FAH Benchmarkathon

So in other words, i havent been lying to the people :) and stanford actually does what it says it does :) oh wait...i already knew that, but its good to see it in [H]ard numbers....good find marty.



and heck...i play warcraft 3, while watching tv on a second moniter and folding and i see no slow down anywhere...yay AIW 9600 :)
 
Ah, great article. Bout time someone went and proved the minimal impact that folding has on your PC performance. Thanks for the linkage marty. :)

 
November 11, 2002 - so might not be as accurate nowadays, bigger proteins, more memory, etc, etc. Prolly still pretty valid though.

 
I find it somewhat funny that the article is 3 1/2 years old and is being treated like it's new. No offense guys.

However, the article is a bit long in the tooth. This mainly has to do with bigpackets and QMD's. The increased memory and memory bandwidth requirements have caused a change.

However, if you are letting the machine run timeless or just regular gromacs, the article still has validity. I'm betting the folding done on the box were tinkers. I don't remember when gromacs made their appearance in any type of numbers.

I just wanted to point out a couple of things before this article is considered the end all be all of things. And no, I don't notice any slowdown in games with a PIII 800 with 384 meg of RAM and a Radeon 8500 once the system is done paging everything. (Well, with the exception of loading new maps) This is running timeless tinkers.

And damn, the version 3 client. That was a long time ago. No service installer or anything with it.

 
Hey, I never read the dang thing nor looked for a date.

Purdy pictures was all I was looking at. I figured a "high end" system was good enough for me!
 
ok...yea...i thought that the "high end" was pretty lean...and i dont read dates...i barely read the article itself :) but it still holds some validity if you consider there "high end" and what most of us are running now...i dont know about QMD (or if they will ever come back) but nuthin i have folded in the last 5 months has caused any system slow down.
 
roftranspo said:
ok...yea...i thought that the "high end" was pretty lean...and i dont read dates...i barely read the article itself :) but it still holds some validity if you consider there "high end" and what most of us are running now...i dont know about QMD (or if they will ever come back) but nuthin i have folded in the last 5 months has caused any system slow down.
Click to expand...

Trust me, QMD's can cause a good bit of slowdown. When they ended and I went back on normal units on my work system (the only one I had that can run QMD's) it was like night and day. Most of it was due to paging from the RAM needed by the QMD's as the system only has 512 meg of RAM. I think the FSB being saturated a bit also had something to do with it.

I have had no trouble with any other units for the most part, though. 364 and 600 pointers would cause a system with 256 meg of RAM to page at times when running WinXP, but that was expected for anything run on that system. 256 just isn't enough RAM for that OS.

 
The two things they ignore are the two that cause performance issues nowadays - memory and bandwidth. Granted, the Stanford guys have done a great job with the client in terms of memory bandwidth, but it still doesn't give up memory, and the use of upload and download bandwidth can completely swamp even a DSL connection for a few minutes.

These are the things holding me back from borging. I can deal if my system is slow and I know the reason, but I don't want to put it on other people's systems and explain that it's harmless and then be disproven. The "report this much memory" option is a good step in this direction, now for the "use this much bandwidth" option.

 
mmm...maybe im just blind to it, but i have never seen a problem with the upload and download bandwidth with the client. But i have never really seen it take more than just a couple of seconds to download anything anyway.
 
On dialup it's just plain painful. And even on 768/128 DSL, it's a half-duplex connection, so uploading even a few MB like FAH does can tie up the whole connection for a few minutes. Try playing games online while FAH uploads and watch your ping times go straight to crap.

 
I would say, without reading it, that this is all based on normal WU's. No QMD's/Big packets and all that mess.

Honestly with ~ 30-50 machines on my T-1 line I never notice them, then again I'm pretty spacy. I think it's when you get 100+ machines, user web traffic (not just one guy) the slow downs get messy. Or slow net connections with people dumping tons of bit torrent traffic....mage..... :)
 
marty9876 said:
I would say, without reading it, that this is all based on normal WU's. No QMD's/Big packets and all that mess.

Honestly with ~ 30-50 machines on my T-1 line I never notice them, then again I'm pretty spacy. I think it's when you get 100+ machines, user web traffic (not just one guy) the slow downs get messy. Or slow net connections with people dumping tons of bit torrent traffic....mage..... :)
Click to expand...

As I said earlier, regular work units are not a problem and for the most part, still follow with what was in the article. Other than 128 meg more RAM, my now main system is slower than their slow system. I notice no problems.

Obviously, you should not be borging others' machines and configging them for QMD's. And for the most part, shouldn't be setting them up for bigpackets either. The actual memory usage should never be much on a borg. I consider that non-issue for the most part.

As far as internet bandwidth is concerned, that is usually a non-entity also. Dial up is one thing but any broadband connection should not be much of a problem. Even with a 128k upload should not be used for long when sending units back. Downloading new units on anything faster than that isn't much either. Especially since in most cases, uploads are larger than downloads. Most regular units shouldn't take more than a few seconds on a 128k upstream pipe. I just don't see internet bandwidth as a problem.

 
marty9876 said:
Honestly with ~ 30-50 machines on my T-1 line I never notice them, then again I'm pretty spacy. I think it's when you get 100+ machines, user web traffic (not just one guy) the slow downs get messy. Or slow net connections with people dumping tons of bit torrent traffic....mage..... :)
Click to expand...
But not all 30-50 machines send back a WU at one time, right? I think if they did, you'd Notice.

And I do all my BT at college, thank you very much ;) The problem was really online gaming - Guild Wars PvP starts sucking with more than say 100 ms lag.

 
You must log in or register to reply here.
Back
Top