So i switched...

[Revv23]

i went from XTREME to xtremesystems.org team..

couldnt find any info on the site about this, but is there any way to keep my points?

[pythagoras]

i went from XTREME to xtremesystems.org team..

couldnt find any info on the site about this, but is there any way to keep my points?

Very nice to have you on the team :toast: Sorry no way to switch points, but that does have its advantages too.

Stanford does keep a running tally of all the points though so your overall ranking dosnt suffer. Like this:

http://vspx27.stanford.edu/cgi-bin/main.py?qtype=userpage&username=pythagoras

Regards

John.

[Ugly n Grey]

Hi Revv!!!! Very much welcome to our team! We have had some interesting folders give us a boost lately (Xciter for a couple weeks, Angra for a number of months) and while the contributions were phenomenol we are always looking to pick up the output.

To answer your question, I don't think it works like D2OL, in that when you switch teams your old points do stay with that team. I wish it were not so, but I am certain it is. I did some searches and Pande group seems to deny requests by users to move points. The feeling is more what's done is done. I can understand that. After all it's as much about the cause as it is about the competition. This teams captain, Pythagoras may be able to add some more to this than I have been able to dig up.

So man welcome!

Cheers
UnG

Edit: John posted before I finished typing :banana:

[Revv23]

hehe

well just finished figuring out how to get it to run as a service (always used the graphical client and hated it..

ive got 2 cpu's on it now, but these computers cant be run 24/7...

ive got to make the rounds to my relatives houses and ill have 4 going 24/7 for us... including my 3.6ghz p4... output shouldnt be too bad...

im also going to get my girlfriends laptop on it and a bunch of my college buddies, they all trust me to work on their computers, so i can get it running in the background easy, they usually like the idea of it, and they let me keep the points hehe...

anyways, never really activley pursued folding before, just had it running on a few computers, i think i can get some good output though now that i am on a team where i actually know some people.

[Revv23]

BTW thanks very much for the warm welcome!!! :toast:


feels good to be among the ranks of such respected XS members.

[Revv23]

okay my p4 cant hang, changing out the hsf, maybe ill take off the spreader, the thing runs 39c idle and 55c load, all that heat builds up within 5 seconds, ive remounted this xp-120 3 times... im a little upset.

this is at 2.9ghz 1.525v btw... this thing on the xp-90 could run 2.65 1.675v under 50c... sheesh.

also, got my buddies laptop all set up...

i misconfigured the net settings on 3 currently folding computers that im not going to have access to for a few weeks, so i should get a nice boost whenever i get my butt over to them...

after that im low on ideas... got to find more CPU's!!!

[Ugly n Grey]

Well the big thing on Intel rigs is how you set them up of course. If ou can get the QMD packets by enabling large workunits and advmethods you can score 450 a day with a 3 Ghz P4.

Those temps you listed are pretty normal for a P4 Rev and nowhere near the chips limits, what are you worried about? If you are heat concerned, you can do what I used to do with the Northy's and undervolt them and underclock them if need be :D

Let me know what I can do to help!

Thanks for your efforts, you're making a great teamate :toast:

[Revv23]

these temps are aweful, i need my fans on HIGH to do this, im talking 7 case fans including a 120mm panaflo, i cant live with that, especially not with a northwood at stock speeds, this thing was doing WAY lower on my xp-90 at stock, but it seems remounts arent working with this xp-120, i though the 120 was better then the 90?

i would live with 50c all day if it werent for the fact that to get 3.6 (where i like it, i noticed the 800mhz difference in windows) temps go up over 65c, from about 45c idle. it just isnt stable at full load. either way, ive got to tear it apart, but ive got all these paper's due for school, its hard to find time to crack her open, as without this thing i would be screwed.

im learning now why D2ol is so much more popular around here. the folding core sucks, it has no MP support, ive been trying to get two instances running as services all night, but the best i can do is get one going as a service and one graphical client in the taskbar.

and there isnt any way to check on the service, at least no good way to see your progress...

oh yeah, intel rigs score more points even though amd's are faster, no way they could take advantage of any AMD features... its like they dont even care about the app, and they are losing may hardcore users because of it.

i cant see any way to set up a farm of folders as the app is made to be in the background of a home pc, not much in the way of dedicating a monitor less box too it.


maybe i need to look into the linux versions? i dont have any experience in linux, but i wouldnt mind setting up a few dedicated folders if its possible.

it would be fun to get one or two of those dual xeon boxes and overclock the crap out of it i hear the C1 rev's that all the d20l guys use get 3.6ghz easy.

but again, will the software all MP support? seems doubtful seeing how things are going ATM.

[Ugly n Grey]

Well allow me to help a little. Folding is tough on a core for sure, If your rig isn't 100% stable it will really suck big time. Setting it up on MP systems as a service and monitoring it is fairly easy, I have one set up on a dual proc Opteron board under XP as services and I will gladly explain what I did. I use the FAH monitoring tool to keep an eye on progress across machines on the local network.

I am setting up a separate thread RIGHT NOW with screenies and how to's and an explantion of what I understand to be what, so hang tight and you will have all the answers.

On a personal note, I only run one instance on my dual core personal rig, I like the balance that provides, there is always an idle proc so the computer seems much snappier than when two instances are running.

[angra]

I just recently read about QMD and bonus scores on intel boxes and switched some of my nodes over.

then I got all sorts of overheating problems. chips reporting 85+! You sound like you've experimented with the bonus work packets - is really high temps a typical symptom?

I have had other thermal issues in my machine room, so it may be unrelated. Also, the temps didnt come down after killing processes, so there may be some other issue.

[Revv23]

Well allow me to help a little. Folding is tough on a core for sure, If your rig isn't 100% stable it will really suck big time. Setting it up on MP systems as a service and monitoring it is fairly easy, I have one set up on a dual proc Opteron board under XP as services and I will gladly explain what I did. I use the FAH monitoring tool to keep an eye on progress across machines on the local network.

I am setting up a separate thread RIGHT NOW with screenies and how to's and an explantion of what I understand to be what, so hang tight and you will have all the answers.

On a personal note, I only run one instance on my dual core personal rig, I like the balance that provides, there is always an idle proc so the computer seems much snappier than when two instances are running.


thank god, the lack of information on this forum has to be hurting this team, just getting the command line client running as a service(which everyone here should be using) was a PITA for a software noob like me, and ive been running all over the net trying to figure what the hell im supposed to be doing, lots of experimentation too...

[Revv23]

I just recently read about QMD and bonus scores on intel boxes and switched some of my nodes over.

then I got all sorts of overheating problems. chips reporting 85+! You sound like you've experimented with the bonus work packets - is really high temps a typical symptom?

I have had other thermal issues in my machine room, so it may be unrelated. Also, the temps didnt come down after killing processes, so there may be some other issue.

well full load is full load for me, once the cpu is at 100% i dont see how it can get any hotter? sounds like an issue with your stuff.


BTW UnG i want to run two instances because of HT, the max cpu time i can use is 50% unless im running two instances...

[Ugly n Grey]

Yeah I have been trying pretty much all kinds of things to up the scores and have done a considerable amount of reading. The QMD packets are very tough calculations, they definitely heat up a CPU and use large amounts of RAM. But I haven't noticed it being any harder per se than a very highly stessed database (say Oracle or DB2).

Before I go on, I will mention I am using Windows XP for F@H at the moment, I have adequate licensing and it's easy to play with (I have nice images). Before I begin F@H on a client or after changing an OC I run 32Meg SuperPi. I find that if a rig is SuperPi stable, it's F@H stable.

So begin testing the QMD packets, I took a standard rig with all stock cooling. On a 2.4c P4 that is overclocked to 3.0 GHz dead on at stock voltage (I wanted a standardized CPU to evaluate other performance metrics against), the chip typically hits 62c while running a single instance of the QMD packet. This is in a SMALL case with one 23cfm intake and one 23cfm exhaust fan. At idle with nothing running the CPU sits at 46c !!!!!! Unreal hot these P4's are.

So no big deal really, I can't seem to get to the point where I am experiencing any kind of throttling or whatnot. Dual Prime 95 makes it run a couple degrees hotter than the QMD packet does.

I think what the problem is partly is that systems are typically not under load 24/7 and most computers are thermally built for long idle periods with short peaks of high stress. DC has the effect of working the unit all the time.

Can you describe your farm and cooling? Usually 2U rackmounts use some kind of semi passive processor cooling or custom built units often use the low profile CoolJag CPU coolers... most of the solutions I've seen are often not adequate unless the rack room temps are maintained AT or BELOW 18 degrees centigrade.

What are those nodes you are running Angra? brand/speed/model?

[Ugly n Grey]

Revv actually at one instance on the virtual core, you are using approximately 85% of your actual CPU capacity. It's just that with two virtual cores, Windows shows them as even, which is actually not true. I run a single instance on an HT proc, I find I get just as many points and I tried both. Virtual cores make a lot of sense for multitasking on the long piped P4's but they don't really add MORE computing power per se, just a way to introduce instructions to the processor faster.

[Revv23]

oh, from what i read up on running dual instances supposedly got 10% better results..either way a guide would be really nice so i can figure out how the hell to get it going, id really like some better monitoring stuff too, ive tried EMIII, and its like childs play, doesnt really give you that much info and its not easy to set up.

i just tried running with my fans on low 39c idle, to wham 63c load...

this is with at least 40cfm on an xp-120... somethings not right...

[Ugly n Grey]

EMIII drives me nuts, all I really want is to know is what the percentage is complete and when/what the packet will score and if it's running or not. I have been running FAHMON and it does this well and simply.

As far as the one or two instances on an HT enabled proc, I am biased towards one. Less chance of error and leaves the computer responsive when you are trying to use it for something else.

F@H aside;
Those are some wild temps for an XP-120... I wonder if the top of the cooler is too close to the side of your case? have you tried with the side panel off to see if there is an impact just so you know?

[angra]

Can you describe your farm and cooling? Usually 2U rackmounts use some kind of semi passive processor cooling or custom built units often use the low profile CoolJag CPU coolers... most of the solutions I've seen are often not adequate unless the rack room temps are maintained AT or BELOW 18 degrees centigrade.

What are those nodes you are running Angra? brand/speed/model?

My farm is a 2 stacks of mostly 1U units of varying types. The cooling is underfloor forced. unsurprisingly, with this power density, the topmost nodes tend to get hotter than I like doin any sort of torture testing. I have looked into alternate cooling methods, but the facilities folks are still grumpy about the 15 tons of AC unit they added primarily for my benefit a year or so ago. It's always cool in there, I want to say mid 60s F, but I am not sure offhand. Maybe I need to buy a big box fan and sit it on top of the rack to pull some cool air up to the top :).

there are 3 node setups. 32 are dual Xeons ~3 GHz. 16 are dual opterons. and 2 are quad opterons. by dual and quad I mean # of cpus, each single core. The purpose of this cluster is to test and experiment with methods we (and others) are developing for having fine-grained parallel applications be redployable onto different configurations without major software rewrites, hence the varying configs of nodes.

Each node has pretty hefty CFM fans on the CPUs.

[Revv23]

hmm my case is huge, ill try it now though...

edit - temps actually worse, my airflow is worked out really well on this antec 1000amg.... just the hsf isnt doing its job...

[Ugly n Grey]

Angra, I was just looking at the pic you posted some time back of the racks and now understanding their configs I can see where you might be getting some air flow issues. The rack doors look like they are the mesh screen type? Yet it looks like you have a duct coming out of the floor directly under the centre one? I doubt the cold air is moving up to the top of the rack before the airflow pressure is dissipated through the mesh. It's naturally warmer at the ceiing in any case and it looks pretty tight from the top of the rack to the ceiling as well. Replacing the mess with Lexcan can help you direct air flow directly up but would only work well with negative pressure at the top of the rack as well.

The quickest, cheapest soltution I can think of is to remove the three tiles right above the rack that have no pipes or whatever running though the tiles and replace them with standard exhaust fans that fit directly in that space. They come in a size meant to hang directly in the ceiling tile frame and can be used to push air up or down (blow/exhaust). If you can put some kind of simple shroud around the tops of the racks, you can ensure the exhaust function services primarily warm air removal at the tops of the racks.

I have no idea of that helps or give you a few ideas, but there it is :)

Revv... Crud. Don't cook your system running ANYTHING as it's not worth it. I think you may need a little BIG TYPHOON action :) quiet and cool baby :)

[angra]

interesting idea. I dont think that the room has a dropped ceiling. clearance between racks and ceiling is not tight. however, I could probably find some compact fans that I could put on top, do the lexan thing you suggest (or, heck, anything solid to keep airflow directed), and pull air through bottom to top.

my student-admin is currently hiking through peru or somesuch. when he comes back I will get him to look into this idea. I have a small budget availabel for parts on this, so I can probably afford some fans.

[Revv23]

so i reseated,for the third time, and i realized AS5 was only touching in the edges of the CPU, i i lapped her down and now im at 48c full load with my fans on low.

i for one am a happy camper... time to crank up the clocks.

edit - maybe i should pull off the spreader? Its a D1 northwood so it shouldnt be epoxied, any comments on the benifits of this?

[Ugly n Grey]

You are talking to a very biased guy, I don't have a single chip I use for overclocking with an IHS in place... I figure they just ship those chips with an IHS to protect the core... :D For me the IHS is just a souveneir of yet another CPU I tried to blow up :cheer2:

@ Angra: I have a little experience with operations centers, if I can offer any useful advice please do feel free to ask me.

[angra]

@ Angra: I have a little experience with operations centers, if I can offer any useful advice please do feel free to ask me.


yes indeed I will and thank you for the offer and advice so far. of course if you wish to experiment with programming parallel applications, feel free to ask me as well. It is nice to have a cluster handy whose sole job is to be a place to experiment, rather than to run production apps all day long (FAH is an experiment, right? :) ).

revv, please accept my apologies for hijacking your thread somewhat :)

[Revv23]

hey its all your angra...


well my temps are about 52 loaded up now, 39 no load, still not what id like, but now im at 3.65ghz, that extra 800mhz should come in handy... i couldnt get it to run through 3dm01, but 32m pi is okay.. so whatever, i think i need a new copy of 3d01...

im wondering this would it be better with HT off? that way i can have one app using all 100% right? in theory this would be better wouldnt you guys say?


btw UnG, you have any experience with a clip on hsf like an xp-90/120? id like to take the spreader off but im afraid of the mounting, number ove because i would have to cut down the holding plate, and number two because to mount these things, you clip one side of the hsf, and then press down the other, not exactly even pressure...

hmmm...

[Ugly n Grey]

Yeah I have done that before rev with a clip on, but I always, always used a shim that prevented the core from being damaged. I never was able to buy one, so I had a couple made in town at the mill in the high school. I generally do not recommend this with a clip on type, though some guys do it. If your proc is the only one you have, I wouldn't do it.

HT: turning it off doesn't actually improve anything. The visual representation shown by Windows screws people up a bit. When program manger is showing that one virtual CPU is not busy, that really normally represents 10-15% of the chips capacity for loading instructions as still being available. Turning off HT doesn't let you use that capacity, it just removes the idle virtual processor from view and from use. So your machine won't feel as snappy with it off.

And those temps are pretty good for an Intelly :toast: Congrats :)

[Revv23]

im hoping for AS5 to settle, im still not that happy with them, but whatever its been folding for 5 hours now at 3.65 so i cant complain too much

hmm maybe i could cut use the 9700 shim or something...

ill put some more thought into it

thanks for the info UnG.