EggXpert

but... but.... I'm already folding for Tech Report....

Ronin:

Ermm....all the research I've done shows the contrary...can you support your findings aside from point values?  I'd be interested to research further.

Sure...

"I'm running a 2610 project at ~12min/frame on my e6600 @ 3GHz. That equates to ~1825PPD."

"have a Core 2 Extreme QX6700, running Windows Vista 32bit. From what I can compute, about 2600 PPD, using 2610 project as a reference. "

Source: F@H Forums

"We balance the points based on both speed and the flexibility of the client. The GPU client is still the fastest, but it is the least flexible and can only run a very, very limited set of WU's. Thus, its points are not linearly proportional to the speed increase.... We have picked the PS3 as the natural benchmark machine for PS3 calculations and set its points per day to 900 to reflect this middle ground between speed (faster than CPU, but slower than GPU) and flexibility (more flexible than GPU, less than CPU). " 

Source: PS3 Faq 

 Hope this helps..

 Edit: I think the confusion arised here where clearly the PS3 is doing much more than the computers on a floating point operation scale.. But keep in mind 95% of those computers don't have E6600 or Q6600 processors in them.. They're just classic old computers that people haven't upgraded yet and have been sitting around for half a dozen years.. I have 5 computers and none of them individually are anywhere near as fast as a PS3. In fact, all of them combined give barely 440PPD.. Since none of them can run the SMP client which gives a huge bonus for having a computer with a ton of RAM (the kind of workunits the PS3 is terrible at) and submitting the workunits within the 72 hour SMP deadline.

 

my e6600 (not over clocked) is doing a frame at ~ 14.5 min (for the same 2610 project).  Don't have a PS3... and since I don't think they can run the 2610 project... ya really cant compare them directly, but if they aren't taking advantage of the SMP that obviously they could were the code set up for them... then they clearly can't push the workload.  In addition, the big reason that it appears like the PS3 is pushing so many points out.. is not very many peeps that are running it on the PS3, apparently, can figure out how to change their name/group to something other then what it comes standard for the PS3.

 

Just noticed another thing while reading the PS3 page on F@H.... it says early in the article that they only expect users to run the PS3 for folding about 8 hours per day... so that 900 ppd is, I would assume, using that 8 hours as their day length.  Now if you were to run a 24/7, or nearly so, you would perhaps see your ppd nearly triple... unless of course I am wrong.. in which I have been in the past...  

Fratricide:

Just noticed another thing while reading the PS3 page on F@H.... it says early in the article that they only expect users to run the PS3 for folding about 8 hours per day... so that 900 ppd is, I would assume, using that 8 hours as their day length.

It's not.. the 900 is based on 24/7 folding..   It takes 8 hours of processing for the PS3 to complete a single PS3 size workunit.  3 work units = 900 points per day (ppd).

tesla120:

but... but.... I'm already folding for Tech Report....

 

switch now or buy your pc parts from another vendor haha just kidding. But really....switch teams

*sigh* Ill have to think about it.... maybe after my next 2 WUs are done....

ive totally got into folding and for poeple concerned about low end systems, dont

my old P4 and celeron systems are turning WU's just fine. even if your not getting fantastic scores your still helping

sure we like to see great scores but it isn't really a race, every bit helps.

my X2 media server is handling SMP client within deadlines as long as i keep the DIVX work down, thats the main thing that kills wu"s

so i am doing divx on my P4 since the X2 is doing the deadline folding

everybody should start there should be hundreds of us neweggers and eegxperts doing this now 

Any CPU that supports SSE will work well.. This includes any Pentium 3 or later, and the socket Celerons (the generation after the Slot P2/P3s). The Slot Athlons (600 Mhz or so) don't have SSE, so don't even bother with them since they're not even worth the power to run.  btw, if you have a P4 with multithreading do not try to run the SMP client and do not try to run 2 clients on the chip..folding@home does not get a speed boost from hyperthreading since hyperthreading does not boost floating point operations per second. So running 2 clients on a P4 will cause both clients to take twice as long to finish a workunit! You'd want to run just one regular client on a P4.

Folding@home loves multiple cores because each core has its own dedicated SSE unit so the more the better! The current Beta SMP client is designed for part-time quad-core processors however many have found that if you leave your dual-core running 24/7 then it will complete the SMP workunits with a bit of time to spare. It appears the future plans are to have an SMP client with workunits that specifically target 2-core, 4-core, 8-core, and 16-core systems using anywhere from 32 megabytes to 4 Gigabytes of memory depending on the resources you allow folding@home to use.. So no matter how high end of a system you get, folding@home will be able to make full use of it.
 

If you run the smp on a dual core there isnt much else you can do with that rig if you want to complete the unit on time. My d805 and opti165 only have about 18 hours to spare.

You only need 512MB per core btw

yeh im looking at 12 to 18 hours to spare, fortunately simple streaming and playing of video does not seem to slow it down to bad, just the high cpu stuff like divx, but i canpass that on to the P4

Where do I see the deadline? Im running the smp client.

Grab FAHMon, it's a great program for monitoring multiple computers. It will give estimates of when the workunit will be finished and will tell you when the preferred and final deadlines are. You can use it to monitor multiple clients on the network, you will need read to all the FAH directories for it to work. You can also check unitinfo.txt for deadline information.

 

 

Thanks for making me aware of this - just signed - my questions to you concern the processor useage - I have a dual xeon dual-core file server (basically, a quad-core) both processors running @2.4 - right now I have it set to higher processor useage, and thread affinity set to processors 2&3 (my applications processor).  What would the recommended settings be to maximize effectiveness of my server for folding without interfering with its file server role?

How much RAM does your server have? If this is an absolute misison-critical server that needs 100% uptime for years straight, then I wouldn't recommend you run folding@home on it at all. It won't ever cause your server to crash, but it will raise the heat generation by quite a bit and combined with other factors may not be a good idea for a server that needs to go years at a time without a reboot.. (Such as a webserver)..

 
If you want to minimize the amount of memory that folding@home uses, I'd suggest you install four seperate copies of the regular console client as a service, set BigWU (BigWorkunits) to NO and then just forget about it..  folding@home will use somewhere between 8-32 Megabytes of RAM (total for all 4 copies). The problem with this method is that you won't get many points for it and your scientific contribution is much lower than it potentially can be..

If you're lenient on the memory usage of folding@home, I'd suggest you do the same as above, but set BigWU to yes..  folding@home will use somewhere between 32-512 MB of RAM total (for all 4 copies installed).  IF you do this or the above, then make sure you actually install four separate copies, each with a unique MachineID (so they don't get assigned the same work), and each in their own seperate directory.  

I do NOT recommend you use the SMP client at all on a server. It's still Beta, not only is the client in Beta, but the Workunits and SMP servers are *also* in Beta.. It's quite common for individual threads in the SMP client to crash and close down leaving the client idle for long periods of time.. You need to closely babysit the Beta SMP client because it still has so many bugs in it.. Which is why I can't recommend it for servers.  But if you keep an eye on it, it's the best way to earn the most points and make the most scientific contribution.. You *must* enable BigWU for the SMP client to work and it will use anywhere from 256-2048 megabytes of RAM.

The folding@home client only runs at idle priority (unless you change it which is not a good thing to do ever)..  So it won't get in the way with CPU usage, but it might get in the way with memory usage.. Just something to watch out for. 

Hope this helps..

Happy folding..

Let me clarify a bit - this server is for home use, and is only being used as a seldom-accessed FTP server and a daily use file server.  It is running Server 2003 R2 with 1GB of RAM.  I have set up the Windows Graphical client for folding.  It is only running 8MB onboard Video (it is a server, after all).  Is that the best folding client to run under such a setup?  I am not too worried about folding interfering with my file server setup so long as it is set up correctly, but I am unsure of what the correct setup is in the folding.  This is the first time I have run a shared processing program, so I want to apply as much of my processor as I can to the folding, but I am unsure how to do that without it interfering with my other minimal tasks the server runs.

In addition, would a server with dual P3 processors with 3GB of RAM be effective for folding if that was the only application running on it?  Not to sound like a needy n00b full of questions, but I am.  again, the other server would be running only 8MB of vram.

Any advice on setup for optimization of folding setup would be much appreciated.

my 1 cent

my X2 media server is cranking out the SMP client, it finishing one every 2-3 days (4 day deadline) i had to move DIVX task off the machine but serving media to my other computers doesn't hurt production hardly at all.

dont know about the P3's, no experience there 

18M:18s/frame - not a clue what this translates to - any hints?  I tried searching the folding forums, but got lost in there without finding anything

rhofmann:

Let me clarify a bit - this server is for home use, and is only being used as a seldom-accessed FTP server and a daily use file server.  It is running Server 2003 R2 with 1GB of RAM.  I have set up the Windows Graphical client for folding.  It is only running 8MB onboard Video (it is a server, after all).  Is that the best folding client to run under such a setup? 

Negative . The graphical client will only use one core.. So you're only using 1/4th of your total CPU capacity.  

rhofmann:

I am not too worried about folding interfering with my file server setup so long as it is set up correctly, but I am unsure of what the correct setup is in the folding.  This is the first time I have run a shared processing program, so I want to apply as much of my processor as I can to the folding, but I am unsure how to do that without it interfering with my other minimal tasks the server runs.

It won't interfere.. If you're sitting at the server or logged into the server a lot, then go ahead and run the SMP client.. But it will sometimes just stop and you'll have to manually shut it down and restart it.. And perhaps even send a copy of the log over to the folding@home forums so they can help diagnose the problem.. So the SMP client is very high maint. but if you're willing to put up with it, then go for it.. 

rhofmann:

In addition, would a server with dual P3 processors with 3GB of RAM be effective for folding if that was the only application running on it?  Not to sound like a needy n00b full of questions, but I am.  again, the other server would be running only 8MB of vram.

All questions are welcome, it's most likely that other people have the exact same questions.  Since the P3 supports SSE, you should run two seperate installations of the regular console client, setting a different machine ID for both, and installing both as a "Service" so they run in the background and stay out of your way. Also since you have a ton of RAM on the dual P3, enable BigWUs (Big Workunits) in the configuration as well. 

rhofmann:

Any advice on setup for optimization of folding setup would be much appreciated.

Well, the P3 setup there isn't much you can optimize.. Usually I recommend people just "set it and forget it" running the console client as a service (in the background) where it's never in your way and uses the least amount of RAM possible.. The GUI client (with the fancy molecules) actually uses up a bit more RAM than the console client which is why I usually don't recommend people run it.. People do like to watch the molecule, so if that's your cup-of-tea then go for it.. It will get boring pretty fast, and when it does, switch to the console client to run folding@home with less memory usage.

If you want to see what the molecules look like, grab JMol which is a seperate 3rd party application for visualizing the workunit you're doing..  it will allow you to click&drag the molecule around so you get a good look at it..  The GUI client molecule display isn't very good in my opinion..

Edit: Added Jmol link.. Note that JMol does not work for some SMP workunits since they're very different.. 

Edit #2: Clarification "you're only using 1/4th of your total CPU capacity." -> Which means that your folding@home client will use anywhere between 0%-25% of your total CPU power.. If you're using 100% of your CPU power doing other things, then folding@home will step out of the way and use 0% (since it is idle priority).. But when the server is idle, on your system, the folding@home won't be able to use more than 25% of your CPU power because the graphical client only runs one core and you have a quad-core system

 So the choices for your dual-dual core (quad core) server is:

    1) Run the SMP client which will use all 4 cores to run 1 workunit (best points per day, but you HAVE to keep an eye on it, it can't be installed in the background as a service and will always run with a console window so you can watch it to make sure it didn't trip over a bug or something)

    2) Run 4 installations of the console client as a service (each in their own directory, different machine IDs, etc.)

    3) Run 3 Installations of the console client as a service (each in their own directory, different machine IDs, etc) AND one copy of the graphical client (which you already have installed now). btw, for the console clients, make sure you use Machine IDs 2,3, and 4.. Since the graphical client always uses Machine ID 1 and is for systems with a single 1-core CPU.

 Edit #3: The GPU client and the graphical client are TOTALLY different.. Don't be confused by them.. To run folding@home on a graphics card (GPU client), you'll need an ATI Radeon X1600 with 256 Megabytes of RAM or better.. 

The graphical client is  the same thing as the console client (both are CPU clients) EXCEPT the graphical client has a fancy graphical user interface and a visualization of the molecule.. The graphical client doesn't use your graphics card at all for anything other than showing you pretty pictures of the molecule. So there is not much difference between the graphical client and the console client, they both run on only one CPU core.. This is why you need multiple installations for multi-core systems unless you use the SMP client (Multi-core CPU client).