New 6.22 Core crashing...beware.

[relic]

All 3 systems, 2 are Vista x86 and 1 is Vista X64, (all MPICH not Deino) are having problems.

All were rock stable on dual 5.91 SMPs and then stable with single 5.91 SMP + GPU2, but 6.22 is crap. None will run 1 x SMP 6.22.

All are very mild OCs (3GHz)

I also tried my t8100 LT with the same poor results. (no OC)
Switches are -smp 4 (-2 for the t8100) -local -verbosity 9

Avoid the new client until they get it fixed.
100% failure in my testing, no systems are stable with the new client.

 

[FuSoYa]

Seems to be hit and miss sometimes with new clients. I have had the opposite results with 6.22b2. I have it running 1xSMP on a Core2, 2xSMP on 2 systems and SMP/GPU on 3 systems, so far all stable for about 24 hours (early to say still I guess).

One of those SMP/GPU systems I could never previously get stable with 5.91 and was running GPU only, but it is so far running smooth with both clients up. I am using the Deino MPI client, so maybe that's the difference? All are Windows XP x86.

 

[relic]

Seems to be hit and miss sometimes with new clients. I have had the opposite results with 6.22b2. I have it running 1xSMP on a Core2, 2xSMP on 2 systems and SMP/GPU on 3 systems, so far all stable for about 24 hours (early to say still I guess).

One of those SMP/GPU systems I could never previously get stable with 5.91 and was running GPU only, but it is so far running smooth with both clients up. I am using the Deino MPI client, so maybe that's the difference? All are Windows XP x86.

Deino is a lower PPD but supposedly more stable, however it is x86 only no x64.

 

[Xilikon]

1- the # of threads after the -smp flag do nothing for now. It require the A2 core, which isn't available for Windows for now. You should use the -smp flag without the # to avoid issues.

2- Clients != cores so while it's folding, the client version have zero effect on stability. the core is the same as before so you have a real problem. Make sure your OC is truly stable.

3- Can you paste the logs around the crashes to see what would be the issue. Also, define what you did to upgrade the new client ?

 

[relic]

Xil,

The systems are rock stable and had no problems with 2 x 5.91 clients.
This is a problem with 6.22 (Even 1 client crashes 100% on all systems)


Using SMP without the # cores was my original attempt which was completely unstable 100% crashes. Using -SMP -4 or SMP -2 had no effect better or worse.

I did not upgrade. I uninstalled the old client, rebooted, deleted the old directory and installed the new 6.22 client.

No interested in fanboi BS, 6.22 is crap alpha level code. I'm back running 5.91 and all is well.

 

[Xilikon]

You will have no choice to upgrade to 6.22 unfortunately since the current client expire saturday.

Can you show the logs and I'll see ? If the problem is genuine, i'll kick the right asses on FCF and get a new version.

 

[relic]

You will have no choice to upgrade to 6.22 unfortunately since the current client expire saturday.

Can you show the logs and I'll see ? If the problem is genuine, i'll kick the right asses on FCF and get a new version.

I have plenty of choices...f@H is not the only DC project available.
We've been here before and 1/2 of [H] moved to a different project because of it.

 

[nomad8u]

I have plenty of choices...f@H is not the only DC project available.
We've been here before and 1/2 of [H] moved to a different project because of it.

It's too bad really that they seem to have a knack for pissing their donators off. This seems like rather last minute to be releasing a new client. Seems to be the norm for these guys though.

 

[MasterOfPupets]

I'm having no problems with stability, but I've got 2 WUs now completed that won't upload. The only thing remotely resembling an error code is "Got status 503" but I don't have time right now to trouble shoot it, I will tomorrow.

 

[APOLLO]

I'm having no problems with stability, but I've got 2 WUs now completed that won't upload. The only thing remotely resembling an error code is "Got status 503" but I don't have time right now to trouble shoot it, I will tomorrow.

Looks like the SMP collection server is down now.

 

[Xilikon]

Looks like the SMP collection server is down now.

Yes, confirmed... I had a WU which cannot be uploaded

 

[flogge]

I'm at 33% complete on my SMP WU with 6.22 on my laptop with a T5550. Been running straight through on vista 64 for 24 hours with no problem. GPU2 is also running on the 8800M GTS.

 

[BillR]

I have plenty of choices...f@H is not the only DC project available.
We've been here before and 1/2 of [H] moved to a different project because of it.

It's too bad really that they seem to have a knack for pissing their donators off. This seems like rather last minute to be releasing a new client. Seems to be the norm for these guys though.

I spent most of today playing with the new client. The best I can say about it is; total crap.

ATM I have been watching project 3065 do nothing for 2 hours. I have tried every form of setup I can think of even going as far as reading the included useless readme.doc that comes with the program.

So, the conclusion I have come to, at least for the moment is there is only the expiring Beta client for SMP and a shitload of worthless non-science GPU work units being done as a “feel good” measure.

I have done my best to remain positive about FAH and the ongoing changes but there comes a point when you learn, albeit slowly, that shit and Shinola are indeed two different things.

Putting up with incompetent management has been the bane of FAH for years now and personally I now declare the project as user abuse. My feelings are we have surpassed incompetence right into out of control.

Yes, I could switch everything over to standard clients, no argument. The question is, with all the money I’ve spent, why and why should I have to?

 

[relic]

Easy there Bill, we have until Saturday.
After that there's always a lot of other projects that would appreciate a [H]ard DC effort.
We've given this arrogant dipshit Vijay a thousand (ok, perhaps a hundred) chances before....might as well give him one more. Regardless of how phenomenally stupid he is.

We're here for the [H] team, let's wait and see what's best for [H]. Saurday is still a few days away.

 

[APOLLO]

So, the conclusion I have come to, at least for the moment is there is only the expiring Beta client for SMP and a shitload of worthless non-science GPU work units being done as a “feel good” measure.

The GPU WUs we're working on now have no scientific value whatsoever?

Yes, I could switch everything over to standard clients, no argument. The question is, with all the money I’ve spent, why and why should I have to?

Agreed. Stanford has to change the way they operate, and stop making bad calls like introduce an unproven client a week before its predecessor's expiration date. At least there's always the option of not shutting down the clients, but that's a short term solution at best. I don't know how you managed to run a client 3 months after its expiration date...

 

[BillR]

Easy there Bill, we have until Saturday.
After that there's always a lot of other projects that would appreciate a [H]ard DC effort.
We've given this arrogant dipshit Vijay a thousand (ok, perhaps a hundred) chances before....might as well give him one more. Regardless of how phenomenally stupid he is.

We're here for the [H] team, let's wait and see what's best for [H]. Saurday is still a few days away.

No argument on the Saturday thing, I totally agree.

I still have a major issues with points being accumulated in vast quantity by all FAH users that have no basis in science. That alone takes Science into a whole new area of hyprocacy does it not?

There is no pride taken in achieving my current 6 million plus point count when the additional points have zero meaning aside from the fact that I was willing to invest in some new video cards.

 

[oraldlight]

The GPU WUs we're working on now have no scientific value whatsoever?

Huh? I'm ignorant as to exact value of specific projects, but I'm not aware of non-science folding. I thought all this folding is for science?

 

[BillR]

Ok, I stand partially corrected; the new client can be made to run. Let me figure out how and I’ll share. So far there are no useable explanations within the program and most assuredly less information at the “offical” site. The explanations from the experts range from funny as hell to huh?.

In Vist64 before you install turn off user control, you will be forced into a reboot. Back into windows do the install as well as –smp –configonly.

Set user control back on and again, this will force a reboot. Once back into windows make your shortcut and add the –smp flag and –forceasm switch.

Now, right click on your shortcut and choose run as admin. You may get an error or several, shut it down and try again. Eventually it SHOULD download the good old A-1 core and start folding.

But wait! There’s more! Once you start folding, with no errors, do a control C and shut the program down. Then, simply click your shortcut in the normal fashion. It might just work. The put me in the mpixe mode and I’m hitting about 12 min per frame on a 3065 which is almost respectable.

Luck to all, I’ll verify this with another install tomorrow. One a day is my limit at my tender age

 

[BillR]

Huh? I'm ignorant as to exact value of specific projects, but I'm not aware of non-science folding. I thought all this folding is for science?

Humm, Rash use of the word ignorant, you are simply uninformed.

I may be wrong, I’ve been wrong before hell I even got married and everything. As of a week ago a beta tester mentioned that we were still doing trial work and to my limited knowledge that remains the case.

If not I really feel no need to apologize, fucking Stanford should keep us informed. Unfortunately, this is a road we have been down way too many times in the past.

 

[Xilikon]

Doing non-scientific work is just a way to look at things. While it's true that some workunits we are crunching witht the GPU2 is already done before and the recrunching is to validate the results. So far, this helped them uncover a few errors and so far, it's promising. They are however already crunching some real science, especially with lambda WU.

 

[flogge]

I know the GPU1 results were found to be un-useable. They learned alot about the computational capabilities of GPUs even if the folding results were not used. GPU2 is also a trial run in that they can't be sure they can trust the results. It's a Beta test, and they are rewarding people greatly to participate. If the real science after the beta was not going to work, I don't think they would have gone into the GPU2 phase of the project.

 

[Sunin]

Yep it is hit or miss... make sure you delete your old directory and do a complete reinstall. I've read at FCF that numerous people copied it over the same dir 5.91 was installed = issues...

 

[relic]

Yeah, I did an uninstall, reboot, delete directory, reboot, fresh install of 6.22....same problems.

I did some additional testing and posted a band-aid solution in this thread.

I'll shut this old thread down and we can continue there.