How Fast do you fold these: p2125_lambda_5way_melt_4_10011?

[GigaHertzAddict]

I just built a AMD Athlon 64 3000 socket 754 system for my sister. (I use it more than her:rolleyes: ) It takes 43 minutes per frame for a p2125_lambda_5way_melt_4_10011 . I was wondering how other CPUS are doing with the same protein so I know where I stand. Thanks. I'm very curious about the Core 2 duo's too..

[ChasR]

About normal Gromacs production on all cpus. 65ppd/GHz on C2Ds.

[AlabamaCajun]

Ah or should I say "Oh No" a patty melt. They take about 3 days to turn in. Just hang in there and the ribos will come in and work faster. What is good about the PMs is that when they do go in, it makes for a nice point jump.

[CJ-5]

I've got two PMs running on a p4 hyperthread at 3.75 GHz. They average about 64min per frame.:cry: About 4.5 days.

[GigaHertzAddict]

Ah or should I say "Oh No" a patty melt. They take about 3 days to turn in. Just hang in there and the ribos will come in and work faster. What is good about the PMs is that when they do go in, it makes for a nice point jump.

Yeah, these are so slow. I hope they are worth a lot. I guess 43minutes per frame isn't bad huh..

[davekusa]

5-way patty-ham and cheese melt suck monkey butt.

Just work through them. Better units are in the mix

[AlabamaCajun]

5-way patty-ham and cheese melt suck monkey butt.

Just work through them. Better units are in the mix
Want fries with that :p

[eystan]

Could you make that cheese fries with bacon ?? Cuz my main rig is folding a 2124(about 4 1/2 days) and the minimac(it take's about 6 1\2 days) is folding a 2125 yummy patty melts !!!

[Sleepy_Steve]

Yeah, can I get mine suppersized??

:cry: for my poor 1ghz rigs

[GigaHertzAddict]

I've got two PMs running on a p4 hyperthread at 3.75 GHz. They average about 64min per frame.:cry: About 4.5 days.

If I understood you correctly you are folding two instances on your CPU?
If you disable one of them, how fast does your CPU fold that protein alone?

64min is painful.

[CJ-5]

There is less than 1 frame left before the 2125 finishes. If I get another when the client downloads a new wu, I'll give it a try. The other instance is a 2124.

Edit: FAH downloaded a 1498 in place of the 2125.

[Sorin]

On my processor at 2.5GHz, about 41:xx to 43:xx per frame, depending on what else I'm doing on my computer at the moment. Takes about 69-72 hours. I hate them.

[deadlysyn]

When will these patty melts go away? I think my systems have been working on them nonstop for the past week maybe week and a half. At 3 days on one, and anywhere from 7-10 days on the other, I can only hope for this nightmare to end. Apparrently a PIII doesn't like to play well with patty melts, but the AXP averages about 112 PPD with these things. I don't think there is a WU out there I would refuse to take over one of these.

[ChasR]

Production on these units is normal for Gromacs WUs, which make up the vast majority of WUs available. To say you hate them is to say you hate Folding@Home. Rather, you should be saying you love bonus WUs and wish there were more.

[jerseytomato]

One of my poor old PIII@.900Ghz at work has been chugging away on PM's. If I recall it was doing one for 290 hrs. (or was it 390 :shrug:)

[davekusa]

I've got a P2 thats working on a 2126. 50 DAYS!.

[Rav]

36 PPD/Ghz for my 920, around 50 PPD/Ghz from my 754 64's. Over 65 hours for the 4Ghz 920 and almost 100 hours for the 2Ghz 3000+. Seems very slow even for plain old gromacs. They should at least break them into smaller units, since some machines that don't run 24/7 have trouble.

[harlam357]

It's all about WU optimization guys! I know some of you guys only have a couple rigs going, so that makes it real tough. But if you have several machines at your disposal, particularly if they range from XP/PII/PIII up to A64/C2D, you can level things out by keep the bonus WUs on your high end hardware and let the slower machines run the patty melts. :thup:

(insert message from ChasR here that by doing this I'm slowing down the science) ;)

There is some truth to that... but when an A64 rig makes ~150ppd on patty melts and my XP rigs make ~110-120ppd, I'm really not slowing them down enough to make a large impact on when they are returned. And also on the flipside the bonus WUs get done much faster. So again, it's all about balancing the farm. :D

http://img.photobucket.com/albums/v663/harlam357/smileys/thharlam_santa.gif

[leelegend]

my sons Athlon XP2500 barton (1.83) takes 57 minutes a WU

lee

[GigaHertzAddict]

Just some food for thought... I checked my Celeron D 256K @3.4GHz and it was also folding the patty melt.. and guess what? It's just as fast as my AMD 64 @2.4GHz Not bad. I'm itching to get a New Celeron D 512K and overclock that to like 4.5GHz.

[Vipersfate]

Each and everyone of my computer is running on a melt.

Athlon 3700 San Diego -> 46:xx a frame
Celeron 2.6 -> almost exactly an hour a frame
PIII Coppermine 1Ghz OC'd to 1.125 -> 2:28:xx a frame...arg..

I hope I can get some ram and a PSU so I can get the P4 rig back up.

[dropadrop]

About 31 minutes per frame


[19:29:45] Protein: p2125_lambda_5way_melt_4_10011
[19:29:45]
[19:29:45] Writing local files
[19:29:45] Using table 1
[19:29:45] Extra SSE boost OK.
[19:29:45] Writing local files
[19:29:45] Completed 0 out of 20000000 steps (0%)
[20:01:01] Writing local files
[20:01:01] Completed 200000 out of 20000000 steps (1%)
[20:32:14] Writing local files
[20:32:14] Completed 400000 out of 20000000 steps (2%)
[21:03:29] Writing local files
[21:03:29] Completed 600000 out of 20000000 steps (3%)
[21:34:51] Writing local files
[21:34:51] Completed 800000 out of 20000000 steps (4%)
[22:06:09] Writing local files

[deadlysyn]

Production on these units is normal for Gromacs WUs, which make up the vast majority of WUs available. To say you hate them is to say you hate Folding@Home. Rather, you should be saying you love bonus WUs and wish there were more.
I'm not necessarily saying I hate anything about F@H, I am just saying that I seem to do a lot better with Ribo's, GBGromacs, Gromacs, and Ambers. It just feels like I get more done in less time with any other WU. I am not really sure how accurate EMIII is, but before the Patty Melt craze, it was showing me around 200-220 PPD. On EOC it was showing about 198, but with the patty melts, EMIII was showing around 140. My EOC stats had dropped down to about 108. If they were in more of a mix rather than back to back to back, I think they would be more tolerable.

[davekusa]

I feel you deadlysyn. We quantify our value to the project with points. I also feel the need to "produce".

[Vipersfate]

After looking at the estimated finish time on all the rigs currently working for me, they will all finish within 6 days of each other.

1150 points in 6 days

[ChasR]

@Harlam
Nothing at all wrong with Swapping WUs. In fact, I do that regularly when I'm not busy. Trashing a WU perceived to be slow to try to get a better WU is what is not just bad, but terrible for the science.

@Deadly
:D I know none of us hate F@H. Fact is, the p2124 and p2125s were benched correctly. Later generations may have slowed a few percent but nothing major. The primary issue is really a lack of bonus WUs not the speed of these. A secondary issue is that AMD cpus are less superior to the benchmark P4 Northwood on Gromacs than on any other core. Everyone with an AMD cpu sees Gromacs WUs as slow when they really should see them as not as much faster than the benchmark machine as on other cores. There have been plenty of WUs far worse than the patty melts.

[JetMech]

I'm getting 303ppd for two instances of 2124 patty melt on the E6600 running stock. Roughly 62.5 hrs/unit. The 2125's I have running are on dual core machines both with 2 instances. One is on the Merom@2.0 and the other on the D820@3.15. The Merom is kicking the crap out of the 820 (42min/frame and 49min/frame respectively). The Merom has 1 gig of ram more than the 820...hmm. All in all though they are doing the same amount of work on the patties as any other gromac as Chaser said. It's just that if you are one to check the stats every update you will see no work done and get the feeling that you are failing. The slope of the graph is your worth indicator if you must have one. As long as it's not going downhill to the right you're good to go.:beer:

[Rav]

My last couple PPD, thanks to MELTS:

12.03.06..........0......0
12.02.06........983......2
12.01.06......3,841......8
:temper:

Oh well, the sooner we finish them the sooner they will be gone.

[PaciFIST69]

39 min/frame on main rig. Just checked the 2600, poor thing is slowly chewing on a patty as we speak, about 60 min/frame and its only on <12 hours a day :cry:

[deadlysyn]

Well as far as WU Swapping goes, I would be more than up for it, except I have only 2 systems working on exactly the same WU. In a situation like this, all I can really do is let them fold. I could delete them, as I have in the past, until I heard about what it does to the project. I just don't understand why it is constant patty melt after patty melt. I don't think I have seen any other WU in a week or maybe two. Maybe they have something to do with another possible breakthrough??

[harlam357]

Personally I think Stanford's assignment servers treat some IP addresses nicer than others based on how quickly & how much work is returned from said IP address. So the more & faster machines you have going, the more likely you are to get the better WUs.

This is complete theory on my part as I have no evidence or justification for such a claim other than personal observation (i.e. I'm still getting the occasional Ribo).

[dropadrop]

Personally I think Stanford's assignment servers treat some IP addresses nicer than others based on how quickly & how much work is returned from said IP address. So the more & faster machines you have going, the more likely you are to get the better WUs.

This is complete theory on my part as I have no evidence or justification for such a claim other than personal observation (i.e. I'm still getting the occasional Ribo).

I've thought they might select according to how much ram you allocate, what switches (-advmethods -forceasm) and possibly by what your cpu supports (original Intel, sse2, sse3 ect).

Having a fast connection can ofcourse also help, especially as "good" packages are usually only available for a short period. Being fast will make sure you get more of them..

I guess it would also be smart to share large packages to people with a fast connection.

These are just guesses on my part, have'nt looked into it very much, and I allways feel I get bad packages (except now for once!)

[dropadrop]

Personally I think Stanford's assignment servers treat some IP addresses nicer than others based on how quickly & how much work is returned from said IP address. So the more & faster machines you have going, the more likely you are to get the better WUs.

This is complete theory on my part as I have no evidence or justification for such a claim other than personal observation (i.e. I'm still getting the occasional Ribo).

Have a look at this thread:

http://forum.folding-community.org/ftopic17034-0-asc-0.html

[ChasR]

You can find the paramaters used by the AS on the server status page. if you meet the qualifications shown there, assignment is random. The AS algorithim tends to assign you to the same work server you just returned to so you can get "stuck" on a server.

The min performance fraction is used mainly to filter out rigs not on all the time. In truth though, most PIs don't even set the min PF. Reported memory filters out rigs with limited ram. Connection speed isn't tracked. One of the whines heard most often is that "my slow machine always gets the bonus WU and my super duper machine gets stuck with the p7xx crap WU". That is randomness at work.

[dicecca112]

Project : 2125
Core : Gromacs
Frames : 100
Credit : 392


-- 1 --

Min. Time / Frame : 25mn 10s - 224.30 ppd
Avg. Time / Frame : 25mn 10s - 224.30 ppd


3.4 C2D, that might be a little high seeing I just got off an hour of oblivion

[Hack30]

30 min per frame on qx6700 quad core


stock no oc

[GigaHertzAddict]

30 min per frame on qx6700 quad core


stock no oc

Dam that's sweet... Do you have all 4 cores folding? If so 30 minutes is Amazing. I can't wait till they come out with the cheaper Quads..

[jws2346]

I'm not even doing the "melt" thing, I'm doing a p2133 and according to Electron microscope it's taking 47:03 mins per frame. I have a 3000+ like at the begining of this thread, but instead of skt 754 I'm skt 939. (I have a slight OC of 2.02 GHz) WTF am I doing wrong? I have tried to trim all the "fat" from stealing cpu resources, what did I miss? :confused: