MPICH

[lehmann]

I followed the guide all seemed to work except its not running all cores only using 25%
Now in the config username Liz_Lehmann , team33, passkey what is a passkey the guide said fill it in? Im running 32bit XP

 

[Axdrenalin]

http://folding.stanford.edu/English/FAQ-passkey

 

[lehmann]

sense that doesnt seem to matter. any clues why smp isnt working

 

[The_Heretic]

Physically add the -smp flag to the command line/shortcut. I realize it's supposed to be the SMP client and that seems silly but... I did the same thing when I first set it up.

 

[lehmann]

ok but you will have to explain what that is and where

 

[The_Heretic]

For instance I set up a shortcut to the .exe for the SMP Folding client. Like we used to do in the old days we could add flags after it for things like -advmethods, -verbosity to log more often, etc. In this case just add -smp to utilize it. exmple of mine is below (my shortcut)

"C:\Program Files\Folding@Home\Folding@home-Win32-x86.exe" -smp -verbosity 9


after the Win32-x86.exe" add a (space) -smp flag

 

[lehmann]

it says smp in target path is not valid you would think after 10 years stanford could make a half harted attemp at a good installer

 

[The_Heretic]

Copy and paste your shortcut text here please and lets have a look

you would think after 10 years stanford could make a half harted attemp at a good installer

You'd think they would make their SMP client automatically do SMP (don't get me started on those guys)

 

[lehmann]

Shortcut to Fah6.exe -smp flag____________________ im getting a headache lol all for a few more points. i must have something wrong in the config

 

[Tigerbiten]

The client needs to have downloaded FahCore_a1.exe or FahCore_a2.exe to be running a SMP work-unit.
Any other core means its running running a standard work-unit, hence only 25% useage.

Due to running the Linux client inside VM's, its been over 2 years since I last installed the Windoze client.
Better PpD that way.

Luck ..............

 

[Zero82z]

You'll have to create a new shortcut to the Folding@Home .exe file, and then add the -smp flag to the end of its target. Navigate to the directory containing F@H, right-click on the executable, choose Send To>Desktop (create shortcut), then edit the shortcut and add the flag.

 

[lehmann]

im redownloading netfram 2.0 just to make sure

 

[lehmann]

Im finished trying this. friggen junk. ill try the vm thing now. im guessing more headache

 

[Kendrak]

Im finished trying this. friggen junk. ill try the vm thing now. im guessing more headache

nope, easy as pie.

 

[The_Heretic]

Yeah their dang SMP client is not only a pain to set up, but tends to crap work units here and there. They definitely need to work on it.

 

[Kendrak]

Yeah their dang SMP client is not only a pain to set up, but tends to crap work units here and there. They definitely need to work on it.

That is why the new client is on it's way. We should be seeing it sooner rather than later.

 

[zero2dash]

I have no problems with the MPICH SMP console client; none at all.
If you set it up (following Xilikon's guides) it works just fine. The only thing his guide doesn't mention is replacing the exe with the drop in updated exe off Stanford's page (and that link is directly above the SMP clients links and pretty hard to miss).

 

[lehmann]

I have no problems with the MPICH SMP console client; none at all.
If you set it up (following Xilikon's guides) it works just fine. The only thing his guide doesn't mention is replacing the exe with the drop in updated exe off Stanford's page (and that link is directly above the SMP clients links and pretty hard to miss).

shouldnt that be below?

 

[AgrFan]

save yourself the headache ... don't bother with the Windows SMP client ... either run notfred or CPU/GPU2 clients until the SMP2 client is released ... consider yourself warned.

 

[zero2dash]

shouldnt that be below?

Sorry, yeah it's below.

save yourself the headache ... don't bother with the Windows SMP client ... either run notfred or CPU/GPU2 clients until the SMP2 client is released ... consider yourself warned.

And why would that be?
I really don't get this...all this animosity towards the SMP clients, anything other than a VM. Some people around here act like the SMP clients don't work, they never work, they never will work, and you'd be better off spending more time farting around with the VM player making a VM client or just not Folding at all. Really, it just absolutely fucking baffles me.

If people want to say "you get more PPD off the VM SMP than the Windows SMP", I can understand that, and that is the truth (although the difference from one to the other is not as huge of a leap as some people would make you believe). Then there's other comments like this one..."consider yourself warned". Why? Does the Windows SMP client murder death kill any desktop it's running on, wiping your entire HD and copying itself to your address book and mass-emailing itself to all your friends so their systems become infected as well? Come on man.

It took less than 5 minutes to set up the MPICH SMP client, I've done it on 3 computers, and all 3 computers are still running the MPICH SMP client since day 1 and the client has never failed. I don't get all this FUD in regards to the Windows SMP client(s). The VM solution is not for everyone. The SMP client does not take a gig of ram off my machine. Between the two cpu clients, I prefer the SMP one. I really wish people would quit acting like Folding SMP is a lost cause and not worth the donation. DEIMOS may be a shit client, but I can vouch for MPICH. I have to use MPICH because I'm under x64, but it has worked flawlessly for me.

 

[APOLLO]

And why would that be?
I really don't get this...all this animosity towards the SMP clients, anything other than a VM. Some people around here act like the SMP clients don't work, they never work, they never will work, and you'd be better off spending more time farting around with the VM player making a VM client or just not Folding at all. Really, it just absolutely fucking baffles me.

The SMP client for Windows does work, I had it running on half a dozen machines at one point with over a dozen clients. Mind you, that was with the older 5.x client and don't have any experience with the newer version that is apparently more troublesome.

If people want to say "you get more PPD off the VM SMP than the Windows SMP", I can understand that, and that is the truth (although the difference from one to the other is not as huge of a leap as some people would make you believe).

Agreed for the A1 WUs but only Linux SMP will grant you access to the A2 WUs and there is where the big difference lies. You are looking at half the frame times with nearly the same point value. That is comparing Linux A1 to Linux A2. Windows A1 to Linux A2 is an even bigger difference PPD-wise. In the past I would have agreed with you, but since the A2 WUs have been released it's a very different SMP ball game.

Then there's other comments like this one..."consider yourself warned". Why? Does the Windows SMP client murder death kill any desktop it's running on, wiping your entire HD and copying itself to your address book and mass-emailing itself to all your friends so their systems become infected as well? Come on man.

I believe the comment was made to spare people a lot of headaches. It doesn't mean that's inevitable, but if you do a search you'll see that many have complained and given up on it after running the client for a while. I even gave up on it on a mchine I could not run Linux since it was 32-bit. I eventually replaced the Windows SMP client with an AGP Ati client and that produces triple the PPD of the SMP client anyway.

It took less than 5 minutes to set up the MPICH SMP client, I've done it on 3 computers, and all 3 computers are still running the MPICH SMP client since day 1 and the client has never failed. I don't get all this FUD in regards to the Windows SMP client(s). The VM solution is not for everyone. The SMP client does not take a gig of ram off my machine. Between the two cpu clients, I prefer the SMP one. I really wish people would quit acting like Folding SMP is a lost cause and not worth the donation.

TBH, unless you are running a 32-bit machine or very low on memory resources, you can at least try the Linux SMP client in a VM. The only other situation I can think of that might describe a bad scenario to run VMs is a machine that sees a lot of CPU-intensive use.

 

[zero2dash]

VM seems like a bit PITA.
I'm not aversed to PITA projects, I just don't necessarily seek them out.

Maybe I'll give it another look. It just looked like the initial configuration was such a pain.

re: cpu intensive stuff, I do actually do some cpu intensive stuff (graphic design), but I rarely ever notice any hiccups in CS4 even with the MPICH SMP client running at the same time. I'm sure I'm nitpicking here on ram use of a VM (complaining about 1gig taken up, since my (primary) q9450+gtx260 system has 8gigs of ram), but that system is my design workstation, basically.

 

[APOLLO]

VM seems like a bit PITA.
I'm not aversed to PITA projects, I just don't necessarily seek them out.

Maybe I'll give it another look. It just looked like the initial configuration was such a pain.

Here's where we totally agree. It IS a pain to install a VM with Linux for the first time. About 10-20 times more of a pain than Windows SMP will ever be. But, it's only a pain the first time or first few attempts because it's extremely different than anything you've ever done before. It gradually becomes easier the more times you do it.

First, you have to get a handle of VMware although that is somewhat easier now with Player. Second, you have to get accustomed to Linux; that's where the big hurdle is to overcome, at least it was for me. The upside is: you get to learn a completely different OS and working environment, hone your computing skills and be rewarded for all the effort with much greater production.

re: cpu intensive stuff, I do actually do some cpu intensive stuff (graphic design), but I rarely ever notice any hiccups in CS4 even with the MPICH SMP client running at the same time. I'm sure I'm nitpicking here on ram use of a VM (complaining about 1gig taken up, since my (primary) q9450+gtx260 system has 8gigs of ram), but that system is my design workstation, basically.

If you work with a lot of CPU-intensive applications or are utilizing large amounts of RAM, I won't sugar-coat VMs - they take everything you allocate for them. Meaning, if you allocate 2 cores and 1 GB of RAM, those hardware resources have effectively disappeared from your Windows environment. It is like instantly being transported to a much older machine with less cores and less memory available. However, you can always shut down the VM for the time the additional resources are required.

 

[Zero82z]

Here's where we totally agree. It IS a pain to install a VM with Linux for the first time. About 10-20 times more of a pain than Windows SMP will ever be. But, it's only a pain the first time or first few attempts because it's extremely different than anything you've ever done before. It gradually becomes easier the more times you do it.

First, you have to get a handle of VMware although that is somewhat easier now with Player. Second, you have to get accustomed to Linux; that's where the big hurdle is to overcome, at least it was for me. The upside is: you get to learn a completely different OS and working environment, hone your computing skills and be rewarded for all the effort with much greater production.

That's also only if you install your own custom distro. If you use notfred's, it is just as easy and simple to set up as the Windows SMP client.

 

[APOLLO]

That's also only if you install your own custom distro. If you use notfred's, it is just as easy and simple to set up as the Windows SMP client.

For most people, yes. I was never able to get notfred working with my systems. There were always crashes and I tried it on several different machines with varying hardware configurations. Another reason I don't like notfred is its pure terminal environment. I prefer a minimal GUI for other tasks, if ever there was need of it. My distro isn't custom; it's an Ubuntu-based distro that runs a very low resource GUI. Resource-wise, it's not too far off from running notfred but you get a GUI.

 

[Zero82z]

By custom, I didn't really mean the distro itself, I meant that you're customizing your own VM installation. I suppose I should have worded that differently. I never had any major issues with notfred's, but I prefer the control that a full Linux installation gives me.

 

[APOLLO]

I never had any major issues with notfred's, but I prefer the control that a full Linux installation gives me.

I'm the only person in recent memory that has experienced these kinds of issues with it. I still can't figure out why. I thought that maybe it was just a bad version I downloaded but I tried a more recent version afterwards and although it went smoother, it still eventually crashed. No distro (and I tried tons of them) has ever caused me that many problems. For a folding-only version of Linux, I accomplished less folding (never completed a WU) on notfred than with any other distro.

I would like to emphasize that this is just me, so please no one take my experience as doctrine. I don't want to apply undue criticism on something that seems to have affected my installations and no one else's.

 

[AgrFan]

I really don't get this...all this animosity towards the SMP clients, anything other than a VM. Some people around here act like the SMP clients don't work, they never work, they never will work, and you'd be better off spending more time farting around with the VM player making a VM client or just not Folding at all. Really, it just absolutely fucking baffles me.

If people want to say "you get more PPD off the VM SMP than the Windows SMP", I can understand that, and that is the truth (although the difference from one to the other is not as huge of a leap as some people would make you believe). Then there's other comments like this one..."consider yourself warned". Why? Does the Windows SMP client murder death kill any desktop it's running on, wiping your entire HD and copying itself to your address book and mass-emailing itself to all your friends so their systems become infected as well? Come on man.

It took less than 5 minutes to set up the MPICH SMP client, I've done it on 3 computers, and all 3 computers are still running the MPICH SMP client since day 1 and the client has never failed. I don't get all this FUD in regards to the Windows SMP client(s). The VM solution is not for everyone. The SMP client does not take a gig of ram off my machine. Between the two cpu clients, I prefer the SMP one. I really wish people would quit acting like Folding SMP is a lost cause and not worth the donation. DEIMOS may be a shit client, but I can vouch for MPICH. I have to use MPICH because I'm under x64, but it has worked flawlessly for me.

I'm glad your experience with the WinSMP client has been a positive one. Unfortunately my experience with the earlier versions was not. I did experience slow performance and lost work due to server and network glitches as have many others. Search the forums and read the posts. Having a sticky describing the 'qfix' process to return crapped out units should be enough of a warning flag. Pande Group states on the download page the SMP clients are beta software and are meant for people willing to deal with issues. The OP needs to be aware of this fact.

I run notfred off a USB stick. I do not use VMs since running notfred off a USB stick works perfectly fine for me. Other than the recent dual-core issues with the 2.10 core I've had zero problems with the Linux SMP client. The SMP client was never meant to be run on Windows in the first place. It's history has been well documented.

If you're getting close-to-equal production from WinSMP (A1 WUs) and VM SMP (A2 WUs), you need to go back and check your numbers. That is simply not true. The A2 units scale much better and produce twice as many points. This is also well documented. Yes, it's all about the points here.

Finally, I find it hard to understand why someone would be willing to go through all of the hassle of getting a fundamentally flawed client running when a totally redesigned version is waiting in the wings to be released. That just doesn't make sense to me. If you think that's a good idea, OK. But being a software developer myself, I don't. My plan is to continue running notfreds (1 dual) and CPU clients (2 duals, 2 singles) until SMP2 is released.

I will probably move all of my boxes back to WCG once the DDDT-2, FLU-2 and CEP-2 projects are released over the next few months anyway. Yes, it's all about badges. But that's another story ...

 

[Mr. Pedantic]

I really don't get this...all this animosity towards the SMP clients, anything other than a VM. Some people around here act like the SMP clients don't work, they never work, they never will work, and you'd be better off spending more time farting around with the VM player making a VM client or just not Folding at all. Really, it just absolutely fucking baffles me.

Because the WinSMP client is, at best, finicky. At worst it's just a downright pain. The Notfred's SMP client is much easier to set up, it doesn't just dump WUs left, right, and center whenever you want to turn the computer off, and it gives over double the ppd.

 

[zero2dash]

Finally, I find it hard to understand why someone would be willing to go through all of the hassle of getting a fundamentally flawed client running when a totally redesigned version is waiting in the wings to be released. That just doesn't make sense to me.

Because I can run MPICH SMP and not lose ram. The computers I have the clients on are in use 85% of the time the client is running, so a performance hit like losing ram would be clearly noticeable. It'd be different if these were headless unused boxes just running the clients and nothing else; if that was the case, I'd bother with the VM's. Sure I can install the VM, shut the VM down when I need the ram, and then start the VM up again when I don't need the ram. I can also then hear my wife and daughter complain about the computer being slower than normal or getting pissy that I keep kicking them off to start or quit the VM.

Very simple with MPICH...it just runs 100% of the time, neither of them notices it. Different strokes for different folks.

 

[The_Heretic]

I run the WinSMP client on one box (WinXP 32bit) because when I run Notfred VM on it it sucks my GPU clients down to silly numbers, despite trying Riva tuner, and apps to reassign affinity.

 

[APOLLO]

Because I can run MPICH SMP and not lose ram. The computers I have the clients on are in use 85% of the time the client is running, so a performance hit like losing ram would be clearly noticeable. It'd be different if these were headless unused boxes just running the clients and nothing else; if that was the case, I'd bother with the VM's. Sure I can install the VM, shut the VM down when I need the ram, and then start the VM up again when I don't need the ram. I can also then hear my wife and daughter complain about the computer being slower than normal or getting pissy that I keep kicking them off to start or quit the VM.

In that scenario I can fully understand. VMs are an enormous resource black hole. If you and your family really needs the RAM and cores, better stick with Win SMP then. Hopefully a newer better version will be released soon and maybe Stanford can extract some additional PPD out of the client as well.