Quick question about getting started

[Xee]

Well, as the title says, I got a quick question about getting started.

Seeing as how I pretty much just use my computer for browsing 90% of the time, and rarely use my PS3, I'd figure I'd might as well use their processing power to fold. I've read through the guides and should have my PS3 up and running shortly but am befuddled as to which client I should download for desktop; I've read the descriptions of all of them but am still not sure which one I need. I'm running Vista Ultimate 64-bit and won't be able to run 24/7 if that matters. Thanks in advance.

 

[Zero82z]

If you want the absolute minimum maintenance cost, you should download and run four instances of the standard uniprocessor client and set up one for each core of your CPU. You can also download and run the Windows SMP client, which will give you more points output, but you can't run it behind the scenes (you have to keep a window open, although you can always minimize it or shift it to a secondary monitor), and it's a little bit less reliable.

If you really want to go crazy, you can set up two Linux virtual machines and run the Linux SMP client which is much more efficient and reliable than the Windows client. However, you'd have to start up the VMs every time you boot your PC, which may be too much of a hassle for you. If you're willing to do it though, it will really give you a nice boost.

If you need clarification about anything, or help setting things up, just ask and either myself or someone else will walk you through the whole thing .

 

[flogge]

Best productivity will be with 2 Linux VMs per capreppy's guide. These however do take some maintenance. Your 4850 can also produce about 2-2.5k ppd. That is the easy one. Just download the GPU systray client and set it forget it.

 

[Xee]

Thanks for the suggestions. I'll definitely give the 2 VM option a try.

Your 4850 can also produce about 2-2.5k ppd. That is the easy one. Just download the GPU systray client and set it forget it.

Only thing is that under load my 4850 sounds like a noisy vacuum cleaner; very annoying. Really wish I would have gone with one that had an aftermarket cooler on it.

 

[flogge]

If it's amajor problem you can cut down the productivity a bit to keep the heat down. Just go into the advanced tab on config and reduce the CPU use percentage.

 

[Chaoticon]

Thanks for the suggestions. I'll definitely give the 2 VM option a try.



Only thing is that under load my 4850 sounds like a noisy vacuum cleaner; very annoying. Really wish I would have gone with one that had an aftermarket cooler on it.

Use something like rivatuner and keep the fan at 40-60percent and temps below about 70c, if the card has fan control - Vacuum sound solved. I would just run two normal clients and a gpu client 24/7(not sure why you wouldnt). Dont do smp clients if you arent willing to keep the computer on for long periods.

 

[Zero82z]

If it's amajor problem you can cut down the productivity a bit to keep the heat down. Just go into the advanced tab on config and reduce the CPU use percentage.

That only affects CPU usage though; there's no way to decrease GPU usage.

 

[Xee]

Use something like rivatuner and keep the fan at 40-60percent and temps below about 70c, if the card has fan control - Vacuum sound solved. I would just run two normal clients and a gpu client 24/7(not sure why you wouldnt). Dont do smp clients if you arent willing to keep the computer on for long periods.

I can't run it 24/7 because when I sleep my head is about 3 feet from the tower; the noise plus the fact the LEDs light my room up like a disco ball makes it impossible to fall asleep.

I'm an early riser though so I can probably leave it on at least 12 hours a day. That enough to warrant the SMP route? I'm about a third of the way through the process of setting them up so it'd be good to know before I go any further.

Thanks for the suggestion about GPU client. I'll change the fan speed and monitor the temps for a while and see how it goes.

 

[Zero82z]

I'm an early riser though so I can probably leave it on at least 12 hours a day. That enough to warrant the SMP route?

Yes. Running two VMs, you should finish a pair of units every 2 days or so. It takes about 16 hours for one of the VMs running on my Q9550 to finish a unit while running at 3.8GHz, so with your CPU at stock speeds it should probably take something like 20-24 hours per unit (and since there are two VMs running simultaneously it's really more like 20-24 hours per two units).

Also, make sure you follow the instructions in this thread if you do go ahead with the GPU client: http://www.hardforum.com/showthread.php?t=1419852
You'll need that to get your CPU usage down enough to run a second VM.

 

[Xee]

Yes. Running two VMs, you should finish a pair of units every 2 days or so. It takes about 16 hours for one of the VMs running on my Q9550 to finish a unit while running at 3.8GHz, so with your CPU at stock speeds it should probably take something like 20-24 hours per unit (and since there are two VMs running simultaneously it's really more like 20-24 hours per two units)

Hmm. Maybe I should look into OC'ing; that's pretty much the only reason I went bought this specific mobo and paid extra to get the heatsink lapped. Not having done it before though, I've just been too nervous to experiment; I don't really have the time or money for repairs if I end up nuking something.

 

[Kendrak]

Hmm. Maybe I should look into OC'ing; that's pretty much the only reason I went bought this specific mobo and paid extra to get the heatsink lapped. Not having done it before though, I've just been too nervous to experiment; I don't really have the time or money for repairs if I end up nuking something.

Keep an eye on temps and you will be fine.

It is very hard to kill a mobo / CPU now. Not so a few years back. Baked one of my Socket A Athlons back in the day. Mobo wans't too happy either

 

[Zero82z]

Hmm. Maybe I should look into OC'ing; that's pretty much the only reason I went bought this specific mobo and paid extra to get the heatsink lapped. Not having done it before though, I've just been too nervous to experiment; I don't really have the time or money for repairs if I end up nuking something.

It is effectively impossible to kill anything while overclocking as long as you don't use voltages that are too high. Just to give you an example, my CPU is running at 3.8GHz with the CPU voltage set to 1.325V, and my northbridge, FSB termination, and CPU PLL voltages each raised by 0.1V. All those voltages are well below the maximum safe values, so there is no risk of me frying anything. And as for temps, Intel CPUs have built-in thermal protection so they will automatically reduce their speed or shut down when the temperature gets too high.

 

[Xee]

If you need clarification about anything, or help setting things up, just ask and either myself or someone else will walk you through the whole thing .

One question. I'm supposed to add the three lines to the affinityinput.txt file and chage the "PAIR0+PAIR1" values to match my system. What would it be for mine?

*\vmware-vmx.exe := PAIR0+PAIR1 [assign=2,priority=Idle]

*\FahCore_11.exe := PAIR0+PAIR1 [priority=High]

*\FahCore_14.exe := PAIR0+PAIR1 [priority=High]

Valid CPU Options

In the Valid Profile Lines, use the values appropriate for your CPU

Intel & AMD Dual Core:

CPU0, CPU1, PAIR0

Intel QuadCore:

CPU0, CPU1, CPU2, CPU3, PAIR0, PAIR1

AMD QuadCore:

CPU0, CPU1, CPU2, CPU3, QUAD0

Intel Core i7:

CPU0, CPU1, CPU2, CPU3, CPU4, CPU5, CPU6, CPU7, PAIR0, PAIR1, PAIR2, PAIR3

 

[flogge]

Intel Core2Quad would be Pair0 and Pair1. Alternately you can use CPU0, CPU1, CPU2, CPU3 if you need to split them up more than that. However using the pair term will keep the cores on the same dies owrking together, remember a Core2Quad is 2x dual-core CPUs.

 

[Chaoticon]

I can't run it 24/7 because when I sleep my head is about 3 feet from the tower; the noise plus the fact the LEDs light my room up like a disco ball makes it impossible to fall asleep.

Can't sleep with white noise hmmm, earplugs? Come on its for saving lives! Just kidding, 12 hours should be enough, and is better then nothing.

 

[Zero82z]

One question. I'm supposed to add the three lines to the affinityinput.txt file and chage the "PAIR0+PAIR1" values to match my system. What would it be for mine?

Those three lines that you quoted are exactly what you need to add. Just copy that text exactly and it should work fine.

 

[Xee]

Thanks for all the help. The PS3 has been working for about 16 hours and I just got the desktop going as well. A few questions though:

1. Is there a way to minimize the two VMPlayer windows to the tray or hide them entirely or do I need to download a 3rd party program for that?

2. I've followed the instructions to change my name and team and both show up correctly when I enter their respective IP addresses in my browser. However, FahMon still shows my username as notfred2630 for both clients. Any idea why?

3. I still have my old computer with the following specs that I can probably leave on 24/7 (it's in another room). Is it fold-worthy? If so, which client should I use for it? Thanks.

AMD 3500+, 1 GB RAM, nVidia 6800 GT OC

 

[flogge]

In order to change the clients username and whathaveya on a VM, you have to make the config changes in the web interface, looks like you already did. Then you have to remote reboot the client for them to take effect.

 

[Zero82z]

3. I still have my old computer with the following specs that I can probably leave on 24/7 (it's in another room). Is it fold-worthy?

Not really. You could give it a shot, but I doubt you'd get more than 500PPD or so from the uniprocessor client, if even that. IMO, it's not worth the power requirements.

 

[Xee]

In order to change the clients username and whathaveya on a VM, you have to make the config changes in the web interface, looks like you already did. Then you have to remote reboot the client for them to take effect.

Yup, did that. The correct name and team shows up in the VM windows as well. I even ran FaHspy and it correctly showed the name and team. -- Is there any way to make absolutely sure I'm contributing to the [H]orde with it?

Not really. You could give it a shot, but I doubt you'd get more than 500PPD or so from the uniprocessor client, if even that. IMO, it's not worth the power requirements.

Thanks for the tip. Consider the idea quashed.

 

[flogge]

Check your processor count on the stanford stats page. It should show a CPU for each VM and 1 for each GPU after the units get turned in. Also check your folding log file after you start it. It shows the team and username at the beginning

 

[Xee]

Hmm. Do you mean that only once a WU gets turned in is the CPU counted? It currently only shows one and I know my desktop has yet to finish one while my PS3 has already done two:

http://fah-web.stanford.edu/cgi-bin/main.py?qtype=userpage&username=Xee

 

[flogge]

Yes, stanford won't recognize a processor until a WU has been turned in and the page has updated. If the page shows 1 CPU then that is your PS3.

 

[Xee]

Well, you were 100% right. My desktop finally finished what it was working on and my stat page got updated to 3 CPUs. An added bonus is that FahMon is now showing the correct username and team. Glad everything is working as it should be. My next project is try and do some OCing and see if I can speed things up a bit.

Use something like rivatuner and keep the fan at 40-60percent and temps below about 70c, if the card has fan control

Well, I gave it a go. Even at 90% my GPU temp kept climbing; it hit 90c before I finally closed the client.

Anyway, thanks to all those that helped.

 

[Zero82z]

Anyway, thanks to all those that helped.

I'm just glad that you've got it up and running and everything is working out for you .

 

[jfb9301]

Thanks for all the help. The PS3 has been working for about 16 hours and I just got the desktop going as well. A few questions though:

1. Is there a way to minimize the two VMPlayer windows to the tray or hide them entirely or do I need to download a 3rd party program for that?

free 3rd party: TrayIt!

http://trayit.en.softonic.com/

I've been using it for years. To me it is an absolute MUST, for even more than just folding.

 

[Xee]

free 3rd party: TrayIt!

http://trayit.en.softonic.com/

I've been using it for years. To me it is an absolute MUST, for even more than just folding.

Awesome. That worked a treat. Pretty small footprint as well.

Also, I found out why the GPU client was raping my video card. Apparently my computer was long overdue for a cleaning. Today I hauled it outside, took the air compressor to it, and slaughtered a bunch of dust bunnies. I found more than a couple dust clumps inside the vid card alone. After putting everything back together the GPU idle temp went down a whole 15C and at 60% fan speed I'm able to run the client at load and the temp stays below 70C. The only thing I've noticed is that when I run the two SMP clients alongside it I only use 80-85% of the GPU.

Speaking of the clients, their PPD's have become vastly different. One is at 906 and the other is at 2,143:

Is this normal?

 

[Zero82z]

The only thing I've noticed is that when I run the two SMP clients alongside it I only use 80-85% of the GPU.

Did you do this? http://www.hardforum.com/showthread.php?t=1419852

http://img14.imageshack.us/img14/9854/screenshot7112009100013.png[/url]

Is this normal?

One of your SMP clients is running a workunit that uses the a1 core that the Windows SMP client uses instead of the far superior a2 core that the LinSMP client runs. If you like, you can delete the queue.dat file in the F@H directory of that VM and restart it, and it should grab a new unit; preferably an a2.

 

[Xee]

Did you do this? http://www.hardforum.com/showthread.php?t=1419852

Do you know what "Start In" directory he's referring to? I've looked through any folders related to the GPU client but can't find any "ati" dll files.

One of your SMP clients is running a workunit that uses the a1 core that the Windows SMP client uses instead of the far superior a2 core that the LinSMP client runs. If you like, you can delete the queue.dat file in the F@H directory of that VM and restart it, and it should grab a new unit; preferably an a2.

The only queue.dat file I have on my computer is for the GPU client. Guess I'll just have to tough it out.

 

[Zero82z]

Do you know what "Start In" directory he's referring to? I've looked through any folders related to the GPU client but can't find any "ati" dll files.

The GPU client directory will be located in the AppData folder for the user account it was installed from; something like this: C:\Users\*Username*\AppData\Roaming\Folding@home-gpu

The only queue.dat file I have on my computer is for the GPU client. Guess I'll just have to tough it out.

You can access the VM's file system over the network. If you navigate to the same directory the log file is in (which you must know since you set up FahMon to monitor it), you'll see a queue.dat file in that folder like this:

If you delete that file and restart the client (this works for all the different F@H clients), the current work will be ignored and a new workunit will be downloaded.

 

[Xee]

If you delete that file and restart the client (this works for all the different F@H clients), the current work will be ignored and a new workunit will be downloaded.

Found it. I deleted it and restarted the client and it just started spamming the message "failed to open queue.dat". I figured this was because I had it set to suspend when I exited; so I set it to "power off" and restarted. It booted up normally and I entered in "root" when it prompted. When I checked FahMon though, it appears to just be hanging:

[14:49:00] 
[14:49:00] *------------------------------*
[14:49:00] Folding@Home Gromacs SMP Core
[14:49:00] Version 1.74 (November 27, 2006)
[14:49:00] 
[14:49:00] Preparing to commence simulation
[14:49:00] - Ensuring status. Please wait.
[14:49:17] - Assembly optimizations manually forced on.
[14:49:17] - Not checking prior termination.
[14:49:18] - Expanded 733236 -> 3855925 (decompressed 525.8 percent)
[14:49:18] 
[14:49:18] Project: 5101 (Run 0, Clone 108, Gen 96)
[14:49:18] 
[14:49:18] Assembly optimizations on if available.
[14:49:18] Entering M.D.
[14:49:24] Protein: CMF in water
[14:49:24] Writing local files
[14:49:24] rom checkpoint)
[14:49:24] Read checkpoint
[14:49:25] steps  (21 percent)
[14:49:25] Extra SSE boost OK.
[14:49:25] s
[14:49:25] Completed 433765 out of 2000000 steps  (21 percent)
[14:49:25] Extra SSE boost OK.

 

[Zero82z]

Try opening the FAHlog.txt file directly and look at the output. From the text you posted it looks like it resumed the same workunit, but that shouldn't be happening. Fahmon might just be reading from a cached version of the log file before you restarted the VM.

 

[Xee]

Try opening the FAHlog.txt file directly and look at the output. From the text you posted it looks like it resumed the same workunit, but that shouldn't be happening. Fahmon might just be reading from a cached version of the log file before you restarted the VM.

Well, it seemed like it downloaded the a1 core again and picked up where it left off with the old one:

[14:48:51] - Ask before connecting: No
[14:48:51] - User name: Xee (Team 33)
[14:48:51] - User ID: 35EDF5DC245120A7
[14:48:51] - Machine ID: 1
[14:48:51] 
[14:48:51] Loaded queue successfully.
[14:48:51] 
[14:48:51] + Processing work unit
[14:48:51] Work type a1 not eligible for variable processors
[14:48:51] Core required: FahCore_a1.exe
[14:48:51] Core not found.
[14:48:51] - Core is not present or corrupted.
[14:48:51] - Attempting to download new core...
[14:48:51] + Downloading new core: FahCore_a1.exe
....
[14:48:55] Verifying core Core_a1.fah...
[14:48:55] Signature is VALID
[14:48:55] 
[14:48:55] Trying to unzip core FahCore_a1.exe
[14:48:55] Decompressed FahCore_a1.exe (3625104 bytes) successfully
[14:48:55] + Core successfully engaged
[14:49:00] 
[14:49:00] + Processing work unit
[14:49:00] Work type a1 not eligible for variable processors
[14:49:00] Core required: FahCore_a1.exe
[14:49:00] Core found.
[14:49:00] Working on Unit 02 [July 12 14:49:00]
[14:49:00] + Working ...
[14:49:00] 
[14:49:00] *------------------------------*
[14:49:00] Folding@Home Gromacs SMP Core
[14:49:00] Version 1.74 (November 27, 2006)
[14:49:00] 
[14:49:00] Preparing to commence simulation
[14:49:00] - Ensuring status. Please wait.
[14:49:17] - Assembly optimizations manually forced on.
[14:49:17] - Not checking prior termination.
[14:49:18] - Expanded 733236 -> 3855925 (decompressed 525.8 percent)
[14:49:18] 
[14:49:18] Project: 5101 (Run 0, Clone 108, Gen 96)
[14:49:18] 
[14:49:18] Assembly optimizations on if available.
[14:49:18] Entering M.D.
[14:49:24] Protein: CMF in water
[14:49:24] Writing local files
[14:49:24] rom checkpoint)
[14:49:24] Read checkpoint
[14:49:25] steps  (21 percent)
[14:49:25] Extra SSE boost OK.
[14:49:25] s
[14:49:25] Completed 433765 out of 2000000 steps  (21 percent)
[14:49:25] Extra SSE boost OK.
[15:03:27] Writing local files
[15:03:27] Completed 440000 out of 2000000 steps  (22 percent)
[15:46:25] Writing local files
....

 

[Zero82z]

That's odd, it should have downloaded a new unit. You could try deleting everything in the F@H directory that you can and then restarting, or you could just let this unit finish and hope that your next one is an a2.

 

[Xee]

That's odd, it should have downloaded a new unit. You could try deleting everything in the F@H directory that you can and then restarting, or you could just let this unit finish and hope that your next one is an a2.

lol That's what I plan on doing. Got my fingers crossed.

One last question (I hope), is there any specific way I should go about shutting down the clients (both the SMP and GPU) when I turn the computer off to lessen the risk of messing something up?

 

[Zero82z]

lol That's what I plan on doing. Got my fingers crossed.

One last question (I hope), is there any specific way I should go about shutting down the clients (both the SMP and GPU) when I turn the computer off to lessen the risk of messing something up?

Which GPU client are you using? If it's the tray client, just right-click and select quit. If it's the console version, just close it. If it's running as a service, you can go into Services and stop it manually.

As for the SMP client, the easiest thing to do would be to just suspend the VMs and start them back from where they left off. If you're using VMWare Player, just click the X in the top-right corner and it will suspend the VM. If you're using Server, go into the web interface, select the VM, and in the menu on the right-hand side there should be an option to suspend it.

If you want to shut down the VMs completely (assuming you're using notfred's), you can just power them off and they'll pick up from where they left off. I also recommend configuring them to save checkpoints every 3 minutes.

 

[Xee]

If you want to shut down the VMs completely (assuming you're using notfred's), you can just power them off and they'll pick up from where they left off. I also recommend configuring them to save checkpoints every 3 minutes.

Ah, ok. I just thought I read somewhere that there was something detrimental to setting checkpoints at such a small interval. Can't find the link though.

Also, in regards to the GPU fix you posted above, the only dll files in that directory are amdcalcl.dll and amdcalrt.dll. All the ati drivers are in the following directory:

C:\ATI\SUPPORT\9-5_vista64_win7_64_dd_ccc_wdm_enu\Driver\Packages\Drivers\Display\W76A_INF\B_80160

Should I just skip the step that asks me to delete them and continue on with the guide?

 

[Zero82z]

Should I just skip the step that asks me to delete them and continue on with the guide?

Just delete the two that you have and proceed.