computer is insane

blinky

Never heard any problems with Prime95. It runs fine on my AMD64 3000+ for 12h+ without any errors (I shut my PC down overnight). It does give rounding errors if I push the overclock too far. Many people on the DFI forums have run it for several weeks without any errors.

I'll assume you are running it at default and are not overclocking at this stage (I don't even know if it has working PCI locks on that board. Not all the VIA chipsets do).

Have you tried running Memtest and seeing if it finds any errors?

The AMD64 while using less power than the Prescott P4's do in general need good quality power supplies. Does the K8V have a 20pin or 24pin ATX connector?

DFI and OCZ both state that if a board has a 24 pin connector that only a genuine ATX2.0 24 pin PSU should be used (ATX2.0 = 2 x 12 volt rails and 24 pin connector). A 20 to 24 pin connector should not be used. DFI also recommend a minimum of 480W. Make sure you have all the power connectors plugged into the board. Also cheap PSU's can cause problems.

My hunch is either faulty RAM or unsuitable PSU. Memtest should tell you if the RAM is okay or not. For a quick test, you could run the memory at a slower speed. It impacts very little on the AMD64's performance.

There is a dedicated thread over at anandtech for K8V owners. Probably worth a look.

T4i

Defo looking like a mem problem then, half hour isnt that long so if its giving you rounding errors thats looking bad.

What is your RAM make/model? If you know.....

Running memtest sound like a good idea right now, least that will give you a definate answer to any RAM problems. Memtest runs without windows so it completely rules out software issues.

You need to run it for a good 12+ hours......24hrs is a good indication for good/bad memory.

I have 2 sticks of 512MB Kingston 333mhz in dual channel. Its supposed to run at 2-2-2-5 but I get errors in Prime95 at those settings, I've found running the RAM at 2-3-2-5 or 2-3-2-*11 (*Nforce2 best bandwidth settings) it runs flawlessly for days....

Have you got anything that will give you PSU realtime voltages? Asus probe may do it, the nforce2 version of asus probe tells me exact voltages on all the different rails.

If you do have something that will monitor them start it and write down the voltages at idle then start a torture test and write down the voltages under 100% load. If your rails have a 10% or more tolerance this is bad.

What make is your PSU? If theres one thing that ppl skimp on its always the PSU. Even though the PSU is the supplier to all your p.c equipment.

Ferris

hi, thanks for your suggestions.

how would i go about testing graphics card memory?

as for RAM, before i sent my PC back to the OEM under the hardware warranty, i conducted tests (8+ hours) with Memtest86+ and the M$ Windows Memory Diagnostic, which both came back with about 5 or 6 errors (for each program, the failing addresses each had the last four characters in common, e.g. all in the form xxxxabcd or something). combined with the Prime95 failure, i assumed this would be enough to show a hardware failure. i sent them screenshots (digicam) of the memory diagnostics as proof.

however, they did not get any BSOD running it in their workshop (which is typical of an intemittent problem!), and they insisted the memory is fine. i get the feeling i have been duped. i may have to return and threaten trading standards action... the problem is, i have no way of knowing whether its a RAM, motherboard, or CPU problem (i think the athlon64 has a memory controller on-board the CPU?)... grrr...

PSU: the k8v has a 20 pin connector, according to the manual. the PSU is branded "HIPER", whoever they are...

voltages: speedfan tells me voltages, they all seem ok; the 12v is a little worrying at 11.13-11.31 (although it comes up as 11.84-11.94 in the BIOS)...

i have just thought of something... when i first did the reinstall, it crashed while i was trying to put sp2 on, and i had to do a system restore, and then reinstall sp2, drivers etc. i think i might do a complete fresh reinstall, and if that doesn't clear it up, i'll storm back to the system builders and complain...

this is really turning into a new epic... thanks for your help

Ferris

oh, one thing i haven't mentioned... the system is now a year old, and for the first 7 months (may-december 2004) it ran as completely solid as a rock, probably the most solid system i have ever used anywhere. i haven't changed any of the hardware, which is why i figured a reinstall ought to have fixed it. it is not overclocked, everything (voltages, memory timings etc) is set to BIOS automatic-detection.

so i just wonder how it is that it suddenly went from rock-solid to constant crashing...

blinky

Hmm, points to PSU. It might have been straining when it was stable and now it can't deliver the power requirements.

Increasing the VDimm, might help. Many modules need 2.6-2.8V

T4i

If memtest gives errors you defo have something bad.

If you have a crash during SP2 install then thats very likely to screw things up.

Two big problems with XP......SP2 bad install and java.

Both can totaly screw XP up. If your comp crashed whilst installing SP2 that points back to RAM problems.

Hipers are not the best PSU. Get a Tagan.

blinky

The Hiper Type-R is okay.

Ferris

curiouser and curiouser...

last night i decided to run a diagnostic using the tool that the shop i am in dispute with recommended to me. it is called Microscope 8 (apparently its a professional standard one, although mine is only a trial version). it found no errors with the memory, but when i tested the video ram, it quickly came back with the following error:

VIDEO BANK FAILURE:
BANK 0/64K, A000:FFFE
EXP=0, ACT=780, XOR=780

so that would suggest there is something wrong with the memory on the graphics card. but surely that wouldn't cause something like Prime95 to crash?

looking back, i think the sp2 crash might have been caused by a driver conflict; i say this because when i did it in the other order (sp2 THEN drivers) it seemed to work much better (the first time round it wouldn't even shut down without blue-screening...

bbb_uk

Ferris wrote:

...so that would suggest there is something wrong with the memory on the graphics card...

Then in that case it is a hardware problem and tell the warranty people that you ran the test using the software they recommended and it came back as memory failure of the graphics card.

Remember though that you would be best to run this test again and see if it fails on the graphics memory. A intermittent memory test problem (especially if its not in the same location) could mean its the motherboard at fault (memory controller, etc).

I believe most warranty companies (computer related or not) would rather swear blind there is isn't a problem with the product under warranty and it is something else!

T4i

Ferris wrote:

voltages: speedfan tells me voltages, they all seem ok; the 12v is a little worrying at 11.13-11.31 (although it comes up as 11.84-11.94 in the BIOS)...

11.13v on the 12v rail is worryingly low. The BIOS will report better results as the mobo is very much idle during BIOS. Its only after the POST when the power will drain.

I agree with bbb_uk, they have supplied you with there testing standard of software and its reported errors, there fault for giving you it. Run it 3 times and take screenies of all reuslts, save the pics for trading std's if needed.