fine for 15mins so I stopped it and tried the Mixed bench(ram intensive, same one I originally tried), ran for 5 mins and then boom, error.

Yes I know I should run it longer. What is the max CPU voltage I should try, I'll work my way up to that point.

 

[BlueWeasel]

What heatsink and fan are you using for the CPU? The most I'd push the voltage would be 1.85v, and I'd really only feel comfortable at that voltage with a high performance HS and fan combination.

However, you could run it at 1.85v with the retail AMD HS and fan for testing purposes. But if it takes 1.8 or 1.85v to run stable at 3200+, I'd really consider getting a better heatsink.

 

[SithSolo1]

SLK900a w/ 64CFM Enermax fan.

Uped it to 1.775 volts. Blended test still failed after 5 mins(also failed after this same amount of time on 1.7 and 1.725), Small FFTs test ran fine for 20mins before I stopped it. Max CPU temp under load was 48C.

I think the memory needs more Volts but the mobo just won't go any higher I knew I should have spent the extra money and gotten the ABIT.

 

[boshuter]

Try loosening the ram timings and run it again....... maybe it's not the psu that's failing.

 

[Duvie]

definitely try 2,3,3,11...I could not get stable any configuration of cas 2,2,2, on my wifes board....

I think the vdimm should be fine since it is running stock 1:1, right???

 

[Duvie]

Also I have never had one fail prime so fast...that is pretty flaky....It may take a bit....usually when I run a system that can't pass prime for 5 minutes it takes up to .05v to get it stable for 12 hours...That flaky may take .1v to get it stable...

Also my board undervolts severely so I imagine your does as well.....Check because default or even the 1.7v you are setting in the bios may only be 1.65v.....Verify this with a program like speedfan...

 

[SithSolo1]

Originally posted by: boshuter
Try loosening the ram timings and run it again....... maybe it's not the psu that's failing.

A) I never said the PSU was failing
B) If the ram is rated for 2-2-2-6 @ 216mhz why wouldn't it run 2-2-2-11 @ 200mhz?


I checked the volts with CPu-Z, said it was fluxing from 1.692 to 1.712 when I had it set to 1.7. Dunno how accurate CPU-Z is though.

 

[Duvie]

CPU_z is accurate so your board is pretty tight on actual and bios setting....

Like I said try the cas 2,3,3,11....It is funny but sometimes it is the chipsets fault...i had ram rated cas 2,2,2,6 as well and I needed to run 2,3,3,8 and that was just at spec speed.....Also running dual channel is historic for requiring looser timings and more vdimm....i still think 2.7v should be enought though...

Use a program like memtest and loop test 5,6,8 for an hour and if you pass it is highly unlikely it is the reason your bailing on the prime test so fast...

 

[SithSolo1]

Damnit, I knew it was the ram.

Set the ram 2-3-3-11 @ 200mhz vdimm 2.7

Ran memtest86+ V1.0 for two full passes. Got 64 errors all in test 5, 28 first pass and another 36 the second. I can't RMA it either as I am not the orginal owner and even if I was Mushkin is out of the Black Level 2 PC3500. There goes the video card budget.

 

[Duvie]

OK...now lets settle down a bit....We want to try to isolate it andmake sure it is not the chipset fsb or the volatage the factor here....

Drop the fsb back down to 166fsb and use the 4:5 ratio to get 400ddr again and use same cas 2,3,3,11 timings....now run memtest again and see how it does...

after that pull 1 stick each and run it.....

 

[SithSolo1]

Well before seeing your reply I tried each stick by it's self at 2-3-3-11 @200mhz three times on test 5, no errors.

Dropped back down to 2-3-3-11 @ 166mhz and did both sticks by their selves, both sticks in single channel, both sticks in dual channel, three loops on test 5 each time, no errors. I'll try what you said tomorrow as I need some sleep and have to get up in 3 hrs. I don't have a 5:4 ratio but won't 125% of the FSB be the same thing?

 

[Duvie]

You are right then it may be a voltage of the vdimm....the single method suggest that....

Also does your board have a setting for the vcore of the nb chipset??? I noticed I have the version like yours but the MN31N with integrated video for my wife....If you have it place this to 1.6v or 1.65v....


Actually you need a 5:6 ratio or 120% of fsb.....166/5 = 33mhz x 6 = 200!!!

 

[BlackMountainCow]

I'm experiencing about the same prob with my Shuttle AN35N-U. At first I had 2 Kingston HyperX 3200 256MB sticks, but they even refused to boot and post unless I set them to CAS 3 though they were rated CAS 2. So I got two TwinMOS / Winbond CH-5 sticks. They are much more stable but still I can't run Prime95 without fail. The blend test usually fails after 5 hours. Memtest86+ 1.0 gives me dozens of error when I pass the 70th pass of it (like 10 hours testing). I also have a 2500+ to a 3200+. Right now I'm trying to play with all the timings. Testing 2.5-4-4-8 right now.

I really got to say that this has been my last Shuttle Board ever!! The Kingsont Hyper X ran fine in my Abit and so do the TwinMOS. Both were 24 hours Prime95 and memtest stable with the same PSU, CPU and gfx.


BTW: SithSolo1: I read several comments in forums that Mushkin indeed has probs with the shuttle board. As well as Corsair. But it's definitly the mobo's fault. U might wanna check out this link. These guys know their stuff, especially Szandor.

 

[myocardia]

Sith, just keep that awesome ram, and buy yourself an Abit NF7-S, the best overclocking board for Athlon XP's being made at the moment. Newegg is selling them for $100 shipped. This sentence is a link to that board, on newegg.

 

[SithSolo1]

I had already decided if push came to shove, less expensive items go before expensive ones do. I've got enough to pick up the NF7-S so I'm probably gonna do that, but I want to tinker around a little more first.

 

[boshuter]

A) I never said the PSU was failing

Sorry about the typo... thought maybe you could figure out I meant CPU.

B) If the ram is rated for 2-2-2-6 @ 216mhz why wouldn't it run 2-2-2-11 @ 200mhz?

If you have to ask that question.....

 

[myocardia]

Originally posted by: SithSolo1
I had already decided if push came to shove, less expensive items go before expensive ones do. I've got enough to pick up the NF7-S so I'm probably gonna do that, but I want to tinker around a little more first.

Well, if you know the ram will run tight timings in single-channel mode, why not just run both sticks in single? Just put one stick in slot #1, and the other in slot #2, instead of slot 3. Now, you're running 512MB in single-channel...

 

[SithSolo1]

Bump for update