Folding@Home Which client?

[SemperFi]

I just upgraded to windows 7 64 bit and installed an new AMD dual core and unlocked the other 2 cores.

I had vista 64 and ran the GPU for the radeon 3300 which is onboard the Asus board I have. I kind of figured I should use the processor instead of the GPU.

What I am looking for is which should I use? The GPU, the multi core beta client or the normal client with 4 instances running? Can or should you run the GPU along with the regular client?

I was going to use the multicore client but after seeing it has expired I wasn't sure if I wanted to use that one. Whats the consensus if there is one, here?

 

[TonyH]

I'm running a multicore client along with the GPU client on 2 of my systems. I feel that I'm getting my best production from that combination. I would estimate somewhere between 4 to 6k ppd from each one of them. I know it's not much by some standards, but for a couple of old quad cores with 8800 GPUs it ain't bad.

There is an updated multicore client but you have to use the old client to run the installation then drop in the updated client afterward.

 

[GLeeM]

If you can run VMWare Player, notfred's Linux virtual machine is easy to set up and gets better ppd than WinSMP client.

There is also linuxrouter's virtual machine here: http://www.evga.com/forums/tm.aspx?m=4464
that can do the same Linux client WUs (big or regular).

If you have any questions, please ask!

 

[SemperFi]

Thanks GLeeM. I found a thread on the folding forums with vmware how to. I just got it running.

Hey I got another question. I have a windows home server box. I figured putting on a client on it is a no no. What I was thinking is I don't use the video and was wondering about putting on a gpu client on it.

 

[SemperFi]

Hey I just looked in the log file an it says:
Arguments: -local -forceasm -smp 4

Warning:
By using the -forceasm flag, you are overriding
safeguards in the program. If you did not intend to
do this, please restart the program without -forceasm.
If work units are not completing fully (and particularly
if your machine is overclocked), then please discontinue
use of the flag.

Where do I remove the flag at or is it alright at current setting?

 

[biodoc]

You can leave the flag in as long as you have a stable overclock and/or sufficient cpu cooling in your rig.

During configuration if you answer "no" to the following question, then the app will use the SSE instruction sets in your cpu.

Disable highly optimized assembly code (no/yes) [no]?

If you answer "yes", then your performance/PPD will drop like a rock.

FAH apps have built-in error checking which ensures they are collecting quality data from all of us. If an error is detected, the WU will end with an EUE (early unit end) and if you restart the app, it will automatically disable the use of assembly optimizations.

The -forceasm flag forces the use of assembly optimizations. If you are seeing EUEs, then you need to back off a bit on the overclock or deal with poor cpu/case cooling.

You can monitor your cpu temps with coretemp.

Since FAH smp is computation intensive, some of us use it to test for stability when we are overclocking (better than PRIME IMHO).

 

[SemperFi]

Ok, I am with you. I finished my stability tests yesterday afternoon with prime 95 and my cpu temp maxed and stayed at 42c. Right now it is reading 39c and it has been running 6 hours. So should be in good shape I would say.

 

[Golgatha]

If you can run VMWare Player, notfred's Linux virtual machine is easy to set up and gets better ppd than WinSMP client.

There is also linuxrouter's virtual machine here: http://www.evga.com/forums/tm.aspx?m=4464
that can do the same Linux client WUs (big or regular).

If you have any questions, please ask!

I'll 2nd Notfred's VM client. Also a GPU client for each video card works well too since the GPU client uses very little CPU resources.

 

[ksherman]

Also using the NotFred client. Runs great!

 

[GLeeM]

Hey I got another question. I have a windows home server box. I figured putting on a client on it is a no no. What I was thinking is I don't use the video and was wondering about putting on a gpu client on it.

Looks like these questions didn't get answered.
You probably could put a GPU client on it - what GPU does it have?

As far as the SMP client on a server, it uses quite a bit of resources, but runs at the lowest priority so at least the CPU would give way pretty quickly to any other program/service, but if the SMP was using too much ram that would take a little longer to free up (by having to write to disk page file).

Does anyone know if WHS can run F@H?

 

[Uppsala9496]

I'm going to give the vmware linux client a run on my XP machine to see what it can do. Should be an interesting experiment.

 

[SemperFi]

Wow I am impressed. It has been running about 27 hours and already sent one back and is 38% finished with the next one.

I was wondering about running cpu and gpu on the same box. I either asked on here a few months ago or thought about asking a few months ago. I don't think I asked as I don't remember getting an answer. Anyhow I may as well get it going on here also.

The GPU on the WHS is an onboard ATI Radeon HD 3200. I just looked and I haven't installed drivers for it yet, it is using standard drivers. I guess I had better look as there may not be WHS drivers for it. I do have 4 gig of ram so that shouldn't be an issue. AT the moment it is just being used for backup and to serve pictures. I do plan on serving music and video I just don't have it completely setup yet. I will probably just put one or two of the normal clients on the server. It has a dual core. If I only setup 1 that should keep the server responsive. I may setup 2 as it has a lot of idle time at the moment. I guess I will just have to do some testing.

 

[Uppsala9496]

Damn, can't run the notfred vm client on my laptop since it is a T4300 chip that doesn't support VT.

 

[SemperFi]

Can I setup the F@HMon to watch other clients on my network. I have 2 other desktops that are on 24/7. One of them is the WHS. I just installed 2 clients on it.

I should have done this a couple of weeks ago and got in on the Holiday Race.

 

[Uppsala9496]

You can still get in on the holiday race.

 

[GLeeM]

Yes it is not too late to join the race!

FAHMon can monitor those clients. I think either with the IP address or for sure by selecting the network drive and folder.

 

[SemperFi]

I did get the FAH monitor to work on the other clients. I just went and shared their folder and it was rather painless after that.

As far as the race I guess I should go sign up I have another box at work that sits idle most of the day I can hook up also. I used to do it a year or so ago. I just have to carry it home on a pen drive as I am not connected to the net at work.

 

[GLeeM]

I just have to carry it home on a pen drive as I am not connected to the net at work.

I do this too!

It really helps if you can get bigger point WUs, that way you don't have to carry it home as often.
If you have to use the uniprocessor client, I have been getting 905 point WUs from server 171.65.103.162 or .160 by configuring it for "work results greater than 5 MB in size" = yes and using the "-advmethods" flag/switch (which can be configured or added to the shortcut target).
If this doesn't make sense, please ask for detailed instructions!

 

[SemperFi]

Gleem, the only thing I don't understand is getting WUs from a particular server. I didn't think about getting larger WUs. Thats why I don't do it much anymore. I had to bring it home every other day.

 

[GLeeM]

the only thing I don't understand is getting WUs from a particular server.

By using or not using the -advmethods flag you can usually get sent to certain servers. Also, by answering yes or no to >5MB up/download can influence which server you get sent to.
This is not guarenteed, just usually happens.

You must make sure to have different MachineID (1,2,3,4 etc. you can go as high as 8) clients otherwise you can (will) get the exact same WU and you can only get points for it once.

 

[SemperFi]

Ok I see a possible problem. On machine ID that is on just one machine right? I was on my WHS last night and went through the config again and I got a note, I forgot which setting it came upon but it was around installing as service. It said I had the same machine ID. I went through the other instance on the WHS it gave the same notice but one was set at 1 and the other 2.

So what I am asking is on my whs I can have one set at machine 1 and the other 2 then on the kids desktop the ID can be 1 and 2 for each instance or should it be 3 & 4 respectively?

Another question about the large WU. I looked at that machine at work today and saw it only has 512 meg of memory so I left it at normal WU. I was afraid the big WU on that small amount of memory might be a little much. What do you think? I now have a cell phone I can use as modem so it will be a bit easier to sent units. We are shutting down at work from 24 to Jan 4. I may just leave it run and go in and upload every couple of days.

 

[GLeeM]

If you a thousand computers, you can have one thousand number 1 IDs.You can only have eight clients per computer, well at least it used to be that way. With the newer clients the number may be 16. And a virtual machine doesn't count, it is a different OS.You won't always get a BigWU, but when you do they seldom go over 100MB ram used.You could let it run for a while on the computer you download it on and see how much it uses. Each WU has a Project Number and WUs within a Project will behave similarly as far as memory usage, up/download size, etc.

 

[SemperFi]

Thanks gleem that is what I thought. Any idea why when I went through the config on my server which I installed 2 clients as a service using ID 1 and 2 it says I have machine id's set the same? I know they are different and can ignore the problem but just wondering why.

 

[SemperFi]

Ok I think I have a problem now. In the VM Player window it says attempt to access beyond limit of device. Then the next line says hda1: rw=1 want 2096536, limit 2088387. There is a whole window full of them then only thing different is the want is not always the same. I assume I just need to allow more space for the VM, but where do I do that at?

Then the last line in the VM Player says hrtimer: Interrupt too slow forcing clock min delta to 241521330 ns. I don't have a clue on what this problem is. Any of you Linux guys help me out here.

 

[GLeeM]

Any idea why when I went through the config on my server which I installed 2 clients as a service using ID 1 and 2 it says I have machine id's set the same?

If it says you have same IDs, you probably still do. Do you have two different directories, one for each client?

As to the VM thing: Is it still using CPU cycles and increasing frames in the log?
If it is NotFreds, then it is a known thing that the vm window will show full of stuff like you are saying, I think, but it keeps working.

Let me know, I'll try to look here more often.