Build Log: 100,000 PPD by 1/23/2011

[Biker]

I'm at 2.3kwh with 11 machines....

I wish 1366 32nm chips were a bit cheaper.....

[DimmuKlokSC2]

I applaud all of you for your dedication to the cause.

[disturbed13]

I'm at 2.3kwh with 11 machines....

I wish 1366 32nm chips were a bit cheaper.....

god dont we all

[shoota]

I'm at 2.3kwh with 11 machines....

I wish 1366 32nm chips were a bit cheaper.....

i wish they were ALOT cheaper

[EtaCarinae]

trn yesterday you nearly caught me. You were at only 22'000 point from me.
693'000 for you and I was at 715'000.
We are having a hot fight. Need to launch my backup forces in the battle.

[trn]

trn yesterday you nearly caught me. You were at only 22'000 point from me.
693'000 for you and I was at 715'000.
We are having a hot fight. Need to launch my backup forces in the battle.

Bring it on! I still have a litte more to get up and running . I don't think I'll average more than 700,000 per day though. I was taking a shot at your WU per day record also. I think I'm getting about 2500 per day so still short there.

I'm hoping for good things in Q2 or Q3 of 2011, octocores

[disturbed13]

Bring it on! I still have a litte more to get up and running . I don't think I'll average more than 700,000 per day though. I was taking a shot at your WU per day record also. I think I'm getting about 2500 per day so still short there.

I'm hoping for good things in Q2 or Q3 of 2011, octocores

foaming at the mouth
and how is that dualie coming?

[karbonkid]

I like files and folders to be more BSD like. If boinc ran native on BSD without Linux emulation I would be running FreeBSD on my *nix crunchers. Hell, I still might even give FreeBSD a go, I hear the linux emulation no or almost no effect on boinc performance.

Sorry if this is old news, but I just saw this thread and reading from the beginning. Try Slackware or Arch...

Awesome project!

[Jaco]

I'm hoping for good things in Q2 or Q3 of 2011, octocores

or even better ... decacores

[EtaCarinae]

or even better ... decacores

Oh yes! A dualie with two decacores + HT = 40 threads in one shot.
And everything running at min 4 Ghz.
This would make a 200'000 ppd machine.
And in terms of Mips/watt that would be .............

Oh! sorry, here I am dreaming again.

[D_A]

I don't think they'll bother with decacore machines. More likely to go to 12 then 10.

[trn]

foaming at the mouth
and how is that dualie coming?

Badly, the asus mobo wouldn't run stable with A0 westmeres so I bought two more gigabyte boards, another ud3r rev 2.0 and a ud5 rev 2.0. I only had time to set one up and running at stock.

Sorry if this is old news, but I just saw this thread and reading from the beginning. Try Slackware or Arch...

Awesome project!

Thanks! I love Archlinux and am runnin it on 4 crunchers. I have ubuntu server on another 3 but that works ok so no real motivation to reformat them especially because I don't know how to carry over boinc data in linux like windows.

or even better ... decacores

Probably not for a desktop socket until 22nm in... Early 2013? 1366 westmere dies are full with 6 cores, I read socket 2011 dies are roughly the same size but the individual SB cores are a little smaller and laid out differently so they can jam in 8 cores. I'm sure there will be a server EX socket version but that will be out of reach $$$ for us non-billionaires. We can dream though...

And really the most important thing we as consumers need is for AMD to bulldoze Intel with some serious competition so Intel acctually has a reason to bring out better products at much lower price points. Intel could have dropped 45nm all together at this point and be selling locked down hexas for $300. The 32nm proscess is done and I'm sure their investment in the process is already well paid off by all the $1000 dollar chips already sold. But why drop your price pointa without competition, they really could care less about the enthusiast community when every quarter sets another record profit in a stagnet economy.

[Rise]

just wanted to say congrats on reaching your goal trn! looks like you are pulling 100kppd+ daily and well before your 1/23/2011 target date

[shoota]

yeah you should probably edit your target to 150,000 ppd

[trn]

Thanks guys but these last two weeks have Been like cheating. Ive been away from home so all my crunchers are running without any work or gaming interruptions. Also I've been running HCC exclusive which is another false boost, I have a 980x at only 4.0GHz that is averaging over 8,000 ppd with Linux. I prefer to run a mix of all projects, so without these boost i really haven't reached my goal of a solid 100,000PPD consistently. Thankfully I still have one more hexacore to build and one stock machine to OC which ought to put me very very close to my goal; a loop of SB's in janurary might give me a solid increase still awaiting some crunching results on those

(and that 1TFlop barrier is looking like it needs to be broken also )

[disturbed13]

999 Gflops
lol so close, i say you will end up with 1.1 Tflops

[shoota]

Thanks guys but these last two weeks have Been like cheating. Ive been away from home so all my crunchers are running without any work or gaming interruptions. Also I've been running HCC exclusive which is another false boost, I have a 980x at only 4.0GHz that is averaging over 8,000 ppd with Linux. I prefer to run a mix of all projects, so without these boost i really haven't reached my goal of a solid 100,000PPD consistently. Thankfully I still have one more hexacore to build and one stock machine to OC which ought to put me very very close to my goal; a loop of SB's in janurary might give me a solid increase still awaiting some crunching results on those

(and that 1TFlop barrier is looking like it needs to be broken also )

and this is how i know you're addicted

[trn]

Aurgh.... Had to bring down the farm today. Flow issues causing the CPU's to overheat.

Got home from being away for two weeks, and I noticed this morning that around half the farm was down. Initially blamed it on a brownout; some of the crunchers run on UPS's. Restarted them and just assumed everything was OK, then remote logged into one of the 3.7GHz L5640's which should have probably been in the 50C load temps... temps were reaching the 80's

Return lines to the RAD's are cold as can be but some parts of the tubing are really soft and warm to the touch. I'm guessing that whatever is turning my water brown finally jammed up the CPU blocks and reduced flow.

Time for a complete tear down, flush and look at my components to see if I can diagnose why the water keeps getting browner and browner with time. Unfortunately not what I was in the mood for, jet-lagged and sick and I was feeling like having a relaxing weekend. I hope I didn't burn up and CPU's

I'm the contaminate issue has to be with something in my loop since it keeps getting worse and worse with time. I have a large mixture of materials, PVC, nickel plated steel, a small bit of galvanized steel, 2 part epoxy, vinyl tubing, brass, copper, whatever the hell is in my water circulator pump, PVC cement, and distilled water with PT Nuke. I'm guessing that something like the 2 part epoxy or PVC cement is being disolved and deposited all over the inside of the loop. Anyways... what a serious PITA. Does anyone know if PT Nuke has any sort of corrosive properties to plastics or anything like that?

[Jaco]

Sorry to hear that trn , that looks like a major downtime.
I don't PT nuke will cause this ...

[OldChap]

I liked the principle of this so much I just ordered a rad. About 32 inches square and 1.6 inches thick it is already refurbished (new core) so over the next few weeks I can have some fun too.

I won't ever have the number of rigs on it that you have but I hope this will simplify the cooling arrangements that I have just now.

Now then on the subject of blocked up system how about you break this into two ...the computer loop and the rad loop and in between put a plate heat exchanger. It is what I have planned so that I can disregard dissimilar metals and can mess with the rad side at my leisure without having to worry about contamination. A little less efficient and needs another pump but worth it in my book.

I got mine from http://www.dudadiesel.com/heat_exchangers.php

[trn]

Thanks for the tips, I think i'm going to take my time fixing this so I can get it right this time and be setup for the long-haul. I plan to disassemble the WC system this week and have a good look at the blocks, pumps, and all the materials and see what I can do to solve the problem.

[D_A]

I'd hazard a guess there was some oil inside that main pump but the amount of dissimilar metals in the system would concern me over the long term.

[OldChap]

That brown look does remind me of the blown head gasket type oil/rust/coolant mix you can get. check out how much you would have to pay to have those rads professionally cleaned or re-cored...I have a feeling in my water that the issue lies there.

[trn]

When I removed some Vinyl tubing from the loop, the brown gunk dries up to a very very light colored brown; almost white when dry. The thing that I don't get is that the issue seemed to keep getting worse and worse instead of leveling off. And that would be a lot of oil! the brown coating is on everything in my loop; and i'm guessing it finally got bad enough to gunk up the cpu blocks and restrict flow. I'm hoping when I tear down I can find the culprit. I really really wished I had built in an easy method of flushing and draining and filling the system now. This is going to take a complete tear down; and if the blocks are jammed up its going to take a bit of boiling in vinegar water to clean them out or something like that.

[jspace]

I learned that regular epoxy doesn't stand up to water very well the hard way. I have and MD-20 which has 3/4" male threads on it. Since I use 7/16" tubing I needed some way of getting the tubing on more than the first 1/4". What I did was take some 1/2" brass barbs and epoxy them to the poly barbs. It worked pretty well and I was proud of myself for figuring out a more elegant (and compact) solution than poly 3/4"-1/2" reducers.
Fast forward a few months and there is a puddle under the computer and the horrible sound of bubbles moving through everything. Fortunately the pump was on the floor of the case so no magic smoke got out, but it was annoying nonetheless.
My point: spend the extra dough and get epoxy that works for aquariums and stuff.

I agree with D_A on the dissimilar metals thing.