Jump to content

Fah And Hyperthreading


sykocus
 Share

Recommended Posts

this is an interesting paragraph i saw on the EM III frontpage. Honestly i'm not trying to tell anyone to change what they are doing, it's just something i had never thought of

 

Folks, for many the HT (HyperThreading) issue is one that they just don't understand. While Stanford allows you to run more than one instance of the [email protected] client on your computer, you are not really helping them if the number of clients you are running is not equal to the number of REAL CPU's you actually have. While running two instances of the [email protected] client will return two work units more quickly than doing one at a time, that is not what is important. Returning one work unit 70% faster and starting the next generation of tests on the work unit, is what Stanford wants. Speed in working through each protein generation is very, very important. Sometimes quality is better than quantity. Quality in this case is work being returned more quickly. While it is well known that an HT CPU can run 2 work units because the OS treats it like two CPU's, and you can gain a possible 15-30 percent increase in points for that computer, it also means that each work unit is returned more slowly. In simple terms, if the project has 300 generations needed to test a theory on a protein model, and running two instances at once delays the return of work for 1 day each time, you end up with a 300 day delay. That translates to about a 10 month delay in examining your data for the final result. The bottom line is simply this: Run 1 instance for each CPU you have. An HT CPU is not two CPU's, it is one. Let's work to advance the science and spend a little less time worrying about the number of points you get.

 

This is a direct quote from Dr. Vijay Pande:

1) If you care primarily about points, running 2 procs on HT is still the best bet. We are grateful for all contributions and if people choose to run 2 procs on HT, our approach is that all contributions are welcome.

 

2) If you care about the science foremost and are interested in our recommendations, then do not run 2 procs on HT, but please just run one process. That won't be best for points, but is best for the science.

 

3) If your machine cannot make the deadlines, then one should run the timeless WUs.

 

I hope that clears these issues up and thanks to all for their contributions.

 

Thanks for listening and please do the right thing.

Larry

Share this post


Link to post
Share on other sites

I do not understand them, I said this the other day and I will say it again they have some WUs with 60 day deadline on some of those 1400s, ok I was giving them 2 in 2 to 3 days, I mean whats the deal, hyperthreads start at what 2.4? thats plenty fast enough to give there WU in a couple days, I think what is happening in my personel opinion is that there are people that probably can not run there machines 24/7, some may play games a lot or forget to turn folding on, maybe they barely make the deadline, I realize a lot of people do not have dedicated folding only machines that run 24/7 and are trying to help and maybe they barely get the WU in time, at least they are are trying to help out, I am not changing a dang thing, if its not fast enough for them what I am doing I can save a lot of money and time and sell all of my machines, like the last 2 days I have been getting nothing but 872s and 873s and do around 50 a day, there not worth any points 33 I think. If just got to have it the way they want it I guess they need to just do all there own folding. I have been spending a lot of money, machines, electricity etc. Christmas will be 1 year 24/7 that upsets me. They keep complaining I am going to pull the plugs on all of mine!

 

When I can do 2 of these in just a little over 2 days, if thats what there complaing about they need to buy them some super computers. :angry:

 

 

 

[14:04:29] Project: 1409 (Run 0, Clone 46, Gen 3)

[14:04:29]

[14:04:29] Assembly optimizations on if available.

[14:04:29] Entering M.D.

[14:04:35] Protein: p1409_polyQ26 in water

[14:04:35]

[14:04:35] Writing local files

[14:04:36] Extra SSE boost OK.

[14:04:36] Writing local files

[14:04:36] Completed 0 out of 2500000 steps (0)

[14:40:23] Writing local files

[14:40:23] Completed 25000 out of 2500000 steps (1)

[15:16:09] Writing local files

[15:16:09] Completed 50000 out of 2500000 steps (2)

[15:51:58] Writing local files

[15:51:58] Completed 75000 out of 2500000 steps (3)

[16:27:46] Writing local files

[16:27:46] Completed 100000 out of 2500000 steps (4)

[17:03:33] Writing local files

[17:03:34] Completed 125000 out of 2500000 steps (5)

[17:39:20] Writing local files

[17:39:20] Completed 150000 out of 2500000 steps (6)

[18:15:06] Writing local files

[18:15:06] Completed 175000 out of 2500000 steps (7)

[18:50:51] Writing local files

[18:50:51] Completed 200000 out of 2500000 steps (8)

[19:26:36] Writing local files

[19:26:36] Completed 225000 out of 2500000 steps (9)

[20:02:24] Writing local files

[20:02:25] Completed 250000 out of 2500000 steps (10)

[20:38:10] Writing local files

[20:38:10] Completed 275000 out of 2500000 steps (11)

[21:13:57] Writing local files

[21:13:57] Completed 300000 out of 2500000 steps (12)

[21:49:43] Writing local files

[21:49:43] Completed 325000 out of 2500000 steps (13)

[22:25:29] Writing local files

[22:25:29] Completed 350000 out of 2500000 steps (14)

[22:59:51] Writing local files

[22:59:51] Completed 375000 out of 2500000 steps (15)

[23:36:00] Writing local files

[23:36:00] Completed 400000 out of 2500000 steps (16)

[00:12:10] Writing local files

[00:12:10] Completed 425000 out of 2500000 steps (17)

[00:48:18] Writing local files

[00:48:18] Completed 450000 out of 2500000 steps (18)

[01:24:27] Writing local files

[01:24:28] Completed 475000 out of 2500000 steps (19)

[02:00:37] Writing local files

[02:00:37] Completed 500000 out of 2500000 steps (20)

[02:36:29] Writing local files

[02:36:29] Completed 525000 out of 2500000 steps (21)

[03:12:39] Writing local files

[03:12:39] Completed 550000 out of 2500000 steps (22)

[03:48:48] Writing local files

[03:48:48] Completed 575000 out of 2500000 steps (23)

[04:24:58] Writing local files

[04:24:58] Completed 600000 out of 2500000 steps (24)

[05:01:08] Writing local files

[05:01:08] Completed 625000 out of 2500000 steps (25)

[05:37:19] Writing local files

[05:37:19] Completed 650000 out of 2500000 steps (26)

[06:13:29] Writing local files

[06:13:29] Completed 675000 out of 2500000 steps (27)

[06:49:39] Writing local files

[06:49:40] Completed 700000 out of 2500000 steps (28)

[07:25:49] Writing local files

[07:25:50] Completed 725000 out of 2500000 steps (29)

[08:02:03] Writing local files

[08:02:03] Completed 750000 out of 2500000 steps (30)

[08:38:13] Writing local files

[08:38:13] Completed 775000 out of 2500000 steps (31)

[09:14:23] Writing local files

[09:14:23] Completed 800000 out of 2500000 steps (32)

[09:50:34] Writing local files

[09:50:34] Completed 825000 out of 2500000 steps (33)

[10:26:41] Writing local files

[10:26:41] Completed 850000 out of 2500000 steps (34)

[11:02:50] Writing local files

[11:02:50] Completed 875000 out of 2500000 steps (35)

[11:38:59] Writing local files

[11:38:59] Completed 900000 out of 2500000 steps (36)

[12:15:09] Writing local files

[12:15:09] Completed 925000 out of 2500000 steps (37)

[12:51:19] Writing local files

[12:51:19] Completed 950000 out of 2500000 steps (38)

[13:27:29] Writing local files

[13:27:29] Completed 975000 out of 2500000 steps (39)

[14:03:39] Writing local files

[14:03:39] Completed 1000000 out of 2500000 steps (40)

[14:39:48] Writing local files

[14:39:48] Completed 1025000 out of 2500000 steps (41)

[15:15:55] Writing local files

[15:15:55] Completed 1050000 out of 2500000 steps (42)

[15:52:05] Writing local files

[15:52:05] Completed 1075000 out of 2500000 steps (43)

[16:28:12] Writing local files

[16:28:13] Completed 1100000 out of 2500000 steps (44)

[17:04:21] Writing local files

[17:04:21] Completed 1125000 out of 2500000 steps (45)

[17:40:29] Writing local files

[17:40:29] Completed 1150000 out of 2500000 steps (46)

[18:16:38] Writing local files

[18:16:38] Completed 1175000 out of 2500000 steps (47)

[18:52:48] Writing local files

[18:52:48] Completed 1200000 out of 2500000 steps (48)

[19:28:57] Writing local files

[19:28:57] Completed 1225000 out of 2500000 steps (49)

[20:05:05] Writing local files

[20:05:05] Completed 1250000 out of 2500000 steps (50)

[20:41:12] Writing local files

[20:41:12] Completed 1275000 out of 2500000 steps (51)

[21:17:20] Writing local files

[21:17:20] Completed 1300000 out of 2500000 steps (52)

[21:53:28] Writing local files

[21:53:28] Completed 1325000 out of 2500000 steps (53)

[22:29:35] Writing local files

[22:29:36] Completed 1350000 out of 2500000 steps (54)

[23:05:46] Writing local files

[23:05:46] Completed 1375000 out of 2500000 steps (55)

[23:41:56] Writing local files

[23:41:56] Completed 1400000 out of 2500000 steps (56)

[00:18:08] Writing local files

[00:18:08] Completed 1425000 out of 2500000 steps (57)

[00:54:19] Writing local files

[00:54:19] Completed 1450000 out of 2500000 steps (58)

[01:30:29] Writing local files

[01:30:29] Completed 1475000 out of 2500000 steps (59)

[02:06:28] Writing local files

[02:06:28] Completed 1500000 out of 2500000 steps (60)

[02:42:31] Writing local files

[02:42:31] Completed 1525000 out of 2500000 steps (61)

[03:18:39] Writing local files

[03:18:39] Completed 1550000 out of 2500000 steps (62)

[03:54:49] Writing local files

[03:54:49] Completed 1575000 out of 2500000 steps (63)

[04:30:59] Writing local files

[04:30:59] Completed 1600000 out of 2500000 steps (64)

[05:07:11] Writing local files

[05:07:11] Completed 1625000 out of 2500000 steps (65)

[05:43:23] Writing local files

[05:43:23] Completed 1650000 out of 2500000 steps (66)

[06:19:34] Writing local files

[06:19:34] Completed 1675000 out of 2500000 steps (67)

[06:55:43] Writing local files

[06:55:43] Completed 1700000 out of 2500000 steps (68)

[07:31:54] Writing local files

[07:31:54] Completed 1725000 out of 2500000 steps (69)

[08:08:07] Writing local files

[08:08:07] Completed 1750000 out of 2500000 steps (70)

[08:44:18] Writing local files

[08:44:18] Completed 1775000 out of 2500000 steps (71)

[09:20:28] Writing local files

[09:20:28] Completed 1800000 out of 2500000 steps (72)

[09:56:38] Writing local files

[09:56:38] Completed 1825000 out of 2500000 steps (73)

[10:32:47] Writing local files

[10:32:47] Completed 1850000 out of 2500000 steps (74)

[11:08:56] Writing local files

[11:08:56] Completed 1875000 out of 2500000 steps (75)

[11:45:06] Writing local files

[11:45:06] Completed 1900000 out of 2500000 steps (76)

[12:21:19] Writing local files

[12:21:19] Completed 1925000 out of 2500000 steps (77)

[12:57:30] Writing local files

[12:57:30] Completed 1950000 out of 2500000 steps (78)

[13:33:40] Writing local files

[13:33:40] Completed 1975000 out of 2500000 steps (79)

[14:09:51] Writing local files

[14:09:51] Completed 2000000 out of 2500000 steps (80)

[14:46:03] Writing local files

[14:46:04] Completed 2025000 out of 2500000 steps (81)

[15:22:15] Writing local files

[15:22:15] Completed 2050000 out of 2500000 steps (82)

[15:58:27] Writing local files

[15:58:27] Completed 2075000 out of 2500000 steps (83)

[16:34:39] Writing local files

[16:34:39] Completed 2100000 out of 2500000 steps (84)

[17:10:50] Writing local files

[17:10:50] Completed 2125000 out of 2500000 steps (85)

[17:47:03] Writing local files

[17:47:03] Completed 2150000 out of 2500000 steps (86)

[18:23:16] Writing local files

[18:23:16] Completed 2175000 out of 2500000 steps (87)

[18:59:31] Writing local files

[18:59:31] Completed 2200000 out of 2500000 steps (88)

[19:35:45] Writing local files

[19:35:45] Completed 2225000 out of 2500000 steps (89)

[20:11:58] Writing local files

[20:11:58] Completed 2250000 out of 2500000 steps (90)

[20:48:09] Writing local files

[20:48:09] Completed 2275000 out of 2500000 steps (91)

[21:24:19] Writing local files

[21:24:19] Completed 2300000 out of 2500000 steps (92)

[22:00:30] Writing local files

[22:00:30] Completed 2325000 out of 2500000 steps (93)

[22:36:43] Writing local files

[22:36:43] Completed 2350000 out of 2500000 steps (94)

[23:12:54] Writing local files

[23:12:54] Completed 2375000 out of 2500000 steps (95)

[23:49:05] Writing local files

[23:49:05] Completed 2400000 out of 2500000 steps (96)

[00:25:17] Writing local files

[00:25:17] Completed 2425000 out of 2500000 steps (97)

[01:01:27] Writing local files

[01:01:27] Completed 2450000 out of 2500000 steps (98)

[01:37:37] Writing local files

[01:37:37] Completed 2475000 out of 2500000 steps (99)

[02:13:50] Writing local files

[02:13:50] Completed 2500000 out of 2500000 steps (100)

[02:13:50] Writing final coordinates.

[02:13:50] Past main M.D. loop

[02:14:50]

[02:14:50] Finished Work Unit:

[02:14:50] - Reading up to 294624 from "work/wudata_00.arc": Read 294624

[02:14:50] - Reading up to 346124 from "work/wudata_00.xtc": Read 346124

[02:14:50] goefile size: 0

[02:14:50] logfile size: 212410

[02:14:50] Leaving Run

[02:14:53] - Writing 1015298 bytes of core data to disk...

[02:14:53] ... Done.

[02:14:53] - Shutting down core

[02:14:53]

[02:14:53] [email protected] Core Shutdown: FINISHED_UNIT

[02:14:56] CoreStatus = 64 (100)

[02:14:56] Sending work to server

 

 

[02:14:56] + Attempting to send results

[02:16:01] + Results successfully sent

[02:16:01] Thank you for your contribution to [email protected]

Edited by road-runner

Share this post


Link to post
Share on other sites

My argument is that my P4 HT folding 2 instances 24/7 will still turn in work units with a performance factor > .95. I have a PIII Xeon 600MHz running with a performance factor around .88. Now using their rationale, I shouldn't be folding on that PIII, or at least it should only be running timeless work units.

 

If they are needing results back so quickly, then those proteins should only be assigned to the fastest machines out there. As long as I fold 24/7 on my P4 HT, I plan on running two instances. I fold because I want to help science and disabling one instance on the P4 is only going to cost me about 500-700 ppw, so its not a points issue to me. It's that I am convinced I am doing more for science running two instances. The 300-day argument is fallacious, IMO. If I disable one instance, it's going to delay the project a lot more, because that means while one work unit will get returned faster, the other work unit must be assigned to another CPU. If I get a 25% boost in points, doesn't that also mean Stanford sees an increase in number of work units returned?

 

I need someone to offer a better argument than what I've seen so far. If we get one, I'll happily stop running two instances.

Share this post


Link to post
Share on other sites

If they are needing results back so quickly, then those proteins should only be assigned to the fastest machines out there. If I get a 25% boost in points, doesn't that also mean Stanford sees an increase in number of work units returned?

 

I need someone to offer a better argument than what I've seen so far. If we get one, I'll happily stop running two instances.

I agree with you. :withstupid:

Share this post


Link to post
Share on other sites

It wouldn't seem like it would make a difference to me either. I also thought it would make more sense to send more complext wu's to the computers that benchmark higher.

 

at the same time evidentally it is a big enough problem, that the founder of the project has come out with a few statements. It's not just the idle/theoretical conjectue of people on the outside.

Share this post


Link to post
Share on other sites

I think whats he is trying to say is get the WUs back as fast as you can, if you take a H.T. and run 2 ver. we all know it slows the machine down a little but gains at the end, but what about people like me running 2 ver. @ 3.7ghz and another at 3.6. another 3.55 etc. and more high clock speed H.T. I give them 2 WU in 2 1/2 days when the prefered is 43 days and deadline 64 days. I think it is pretty dang quick, if thats not fast enough I am quiting and turning everything off and selling them. I have been doing it to help science for almost 1 year and not for points or contest, if I want something I go buy it, cancer runs in our family and I wanted to help out, now there . me off, if they want me to turn H.T. off well thats going to be the straw that breaks this camels back. I could have saved a lot of money over the last year and bought non H.T.s I am going to sign up on there forums and see what they have to say. they can buy them some non H.T. Pcs or super computers and do it thereself, it seems to me theyare saying if you can not make the deadline dont fold, or set it for timeless WUs, they are steadly lowering the deadlines so maybe everyone will go buy a faster PC. I am done now.

 

:foldon:

Edited by road-runner

Share this post


Link to post
Share on other sites

.

 

its great you have spent a bunch of money on the project. those people who can't or haven't spent a bunch of money are great too.

 

They have requested that people not run 2 instances. they haven't told people who do they are unwelcome. they haven't disabled mulitple processes in the client.

 

I don't understand the math either, but all 5 of us following this thread make up a very insignificant percentage of the systems involved in fah. I think Dr Pande has a much bigger picture of what is going on.

 

you are over reacting about something. what i don't know. It sounds like you are trying to teach them a lesson

 

If you are unsatisfied with the direction the project is heading, i think by all means you should find something else to do with your computers/money. it makes no sense to spend the time and effort you do, in something you don't agree with.

Share this post


Link to post
Share on other sites

And when I am ready to stop believe me sycocus I will. Well I spent a few hours seaching and looking on there forums last night, and all I saw on Pande site was they were encouranging people with H.T. to run 2 instancesin places and not in others, but from what you posted from EM3 I think the final meaning of it is if you can not make the deadline do not do it and run untimed WUs. I am not changing anything, and if he says enough about it and decides to fix it were it will not run on H.T. I can think of a few things I can do with some of these machines besides fold, beleive me, when they stop it I will definately quit if not sooner, I just do not under stand his theory take a AMD 2200 or 2400 or 2600, or some of the slower P2s P3s, I can do 2 faster that it can do one. I just got a 1400 series on my 2400 I am going to see how long it takes, I will post it in about week when its finished.

 

From Pande Forums

 

HT enable is good for first dedicated folding system and other general uses It's not a good idea for ppl who run softwares which rely on CPU only like capturing/encoding video becuz those software are suffered from loosing part of CPU to [email protected]

 

 

That's certainly true.

 

If you disable HT, FAH uses a low prioirty and it does a pretty good job of releasing the CPU when something else comes along.

 

With HT enabled, you loose the benefits of priority. THe OS keeps FAH running even though the "other CPU" is running a higher priorty task.

 

One piece of additional information on the P-IV "HyperThreading" CPUs, the "Task Manager, Processes Tab" function has has an "affinity" setting when a process is given a "right click". The "affinity" setting specifies which CPUs, or virtual CPU, is allowed to work on that process. You may select both or only one, or none. This may be helpful in making the best use of this systm.

 

my HT P4 sees about a 16% boost in production when running 2 instances of FAHConsole with HT enabled over only running one session with it enabled... I haven't tested with it disabled, but i don't think that would help anything, especially if you use your pc for things other than folding...ie when gaming w/o hyperthreading, fah gets basically no cpu time and frames that normally take 6-7 minutes will take well over an hour...with hyperthreading enabled, that shouldn't be the case...

 

All of mine are dedicated folding only if that matters, but I kinda see what there talking now if you have H.T. enabled and playing a game or something folding slows way down, I have always used the affinitys on mine, even after someone said windows done better leaving them alone, I am still going to see how long this 2400+ takes.

Edited by road-runner

Share this post


Link to post
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

 Share

×
×
  • Create New...