Folding Forum

toTOW wrote:You can find it in this thread : viewtopic.php?f=14&t=430

I looked at the page and found the tool after you posted it. If the machine kicks its current WU (2484 [R121, C5, G4]) then I'll run it, but it's been working fine on its current WU for about a day and a half. I think if there is a hardware problem with my machine, it is something that is extremely subtle as the machine currently has an uptime of 23 days, has not had any segfaults or anything with other programs, and it completed the two Core 82 projects just fine. I'd probably have to run StressCPU for a week to pull up any errors if there is a hardware problem.

I also just got a new laptop that runs 32-bit Ubuntu and am running FAH on it to see if I can get a 2484 WU and see how that goes. It is currently processing a pair of SimT Gromacs WUs (4100 and 4108) so it will be at least a half a day before either of the cores finishes. The sad part is that the five-year-old 1.6 Duron would get a higher PPD on the 2484 (~135 ppd) than either of the cores on a 2.00 GHz Core 2 Duo T7250 gets on the SimT WUs (~130 ppd).

All right, my machine has been running StressCPU v2 with one thread for the last abut three and a half days (5050 minutes of CPU time according to top) and has computed about 7.2E14 FP ops without a single error. The WUs that EUEd almost always failed earlier than that amount of time, so I think that my machine is probably not causing the problems. I'll keep StressCPU v2 running just to see, but I have a hunch it will just keep on going without any errors until I stop it.

Any ideas?

EDIT: The laptop hasn't picked up any 2484 WUs yet. Each of the two cores has gotten and successfully completed four 41xx WUs, no 2484s yet.

Mizzou_Engineer wrote:The WUs that EUEd almost always failed earlier than that amount of time, so I think that my machine is probably not causing the problems.

If other folders are completing the WU's then the problem unfortunatley is with your machine. If you can only complete Amber WU's then it could be a CPU/SSE problem as it's the only core that does n't use it. Check out the RAM and keep an eye on temps and voltages is my only other suggestion.

Good luck!

John_Weatherman wrote: If other folders are completing the WU's then the problem unfortunatley is with your machine. If you can only complete Amber WU's then it could be a CPU/SSE problem as it's the only core that does n't use it. Check out the RAM and keep an eye on temps and voltages is my only other suggestion.

Good luck!

Maybe there is an issue. Looking at it on its face, the CPU is a Duron Applebred, which is supposed to support SSE and is basically an Athlon XP Thoroughbred-B with 3/4 of the cache disabled. It isn't AMD's first chip that had SSE (that would be the 180 nm Palomino Athlon XPs) nor is it even the first spin of 130 nm AMD silicon (that would be the Thoroughbred-A.) I haven't checked the errata sheet, but it doesn't look likely there is some problem with SSE. If there was, I would have expected to see the problem pop up earlier with "illegal instruction" errors showing up in MythTV as that does use SSE in the mpeg2 decoding. This machine is my MythTV machine and sits happily in a mid-tower ATX case next to the TV cabinet, which is why I thought it would be a good candidate for folding.

But perhaps there is some error in the interaction between the core binary and the CPU. I wonder if anybody else here is folding with an Athlon XP T-bred or a Duron Applebred, or even any Athlon XP and is getting these cores. I know the chips are getting a bit up there in age, but if somebody else has one, I'd like to see if they can complete the WUs. If they are experiencing the same issues, then it would suggest a problem with the core and the specific line of CPUs as my CPU apparently seems to be passing every *other* stability test I can throw at it. I'm just curious to see what's going on here.

I'm using a 2.138 Ghz T-bred Athlon XP 2600 and I've had the same cores (including the Double Gromacs that use SSE2, which the CPU does n't have) and did n't get any EUE's. One thing I had to do, because I was getting freezes and errors in Windows, was reduce the clients CPU usage from 100% to 98%. That stopped the problems.
Have you tried that? You can stop the client and run configonly if you want to give it a go.

I am using Linux and there is no percentage throttling in the client as there is in the Windows client.

I finally completed a 2484 WU. 2484 (R17, C35, G6) completed successfully after I disabled assembly optimizations in the client configuration, which disabled SSE. I think that pretty well defines the problem as some error in how the cores use SSE on Applebred-based Durons or that my CPU has some very occasional error in handling SSE instructions but not x87 FPU instructions.

Sounds like you'd be best changing the CPU - must be some floating around on ebay - good luck!

Before you decide to replace the CPU, get some canned air. Blow the dust out of the heatsink and make sure all the fans are turning. If there's a filter on any of the fans, clean it, too.

bruce wrote:Before you decide to replace the CPU, get some canned air. Blow the dust out of the heatsink and make sure all the fans are turning. If there's a filter on any of the fans, clean it, too.

I opened the machine up about a week ago to remove one of the RAM modules and put it in my desktop (I didn't need 2 GB in that machine, but my desktop needed more RAM.) I did blow it out with some air, but it was reasonably clean to begin with. I doubt it's overheating as the heatsink doesn't feel all that warm to the touch and although I can't get any values from the CPU's thermal diode in the OS, the board's under-socket diode says 39-41 C at full load. I have this heatsink on the CPU as the board and CPU came to me without a heatsink. It's supposed to be good for up to an XP 3200+, which is almost an 80-watt chip while my Duron is about a 50-watt unit.