• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

Update: Memtest runs fine, but Prime95 fails

BlueWeasel

BlueWeasel

Lifer
See my earlier thread

Basically, I got a new board and have had a time getting my PC2700 and PC3200 sticks to work together. Finally, the RAM appears to be running together @ 208 FSB (2.75v) and my 2600+ mobile is running at 12x208 (1.85v).

Memtest just ran for ~5 hours (10 passes) with no errors. However, earlier today, Prime95 was failing between 1 to 2 hours when running the Blend test (CPU and max RAM tested). I've not tried running the P95 test where minimal RAM is tested.

Since P95 is failing and Memtest passes, then that means the failure is due to the chip and not the RAM, right?

I've got the TT Silent Boost all copper HS with a Panaflow L1A fan and my temps are staying around 50-52C under full load with 1.85v. I may try to push the voltage just a little more to 1.875 or 1.90v to see if that helps, but I'm kinda getting a little hesitant because of the voltage.

Update - 1.04.05
Apparently, it's not the RAM that's causing the P95 errors. Last night, I started the Blend test with the chip running at 11x208. The test was still going strong this morning and I manually stopped it this morning after ~8 hours of testing.
 
If Memtest passes, it's not your ram.

But damn, 1.9v on your CPU?? What CPU do you use? I can't think of any CPU that should be going that high...
 
Originally posted by: Mrvile
But damn, 1.9v on your CPU?? What CPU do you use? I can't think of any CPU that should be going that high...
Click to expand...

I've got a XP 2600+ mobile. Actually, 1.85v doesn't really scare me at all. The general rule of thumb is no more than 0.2v over the stock voltage when overclocking. Sure, the mobiles have a default voltage of 1.45v, but they are basically the "best of the batch" from the same desktop Barton chips (which run at 1.65v).

So, I'm not worried about 1.85v-1.875v, especially since this board (8RDA3I) appears to undervolt ever so slightly.

Also, just check out the CPU Overclock database....there's plenty of the mobile chips running at 1.85v with air cooling similar to mine.

 
is the default multi on your proc 12x? i would run everything at stock settings and see if prime95 fails then. its not your ram, so ovbiously your proc cant handle 12x208 or its defective.
 
Originally posted by: Mrvile
If Memtest passes, it's not your ram.

But damn, 1.9v on your CPU?? What CPU do you use? I can't think of any CPU that should be going that high...
Click to expand...

Not necessarily. Last time I went to tweak my memory timings, sometimes it would pass memtest, then fail Prime95. In other words, it could be your ram. It may also be your board, I know the board I'm running will not run over 200 fsb for anything.
 
athlon xps can suck up voltage like bh-5, just keep em cool. running 2v through mobiles isnt bad (i hope it isnt, i do it sometimes)
 
i would keep a close eye on MAX CORE TEMPERATURE, because that is what will cause errors if the cooling isnt up to par
 
Originally posted by: AkumaX
try doing 11 x 208 w/ prime95, if that passes, well...
Click to expand...

Good point.

Actually, I'm beginning to think that it's because of the RAM. I ran the small FFT test (no RAM tested) today and it ran fine at 11x208 for about 5 hours until I stopped it manually. I then tried running the Blend test (all the RAM tested) and it failed in about an hour (the RAM was at 2.9v).
 
memtest is not really complete enough to detect all RAM failure. The FreeBSD crowd still detects errors in `make world` that go away with new RAM modules but no error has been detected by memtest86.

Like all benchmarks (which this really is) there are very complex patterns and timing that are very hard to artificially reproduce.

If you have two or more RAM modules, try pulling all but one and test them in isolation.
 
Originally posted by: MartinCracauer
memtest is not really complete enough to detect all RAM failure. The FreeBSD crowd still detects errors in `make world` that go away with new RAM modules but no error has been detected by memtest86.

Like all benchmarks (which this really is) there are very complex patterns and timing that are very hard to artificially reproduce.

If you have two or more RAM modules, try pulling all but one and test them in isolation.
Click to expand...

Right now, all 3 RAM slots are full (1 stick of PC2700 512mb and 2 sticks of PC3200 256x2). I just know it's got to be the PC2700 running at 208 FSB with CAS 2.5. I know Muskin RAM is great, but I have a hard time believing this RAM could be stable with such a high overclock. I've tried running the RAM at CAS 3.0, but it apparently likes only 2.5 or lower and won't boot at CAS 3.0.

I'm gonna try to run Prime's Blend test again tonight using 11x208 and see how it goes. I just can't think it's the CPU, as this 2600+ has run 12.5x190 stable at 1.675-1.70v for months, but now can't run 12x208 FSB at 1.85v.

We shall see, though.
 
Originally posted by: BlueWeasel
Originally posted by: AkumaX
try doing 11 x 208 w/ prime95, if that passes, well...
Click to expand...

Good point.

Actually, I'm beginning to think that it's because of the RAM. I ran the small FFT test (no RAM tested) today and it ran fine at 11x208 for about 5 hours until I stopped it manually. I then tried running the Blend test (all the RAM tested) and it failed in about an hour (the RAM was at 2.9v).
Click to expand...

are you running bt or anythhing? that sucks a lot of mem
 
Originally posted by: AkumaX

are you running bt or anythhing? that sucks a lot of mem
Click to expand...

Not sure exactly what you are asking.

When I run the P95 blend tests, I don't use the computer and basically give the P95 executable the highest priority possible. I just leave it running and come back later (with crossed fingers) and hope it hasn't found errors.
 
Apparently, it's the CPU and not the RAM. I ran P95 for ~8 hours last night (stopped it manually) at 11x208, so I conclude it's the chip and not the RAM that doesn't like 12x208.
 
You must log in or register to reply here.

TRENDING THREADS

Back
Top