Milkyway@home - September update

ReaVerUK · 30 Oct 2010 at 16:35

I thought it deleted any unwanted .exe's but wasn't sure.

ReaVerUK · 31 Oct 2010 at 00:43

CGS said:
let me know how you get on dont think i have any exe's other than cuda one for gtx460

OK i've got it crunching away with the stock app for CPU and the fermi app for GPU. I'm not sure it'll be worth it tho. Its looking like ~13Hrs per stock WU compared to the 0.21 app at 2.5-3Hrs

The next question is what is the stock app for 32bit apps? It may be the same for both but i'm not sure. The file i get is milkyway_0.40_windows_x86_64.exe. So does that mean its both versions or 64bit only? :confused:

There doesn't seem to be another 0.40 app listed.

ToxicTBag · 31 Oct 2010 at 07:31

Morning Mr Reaver sah, going by what it says on arkayns page and the mway forums they are in the process of updating the apps so i would have thought there'd be something new out fairly soon.

ReaVerUK · 31 Oct 2010 at 08:01

Mornin
Yeah i noticed that as well whilst checking just nice to know it can be done if need be

me likes to play

ToxicTBag · 31 Oct 2010 at 12:42

Yup i know how that feels

CGS · 31 Oct 2010 at 14:57

ReaVerUK said:
OK i've got it crunching away with the stock app for CPU and the fermi app for GPU. I'm not sure it'll be worth it tho. Its looking like ~13Hrs per stock WU compared to the 0.21 app at 2.5-3Hrs

The next question is what is the stock app for 32bit apps? It may be the same for both but i'm not sure. The file i get is milkyway_0.40_windows_x86_64.exe. So does that mean its both versions or 64bit only? There doesn't seem to be another 0.40 app listed.

Got mine to work too on gtx460 and i7

The win7 32bit uses milkyway_0.4_windows_intelx86.exe this seems very slow rekon 12hrs+ per wu.
Hopefully optimised version will be out soon when they update the apps maybe with sse3+ support
Thanx for help ReaVerUK

CGS · 2 Nov 2010 at 07:18

Just a quick one might be me.
has anyone else noticed wu's are taking longer
my ati4750 was taking around 300secs/wu lots of em now are in the 1000secs +
is this due to a change that im not aware of ?

kolala · 2 Nov 2010 at 08:40

could it be this?..

Posted 1 Nov 2010 15:57:21 UTC
I've updated the Windows and 32 bit Linux applications to mostly fix the very slow workunit problem. The problem ended up being in the math libraries, which affected the Windows applications as well as 32bit Linux.

This should also fix the workunits failing on systems without SSE2 from the previous builds. I had built the BOINC libraries before with SSE2, since it was required for the N-body. However, since it is irrelevant if they use SSE2 or not, it only ended up polluting the non-SSE2 separation binaries.

Non-SSE2 versions will still be quite a bit slower. The SSE2 builds (which also covers 64bit) should be in the range of 25-30% faster than the old 0.19s. The 32bit Linux version is now using crlibm, and will be slightly slower than it should be now as a quick way to get math using SSE2. The 64bit Linux + clang build is still the fastest, but only by < 10% now over the Windows + SSE2 version.

CGS · 2 Nov 2010 at 12:24

kolala said:
could it be this?..

Ive detached and reattached still get 0.23 for ati gfx and 0.45 now for cpu
The ati gfx units now look fine about 200ish secs the cpu units seem about 10hrs + so pretty pointless really for cpu anyway
Think the optimised ati drivers for 0.23 were messing up the cpu workunits and slowing everything down

Update looks like cpu workunits will be around 30hrs now on c2d cpu

BaBa · 2 Nov 2010 at 14:08

Have to show these off while i can, not too many more to come

Must say seti being down has done MW the world of good :cool:

edgedemon · 3 Nov 2010 at 06:02

BaBa said:
Have to show these off while i can, not too many more to come

Must say seti being down has done MW the world of good

Nice tiffy s'ah

Seti being down has helped, but the real difference has been the release of the fermi app for MW, if it were'nt for that, a lot of us would be crunching einstein..

ToxicTBag · 3 Nov 2010 at 09:20

Amazing tiffy BA

kolala · 3 Nov 2010 at 19:12

Its only worth 1/8 of a ba but I'm proud of it anyway

BaBa · 3 Nov 2010 at 20:31

Very nice tiffy's all

Having trouble with my milkymonster build, can only get 3 5970's running in one rig.

As soon as the forth goes in boinc crashes with a driver error (caldd64.dll)

Still cant see anyone running even three so should make it top rig in the world on MW

Going to move it from a K9A2 over to a DFi board tomorrow (dfi has only three pcie slots) to free up a pcie for later.

ToxicTBag · 6 Nov 2010 at 12:02

Nice one Mr Kolala sah!!!
Are you having any luck with getting your 4 5970's running Mr Ba?

BaBa · 6 Nov 2010 at 14:55

Tiffy's coming thick and fast i see and impressive RAC Mr Toxic.

In fact impressive RAC all round I see credit that would have had several of you at the top not long ago, well done all

Cant get 4 5970's running have settled for 3 as soon as the 4th goes in boinc will not run

Have tried every driver under the sun, some will load in windows and show all 8 gpu's but again boinc crashes with a driver error.
Tried XP64 that showed all the GPU's but would not load the drivers

Would try vista but i dont have a copy of 64 bit maybe later.
Only good thing is i can now run it all on one psu (1500w tt)
Will post some pics when its ready to go in the rack

Biffa · 7 Nov 2010 at 01:40

Have you tried 32bit?

BaBa · 7 Nov 2010 at 03:15

Not tried 32bit but with 4gb of system ram and 8gb of video ram not sure how it would cope.
Was under the impression that 32 bit had limit of 4gb combined video/system memory!?

BaBa · 10 Nov 2010 at 16:51

Seems we have some double accounting going on in MW ATM :eek:

My usual daily credit runs between 1.2 and 1.5 million today i will be getting about 3.6!!!!!

Biffa · 10 Nov 2010 at 17:01

Niicce

Heaven help me if you decide to fold, mind you you would have to swap all those AMD cards. Unless you are waiting for the OpenCL client to let AMD dominate!