New To Folding

[satandole666]

pscout recommended that I start my own thread since I had some questions.

The CPU folding is going nicely. I haven't had it running nonstop so I haven't got a lot of "work" done yet. When it was running for about 14 hours straight it said I would finish the WU in just over 5 days. Is that really slow? If so, is it even worth it to bother?

Next question...does the GPU client work with the newest Omega drivers? I tried to get it running. I got past this problem:

"This application has failed to start because d3dx9_30.dll was not found. Re-installing the application may fix this problem."

When it finally opened it didn't seem like it was actually doing anything. It didn't show any progress or anything like the CPU client. It also made my entire system very sluggish. Windows (not the OS, but open folders, etc) became very choppy and the system was nearly unusable by my standards.

Is this the Omega driver thing? If so, that's an easy fix (although I'd rather not if there is a different way to fix it).

Lastly, if I get the GPU client figured out...is there a guide posted to running a GPU and CPU client at the same time? pscout said:

If it's a DC you can run 1 of each ... each instance needs to have a different machineid in the client.cfg file.

I'll take a look at changing those...but I have no idea what to change...etc.

Just a little background...I don't do a ton of intensive gaming. I play CS:S maybe a dozen times a week...other than that I browse the web, play movies/music...etc. I'm more than willing to leave it up whenever the computer is up if I can get all of the kinks worked out.

Thanks.

[S3CUR3]

Here's what Pande says about drivers:
* Catalyst driver version 6.5 [download from ATI or from a Stanford mirror ]. Version 6.10 [see main ATI Folding@Home page or download driver from ATI ] with the newest core appears to be working now (but requires more testing to be sure). Versions 6.6 and 6.7 will work, but at a major performance hit; 6.8 and 6.9 will not work at all. Due to all the complexities of support, we will support only versions 6.5, 6.10, and 6.11. Version 6.12 does not work correctly, but we are expecting that the 7.x series to solve that problem.

Also, I'm not sure what the latest update is, but I know in the past they have essentially said that the GPU will make the user interface sluggish and there's nothing they can do to change it. The best fix is to pause it while you're using the box. Perhaps someone can correct me on this?

[dz_jad]

single core box for the GPU and CPU at the same time??
Open up your Task Manager, (CTRL+SHIFT+ESC) and sort by proc usage....
FahCore_10.exe is GPU, if its a single core, this should be 100% or near it.
Sometimes when I boot up, it doesnt show progress, then I remember that I still have a CPU client on service, so I stop that, then the GPU flies down the road.


Depending on what work unit you have, 5 days is plenty fast. There are some WU's that fold quick, some that go slow. The patty-melts are the slowest ones currently, and there seems to be plenty of 'em.

With the DX error, I just went to the site that the GPU FAQ (on stanfords site links to) and just threw the .dll in the Folding folder.....not sure if that's the way you're supposed to do it, but it works for me...

[pscout]

Ok ... amd noob that i am ... i had to google opteron 165 to find that it is a dual core :o

So, your best folding confg is to run 2 folding instances ... either 2 cpu instances, or 1 gpu + 1 cpu instance.

When you run 2 instances, the client's need to be configured with different machine ID's. The default is 1 ... so if you try to run another instance with machine id =1 and the other is already running, the second one should stop immidiately.

I am not sure if gpu folding on your x1800 will be better points than running a second cpu instance. Depends on your cpu/gpu oc's, and what cpu wu's you get.

I would try out the 1800 first tho to help you decide. You DO need that dx9 driver or else you will get constant EUE's .... wxp sp2 even updated by ms update does not include it.

I have read about people having problems with the omega drivers. Not sure what omega drivers are myself :o

I fold several x1900 AIW's and with the latest gpu client, it works perfectly and installs normally with the 6.11 drivers from ati.
In the first gpu client beta (now expired) i had to force install my 1900 AIW as a x1900 radeon but the x1900 is now supported.
The GPU faq says the x1800's are supported so i assume that includes the 1800 AIW.

If you haven't looked yet, there is a lot of info on the stanford gpu forum -> http://forum.folding-community.org/viewforum.php?f=54. Also look at the GPU FAQ on the stanford download page.

In general, it is best to stay away from the graphical versions of the cpu and gpu consoles ... very few folks here use them for long ... they have problems depending on what other software you run on your box(es).

If you are running the graphical. you can switch. Let us know if you need some help on that.

On how to change your machine ids - here is a post i just did on another thread about editing client.cfg -> http://www.ocforums.com/showpost.php?p=4860015&postcount=12

if you use the -config or -configonly or 'delete the client.cfg and restart' methods to change the client configuration, then you need to answer yes to advanced settings so that you will be prompted for machineid ... otherwize you will get the default "1" which will cause problems if you run more than 1 instance.

[satandole666]

Wow. Thanks for all of the help.

I'll probably end up running 2 instances of the CPU client. I'm running the CPU at 2700 right now...but my x1800 AIW has horrible problems overclocking. It doesn't like any voltage above 1.125. :/

I will do some playing to figure out what will yield the best PPD and then stick with that.

[pscout]

:cool: Good luck with it :)

[satandole666]

Ok, I switched from the GUI to two version of the console client. I've attached a picture, let me know if it is setup right. I set the machine ID and I've attached the -local flag to the shortcuts.

The only question I have is the protein. It looks like both clients are doing the same work. Both are working on project 2124. This is what I took a picture of, along with task manager at 100% (nice to finally see that outside of Orthos! :clap: )

I have to say I don't like the console that much, but I guess it isn't what I like when it comes to this stuff.

http://img170.imagevenue.com/loc589/th_13440_F@H_122_589lo.JPG (http://img170.imagevenue.com/img.php?image=13440_F@H_122_589lo.JPG)

EDIT: d'oh...set team to 5 instead of 32. Fixed now.

[Patched]

It can be the same project as long as the run, clone, gen are different

I also have a X1800 card and a dual core Opteron easily beats its plus no mouse lag and GPU fan noise.

[satandole666]

It can be the same project as long as the run, clone, gen are different

I also have a X1800 card and a dual core Opteron easily beats its plus no mouse lag and GPU fan noise.

Awesome...it looks like I am set then. Thanks a TON for saving the time experimenting trying to find the best PPD.

[satandole666]

Alright. I had my first set of hiccups.

Running two consoles I have received this error 4 times (twice on both clients). The first two happened when I was pretty far into my work. :mad:

down: FILE_IO_ERROR
tch (work/wudata_01.xtc)
Premature end of file when checksumming <43396 bytes left>
- Could not calculate checksum <work/wudata_01.xtc>
Checksum not what expected.

Folding@home Core Shutdown: FILE_IO_ERROR
CoreStatus = 75 (117)
Error opening or reading fram a file.
Deleting current work unit and continuing...

According to this thread:

http://forum.folding-community.org/ftopic16110.html

I should run checkdisk and make sure I have -local on both clients. Well...I do...

Does anyone have an idea how to fix this?

EDIT: According to this:

http://fahwiki.net/index.php/CoreStatus_codes

I have a hardware problem. I've never had an issue with anything...

EDIT2: And it happened again...on both consoles. I've stopped them for now while I try to figure this out.

[pscout]

Sounds like your oc is unstable. Dual folding will find any instability :D

I assume your pci is locked for your oc. I am not an AMD guy so can really help much with the details but you either need to reduce the clock and/or loosen the memory timings or speed. With my intels, orthos stable is not a guarantee of stability but is close. If i eue'ed a new OC, i would just drop my fsb by 5-10. The memory tweaks on ddr2 memory aren't that important on C2D rigs, but matter a bit more on my presslers and i think they do on AMDs.

For folding tho, you need to leave a buffer below max OC for variable ambients, dust, and even new wu types. Valid EUE's are usually pretty rare on most projects but have a higher probability on advmethods but usually still low. The projects normally go through some beta testing before released. File i/O errors are lmost guaranteed to be hardware ... but a checkdisk is a good first step in diagnosis.

I wrote up a post a while back in the benchmarking forum on the approach i use to getting a good OC for folding ... http://www.ocforums.com/showpost.php?p=4631951&postcount=15

It talks about how not to turn in EUE's due to an unstable OC.

Also, you should probably set the -advmethods flag since there seem to be more good wu's there atm.
Unfortunately it will take you a few days to crunch those paty melts and hopefully when they are done the ribos and 1499's are still coming on advmethods.

The patty melts (2124/25) don't stress too much .. the 1499's will give your OC a real test and nice points, so you want your rig stable for them.

[satandole666]

I'm fairly sure this OC is 100% stable. I've tested it to hell and back using everything I can think of to include benchies, extended Orthos, games, everything. I've been using it for weeks now at these settings and haven't had a crash or hang yet. I didn't have any problems running the GUI version for two days either.

I think I did find the cause of this error, however. It was starting with Windows as a service. I do not believe this service had the -local flag enabled. I fixed this in the config and I am starting it manually. I'm going to leave it on overnight and see how it runs. I'll post back with the results.

[pscout]

Are you running the GUI client?

It can cause problems with second instances. Since I stopped running it long ago, and most here have i would suggest getting rid of it.

You can move the wu it is currently running to to another service or desktop console instance so you don't lose the WU, then uninstall it completely.

Till you get it sorted out tho, it is a good idea to backup the wu's and disable the lan while testing. Then you can just restore the backup if you get an eue and test any changes.

Everytime you return an eue, it gets reaasigned to several machines ... so since it is not really a valid eue if it is due to config, it wastes those other rigs time folding the same wu needlessly.

[S3CUR3]

Just a note for a new folder - EUE = early unit end, the problem that you're having. If you already knew that, just disregard this... :)

[pscout]

One other thought ... are you only getting errors with 2 instances running or did you get them running just 1?

[satandole666]

Alright, I guess what I did fixed it. It was the lack of the -local flag when it starts as a service. It went overnight without a hitch.

And no, I'm not using the GUI client anymore...just 2 versions of the console.

[pscout]

:thup: glad to hear that it sounds like you are set then :)

Is a shame you had so much trouble getting it going right ... but looking on the bright side you probably learned a lot about folding along the way :)

Fold on!

:beer: :beer:

[harlam357]

:welcome: to teh FOLD satandole! :D Looks like pete already has you Folding in the right direction. ;)

[satandole666]

Well everything is going fine so far.

I am 93% complete with 2 WUs on my 165. I didn't expect it to take this long...espescially since my comp has been up 24/7. 4-5 days for a patty melt seem about right? Keed in mind I have done a fair number of restarts in between. This is a fresh install of Windows and I've been installing stuff and updating.

Also...my stats haven't updated at all. I figured that points wouldn't be awarded until completion, but shouldn't it update the CPUs I have active? I kind of guessed that having a dual core running two instances would count for 2 CPUs. I also have it installed on my 3000+ as well.

And as a side note...I'll be installing F@H on three more computers today. WOOHOO!

[harlam357]

Right On! :thup:

Nope... active cpus are shown for procs that have turned in work in the last 7 days. So you won't see anything there until those clients turn in their work.

And 3 more cpus for T32! Double Right On! :D You're well on your way to becoming another Folding Addict. http://img.photobucket.com/albums/v663/harlam357/smileys/borgsmiley.gif

[satandole666]

Its actually 4 CPUs. They are 3.2ghz P4s (with HT, one of the newer P4s I believe) paired with 1gig of ram each. It should be a nice addition. I might have a firewall problem but I'll get that figured out.

And I'm already addicted...I just don't have the money or the spare parts to start a farm in my house. :)

EDIT: Quick question. IF the console/core has work and is working (shows progress) does that mean it can contact the server to download work? I noticed problems contacting a server but it still managed to start working afterwards. These are fresh installs, so I guess the question is does the download come with a WU built in?

Last note: First 2 WUs complete...got just under 800 points for them. I'm up to 7.2 ghz of AMD64 (one 1.8 3000+ and a 165 at 2.7) and 9.6ghz of P4 power. Woohoo...there are about 5 more 3.2ghz P4s I can put this on too.

[pscout]

:thup: WTF (way to fold :D ) :)
More Borgs!!!

On the network question: The client will contact the assignment server initially to get assigned a WU from one of the project servers as listed on the Psummary page -> http://fah-web.stanford.edu/psummary.html. It will then download whatever wu it was assigned from the project server.

When the wu completes, the client will attempt to connect to the project server it got the wu from. If it can't connect (due to load, network congestion, server down, etc) it will retry. The completed wu will be held in the queue on the client (queue.dat and results files in the work dir) untill it can be sent.

The client will then contact the assignment server for its next wu and the above process will repeat.

After a number of retries to transmit a completed wu, the client will send it to the collection server. I don't think points are creditied untill it actually gets processed by the original server when it comes back online.

You may see in the fahlog.txt on the client (might need -verbosity 9) periodic retries by the client every 6? hours to transmit any completed wu's from the client's queue.

So if no wu's are to be sent, and a wu is being worked on, there does not need to be a network connection until the current wu finishes.

Upto 9 wu's can be queued on the client. EMIII has a viewer to display the queue contents on any folding instances it is monitoring.

[satandole666]

Alright, that makes sense.

It looks like my new WUs are p1497's. The slower of the two clients will be finishing it in about 21 hours and the faster will finish in about 19. These are going MUCH faster than the last WUs...are they worth less points per WU?

And thanks everyone for getting me addicted to F@H. I've been looking for something constructive to do with my computer (besides gaming, that is) and this is perfect.