That's my plan. Not a couple but likely one more.
Folding@Home
- Thread starter The JoJo
- Start date
That's my plan. Not a couple but likely one more.
CougTek
Hairy Aussie
I saw it too. I'm now barely above you. I sure won't be able to compete with you if you add another card. I don't plan to upgrade my rig before at least Spring, if not Fall 2014. I don't want to add other cards as I find they produce too much heat. I'll probably get rid of two or thrree machines and replace them with a dual or quad socket server.
The heat in my house on the second floor is 100% electric so using the PC with GPUs isn't that bad in the winter time. I know that doesn't work for everyone though. Summer is a pain keeping the room cool but it's only one room with a window AC which isn't too bad.
Would you build your own 2-4P server or buy the guts from someone like supermicro in a 2-4U case? If you're planning to go as far as 4P, will you stay Intel or venture to AMD?
Would you build your own 2-4P server or buy the guts from someone like supermicro in a 2-4U case? If you're planning to go as far as 4P, will you stay Intel or venture to AMD?
I'd just stay with GPU folding rather than go 2-4P CPU. If I were to build a mega-system from scratch I'd get a MB with lots of PCI-e slots, a low-end CPU with at least one core per PCI-e slot, and a PS to exceed the expected power needs of all your GPU's planned then add GPU's as your budget permits. If you start doing multi-GPU systems then water/oil cooling makes it much easier. You can do it with air, but it is much harder to get and maintain stability.
If you are still interested in a multi-processor setup then Intel Xeons are very expensive compared to AMD Opterons but I'm not an expert at multi-CPU setups. A better place to get info is HardOCP. They have many people there that have specialized 4p folding setups and know what they are doing.
If you are still interested in a multi-processor setup then Intel Xeons are very expensive compared to AMD Opterons but I'm not an expert at multi-CPU setups. A better place to get info is HardOCP. They have many people there that have specialized 4p folding setups and know what they are doing.
Last edited:
CougTek
Hairy Aussie
I don't know what dual or quad core I'll end up buying because I don't know when I'll have the funds to buy it. I find that GPU folding produces too much heat overall, which is particularly annoying in summer. While the hardware vs ppd production ratio is better with graphic cards, the wattage vs ppd production is better with big units CPU folding.
Unless AMD really pulls a rabbit out of their hat, I don't plan to opt for an AMD-based SMP server.
I'll probably choose a rackmount server of some kind. I've been impressed with Fujitsu's gears during our server purchasing study (we ended up with HP DL 380p's because my stuborn boss didn't listen to me and now with regret it because the SAN part doesn't do hardware replication between sites like we originally wanted). However,, since I won't be buying a server farm in the six-figures, I doubt I'll be able to deal the price down, so going for a big-OEM solution of any sort have little chance of fitting within my budget. This leaves Supermicro and... that's about it. I've read a scary number of horror stories about Asus' server boards issues so unless I get an incredible deal on one of their 2-P motherboard, I won't go that road.
What I'd dream to get is one of Supermicro's Twin² 2U rackmount server and gradually fill it with processors, RAM and SSD. We'll see.
Unless AMD really pulls a rabbit out of their hat, I don't plan to opt for an AMD-based SMP server.
I'll probably choose a rackmount server of some kind. I've been impressed with Fujitsu's gears during our server purchasing study (we ended up with HP DL 380p's because my stuborn boss didn't listen to me and now with regret it because the SAN part doesn't do hardware replication between sites like we originally wanted). However,, since I won't be buying a server farm in the six-figures, I doubt I'll be able to deal the price down, so going for a big-OEM solution of any sort have little chance of fitting within my budget. This leaves Supermicro and... that's about it. I've read a scary number of horror stories about Asus' server boards issues so unless I get an incredible deal on one of their 2-P motherboard, I won't go that road.
What I'd dream to get is one of Supermicro's Twin² 2U rackmount server and gradually fill it with processors, RAM and SSD. We'll see.
timwhit
Hairy Aussie
My PPD is now higher than Dave's. I will overtake him in just 104 years.
ddrueding
Fixture
Sweet. I'd forgotten to restart in a couple days 
CougTek
Hairy Aussie
WTF? My CPU clients are all stopped and waiting for a new assignment. I don't have time for this as I'm late for work. F*ck!
Internet down?
timwhit
Hairy Aussie
Can you remotely administer it after you get to work?
timwhit
Hairy Aussie
Sweet. I'd forgotten to restart in a couple days
It still looks like you forgot to restart.
ddrueding
Fixture
Running again. Started the 8.1 update before I left for the weekend.
CougTek
Hairy Aussie
I have no way to connect to those systems from work, nor do I want to. When I'm at work, I work. I'm not paid to configure my home computers at work.
When I came back this evening, I restarted one system, delete the CPU slot and recreated it on the other and all of my systems are again fully operational.
When I came back this evening, I restarted one system, delete the CPU slot and recreated it on the other and all of my systems are again fully operational.
timwhit
Hairy Aussie
You don't get a lunch or a 5 minute break?
timwhit
Hairy Aussie
I started folding on an older system with a Q6600 CPU. It appears to do about 5000 PPD. Is this worthwhile or is it just a waste of electricity?
I also setup the folding client on a laptop with a GK107 (Quadro 1100M) GPU, but it doesn't seem to be able to grab a work unit. From 2 minutes of Googling it looks like this GPU is no longer supported. Does that sound right?
I also setup the folding client on a laptop with a GK107 (Quadro 1100M) GPU, but it doesn't seem to be able to grab a work unit. From 2 minutes of Googling it looks like this GPU is no longer supported. Does that sound right?
5000 PPD give 1M in a year. My first year folding I produced 60,000 points with a dual CPU folding.
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.
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.
CougTek
Hairy Aussie
I made graphs showing the monthly production of the top six team members. I've tried to keep everything on scale. I didn't include the other members because their output wouldn't be easily visible on a similar scale graphic. I haven't compiled data before October 2010.
I know, the image is quite large. I thought it could be informative to some to see the point breakdown through time.
Also, at this moment, me and P5-133XL have done almost exactly the same amount of points since October 2010. We are separated by 6.08 million points, just like we were back then.
I know, the image is quite large. I thought it could be informative to some to see the point breakdown through time.
Also, at this moment, me and P5-133XL have done almost exactly the same amount of points since October 2010. We are separated by 6.08 million points, just like we were back then.
ddrueding
Fixture
Thanks Coug, I love me some data. My performance could most charitably be called "inconsistent".
CougTek
Hairy Aussie
Yeah, you're the home run type (with lots of strike outs too).
timwhit
Hairy Aussie
I didn't make the chart. :-(
CougTek
Hairy Aussie
This is a chart with three years worth of data. You've started this month. Also, each line represent a million points per month, meaning you wouldn't reach the first bar of the graph. It doesn't mean that your contribution isn't important (it is). It only means that you aren't sick in the head haven't been invested in the F@h project for as long and as deeply as some other members. But we still need and greatly appreciate your contribution. With your input, as well as StarBlazer, the return of Ddrueding and the arrival of BenR, we've been able to overtake PNY Folds, gain a lot of ground on TKWSN and keep Folding For Ancient at bay. That's something.
I'll make another comparative chart with a few members on another scale. Just give me a moment.
I'll make another comparative chart with a few members on another scale. Just give me a moment.
CougTek
Hairy Aussie
There you go. The scale is different than the first one. The charts are not as telling because the contribution of the members on that chart has been less steady. I omitted Splash because he hasn't posted anything since I've started to compile data. There was no place for Santilli in a six spot chart, but he would have been the next one to make the cut. His best month was ~225000 points, so he would register on the graphics.
The charts will be more informative for Timwhit and StarBlazer in a few months at least. For now, it just shows they've had a nice start.
The charts will be more informative for Timwhit and StarBlazer in a few months at least. For now, it just shows they've had a nice start.
Can you put them all on one scatter chart? It's a bit confusing with the different y axes.
CougTek
Hairy Aussie
If I keep putting pictures on the forum, Handruin will start to charge me for the used bandwidth.
You could use the png format for such graphics. The files are much smaller and appear cleaner.
CougTek
Hairy Aussie
Ok, at the risk of being moderated by the admin, here are the charts you asked :
CougTek
Hairy Aussie
I must have been moderated, because I cannot post the other chart for the moment.
Someone can delete the original ones if they want. I screwed up Fushigi's chart as the scale is wrong.
Someone can delete the original ones if they want. I screwed up Fushigi's chart as the scale is wrong.
CougTek
Hairy Aussie
Done. I won't post any other picture, jpeg or png, for quite a while. Sorry for the bandwidth. Feel free to delete at any time.
mubs
Storage? I am Storage!
OT: Fushigi and Pradeep seem to have completely disappeared from the forum
CougTek
Hairy Aussie
So.... We have our new servers at the job and we need to stress test them to insure stability, plus we want to know what's their maximal power consumption. We don't have our Windows Server licences yet, so I have the green light to install Linux and stress test them with whatever I want. Guess what I'll use?
Since these are not "my" machines, I originally thought it wouldn't be fair if I configured them on my account. However, if I create a new user just for them, we'll lose points for the first 10 units utill the passkey code allows bonus points. IMO, our team needs the points first and foremost. I'll just run them until tomorrow morning under my account. It's just 3 HP DL380p with dual E5-2690v2. Running them for ~15 hours probably won't boost my stats too much. I don't expect to get a million points from that.
You have about an hour to an hour-and-a-half to submit any objection before I start the client on all three.
Since these are not "my" machines, I originally thought it wouldn't be fair if I configured them on my account. However, if I create a new user just for them, we'll lose points for the first 10 units utill the passkey code allows bonus points. IMO, our team needs the points first and foremost. I'll just run them until tomorrow morning under my account. It's just 3 HP DL380p with dual E5-2690v2. Running them for ~15 hours probably won't boost my stats too much. I don't expect to get a million points from that.
You have about an hour to an hour-and-a-half to submit any objection before I start the client on all three.
ddrueding
Fixture
You secured the computing time, it is only fair to run them on your account. If you felt like buying a bunch of time on AWS or secured time on some other supercomputer, that would be your contribution as well.
Technically, the folding EULA requires permission of the owner or your boss, preferably written, to fold on someone elses machine. There have been people fired for misappropriation because they folded on company HW without permission. Also, permission makes it much easier to deal with IT getting through such things as firewalls or proxy servers. Written permission is good protection against office politics or a change in administration causing problems.
CougTek
Hairy Aussie
Just for your information, those Proliant DL380p Gen8, configured as I wrote above, consume approximately 450W each while running the FAH client. I'm not sure I've configured them right because I've had all sorts of issues installing a modern version of Linux on them. That and one of them has a very flaky FC cable that gave me trouble for a while. I've finally configured the config.xml file manually using vi (a PITA). I'm sure the client is running on all three. I'll less sure I'll have the credits for the work that's being done and much less that I'll get Big Units. The points will tell. Without the Big Units and simply doing a regular SMP WU (a 7808 and a 9002 IIRC), the rating was a bit above 80K ppd each. I've had a project 8104 on the third server. I think it's a Big WU, but I'm not sure. I also don't know if it will complete it by the time I stop the clients tomorrow morning.
Bigadv's use the A5_core so it is easy to find out using Psummary: p7808 and p9002 are both A4 cores so they are standard SMP while p8101 is an A5 so is a bigadv.
Even in Linux, you should be able run FAHControl to configure the client rather than manually editing the Config.xml assuming you have a graphical interface and downloaded and installed FAHControl.
Even in Linux, you should be able run FAHControl to configure the client rather than manually editing the Config.xml assuming you have a graphical interface and downloaded and installed FAHControl.
CougTek
Hairy Aussie
I'm perfectly aware of that, but in all three cases, once I've installed FAHControl, following Stanford's online guide, it failed to appear after I clicked on the shortcut. I was already past due to leave the office, so instead of busting my head trying to debug FAHControl, I chose to do what I felt was within my skill set and I manually edited the config.xml file. Looking at my points, it seems to have worked, somewhat (I did between 50K and 70K more than I expected to do on my own systems for yesterday).
I'll stop it in about four or five hours. I won't do any other troubleshooting on it since the hard drives will be formated and another OS will be installed.
Thanks for your input.
CougTek
Hairy Aussie
I've stopped one of the clients around 8h30am and the two others a bit before 10am. 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'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?
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?