AMD's, Folding and SSE

I guess I'm just a litle perplexed.

I know the story, I know the problems.

BUT, I have around 12 AMD machines folding for me. 1 Is a [email protected] T-bird - it has never had a problem, folds with 3dnow and never locks or drops a WU. Unfortunately, folds so dag slow...

Then I have my main(home) rig. This is the rig in my signature. The severely OC'ed one. It has locked all of 2 times in 9 months(unless I was playing with the OC). It has never dropped a WU. My log is so long it's scary - Yes, I have not even rebooted since the last Windows update that required it. I never have had to reboot it(unless required to by update) and it has never dropped a WU. And this is my gaming machine!

And finally there are the rest. Mostly bartons with a few T-breds. While there is no real pattern, these lock or drop WUs 2-7 times a month. 3 of these are OCed. 2-Tbreds and 1 Barton 2.8. The barton only barely. I will have to say that the OCed Barton drops WU's mostly. The others mostly lock up.
None of these Rigs have temp problems. I admit my main rig runs between 33C(Currently) and 40C(summer, forgot to turn on AC that day).

I could of been called a AMD boy 6 months ago. Now I will admit to leaning towards Intel again. Why? Well AMD's even after Intel $$ drops are still more expensive. But, they don't become unstable when using SSE. Personally, I think AMD has a flaw in their SSE Instruction set but maybe no one agrees with me. I've always thought, if your temps are right and your powers is good, you should be stable.

Why do I feel that everyone else is finding it 'Ok' that AMD may have a stability issue with SSE instruction set? It's hard to find another application to *really* test this but...I do not believe it is a problem with using folding and the SSE flag - otherwise why wouldn't Intel's have the same problem? We have all been hoping for a new client that fixes this problem....Do you really think it's the client? I don't.

Ok now I'm rambling...need to go to bed. But, think about it....

I don't remember where I saw it but I seem to remember seeing somewhere on these forums that the flaw is in the programming of the client. I'll do some looking and see if I can find where I saw that.

If they have a flaw, it might be in the later Tbreds and Bartons only. I have two Morgan durons that haven't had any problems with folding and SSE forced.

However, my Barton only threw folding errors with SSE when it got too hot for a week. Once the temp came down, it's been cruising. (see sig. current temps are 20case 42cpu)

With my new Tbred B though, I did notice that SSE makes the CPU hotter than 3DNow!. (sorry if this is old news) I had the 4.0 beta client running with -forceasm before I knew about the -forcesse flag. I upgraded to the final v4 and read the FAQ (go figure) and set the flag to -forcesse, and now the CPU is always at least 1C higher.

If there is an SSE issue, I reckon it rears its head when those CPUs are past a certain temp threshold, whether it be from OCing or not.

I guess we'll know more once the v4 client settles into the install base.

ive got 2 bartons and 2 new locked tbred b's folding here...the bartons have been going without a reboot for about 3 months now with no errors, and the tbreds are still pretty new but one is going 2 months straight and the other is on 3 weeks straight....no errors of an kind.

All are overclocked and temps are decent.....btw do those computers crash in prime?

Nope, these computers are all stable cept with Folding and SSE

AMD T-Bred B's and Bartons have issues, no 2 ways about it. I have 65 bone stock 2400's on good mobos, good PSU's good rails etc, etc. and EVERY single one WILL crash with SSE, just the way it is. I have 4 2400 boxes running in my garage right now and the CPU temp does not reach above 35C and still it will kick the bucket (one at random) whenever it feels like it. No rhyme or reason but a fact we must deal with.

well folding isnt the only thing that uses SSE. You can run the SSE stress test found on the electron microscope download page. Also, I have all AMD's in this house and every one of them can not only pass the SSE stress test, they can also render video without a problem (as many rendering proggies use SSE as well)
The only thing that doesnt work well on my rigs are certain projects of folding. Also be advised that SSE doesnt just depend on the proc as the mobo itself can have something to do with it as well. Ive been spending a quite of bit of time (as many of you other AMD'ers are too) saving log files of failed wu's and noting what happens with different cores and clients. After digesting what ive seen on the forums and what i have in my log files, I am fairly certain the issues lies more within the folding preoject than it does with mobo/proc issues. BTW All of my AMD rigs have gone weeks without a hiccup one has gone 10 months (a dually with tbred b's) without a failure.