Folding Forum

I played with folding for about 3-4 months last year and had many problems with instability.

My computer.
I7 950 quad core with hyper threading overclocked to 4.3 ghz (custom water cooling I can get a slightly higher OC)
GTX 470 soon to be in SLI.
6 gb RAM

Loading the graphics card hasnt been a real problem but there is still a massive amount of processing power. Is there an easy way to load all threads?

Loading a single copy of folding at home ment the cpu would run at about 12% when running 8 copies it would run at 90+% but the computer would be unusable (not cool!)

I guess what i would like would be a client that would load the computers cpu but that could be turned off or dialed back when i need more cpu power? This computer could be dedicated to folding on average 16 hours a day.

Anyone have any suggestions?

Steven

Welcome to the forum, flying freak.

All the FAH CPU clients should yield CPU time to any other program that needs it; by default, they run at the lowest priority, and thus basically every other task gets CPU time if it needs it. The GPU client doesn't always work this way because GPU drivers don't handle multitasking all that well.

What do you mean by unusable? Was the computer very laggy, or was it crashing? If it's the latter, know that FAH pushes hardware harder than basically any other program; an overclock that's stable for gaming may not be FAH stable, especially since most other programs don't use the Floating Point Unit nearly as much as FAH. It's recommended that you run StressCPU2 to test your CPU's overclock; it runs that same Gromacs code that FAH uses, and is considered the best FAH stability test.

Anyway, the current recommendation is to use the v7 client. It's an all-in-one client that can run all your hardware using different slots, which are analogous to separate v6 FAH clients. The best setup for your computer would be one GPU slot (2 when you get your second 470) and one SMP slot. The SMP slot will use all your CPU power to process one WU. As mentioned before, it should back off when other programs are running, but let us know if it doesn't.

Zagen30 wrote: Anyway, the current recommendation is to use the v7 client. It's an all-in-one client that can run all your hardware using different slots, which are analogous to separate v6 FAH clients. The best setup for your computer would be one GPU slot (2 when you get your second 470) and one SMP slot. The SMP slot will use all your CPU power to process one WU. As mentioned before, it should back off when other programs are running, but let us know if it doesn't.

Yes, the v7 client is a good choice. It works very well scientifically, but does have some bugs, which the developers are hard at work on stomping out. So at this point I'm a bit cautious about quickly recommending it. However, at the same time it really does solve many of the lingering issues that have been plaguing previous clients, and since it unifies them all the user no longer has to have the responsibility of choosing the right client. It has a really nice interface with big intuitive control buttons, which is very nice. It will likely be only a few months before v7 becomes a fully recommended polished product. In the state that it's in now, it works well for its underlying purpose (crunching WUs on different hardware) but has a few bugs many of which are cosmetic. So, go ahead and check it out, but I just wanted to let you know that it isn't perfected just yet.

V6 is more stable than V7

ruleeet wrote:V6 is more stable than V7

Hi Ruleeet, welcome to the forum.

There are so many different criteria that must be considered when saying one is more stable than another. What OS? What set of clients? What hardware? What CPU? What GPU? Dual GPUs? How do you define stable?
I would be cautious in a blanket statement that v6 is more stable than v7. So far scientific production is concerned, I think v7 is superior to v6. Dr. Pande's blog says this: "If you are also concerned about v7 affecting the science of Folding@home I have this to say. v7 is more likely to increase your contribution but if you do have a problem, and you report it, you are potentially helping thousands of users donate computing time to Folding@home so it's more likely a net gain. However, beta testers often have to be satisfied with peer recognition and praise rather than points. If you've been on the fence about v7 now is a great time to upgrade. Here are some of the other comments we've been getting: "i am running windows 7 ultimate 64bit and windows 8 developer preview release. everything went smoothly. . .keep up the excellent work." "The upgrade from v7.1.24 to 7.1.33 went flawlessly. . .install took less than a minute and it resumed folding the wu's they finished without issue. . ." "Congrats on the new release guys. with V7 you've made it all so much easier." "Wow, that is one of a big advance. That was what F@H just needed. It's awesome. Easy to understand and just kickin' ass, the slot thing is what I was missing from BOINC." "Awesome! The best client i ever use, perfect for newbies, easy setup. . ."
V7 is not perfect, that's for sure. But in some case it works far better than v6. In numerous places throughout the forums site moderators are saying "have you considered v7?" So it can't be that unstable. They usually say that after helping out that user with a v6 problem. V7 also has greater scientific capabilities than v6, such as new cores for Linux.

I look past the cosmetic bugs. They will be fixed soon. I for one feel that v7 is more "stable" than v6 all around.

Both client versions use the same fahcores (with the exception of core 16 for ATI on V7), and the fahcores do all the work, so effectively, they should be equally stable.

ruleeet wrote:V6 is more stable than V7

Facts please. If you've had an instability with V7, we can probably help you, but you'll have to post enough information for us to know what happened.

I have been running V7 since it was first introduced and I have not seen anything I'd call an instability that's associated with the client. I do see reports here on the forum of WUs that are unstable, but that's either the fault of the WU or the core, and V6 and V7 run exactly the same WUs and the same cores except for the new openCL core that's used for ATi GPUs. That core has had a separate beta testing process from the client. If you do have an ATi GPU, running it would be optional, anyway.

Hey guys thanks for the help and sorry about taking so long to come back iv been having problems with my ISP.

How many ppd should i be getting? im currently only getting about 12k on my gpu and 5-6k ppd on an i7 at 4ghz.

My overclock is very stable, never crashed. Just the computer was very laggy this problem has been solved by the v7 beta .

Another question, I notice the gpu client doesn't load my gpu client nearly as much as furmakr , is that normal?

Steven

flying freak wrote:Hey guys thanks for the help and sorry about taking so long to come back iv been having problems with my ISP.

How many ppd should i be getting? im currently only getting about 12k on my gpu and 5-6k ppd on an i7 at 4ghz.

My overclock is very stable, never crashed. Just the computer was very laggy this problem has been solved by the v7 beta .

Another question, I notice the gpu client doesn't load my gpu client nearly as much as furmakr , is that normal?

Steven

Hi Steven
your ppd are about right for the GPU (GTX 470). And the loading is normal.

For your i7, you should be running the smp client. You said earlier you have an i7-950, this can run 8 cores, which the v7 client would have recognised at installation.
Have you got a passkey and have you completed at least 10 smp WUs with at least 80% within the deadline? This is required to qualify for the Quick Return Bonus.
Note that if you are only folding 16 hours a day this will reduce your ppd for the SMP WUs because of the reduced QRB.

Yes I have a passkey and have to date completed about 200 WU Vast majority of which have been completed successfully.

Is there a way of increasing the load on the gpu? (already set to 100% I just feel its got a lot more in it)

How do i get the quick return bonus? Do I need to add flags?

flying freak wrote:Yes I have a passkey and have to date completed about 200 WU Vast majority of which have been completed successfully.

Is there a way of increasing the load on the gpu? (already set to 100% I just feel its got a lot more in it)

How do i get the quick return bonus? Do I need to add flags?

Only SMP WUs count towards the QRB, not the GPU WUs.

You cannot increase the load on the GPU beyond 100%, but you can overclock it to some extent, dependent on suitable cooling etc.

Once you meet the criteria, the QRB is awarded automatically for each SMP WU returned within the deadline.

I still think something is wrong as far as my cpu folding.
Current work unit is 2% complete predicted 1.39 days left estimated ppd of 1750
( cpu : i7 950, hyper threading enabled mild 4.0 ghz OC very stable)

Your ppd for the CPU does sound low, I get about 8-10,000 ppd on my 2.8 GHz i7 folding 24 hours per day. That is with smp:8. Depending on how much CPU resources it takes to run your GPU units, you might need to run with smp:6 or 7. 6 threads will work with all cores and WU's, some cores and WU's will crash using 7. What project is your current WU from?

Currently working on project #7610 (472, 0, 47)

I took a screen shot showing all my settings this may help? http://imageshack.us/photo/my-images/834/21252235.jpg/

Steven

SMP:

You have two a4.exe running where there should only be one. Pause/stop the client, go into task manager and if there's one still there kill it.

GPU:

Adding advmethods to your GPU configuration will get you WUs that better utilize your 470, at least for now. Keep an eye on your gpu temperatures as you might have to increase your fan speed.