error running bigadv on windows

[muziqaz]

Please do not take any suggestion about unstable machine too personal. That means to all of you. We want correct data and stable machines for this project. Don't really need wrong data at the end of the project So the 1st and the most important thing we look at when seeing EUE is hardware. Yes, you passed one million runs of super pi, yes you passed 24 hours prime stable, yes, intelburn melted your heatsink, or you ran fine with normal SMP. Nothing compares to the bigWUs and their demand for stable ram.
You can have rock solid system with all the stuff you throw at it, but if you fill you ram at the certain point system becomes unstable.
As 7im said, try easing ram timings at first. see what is happening then. or ease the OC a bit.
At the end of the day we all want good results for the project and if someone suggest unstable machine, please consider this without feeling insulted

[B2K24]

I come to the conclusion trying to use all 8 cores -smp + running GPU3 with my GTX 480 was a mistake. I also just realized using setPriority software I had the FahCore_15 (GPU3 core) set to High priority and a3 core set to AboveNormal priority is where I assume went wrong.

Next time I attempt bigadv after stability tests complete will be -smp 7 with no SetPriority software running.
I will also put all Ram timings auto settings in BIOS as with the Corsair Dominators C7's that I have I notice 0 differences anyway.

[7im]

Do NOT change the Windows Priority settings. It leads to bad results. Using only the priority settings in the clients is sufficient. SMP stays at the default idle setting. GPU client gets the "low" or "slightly elevated" setting (same) depending on your client type.

Affinity settings are a different issue, but not really necessary, especially if you use -smp 7. But everyone's situation will vary. Try both -smp and -smp 7 with the GPU running. Go with whichever config nets the most points.

[AussieFX]

The second you overclock your cpu you lose your guarantee that the system is stable and returning correct results.

This is correct unless you stress your system to ensure that it is stable. I personally choose IntelBurnTest as it is the meanest ones out there that I know of.

Are you really serious about that comment?
Try that one on intel or AMD and see what they have to say.

Maybe Cray should pull their finger out and overclock Jaguar, that would make it uncatchable.
What do you suggest we stress it with, intelBurnTest? It wouldn't matter that all the results that come out of it are invalid would it.
I will state it for you again, You're not playing Crysis here and E-Peens are not required

If you want to run bigadv then go and put your hand in your pocket and run the correct hardware.

[PantherX]

...It wouldn't matter that all the results that come out of it are invalid would it.
I will state it for you again, You're not playing Crysis here and E-Peens are not required

If you want to run bigadv then go and put your hand in your pocket and run the correct hardware.

Again, if the hardware gives error then yes, it is invalid. Hence PG has their own stress software that ensures that the data produced by the system is correct: Stress CPU 2.0. If they has something against overclocking, they would clearly state it
However, they do accept overclocked hardware according to this post: (http://foldingforum.org/viewtopic.php?t=10697)

Q: Can I try running these units on my super-overclocked, liquid-cooled quad-core system?
A: No. In our experience, fast quad-core systems tend to come in over the 4-day timeout and would thus 1) not contribute to the scientific goals of finishing these projects quickly and 2) not receive bonuses. Quad-core systems can make important contributions to standard SMP projects, and we'd encourage you to apply them there.

A lot of folding enthusiasts do overclock their system as it helps to do more science. Regarding bigadv, as long as my i7-860 @ 3.8 can do it, I will. If it fails to meet the Preferred Deadline, I will simply shift over to normal SMP2 Folders with dual socket motherboards (SR-2) also overclock their system.
BTW what is this "Crysis" that you talk about I use my system to fold 24/7

[muziqaz]

They have no choice but accept oced results. They always suggested against ocing, but as everyone does that, they have to accept it. and there is no way to see that system is overclocked with todays CPUs(black edition, extreme edition).

[leexgx]

@10e
that setting about how much ram is going to be used for SMP2 is just for reporting to servers when getting an work unit Bigadv only using 1.4-2gb of ram at best when active, norm A3 SMP i think was less then 200mb

@others
(not sure what the first go at me was about overclocking was all about)
overclocking is fine but as i found out some times the OC can fail, like my linux system has been stable for weeks and now it keeps on kernal panic hanging now after 20 mins (just need to dial back the OC on the BLK back 1 point or more voltage) even this system i am on right now is not fully stable seems to BSOD after 4-5 days but does not affect the SMP folding thought

[muziqaz]

I do hope that pg built in code which checks the simulated sequences for errors is good enough. Letting unstable machine to fold is very irresponsible. The Wu might not crash but it can generate wrong data. I hope the OP realizes this and makes sure his machine is stable in every possible way.

[Grandpa_01]

I do hope that pg built in code which checks the simulated sequences for errors is good enough. Letting unstable machine to fold is very irresponsible. The Wu might not crash but it can generate wrong data. I hope the OP realizes this and makes sure his machine is stable in every possible way.

by AussieFX » Tue Jul 06, 2010 7:57 am

B2K24 wrote:
Listen folks this is a BETA which means the less people participate then the less feedback developers get to make improvements.
My system is not unstable. I have had uptimes of over 1 week plus in the past.

The second you overclock your cpu you lose your guarantee that the system is stable and returning correct results. An uptime of one week? Don't make me laugh, this isn't Crysis.
You have to be able to guarantee stability and more importantly correct results. You cannot do that whichever way you attempt to spin it.

I will say it again, bigadv was released for highend multisocket machines not quadcore machines with pretend cores overclocked to high heaven in an attempt to meet a deadline so you can gain some extra points.
If you really want to do something for science use your machine as it's supposed to be used, on standard clients.

If you want to play with the big boys stick your hand in your pocket like others have and build the correct machine for the job.

I'm not having a go at you personally it's aimed more at PG. They are letting the results become compromised to accomodate point whoring and I for one would like an explanation as to why they have dropped their requirements. People invested a lot of money in purchasing multisocket systems for this. Some way to pay them back.

In regards to your snide jab at orion, he can guarantee his machine is stable. He is [strike]using[/strike] donating a correct multisocket platform (as per the requirement,) several of them in fact.

I am guessing neither of you have ever been able to achieve a stable OC. A stable OC is achievable there are many of us that have achieved it your statements about not being able to achieve a stable OC are wrong. You may not be able to do it but others can.

If OCed I7's were compromising the results I am sure PG would have said something about it along time ago. If PG wants people to quit running I7's on bigadv they are the one's that need to say that not you. And both of you have been a bit snide and snobbish with your comments. My OCed I7's will continue to run bigadv WU's and put 3 of them out every 48hrs until PG ask me not to.

[10e]

Are you really serious about that comment?
Try that one on intel or AMD and see what they have to say.

Maybe Cray should pull their finger out and overclock Jaguar, that would make it uncatchable.
What do you suggest we stress it with, intelBurnTest? It wouldn't matter that all the results that come out of it are invalid would it.
I will state it for you again, You're not playing Crysis here and E-Peens are not required

If you want to run bigadv then go and put your hand in your pocket and run the correct hardware.

Until you can find proof that B2K24's system is unstable and causing issues then you are just reaching at straws.

I can tell you that I have 5 BigAdv capable systems based on Core I7 processors that are overclocked to 3.8 Ghz or higher that have not produced errors in any BigAdv units except ones that suffered from stability themselves, either due to a bad WU or bad core. I test them at full bore with LinX 0.63 for over 25 hours WHILE the GPU2/GPU3 client is running concurrently (at normal priority due to CPU starvation when running LinX) and all have passed at suitable temperatures without error, and if there was an error, I re-did my setting and/or lowered my overclock. Additionally I test them at higher CPU speeds than what they fold at, and generally CPU folding on 8 cores produces temperatures at 10 - 12 celsius LOWER than LinX.

I was one of the people who complained about 2684 Linux A3 BigAdvs not being stable, and guess what? They pulled the A3 2.22 core because it has problems. Had I successfully completed the EXACT same unit (run, gen, clone) as B2K24 in Windows, then I'd say there is a system issue, but I haven't.

Vijay himself stated in the thread started by Leganfuh that the code is very complex and errors are possible. Focusing on one potential issue is causing you to suffer from tunnel vision.

And AMD and Intel would probably say the same thing: You are running hardware above spec and are reducing its lifespan, and increasing the possibility of error, not guaranteeing that there WILL be an error. Why do you think they sell multiplier unlocked CPUS? For soccer moms on Facebook?

Now if you are not here to help B2K24 then please stop trying to derail this thread.

[muziqaz]

I have perfectly stable phenom x4 oced to 3.8ghz. Aussie has perfectly stable x6 oced to 4.2ghz. I fold with that system as well. But if I get any kind of instability I drop my oc and rerun the wu. All we are saying that people need to realize that they are running above the specs and 1st thing they MUST do is check the hardware. Not look at the past, what and how many wus were completed with that system. Noone is singling out this issue.
If you run oced system don't get offended when someone tells you your system might be at fault.

[7im]

I do hope that pg built in code which checks the simulated sequences for errors is good enough. Letting unstable machine to fold is very irresponsible. The Wu might not crash but it can generate wrong data. I hope the OP realizes this and makes sure his machine is stable in every possible way.

For the record, PG is NOT against overclocking. They are against unstable machines causing EUEs, corrupting data, and turning in partial work when the WU could easily be completed on a stable system. That wastes time and resources. Unfortunately, guess what kind of systems are typically the most unstable? Or OCers claim to be Windows stable running lame test tools like Prime95. And then OCers come here and complain that Fah is broken because it crashes on their "24 hour Prime Stable" system. Fah is a lot more demanding that just running Windows and a web browser. Get some real stress test tools.

And yes, PG does have error catching code, hence all those error codes documented in the WIKI.

Alternately, there are WUs that finish early by design, and not unexpectedly, we will get a few bad WUs from time to time. Just part of the research we do.

...
Until you can find proof that B2K24's system is unstable and causing issues then you are just reaching at straws.

snip

Now if you are not here to help B2K24 then please stop trying to derail this thread.

No, not entirely guessing without supported reasoning. Again, if the same WU fails in the same place on the same system, the WU is typically the problem. If the same WU fails in a different place, it is typically a system problem.

Easy test. Since B2K24 errored out on that WU posted on page 1, it was likely was reassigned to another user. Let's ask a Mod to check the WU DB and see of that other user completed the WU with an error, or without.

[bruce]

Ok so I woke up to this Project: 2684 (Run 3, Clone 12, Gen 5) seems to be bad luck for me.
Why did it end early?

Partial credit for an error: Hi B2K24 (team 86565),
Your WU (P2684 R3 C12 G5) was added to the stats database on 2010-07-06 05:09:51 for 6914.44 points of credit.

Full bonus credit for no error: Hi xxxxxxxxx (team xx),
Your WU (P2684 R3 C12 G5) was added to the stats database on 2010-07-07 05:09:08 for 60956.7 points of credit.

Now it's time for B2K24 to explain why other people's machines work and his/hers doesn't.

Those who overclock and run FAH must be extremely conservative about how much they push the limits. A single error like this can waste a lot more than can be gained by that last notch on the clocks. Anything that you do that increases the chance of errors probably isn't worth it.

[B2K24]

Now it's time for B2K24 to explain why other people's machines work and his/hers doesn't.

A beta program is a learning process not only for the developers but end users as well. You learn what works and what does not.
In my post previous to this one I posted where I think I went wrong. My machine is stable I just made some mistakes in how I went about things and settings/software I choose to run (SetPriority) which I understand now is not needed at all.

With respect to F@H I'm not just changing a few settings and running bigadv again right now with a desire to succeed. I will do the stability tests with multiple programs to ensure on my end it will be the best I can get it. I appreciate everyone who stuck up for me in this thread and gave me useful feedback to learn from this experience and even though the WU failed I have gained knowledge plus a better way to go about executing this beta program. I thank everyone that gave me useful feedback and ideas, much appreciated.

[PantherX]

@B2K24: Make sure you keep eye on the CPU temps as some stress softwares can really heat your CPU in a short time period. Some recommended stress softwares are mentioned in my mini-guide (link in sig). Wish you all the best.