Want to start F@H...

[bob4432]

how do i configure the program for both cores of cpu and a single gpu?

i d/l and installed the cpu version but only get 1 instance. do i just d/l the gpu version? what about 2 instances of cpu?

assistance appreciated

 

[DragonOrta]

You can only run one client per core, no matter if that client is the CPU client or the GPU client. So you can either run one GPU and one CPU client, or two CPU clients.

 

[bob4432]

Originally posted by: DragonOrta
You can only run one client per core, no matter if that client is the CPU client or the GPU client. So you can either run one GPU and one CPU client, or two CPU clients.

which would be the fastest setup? also, i d/l the gpu one, but when i try to launch the cpu one i get an error about another f@h copy is running. how do i get around this?

 

[Hurricane Andrew]

Not sure about the relative speed of each, but to run two instances via the CPU you need to do the following:

1) Create two seperate folders, for example, C:\Folding1 and C:\Folding2

2) Copy the console version to each directory.

3) Launch the first one, and configure it as you wish, making sure to set the machine number to 1.

4) Launch the second one, configure, and set the machine number to 2.

That's it....you're now folding on both cores!

P.S. Welcome to the F@H TeAm!

 

[GLeeM]

I think that with your X1800XT you should get best points per day if you run one CPU and one GPU.

Yup, as HA says you need to configure one of the clients to a different MachineID number.

If you have ever installed a F@H GUI client you will need to add the -local flag to the shortcuts that start the clients.

Our GPU guru, CupCak3, will help with any questions about the GPU client

 

[bob4432]

Originally posted by: GLeeM
I think that with your X1800XT you should get best points per day if you run one CPU and one GPU.

Yup, as HA says you need to configure one of the clients to a different MachineID number.

If you have ever installed a F@H GUI client you will need to add the -local flag to the shortcuts that start the clients.

Our GPU guru, CupCak3, will help with any questions about the GPU client

will look into changing the machine id for the gpu...thanks

 

[GLeeM]

I see from the other thread that you have installed the GUI.

You will need to add the -local switch/flag to the startup shortcut for any console client.

You can find very good instructions for doing that here at the F@H Wiki.

Just so you don't get confused (I did :roll: ). You can have up to eight MachineIDs per computer.

Did I forget to say "Welcome to the TeAm"!

 

[bob4432]

Originally posted by: GLeeM
I see from the other thread that you have installed the GUI.

You will need to add the -local switch/flag to the startup shortcut for any console client.

You can find very good instructions for doing that here at the F@H Wiki.

Just so you don't get confused (I did :roll: ). You can have up to eight MachineIDs per computer.

Did I forget to say "Welcome to the TeAm"!

thanks for the welcome.

actually i d/l the command client, moved over the cfg file and data file info and work folder into a folder called folding cpu, then moved all the info for the gpu one into a folder called folding gpu. made the gpu machine id 2, cpu one id 1 and both are working now. then put a short to each one on the desktop. seems the gpu one takes quite a bit out of computing, and just out of curiosity, why is it using so many cpu cycles? i thought it worked on the gpu, but if i have it to 100% it will utilize 1 core by itself. i have to run the cpu one at 90% and the gpu one @ 75% or else the machine is very sluggish.

just out of curiosity, how do i see how far along the command line one is?

thanks in advance,
bob

 

[bob4432]

wow - the gpu one says it will take 515days to finish the wu...does this sound right? i have it @ 75%

 

[CupCak3]

the gpu client needs a cpu core for status and data polling. no eta on if or when the cpu usage will be reduced.

the command line client keeps open a command window which prints to screen everytime it completes a frame. I personally really like it and use it on my own GPU client. each frame = 1 % completed

 

[CupCak3]

Originally posted by: bob4432
wow - the gpu one says it will take 515days to finish the wu...does this sound right? i have it @ 75%

it should be fine after it completes a WU

 

[GLeeM]

Originally posted by: bob4432
wow - the gpu one says it will take 515days to finish the wu...does this sound right? i have it @ 75%

It will take a while for it to accurately say how long it will take. Give it an hour or so. It should be less than two days, I think.

The GPU client needs a CPU core because with DX9 the CPU must poll the GPU to keep it supplied with data. (DX10 might not use as much CPU)

You might be better off having the GPU at a higher percent and the CPU lower. I would have to know how long each frame (percent) is taking. You can see this information in the FAHlog.txt file. I also need to know the CPU project number.

If you want to spend a little time experimenting we should be able to tweak your system to get you the most points per day (PPD) and the most science done for F@H.
Edit: And keep your system useable for you.

 

[bob4432]

Originally posted by: GLeeM

Originally posted by: bob4432
wow - the gpu one says it will take 515days to finish the wu...does this sound right? i have it @ 75%

It will take a while for it to accurately say how long it will take. Give it an hour or so. It should be less than two days, I think.

The GPU client needs a CPU core because with DX9 the CPU must poll the GPU to keep it supplied with data. (DX10 might not use as much CPU)

You might be better off having the GPU at a higher percent and the CPU lower. I would have to know how long each frame (percent) is taking. You can see this information in the FAHlog.txt file. I also need to know the CPU project number.

If you want to spend a little time experimenting we should be able to tweak your system to get you the most points per day (PPD) and the most science done for F@H.
Edit: And keep your system useable for you.

no probs experimenting...already looking at a new gpu cooler...d@mn even at the 2d clocks it get to around 70C @ 75%. i live in phx az, and my computer room is ~78F even now.

i also put it on my home server - xp2000 just running @ ~ 35% as it does a couple mysql dbs and is a bit ram starved. been thinking about putting a barton in there, so this might be the reason

also, put it on the laptop but am having issues connecting - the laptop has wifi, ethernet and a egde cingular card...maybe it is getting confused as to which one it is supposed to get data from??? all i get is either the news or cannot connect to server. i did add it to the windows firewall and all should be good, but it is not.

but hey, it doesn't have to be easy becasue i know as soon as i get this running i won't have to mess with it for some time.

fire away with suggestion

edit - for when my computer room gets warmer i went ahead and made another cpu link - i will not run 2 of the cpu and 1 of the gpu, but i will run either 2 cpu or 1 gpu and 1 cpu. anyway, i can't get data from the server from the new link and it is configured the way the others are, so possibly the main server is not up or putting out data. not too worried about it now that the working machine can't get data, i have 3 wu between the 2 machines so i should be good for a couple of days.

without going liquid, any suggestion on a better gpu cooler? i plan to cut a hole on the left side of the case and put in a 80-120mm fan blowing in on the gpu area, don't want to blow up the card....

 

[chris huff]

The Zalman VF900 gets good reviews, and comes stock on a couple of higher end X1950pro's.

You don't mention what case you have, but it sounds like you need to get air out, not necessarily blow more in. So, if you put a side fan in, try it both directions.

If you are not too concerned about noise, you could try a PCI Slot Case Cooler. Though, some of these are getting decent noise reviews.

I had issues with a laptop being able to connect also, but I asked for WU's with no deadline. Turns out, it couldn't connect because right now, there are no deadline less WU's.

On your server, with the console, you can specify how much ram to let F@H use. Might keep it from downloading a memory hog accidentally .

Welcome to the TeAm

 

[bob4432]

Originally posted by: chris huff
The Zalman VF900 gets good reviews, and comes stock on a couple of higher end X1950pro's.

You don't mention what case you have, but it sounds like you need to get air out, not necessarily blow more in. So, if you put a side fan in, try it both directions.

If you are not too concerned about noise, you could try a PCI Slot Case Cooler. Though, some of these are getting decent noise reviews.

I had issues with a laptop being able to connect also, but I asked for WU's with no deadline. Turns out, it couldn't connect because right now, there are no deadline less WU's.

On your server, with the console, you can specify how much ram to let F@H use. Might keep it from downloading a memory hog accidentally .

Welcome to the TeAm

thanks for the warm welcome

the case is an older antec 635 or 835 - about 21-23" tall. i alreayd have 3 80mm fans in the front, 2x80mm in the back, the 1 120mm in the psu and then a zalman 92mm setup on the cpu and also the stock gpu cooler. definately not too concerned about noise. the last time i had all the stuff out of the case i really contemplated about putting in a blow hole, but didn't. probably should have...

i will definately set the ram usage. how littlle is enough? that machine needs an extra 512 stick and a cpu upgrade but i just haven't done it yet..

as far as dealines....how far out are the deadlines? on my main pc and server, if they are out a bit that is no big deal. is there any way to tell it to only get stuff that has a 2wk deadline or so? this way even if/when a rig gets turned off (especially in the summer) i will still have enough time to get them done - don't want to waste cpu cycles and not finish in time.

thanks for the information

 

[GLeeM]

@Heat problem: I bet a fan to move air around or out of the case would solve the problem.

@Home server: Some WUs use more ram than others. Don't try for WUs >5MB download, as these use lots of ram.

Setting the client to 35% will not help free up ram, it all depends on which WU you get.
I read where someone tested different CPU usage percents and there was not much difference between 88% and 25% on most processors (with the client set to idle priority). This was for a system that the user could notice when F@H was running. Most can't notice it running even at 100%. They did not consider heat when testing for best % usage.

@Laptop not connecting: The client uses HTTP to send/recieve WUs. Did you try changing the Use Internet Explorer Yes/No setting?

I think the only time we couldn't get WUs from Stanford was when the whole university network was down or they didn't have electricity.

 

[GLeeM]

Originally posted by: bob4432
i will definately set the ram usage. how littlle is enough? that machine needs an extra 512 stick and a cpu upgrade but i just haven't done it yet..

as far as dealines....how far out are the deadlines? on my main pc and server, if they are out a bit that is no big deal. is there any way to tell it to only get stuff that has a 2wk deadline or so? this way even if/when a rig gets turned off (especially in the summer) i will still have enough time to get them done - don't want to waste cpu cycles and not finish in time.

The client checks how much ram there is and reports this to the Assignment Server so that it can send you a suitable WU. You can set how much ram you want the client to tell the AS you have.

I'm not sure how low you can go before you won't get a WU, but you could try 32 or 64 and see how it goes.

You can see all the projects and the deadlines here.
There are two deadlines, after the Preferred they will re-issue the WU to someone else, and after the Final Deadline you will not get any credit (points).
The WUs simulate a protien forming. Your WU results need to be returned before the next one can be made and sent out to someone. This is the reason for the Preferred deadline. The quicker the WUs are returned the sooner the scientists can see if the project will help find a cure.

This is the reason F@H is making new GPU and SMP clients. The GPU really speeds up the process, instead of 1-2 years of crunching it may only take a few months to finish a project.
The SMP uses up to 4 cores/CPUs to make things 4+ times faster. Actually it's probably much faster than that, because only the fastest computers will have 2-4 cores. Whereas the regular client WUs get sent to all computers even the slowest P3s and those being run 8 hours 5 days a week.
The SMP is, at this time, only for Mac/Intel and Linux.

 

[winr]

Welcome "bob4432" :sun:

 

[bob4432]

thanks all

been carnking away. when i am using the machine i run 2 on the cpu, then when i go to bed i fire up 1 on the cpu and one on the gpu - man the gpu pumps out a ton of heat and slows the machine down even at 75%. it this normal with the gpu setup?

i had to set the server machine down because it was getting pretty warm. as soon as i want to tackle the swap i will be doing that, or possibly swapping out the server completely to a 939 dc so it will basically last me for 8+yrs that way. decisons, decisions....

one more question - do you sign up and give f@h your email address? i signed up under my older username and it still had my stats (bad stats) but i can't remember if there was a place where you kind of had a acct with them other than the name, but i might be wrong, maybe it is just name only.

 

[Insidious]

You don't have to sign up to F@H.

Just give your chosen username and team number when you install/configure.

If you want to post in the Stanford F@H forums, you have to sign up for that.

-Sid

 

[GLeeM]

Originally posted by: bob4432
man the gpu pumps out a ton of heat and slows the machine down even at 75%. it this normal with the gpu setup?

On some setups with 256MB video memory lag is normal.

I think it was Athlex that had no lag with a 1950 256MB AGP card.

You'll have to wait for some numbers to see if it is worth the bother of starting up the GPU nightly or stay with the convienence of two CPU clients.

What kind of frame (percent) times are you getting on what projects?
An average of several frames to the closest minute would give close enough PPD, but average with minutes/seconds would be more accurate.

 

[bob4432]

Originally posted by: GLeeM

Originally posted by: bob4432
man the gpu pumps out a ton of heat and slows the machine down even at 75%. it this normal with the gpu setup?

On some setups with 256MB video memory lag is normal.

I think it was Athlex that had no lag with a 1950 256MB AGP card.

You'll have to wait for some numbers to see if it is worth the bother of starting up the GPU nightly or stay with the convienence of two CPU clients.

What kind of frame (percent) times are you getting on what projects?
An average of several frames to the closest minute would give close enough PPD, but average with minutes/seconds would be more accurate.

not sure about the cpu speeds because i am running the console version, but gpu it says "Working on: Protein GROMACS core" and it is usually 6-8s/frame

 

[CupCak3]

the frame speeds vary per project... now they are usually a 4 digit number. with the project number and frame time you can calculate the points per day your computer will earn