OFFICIAL Holloween Race Roster & Details

[road-runner]

3xQ6600@3.33ghz

Which is better for project 2665?
These projects study how influenza virus recognizes and infects cells. We are developing new simulation methods to better understand these processes.

A. dual smp = 144 wu's per month.
B. single smp = 108 wu's per month.

forget the points.. by the monthly numbers above...
how many months will the project take?
1. with single smp? 108wu/month
2. with dual smp? 144wu/month

it all boils down to simple math... the faster they get through the results, the faster the project gets done.

I agree Mike as you know. Stanford acts as if they need that one result back a minute quicker as they are on the brink of a major break through in science. To be honest I used to think this way also, after years of this I have seen nothing really, people ask me all the time why do you spend all that money on all the computers, time, and electricity, what have they done? You spend a house payment or car payment a month just to power all those. What can I show them? Nothing really that means anything to anyone, I just hope someday it will. I also understand they want them back quickly dont get me wrong, however I think the main reason they say on do 1 WU on a quad is for people who have no knowledge on PCs and go out and buy say a Dell etc. quad core and decide to fold on winblows at stock speed, while stopping the machine half the time to play games and barely make the deadlines if they do at all.

[MikeB12]

yeah, I understand that part completely, Marvin. quicker = better. I just don't see how 10hrs makes a difference to a project of this size. They haven't even reviewed the results before I've spit 10 more out... these projects dont run on a time frame that small. they run on monthly and longer schedules, so in the end the more completed wu's they get back in a period of time, the faster the project reaches the end of it's life cycle. these projects are not running on hourly and half day schedules...

but I definitely understand the basic concept of quicker turnins = quicker results. the concept just doesn't apply to the amount of work I turn in and the time frame at which I turn it in on my hardware. I doubt there is some lab guru sitting there saying, "where is mike's wu, we just gave it too him 20 hrs ago... oh there it is at 30hrs, he's 10 hrs late... but looky looky, he turned in 2 instead of 1. YAY!!!"... that time frame doesn't apply to this scenario.

[Marvin_The_Martian]

yeah, I understand that part completely, Marvin. quicker = better. I just don't see how 10hrs makes a difference to a project of this size. They haven't even reviewed the results before I've spit 10 more out... these projects dont run on a time frame that small. they run on monthly and longer schedules, so in the end the more completed wu's they get back in a period of time, the faster the project reaches the end of it's life cycle. these projects are not running on hourly and half day schedules...

but I definitely understand the basic concept of quicker turnins = quicker results. the concept just doesn't apply to the amount of work I turn in and the time frame at which I turn it in on my hardware. I doubt there is some lab guru sitting there saying, "where is mike's wu, we just gave it too him 20 hrs ago... oh there it is at 30hrs, he's 10 hrs late... but looky looky, he turned in 2 instead of 1. YAY!!!"... that time frame doesn't apply to this scenario.

If someone from PG will reply to you that you're wrong, and offer an explanation I'm unable of giving due to my lack of exact knowledge, will that change how you run the project?

Because I think I'm going to try for that to happen. Not a folding forum mod or admin, but a direct member of the PG who is involved with the analysis we're talking about as I hope their word carries ( obviously ) more weigth then mine.

And if I'm wrong, they can tell me that and I'll go dual smp even at 3ghz

[MikeB12]

they wont do that because they dont support dual smp. it was hard enough to get people to use single smp without everyone complaining about it and saying "this is a bunk client"... to support dual smp would be a whole new nightmare of complaints. at least they've accepted people use it to supplement's a1 core's inefficiencies. even some of the mods at fah run duals. but to publicly come out and put an offical approval on it-it'll never happen for political reasons.

now if you can expedite the the release of a2 core to windows then I'll buy you a humongo bag of the finest greenery...

but on the dual smp subject I think we can pretty much agree to disagree and move on. no need to argue about it... some people run it and some people dont, that's the bottom line.. and everyone is entitled to an opinion.. in the end realistically none of us here have a microgram of influence over how fah runs there junk. if pg really wanted to stop people from running dual smp's they could, just code it into the client to only allow one client per cpu.

[Marvin_The_Martian]

1. Point would not be about supporting a client or not, point is to have him explain how the project works. I'm not sure about the months you mention it takes to run a range of wu's I think it's less then that and therein lies our disagreement. If it were a long enough time anyway, you will not do harm with dual smp because you're not slower then the abovementioned quad dell office pc running stock. But if the project runs for let's say a week or less before the results need to be analyzed and new wu's send out, then anything which introduces lag is bad. I already send a pm, so I can't take that back.

2. Can't expedite that, me talking with people != people listening to me ( well... you know what I mean ) And btw, those green plants, I'm cutting down, been clean a few days

3. Well I was thinking I'm ready to sway your side if they tell me I'm wrong, but I don't need you to feel the same we can disagree on how we fold that's just a personal take I agree. Btw, their cpu core logic is not that good ( if it where, you wouldn't need to run AC with dual smp's ) and they couldn't hard code a limit of one per cpu as in the not to far future we will have six core cpu's. If it where as easy, they would have prevented it already just to make it easier to debug the current clients.

[MikeB12]

actually I wish they would advance forward past this multi instance fahcore junk... seems like by now they would have one executable that could load balance across all cores whether it be a 2,4, or 8 core system. isn't that the true meaning of smp (symmetric multiprocessing).... this multiple instance of fahcore_a1 is not true smp. it's just a bandaid of single core technology where the client launches multiple executables... but I guess that's a ways off unless they pull in some high tech $$ programmers...

[Marvin_The_Martian]

actually I wish they would advance forward past this multi instance fahcore junk... seems like by now they would have one executable that could load balance across all cores whether it be a 2,4, or 8 core system. isn't that the true meaning of smp (symmetric multiprocessing).... this multiple instance of fahcore_a1 is not true smp. it's just a bandaid of single core technology where the client launches multiple executables... but I guess that's a ways off unless they pull in some high tech $$ programmers...

Any single thread can only run on one core, smp means that the threads have bonds together and parallel tasks can be done much quicker since it only has to be executed once for each thread.

The clients will always launch multiple tjreads ( executables as you call them ).

What they need to finish is indeed the core logic and load balancing, which will hopefully allow them to utilize whatever hw the underlaying system has to it's exposal. If it's capable of seeing the cores, their relationship, cache's ect. they will be able to have it download wu's which are known to run the best on that hw, and the client will spawn as many threads as there are cores.

[MikeB12]

Modern professionally produced smp software can launch multiple threads per executable... they don't launch multiple exe's... it's the same exe instance, mutli threaded. I imagine fah would love to have it this way, but obviously there are pg project obstacles we dont understand that prevent making that kind of advancement either impossible right now or just not feasible. maybe we'll see that in the furtue, dunno. now they've got their hands full trying to get a2 windows stable so it reduces the leaks present in a1.

[Marvin_The_Martian]

We're mis communicating.

Fahcore_11.exe for instance is both an executable ( when you look at the file on the hdd ) and it can be a thread ( taskman ). So your fah.exe launches the executable fahcore_11.exe 4 times in a seperate thread. Think we agree on that?

Not even sure if we're disagreeing here or agreeing

[MikeB12]

Same theory just different ways of saying it... it really doesn't belong in the halloween race thread anyway... let's just end it and if we need a new thread for smp hypothetical discussion one can be opened... deal...

I make a motion for the floor:
the next person to bring up hypothetical smp discussion or the validity of dual smp vs single smp in this thread... will be:
chained and locked in Dak's basement for 1 week, then visited nightly after Dak's blood has reached .010 grey goose...

[Marvin_The_Martian]

If you left out the last part...

[skycrane]

Please dont take this the wrong way, but thank god its all worked out, yall lost me about 4 pages ago

now back on topic, i see my production is finally ramping up, left this morning and restet both gso's, just got home and see that both are folding away thanks for the help Marvin..

ive got another question, about how high should your temps get on a pair of gso"s ??? is 79c on the top one and 83c on the lower card good?? and yes they are stock

im able to get the temps down on the upper one to 52c now, had to manualy run fan at 100% it was only at 42
now ive got a 20"box fan sitting 8" away from the case, my temps are 52c on the upper one, 71c on the lower one.....

how do you adjust the fan speed of the lower gpu? im using evga precision tune, and it will only let me do the upper card.

[road-runner]

I just got my GSO's going a minutr ago myself, I see no fan control in rivatuner so I am looking for another way to adjust them myself....

[Marvin_The_Martian]

I posted in the other thread you mentioned it in, what driver version are you guys running?

Some are reported broken 177.84 is definetly good I can tell because I'm using it. I won't use latest drivers no more.

And guys, don't let windowsupdate install any drivers I was stupid enough to let one slip through.

[WFO]

I just got my GSO's going a minutr ago myself, I see no fan control in rivatuner so I am looking for another way to adjust them myself....

Riva tuner works great for fan control and overclocking.

Google is your friend. Here is a complete guide...

http://www.vaguetech.com/index.php?p...rt207#rivaconf

The Cliff's notes version of fan control for folding...

Select the monitor you want to control settings for from the drop down at the top of the main page...



Below the drop down menu for monitors, click the "customize" button. Click the picture of the graphics card. You will get this screen...



You'll have to click: "Enable Low level fan control"/ "Detect Now." Then you can select a fan speed. I run mine @ 100% from startup. In the settings tab on the Main page, you'll also want to check boxes for run at start up and send to tray on close.

Hope that helps.

Darn that Dak!!! What's a guy have to do to get a full days folding in? When I came home I saw him driving away in a car that resembled a Grey Goose bottle. And sure enough when I got inside the power had been shut off and turned back on. All my rigs were down and it took a full 30 minutes to get everything running and reset the OCs. My only full day of folding so far has been my warm-up lap.

Quit pissing me off Dak or I'll crank my OCs up even more, install and extra SMP and order parts to build that other rig that's almost complete!!!

[DAK1640]

Floor fans are always an option for GPU cooling...It's a good idea to remove your side case cover at minimum...

[road-runner]

I know how to work RT, I have fan options on my 8800GT's, however there is not even the option with these 9600GSO in RT...

[Kondik]

Fans on 100% ? Alright .. who said that will buy me a pair of earplugs if I leave my PC 24/7 ? Did someone of you hear GTX280 on 100% ? I know the temps will be extremly low [ ~ 50 ] but I'd have to bea deaf and probably dead to not hear it. Btw I'm keeping my PC on for today If I woke up in the morning exhausted ... Marvin start running I'll fly to Amsterdam for only one purpouse ... to skin you alive with my plastic spoon

[road-runner]

Check it out, no option?

[WFO]

Check it out, no option?

I noticed you've been trying differen drivers without luck. Do the cards overheat when folding? If so RMA and get Evga or MSI. Wish i could be more help. I don't have any Asus cards.

[road-runner]

I noticed you've been trying differen drivers without luck. Do the cards overheat when folding? If so RMA and get Evga or MSI. Wish i could be more help. I don't have any Asus cards.

I dont know if there overheating or not, GPUZ only detects the core clock and memory and nothing else...

[WFO]

I dont know if there overheating or not, GPUZ only detects the core clock and memory and nothing else...

/me wonders if your cards even have fans??? Sorry. bad joke. If riva tuner can't detect them nor GPU-Z, it seems something is very wrong. I hate to suggest Asus support... that or an RMA seem like the only possibilities. You've done everything anyone else would do to troubleshoot on your own.

[road-runner]

/me wonders if your cards even have fans??? Sorry. bad joke. If riva tuner can't detect them nor GPU-Z, it seems something is very wrong. I hate to suggest Asus support... that or an RMA seem like the only possibilities. You've done everything anyone else would do to troubleshoot on your own.

All I have to install is what I download from nvidia right? That is all I done with my 8800s, these are also on server 2008 so there may be something missing?

Edit: They are doing 5200 PPD on a C2D with 1-SMP client and only the shaders moved to 1846...

[WFO]

All I have to install is what I download from nvidia right? That is all I done with my 8800s, these are also on server 2008 so there may be something missing?

Edit: They are doing 5200 PPD on a C2D with 1-SMP client and only the shaders moved to 1846...

Nope. Cuda drivers from Nvidia are all you need.
I haven't used Server 2008 in 6 months. It was extremely limiting in AVs and other software. No problem with your current points for the GSOs. You would have to swap the cards to the rig with the 8800s and the 8800s to the Server 2008 rig to see if Sever 2008 was indeed the issue. With those points, it probably isn't worth the effort. If the cards die... the warranty should cover them.

[road-runner]

Ah hah, found it on the Asus disc... 5317 PPD each at those settings...