Failure on blend test after 1 minute!

[ET_]

Originally Posted by zombie1000

All my timings and speed etc for memory is auto.

Maybe im just unlucky

It may be a feature of the BIOS. Which one do you have?

Luigi

[Thabor]

I recently put together an new system, and I'm trying to track down an instability using MemTest86, and Prime95

MemTest86 seems to pass all tests except for #7, the random number test.

Thinking it might be the FPU I tried Prime95. The blend stress test consistently fails in few seconds, with :

FATAL ERROR: Rounding was 0.4990234375, expected less than 0.4

or

FATAL ERROR: Rounding was 0.5, expected less than 0.4

So I decided to try the in place test instead of the blend. Currently I've passed the 8K, 10K, 12K, and 14k tests, and am chugging away at the 16K test.

Anyone have some idea where my trouble is? Just by way of coincidence, I also have a MSI mother board..

(Results information below)

[Tue Dec 21 19:50:20 2004]
FATAL ERROR: Rounding was 0.4989803823, expected less than 0.4
Hardware failure detected, consult stress.txt file.
FATAL ERROR: Rounding was 0.5, expected less than 0.4
Hardware failure detected, consult stress.txt file.
FATAL ERROR: Rounding was 0.5, expected less than 0.4
Hardware failure detected, consult stress.txt file.
Compare your results to other computers at http://www.mersenne.org/bench.htm
That web page also contains instructions on how your results can be included.

Intel(R) Pentium(R) 4 CPU 3.40GHz
CPU speed: 3415.27 MHz
CPU features: RDTSC, CMOV, PREFETCH, MMX, SSE, SSE2
L1 cache size: unknown
L2 cache size: 1024 KB
L1 cache line size: unknown
L2 cache line size: 128 bytes
TLBS: 64
Prime95 version 23.8, RdtscTiming=1
Best time for 384K FFT length: 17.523 ms.
Best time for 448K FFT length: 21.159 ms.
Best time for 512K FFT length: 23.424 ms.
Best time for 640K FFT length: 29.486 ms.
Best time for 768K FFT length: 35.691 ms.
Best time for 896K FFT length: 42.353 ms.
Best time for 1024K FFT length: 47.223 ms.
Best time for 1280K FFT length: 62.436 ms.
Best time for 1536K FFT length: 75.353 ms.
Best time for 1792K FFT length: 90.292 ms.
Best time for 2048K FFT length: 101.579 ms.
FATAL ERROR: Rounding was 0.4990234375, expected less than 0.4
Hardware failure detected, consult stress.txt file.
[Tue Dec 21 22:32:37 2004]
Self-test 8K passed!
[Tue Dec 21 22:48:37 2004]
Self-test 10K passed!
[Tue Dec 21 23:04:49 2004]
Self-test 12K passed!
[Tue Dec 21 23:22:35 2004]
Self-test 14K passed!

[Thabor]

My apologies, it isn't passing the in place test, but the small FFTs test.

Also, for reference MemTest86 Test 7 is the Random Number Testing, which was what led me to believe the FPU might be at fault.

Hello all, just thought this might be useful and I might be able to get some help...

I have just built a new PC in a shuttle case and have had problems during long gaming sessions, nVidia recommended running this program, and guess what.... I get the same problems that every one else is having in less than a minute, I have tried underclocking (significanlty) but I am still consistently getting this problem, can anyone give any advice on what the problem could be?

Shuttle SN95G5
AMD 64 3500+ (NewCastle core)
1GB Corsair 2-2-5-2 (running in auto, but have dropped it to 3-3-6-3)
Maxtor 250GB SATA HDD
BFG 6800GT OC (running at stock speed)

[Prime95]

Originally Posted by Thabor

My apologies, it isn't passing the in place test, but the small FFTs test.

Also, for reference MemTest86 Test 7 is the Random Number Testing, which was what led me to believe the FPU might be at fault.

So prime95 works on small numbers that fit entirely in the L2 cache, but fails on numbers where main memory is accessed. Sounds like a memory problem to me. Try slower memory timings or slower FSB or a touch more voltage to memory or running with one stick instead of two or try a friends memory stick.

This should give you more info on whether its a motherboard, timing, or memory stick problem.

Isolating hardware troubles is not easy!

[Prime95]

Originally Posted by Newbie

I have tried underclocking (significanlty) but I am still consistently getting this problem, can anyone give any advice on what the problem could be?

Shuttle SN95G5
AMD 64 3500+ (NewCastle core)
1GB Corsair 2-2-5-2 (running in auto, but have dropped it to 3-3-6-3)
BFG 6800GT OC (running at stock speed)

I'd guess its a memory, motherboard, or CPU (memory controller) problem. Now wasn't that helpful

Is this two sticks of memory or one? If two, try running with just one. Lots of motherboards are finicky running two sticks, some are finicky about which slots are used.

Have you tried memtst86?

Hi I have run the same tests but that one in particular instead of the blend test and it passed....

Very weird, no I haven't tried memtest as I don't have a floppy drive and I can't see a way for it to work without one.

Thanks for your suggestions though.

[sdbardwick]

Memtest86 can be run from a bootable CD-ROM. The memtest86 site even has prepared ISO files ready for burning a bootable CD.

[clowns789]

If the XP disk doesn't work than perhaps I can e-mail you the files you can't get off the disk.

[Thabor]

I had both the memory sticks replaced, and now I'm reaching over 3 minutes on the blend test before failing.

As a side note, Memtest86 doesn't seem particularly reliable.
Only the random number test (7) in Memtest86 revealed any problems with the original memory.

It seems likely now that only that test was really bypassing the cache to test the system memory.

When I exchanged the memory the used a tester in the store which verified the problem.

I'm guess the new memory is better, but still has problems. I may ask them to exchange it for something other than Corsair memory next time.

[sdbardwick]

Originally Posted by Thabor

As a side note, Memtest86 doesn't seem particularly reliable.
Only the random number test (7) in Memtest86 revealed any problems with the original memory.

It seems likely now that only that test was really bypassing the cache to test the system memory.

So, Memtest86, by using a variety of tests, accurately identified a borderline memory module, and you maintain that discovery is evidence of a lack of reliability? I'd say it demonstrates the opposite. A less rigorous testing program would not have identifed the flaw. Intermittent failures are the hardest to detect, hence the standard recommendation that Memtest be allowed to run for several complete error-free iterations before classifying a given module as error-free.

The algorithms/methods used in Memtest do disable/bypass the cache. The documentation so states, and as a reality check, imagine how fast the tests would complete with cache enabled.