Sick AMD 3000+
29 Dec 2005 0:56:28 UTC
Topic 190465
(moderation:
I've got an ASUS A7N8X board holding an AMD Athlon XP 3000+. This is the older basic version of the A7N8X. The purpose for the machine is as an Einstein@Home cruncher. I've been quite pleased with the A7N8X (several versions) as crunching platforms.
Installation of the machine went without problems. It's a sick puppy when it comes to crunching for E@H. It's speed through a work unit is respectable 21,000 seconds. But a high percentage of the results are "client error" and "validate error".
To see the errors, you can look
I've searched and found that "client errors" are supposed to be the chip overheating. Well, I'm not trying to overclock the chip at all. I'm trying to just run it stock. The BIOS did not give me much control. Initially I tried "aggressive" timings, but now I've shifted to "optimal" and it's actually gotten worse. I've removed the fan to inspect and carefully reseated the fan to assure that no lapped edge was catching and holding the fan away. It seems to have good contact. I can look at mobo temperature parameters via ASUSProbe and via the BIOS. Both show VERY reasonable temperatures--CPU like 100 degrees F.
The computer is living in a basement that is at 55 degrees which should be ideal for crunchers. And to repeat, I'm not overclocking at all. I've got the FSB set for 166 which should yield the 333 that this particular chip needs.
So, I'm looking for advice on potential causes and solutions for the frequent client error and validate errors on E@H.
Are some chips just "flaky"?? This was an ebay purchase and not guaranteed. Can some chips work in a flaky way like this?
I have another 3000+ chip coming that I will pop and drop in as a check of the mobo. And if I have to, I can buy another 3000+ to replace this cpu.
Or do these symptoms reek of mobo issues? I've got three other A7N8X boxes crunching quite well with 3000+'s sitting right next to this box. (No, I"ve not swapped the CPU's around--probably will for troubleshooting)
I look forward to your thoughts.
Bruce Richardson
Team Richwood Timber LLC
When the going gets weird, the weird turn PRO. -- Hunter S. Thompson
Sick AMD 3000+
)
You have several Exit status -1073741819 (0xc0000005) errors. These are the standard Einstein 4.79 "graphics bug" errors. Read this message and thread. It could be that the other errors are related to this, too - but maybe not.
Hi there: I'll keep
)
Hi there:
I'll keep digging on the "graphics bug" thing. My memory is that I'm using a junker AGP 2.0 card in that one.
The box is actually heating my parents house :-) . The next time I visit, I will:
1. Uninstall Boinc 4.whatever and install the new one.
2. Update the graphics drivers.
Not quibbling, but I try to install BOINC as a service on all my machines so I've never seen the graphics.
If any of you have additional thoughts, keep 'em coming. Thanks.
Bruce Richardson
W9FZ
When the going gets weird, the weird turn PRO. -- Hunter S. Thompson
Don't use the Einstein@Home
)
Don't use the Einstein@Home graphics at all is likely to be the cure ... or to only run Albert work which is supposed to have fixed the known issues. The usual culprit is ATI cards.
As far as the heat, from what little I know of AMD chips that sounds high to me ... OF course, I don't know that much ...
If you wish to do further
)
If you wish to do further stability test here are two programs to try;
Prime95 Torture test
Memtest86,
BOINC Wiki
You might also try manualy
)
You might also try manualy tweaking the memory timings. Set the the various delay values extremely conservatively and see if it helps.
(I've just fought a small battle with an old NF1 mobo and some generic 'Best Buy' (yea, sure....) memory. The RAM is specified at CAS2 at 133/266 (2,2,2,6), and reports these values to the motherboard. Unfortunately, it won't run stably unless I bump the CAS value to 2.5.)
-Greg
RE: Hi there: I'll keep
)
It appears this host is now successfully processing Albert units. Since Albert has the same "graphics bug" fixes as the Beta application, I think you can safely assume that that the "bug" was your whole problem.
In case you were not "up to speed" on the "graphics bug" issue, there is a little more to it than ATI driver problems. That is, Einstein 4.79 was built to the BOINC library that was in effect when BOINC 4.19 was the standard. The BOINC library was changed by the time BOINC 4.45 was released. My guess is your problem was more related to the library issue than it was to your graphics card.
Ah yes..good old ATI..the
)
Ah yes..good old ATI..the 'we'll have a new driver to fix these problems SOON, Honest, we really mean it this time' people.
Their motto should be 'we're so fast and cutting edge, we don't even work well'
ROFL. sorry..it's SO hard not to 'moon' ATI. I'll be good now.
Roger on all. Yes, I had
)
Roger on all. Yes, I had noted that it seemed to be humming along on the new WU's. I'm still away and have not visited the machine and not had a chance to effect the changes I intended to effect (new drivers and new BOINC front-end).
I think you are right, even if I had done those changes, it may not have perked right up.
And as I'm recalling this machine (named Heinlein), though I normally use AGP cards, I think I used a PCI vid card--which most likely was an ATI.
THAT explains why this machine (which is SO similar to Asimov and other hosts I've set up) kept "coughing".
Well, I'm glad the Albert WU's came along. Helped resolve a problem that was stumping me till the other postings in this thread started steering me towards potential root issues.
Bruce Richardson
Team Richwood Timber LLC
When the going gets weird, the weird turn PRO. -- Hunter S. Thompson