I dont OC my CPU to start that straight off, mainly because it's a Palomino and I have a KT333 with no 5/1 divider.

My memory timings are as tight as can be, stably.

How much faster is the text based client? Is it?

Thanks!

EDIT: Does an increased molecular draw rate improve performance? I cant find anything on the main F@H website.

I don't think its much faster, estimated CPU use time for the graphics portion is supposedly between 1-2% of overall and close to nothing when its minimized.

The difference is that with the text client you can add some of the tags that may force you to optimize methods in case the graphical client decides that it doesn't want to utilize those same methods for whatever reason...

I know you can add the tags to the text client, but some have suggested that you can simply add the same tags to the shortcut for the graphical client and thereby eliminating a major speed factor between the two. In terms of raw efficiency, I don't believe there is more than a marginal difference.

The two most commonly used switches however, are -forceasm and
-advmethods. These are usually used to force CPU algorythms that are optimized for higher end P4's and AMD chips. It also allows a preference for Gromac WU which are optimized for P4 chips. AMD chips work better overall due to their higher processes per clock cycle.

Judging from the discussion on the folding farm thread, the most surefire way to increase your folding speed is to buy more hardware.

Seriously though, aside from the -advmethods switch for P4's (which I swear by religiously), I think if there are any "hidden" ways to improve your performance by anything more than marginally, people be raving about it all over the forum by now. (Unless some of the top folders have any secrets they'd like to share with us, aside from "brute force"...) Since you're using an AMD though, you've already got a leg up on some of us here.

Happy folding!

Yeah, I just noticed that. My 1.4GHz AMD can process almost as much information per second as your 2.2GHz P4 (dont know if it's a Northwood though)... Cool!

BTW, with SSE enabled with the text client how much faster is it then without SSE?

Yup, its a Northwood, wow, I seriously should switch over the XP1700 downstairs ... SSE is usually enabled by default for upper class processors. -forceasm is the switch I believe. It turns it on if its available, but if its available, its usually turned on already.

RPC180:
Yep. My estimate was just off of MHz and IPC, nothing to do with the cache or FSB.

Mine (XP1700+): 9x1467 = 13203 IPS (Instructions per Second, I think...)
Yours (2.2GHz): 6x2200 = 13200 IPS

Only 3 IPS difference, but hey, not bad for a 800Mhz speed loss. I dont know how much of a difference your additional 256K of L2 cache helps though. I think this might be why most people base their "farms" off of AMD processors, because they have the best bang for the buck, or so it seems to me.

============================================
Everyone:
Can someone answer my question on the Molecular Draw Rate. please?

And also, on the SilentPC F@H stats page, for individual users it says stuff like "8 processors active". Does this mean that they have 8 processors working on this stuff?

umm Atomic, your off by 1*10^6. its 13,203,000,000 IPS for the athlon and 13,200,000,000 IPS for the P4. 3 millions pretty close to 3, right? I think Athlons have more of an advantage though- I think these things are pretty heavy on floating point calculations, the P4 has 1 FPU, while the Athlon has 2... leads to even larger performance difference.

oh yeah, forgot about that The draw rate is just a snapshot of what's currently being done it happens every so often, set by the redraw rate. It doesn't help processing go faster as the rate of processing remains stable (save for the 1-2% amount of CPU/GPU power used to redraw), so you can set it for whatever you like and not take a hit in any real sense. Of course, minimized it doesn't take up any real CPU/GPU cycles.

Thanks for the responses guys!

Hello:



Actually, Athlons have THREE floating point units!

TO respond to the second part of the orginal post, I think the moecular draw rate slider only affects how often the graphical app displays the protein folds. I turn mine down on the 2 PC's that run the graphical client. Plus I always have it minimized to the tray anyway.

-Mike

*cackles*

Aw cause of all you AMD fans I went and OC'ed the P2.2 ... got to 2.59 before it crashed AND lost a gromac's core at step 70/100!!! Learned my lesson ... keeping stock and quiet until I spring for more stuffs

These offers are getting fairly common around here as of late!

I am running all AMD's and the only tweak I have done is that I've clicked the "Slightly Higher" option on the Advanced Tab. This raised my CPU temp by a few °C so it must be helping a bit (in theory). I tell you what though, My 2800+ Barton's plow through the WU's more than my other XP's.

Booya. I finished my first WU and when I came home it had already finished 30% of the next one! Not bad for being away 6 hours...

I finally got the old AMD 800 reconfigured for folding. It ain't much, but it's good to be contributing again. I gotta sneak back into my friend's place and find out why his computer hasn't contributed lately.

Does anyone have a simple answer to the 'best' options with regard to
Tinker v Gromacs
ADVMETHODS
FORCEASM
AMD v Intel

I am running 3 different processors, all are console mode with -advmethods -forceasm.

The XP2700 seems to churn between 70 points per day whatever it gets.
The Thunderbird 1400 seems to do about 35 points a day whatever it gets.
The (mobile) pentium 3 1.2 I cant understand (I guess it is female ): on Gromacs (e.g. project 540, BBA5 in water etc) it churns through 33 pointers a little faster than the Thunderbird: about 35-40 points/day. But it also gets Tinker stuff (currently project 676, a 20 pointer) that will take 36 hours!

WTF

I thought -advmethods was supposed to stop the Tinker stuff? Except I have -advmethods and I get Tinker stuff.
I thought Pentiums were supposed to be completely outclassed by Athlon with Gromacs, but hold their own with Tinker? My Pentium seems OK on Gromacs but is useless on Tinker.

I am completely baffled by this damn Pentium: is there anything I can do to get it more predictable?

Pentiums do better with Gromacs, and terribly with Tinker. Gromacs is newer, and takes better advantage of SSE.

I've gotten 1 Tinker in the last 30-40 WU's after using -advmethods. Have you tried tallying how many Tinkers vs. Gromacs you've gotten? Maybe it's a fluke, assuming you have everything configured correctly.

Thanks for the correction on that. A supplementary question, is it worth using -advmethods on an Athlon then? Or at least my Thunderbird which doesn't have SSE?
Just checked queue.dat - the current unit is Tinker and 1 out of the previous 9 (total 2 in 10).