Judas

Tipstaff, no i haven't gotten it working..

Disabled crossfire..

Set GPU 0 and GPU 1 or 2 or 3 with the other monitor plugged in with extended desktop results in a conflict. and automatically shuts down one client and totally freezes up the other previously running one.

__________________

Tipstaff

Ouch. That sounds like the client is not directing itself to use the second GPU, and instead is trying to use the same GPU that the other client is using.

Hmm.. Clients are all in seperate directories, right? Just to lay it out (you probably already know/tried this, but I may as well lay it out for anyone else wanting to try it, so please forgive the way I'm going to desribe this.. lol):

- I'm running the Multi-GPU client (THIS one) in seperate folders called F@H GPU 0, and F@H GPU 1.

- I have 2 shortcuts to each client within their respective folders. Each is setup like this: One says

"C:\F@H GPU 0\FAH5.91beta-console.exe" -local -configonly
"C:\F@H GPU 1\FAH5.91beta-console.exe" -local -configonly

which will run the client locally for configuration, and the other

"C:\F@H GPU 0\FAH5.91beta-console.exe" -local -gpu 0
"C:\F@H GPU 1\FAH5.91beta-console.exe" -local -gpu 1

which will run and direct each client to their respective GPUs, but each from their local folders only.

One thing I've found is that sometimes the IDs that Windows shows can get moved around. Say you look at your monitor setup, and it shows monitors 1 through 4 (1 will be active, 2-4 will be greyed out). When you connect the 2nd monitor to the other card and extend the desktop it will show up as 1, 3, 2, 4. However, one time after rebooting Windows reorganized their priority so that it showed up as 3, 1, 2, 4 (I don't know if this is a glitch in the Catalyst software, but sometimes I'll get a message from it that says it's "optimizing the displays for multi-monitor support" or some such). Windows didn't switch monitor 3 to my primary desktop monitor, but just reoranized the layout. Kinda wierd. Thing is that when you try to run the clients when it's setup this way the exact thing you described happens, and the only way to correct it is to remove the extended display option, and reenable it so that Windows puts the order back as 1, 3, 2, 4.

BTW, just incase anyone was wondering, Catalyst 6.11 drivers work just fine with the client.

__________________

Judas

yeah i've checked that.... it seems to detect that it's using the right GPU for each client.. but then it conflicts on System ID or something.. i posted screenshots back a ways in the thread on how i set it up.... i'm going to try the -local again.

__________________

Tipstaff

BTW, don't bother loading it as a service. However, if you do, then the -local setting needs to be added manually in the registry (I think you also need to setup the -gpu x as well). Otherwise it will load each client to run on the same GPU.

Of note, for best peformance, when you setup the client have it set to use 95% of the CPU power, and have it run in LOW priority. Since it's best not to run the GPU client while gaming you are most likely going to quit the client to do that, so having it in IDLE mode is kinda useless.

One thing I've yet to try setting is for it to run large WUs, or even the advanced WUs as the cores that run on the GPU are not the same as the CPU WUs. As such I have no idea what effect those 2 settings might have.

__________________

Blazer

How will the winner be determined? will it be a randam drawing? or how many units you have processed? or your score after december 1st?

__________________
ASUS Maximus Formula, OCZ 780MXS ps,
Core2 Duo 8400 @ 3.80GHz,Corsair Water cooling kit, 4x2gb Mushkin DDR2 8000, 5 hardrives totalling 1.75 TB, Sapphire Theatrix, SyncMaster 2343, Logitech MX-518, 2 Sapphire Radeon HD 4870 Crossfired , running on Windows7 64 bit, all in an Antec P160w Case.

Tipstaff

Random draw. So, as long as you complete at least 1 work unit, and the F@H site shows it, then you're in.

__________________

Erroneus

Here is a tip, dont bounce your laptop around :O

This morning i dropped midly my laptop on to my bed, nothing hard or anything, just a small drop. IBM's harddisk monitor application must have kicked in and locked the drive for 30 sec or so. In those 30 sec, I'm guessing FAH needed to write some data, but it couldn't, so it ended my WU before time

DOH!

__________________

Peter123

I noticed that there is a new GPU client out (still beta).
It now officially supports all X1600 and higher cards!
The original GPU client will apparently expire Dec 1st.

Cba - Andy

Oh bum, uve got 2 finish 1 Wu, if only i started sooner. I must of got assigned 1 of the big WU's 20000 frames and wont be done till the 2nd 11000 done

__________________

BWX

I got one like that- it only took a couple days.. get fahmon 2.1.3 and let it load the same time as the f@h client and wait.. it will give you a different time that it will end (much sooner I would think).

__________________

Peter123

Congratulations Dagski!

Lelisevis

LOL The folding people must think im special, they keep throwing some crazy projects my way, latest one is 20,000, exactly the same with my poor old desktop. It'll take me about a month to complete the 2 of them.

dagski

i have linux(RedHat9) folding client on a PIII 1GHz, i keep getting gromacs wu's @ 2000000 each. but when its done(after 3 days or so ), the FahCore hangs up, & never get to send the result. do u guys know a workaround or something? btw,i ran the client with no option.


@peter123: thanks!

__________________
Recursive traversal of loopback mount points.

Tipstaff

I take it that you have alot of outstanding WUs then that haven't been sent in? One thing you might want to try is adding the "-send all" command line switch, so that when you run the client it will send all completed WUs. It should exit when it trys to send an incomplete WU.

The problem you are having could be due to the server the client is trying to send data to is either busy or offline (take a look at THIS page, and see if the server is online/offline), or it could be due to some firewall issue or incorrect internet TCP setting. It should be noted that even if you can receive WUs perfectly fine the client may not be able to send them as F@H uses 2 TCP ports: remote port 80, and local port 8080. It could be that something is blocking one of the 2 ports.

If anything I would post within the F@H forums as one of the team members will be able to help you get things sent, and possibely figure out how to fix this from happening again.

__________________

dagski

i can access ports 80 & 8080 thru my proxy perfectly well.this is what i got after i killed the client & query the queue.

suffice to say. the server(171.65.103.160) is currently full.

__________________
Recursive traversal of loopback mount points.