TR Forums

Ive been planning to start folding for a while now but cant find a clear answer. Which client should I be using for my e8400? The beta, regular or the smp client?

Both the SMP and GPU clients are "beta".

SMP. If you are after points you should run the Linux SMP client under VMware too.

Edit: If a "lowly" E2180 should run SMP, then no reason for the E8400 to be running another.

That depends on a number of factors.. one big factor is the average amount of time you keep your computer on each day. If your computer is on for more than eight hours per day, then the smp client would be a good choice since its work units typically consist of deadlines of 3-4 days. Given their tight deadlines, the smp work units are worth a great deal more in points versus the regular work units. Another thing to consider is that the smp client is still in beta so that means there are a few bugs yet to be worked out, which means the end-user may have to keep an eye on the client on a regular basis to ensure it's working properly. If, however, your computer is usually only on for a few hours a day, or you want a client that you just install and forget about it, the regular client would be the ideal choice. Like I'd mentioned earlier, the work units for the regular client isn't worth as much in points as the smp client due the fact their deadlines are usually stretched out to weeks or even months, which are more suitable for the casual folder, with reduced system uptime.

Feel free to post any further questions you may have and the folks around here will be more than happy to help you along.

thanks for your replys. I'd say the average uptime is anywhere 5-8 hours a day. I think i'll try the smp client for now and see how i do with the first couple deadlines.

Flyingfox im just curious why i would need to run a seperate client under linux. Couldn't i just run 2 under windows?


And i hear they've been working on a gpu client for a while now thats supposed to be pretty productive. Any word on that one? i'd like to put my 3850 to work

What OS are you running, by the way?

The gpu client doesn't support the ATI 3xxx cards, heck it doesn't even support the 2xxx cards yet.

im running vista and im all up and running now
If im turning off my computer and need to restart do i just close the cmd window? And if i start fah.exe again does it just resume where it left off?

They're supposed to be making a new gpu client for 3xxx cards. You cant download it yet but it seems like theyve been working on it forever

Well, that was quick! People often run into issues when installing in Vista, but it seems you breezed through that one.. I suppose you did your homework before the installation? The best thing to do to stop the client is by pressing ctrl + c. When you start up the client again, it should resume from the last check point... I set the save checkpoint to every five minutes so there isn't much work that is lost when the client shuts down.

Yea i read a couple guides and it went smoothly. I set my intervals to 3 minutes.
I think im running too hot though: 60c at 3.6 with the stock cooler

so if i want to start it up again just start fah.exe?

edit
how do i know when my deadline is?

Welcome to the team!

I think your available uptime is going to be the mitigating factor. While the SMP WU's provide better points, they also have short deadlines.

I just brought an E3110 online this past weekend, and am running SMP on it under Vista My current 3050 WU is going to take ~21 hours total, and will come in under the prefered deadline by about 3 hours, but this machine is folding 24/7. The final deadline is 1 day, 20 hours out right now. If the machine was folding 8 hours a day, it wouldn't make it.

thats too bad. I want to give it a go from the start of a new deadline to see if i can finish it reasonably, before going to the regular client. Where can i see the deadlines?

oh and i downloaded a patch along with the latest client, should i just install that over it?

You can look up the points and dealine for each project at Stanford- http://fah-web.stanford.edu/cgi-bin/fahproject. Or run a monitoring application like FahMon - http://fahmon.net/

60C is well within spec, but a little hot for my liking. With a good aftermarket cooler, you can bring the temp down to the mid 30's. I have a Zalman CNPS9000 on an 8400 @ 3.6 and the load temp is about 37C, which is quite a difference from your 60C.

I've a made a shortcut of fah.exe and put it on the Quick Launch bar to make it easier to start up the client. There are severals way to check the deadline of a particular unit.. one being the project summary page on Stanford's website.. another by accessing the 'unitinfo.txt' file in your F@H folder.. and another is by using a utility called FahMon.

Welcome to Team TR Rome32! I hope you enjoy catching the "folding bug". It can be addicting.

Firstly I would like to say welcome to the team.

I'm using an E8200 and I have it running on a VM with the linux SMP client. It's chewing through the WU like no tomorrow. But if you don't want the hassle of installing a VM or don't have enough memory to keep it running then just run the Windows SMP client if you can get it to work.

I getting about 62 to 66 degrees celsius under load so I think it should be ok.

rome32 wrote:

Flyingfox im just curious why i would need to run a seperate client under linux. Couldn't i just run 2 under windows?

As I said, it is for points. The Linux SMP client now is the best performing in terms of speed (and so points per day, ppd, meaning more points for the same amount of time). Now, if you have Windows as your main OS, how can you take advantage of the better Linux SMP client? Run a virtual machine with the Linux as the OS of course. It basically allows you to run Linux inside Windows, behaving like a separate machine altogether.

flybywire wrote:

60C is well within spec, but a little hot for my liking. With a good aftermarket cooler, you can bring the temp down to the mid 30's. I have a Zalman CNPS9000 on an 8400 @ 3.6 and the load temp is about 37C, which is quite a difference from your 60C.

"CPU" temp or core temp?

CPU temp.

That's probably the difference between you two -- he was probably measuring at the core and not the cpu temp.

thanks for all the responses!
the temperature i was measuring was core temp. I know it's within the specs but id like it a little cooler. I was planning on getting an aftermarket heatsink anyway (particularly the new noctua cooler for the performance to noise ratio)

so i switched over to the regular client because i probably wouldn't be able to keep up with the smp deadlines. I know i wont get as many points but its better than nothing i guess.

Its all running again but at around 50% cpu usage (1 core). I suppose there's no way to let it use both cores, so do i install it again under a new directory and let it do 2 proteins at the same time?

Correct. When you run the installer, hit the yes for advanced, and for the 2nd installation, change the machineID to 2 (or something different than the first one).

Im all set then
thanks everyone

edit.. hmm the first client says it extra sse2 boost and the second one only says sse. is it not detecting it properly?

Your first client is running a DGromacs WU which makes very good PPD with SSE2 enabled .. for a single core client. Your second client is most likely a regular Gromacs WU or if you're lucky a big packets (double points bonus) Gromacs WU which at best only uses SSE boost. How are your temps running?

http://fah-web.stanford.edu/psummary.html

that must be it. thanks for the link

for some reason temps are a little better than before, reaching 58c