Folding@Home

[Handruin]

I'm considering chipping in, but I have no clue where to start or if it will even be worthwhile. I have a computer I use for gaming that has a 3770K but it's not idle much during the day. I do tend to leave it on the rest of the time. It has a Radeon 7800 in it.

I also have another computer with a E8400 and Radeon 4800 that's not being used except a few hours a day.
Both computers are running Windows 7 64Bit OS.

Would these produce anything worthwhile, and if so what do I need to load on them?

Any contribution you're willling to give would be beneficial to the project and to this team. You might be able to get 20K->30K points per day out of the 3770K. The GPU support whitelist has a few hits on the cards you mentioned. You can try installing the client software (see below) and see if they detect them properly.


To get started,

1. Download the F@H client software here.
2. Register to get a passkey here. Don't share this, it's for you only.
3. Install the client software on all machines you wish to use. Enter team number 10047 to help storageforum fold. Enter your passkey.

 

[P5-133XL]

This will at least get you started:

F@H Home Page


1. Download the Windows v7 client
2. Get a passkey (Make up a username using Alphanumerics only)
3. Install the v7 client using your username, Team 10047 (StorageForum), your passkey (Emailed to you - Use copy and paste).

Folding FAQ index
Folding support Forum

 

[Handruin]

Striker is on the charts! Thanks for joining, we're happy to have you contribute to effort!

 

[Striker]

Thanks for the help.

 

[Handruin]

I wrote my buddy (BenR) who joined but has low points and asked if he could get a passkey to see if it might help. I also sent info to another friend to help him join our team. Hopefully he's interested enough to add one more to the effort.

 

[timwhit]

The 1100M should be foldable under both CUDA and OpenCL folding cores (Core_15 and Core_17). It's basic architecture is Kepler (GTX 5xx) based which is supported. The big issue with GPU folding in notebooks is with overheating issues. so be watchful. Also make sure you have modern drivers with OpenCL support. I would expect 6000-10,000 PPD.

I just got back from vacation last night, so I wasn't able to set this up until now. I added the GPU slot and set client-type to advanced on that slot. It pulled an 8900 WU, but it looks like it's going to take about 3 days to complete this WU. It doesn't really seem worthwhile, as it will only be producing less than 3000 ppd. Thoughts?

 

[P5-133XL]

FAHControl's estimated PPD is not accurate especially with Core_17: HMF.net works much better. My observation is FAHControl's PPD estimates start out inaccurate then waffle back and forth between way to high and way to low with it actually in the middle.

Core_17 has been released to general-release so you do not need advanced anymore. For low-end or part-time GPU's it is actually significantly better PPD-Wise to get the older Core_11/15 projects. PG knows this and has claimed to be working on a solution on the problem.

 

[CougTek]

I'll have another run at it in a few weeks when we'll have to convince our boss to insonorise our server room.

I did wait too long before restarting. Long story short : my boss entered our small room (where the rack is...three feet behind me) and told us "Hey, the noise's not that bad!". The SAN is not even powered up, so even at idle, the base noise level will be significantly higher than it was with onyl three idle server. It was obvious that he wasn't going to improve our office's sound-proofing, so I fired up just one of the clients and the remaining two a few hours later. With the SAN plugged, it's going to be that noisy or worse anyway. The guys in the offices next to ours asked us if we've opened a barber shop since it always sounded like a bunch of air dryers were working.

Shortly before the end of the day, the building owner strangely came by to check what he could do to improve the isolation of the server room's doors.

Now I'll just finish the BigWU that have been started and then I'll stop for good. I may not be Mr. Nice Guy and I often bend or break non-official rules, but I'm the guy that gets the job done. Thank you Stanford for my new sound-proof doors.

 

[CougTek]

Plus, spending some additional time working on the server helped me find a and fix a networking problem one of them had. I've also learn how to configure an open-ssh server on Linux for the first time (so that I can monitor the FAH instances from home). So it wasn't completly futile on a technical point of view.

 

[P5-133XL]

Right now PG is having problems with their network and it is not assigning WU's. The point is that it is not anyones config, computer, or network it is them so don't waste your time trying to figure out what the problem is. The client will periodically query PG till it is resolved and then automatically get a new WU without intervention on your part.

 

[Handruin]

Thanks for the update.

 

[timwhit]

What does PG stand for?

 

[LunarMist]

What does PG stand for?

http://www.pg.com

 

[P5-133XL]

No not Procter and Gamble: Pande Group is the group of researchers that is doing F@H under the auspices of Vijay Pande.

 

[Handruin]

What does PG stand for?

I believe it stands for Pande Group.

http://en.m.wikipedia.org/wiki/Vijay_Pande

 

[P5-133XL]

The assignment servers should be issuing WU's again...

 

[Handruin]

Thanks for the help.

It looks like your passkey has kicked in and you've completed the 10 work units! Now you're points have tripled! Thanks for continuing!

 

[Handruin]

Welcome back Bozo! Glad to have you producing again. I like seeing the longer list of team members in the stats! It has been a while since we've seen this many producing at the same time.

 

[CougTek]

According to my stats, the last time we were that many was in November 2011. We had ten members contributing at various levels back then.

 

[CougTek]

...and there were nine in December 2011. Missed that.

 

[timwhit]

I would like to configure one of my systems to fold when idle. I moved the slider to "Idle," and then left for 20 minutes to go to lunch. I came back and it hadn't done anything. Is there any trick to getting this to work correctly with Linux?

 

[P5-133XL]

Setup a screensaver: It's the timer that folding uses to detect idle. In Windows, The monitor power down is also used as a signal, but I don't know if that is also true for Linux (I don't have a Linux GPU folding machine to test it out with).

 

[timwhit]

This is how my current setup looks:



I'll try modifying it and seeing if it works.

 

[timwhit]

I setup a screensaver, but it doesn't appear to make a difference. I also tried restarting the screensaver and FAHClient services. I guess my only option is to remember to enable it. Not really ideal.

 

[timwhit]

Congrats to CougTeg for taking the lead.

 

[P5-133XL]

Congrats.

It happened much faster than I thought it would but it was inevitable.

 

[CougTek]

Thanks, but you all know it's not because of my own machines. Let's just say that I'm doing this while performing tests on the new servers setup (replication, remote access and monitoring, backup recovery, making sure the room's cooling system can keep up, etc). Since so far, I'm the only one who's configured the new servers, the more tests I do, the better I'll be when the production systems will shift to the new hardware. I also have a particular reason not to care about what it might cost the company (honestly, for the two weeks duration the VM will run, it will only cost them ~30$ so it's not the end of the world).

I may add up to four other 18-core VM tomorrow. So far, the room's temperature control easily manage to keep the temperature down with 3 servers running at ~80% load. However, we have 5 servers plus two SAN (one with 40x 2.5 15K drives and the other one with 32 2.5" 15K drives). Those are not yet powered up. I suspect they will generate a significant amount of heat. Of course, our servers won't get anywhere near an 80% load with our typical operations going on, but in order to simulate the added heat from the soon-to-be-turned-on SAN, that's the best I can think of. My take is if everything goes smooth with 5 busy servers, it will go well with 5 semi-occupied servers and two medium-size SAN.

My production will decrease tremendously somewhere between January 5th and 6th.

 

[timwhit]

You'll have a very nice lead by then.

 

[P5-133XL]

As far as I can tell, the entire folding network just shut down. Santa is giving folding computers worldwide the gift of relaxation. For them it has become a SPA day.

 

[P5-133XL]

As far as I can tell, the entire folding network just shut down. Santa is giving folding computers worldwide the gift of relaxation. For them it has become a SPA day.

Whatever happened, fixed itself and it is all working now.

 

[CougTek]

So since yesterday's afternoon, I have a farm composed of 10 18-virtual-cores VM working full-time for the FAH project. I expect to average ~1.4 million ppd. That's good. However, given the insane hardware I've (temporarily) put to use, I have a hard time figuring out what a member like Jim Gluftsis (Apple Inc. team) is using to achieve close to 10 million points per day. He would need something like 50 R9 290X to get that and the power needed would be aweful.

 

[CougTek]

We'll be close, but it doesn't look like we'll be able to overtake NCSU Wolfpack before the end of the year (the end of the year here, on the East Coast, at least). We'll end 2013 at our highest ranking since about 2005-2006. We've been 117th overall earlier this year too, before we lost 6 positions to other teams, so of which we've been able to re-overtake since.

We'll do almost 40 million points this month. It's been a great year for us and a great month too. I don't think we'll have such a good month in the first half of 2014 either. It will have to wait until Handruin gets tired of making money with his R9 290X and another member or two upgrade their hardware too.

As for me, I'll enjoy the team's top spot for most of the year I think, but I've calculated that when Handruin focus all his ressources on the FAH Project again, he should be able to overtake me in 250 to 300 days. And that's only if P5-133XL doesn't upgrade one of his GTX 460 with something more recent. He's not very far from me and could easily pass me again with a modest upgrade. One thing is sure, I won't lead the team for some 9 years like he did.

 

[CougTek]

So that's it. The last of my big units has been uploaded to Stanford's servers. All the temporary virtual machines have been shut down and will be deleted tomorrow.

We should still gain some positions, albeit slowly. Our global production, without the temporary servers, should place us around the 80th rank by daily output. With the servers, we were 38th.

 

[P5-133XL]

For anyone running AMD GPU's, there are reports that the new 14.1beta1.6 drivers include a new version of OpenCL which is adding a bunch of PPD. To give numbers, an extra 29K on a R9-290X, or 15K on a HD7870.

 

[CougTek]

Has there been any significant improvement regarding AMD GPU support on Linux with the FAH client? Last time you wrote about it, AMD GPU support for Linux was suppedly flaky. I've stopped using my Radeon HD 7950 because one of the fans on it became noisy. I've also stopped using the Radeon HD 7870 because the Windows 2012 Server beta licence has expired on the system it was installed on and I've been too lazy to reinstall another OS. I'd like to use Linux on it, but only if the FAH GPU client for AMD is reasonnably reliable. Otherwise, I'll install another Windows OS that will be good for 60 days.

 

[Handruin]

For anyone running AMD GPU's, there are reports that the new 14.1beta1.6 drivers include a new version of OpenCL which is adding a bunch of PPD. To give numbers, an extra 29K on a R9-290X, or 15K on a HD7870.

That's a decent improvement. I'll be happy once I switch over from mining to folding to get the extra PPD on my 290Xs.

 

[P5-133XL]

Has there been any significant improvement regarding AMD GPU support on Linux with the FAH client? Last time you wrote about it, AMD GPU support for Linux was suppedly flaky. I've stopped using my Radeon HD 7950 because one of the fans on it became noisy. I've also stopped using the Radeon HD 7870 because the Windows 2012 Server beta licence has expired on the system it was installed on and I've been too lazy to reinstall another OS. I'd like to use Linux on it, but only if the FAH GPU client for AMD is reasonnably reliable. Otherwise, I'll install another Windows OS that will be good for 60 days.

I haven't yet heard of anyone testing to see if 14.1beta works for Linux. It is the first driver version with a different OpenCL so maybe that's the one to make a difference. Unfortunately, I don't know if the beta 14.1 is even available for Linux yet.

 

[CougTek]

It seems to be available. I'll give it a try later this evening.

 

[timwhit]

Dave, reminder: you are not folding again.

 

[CougTek]

Dave is folding plenty. He's folding diapers all day long.