Message boards :
Number crunching :
v0.18/v0.19 issues here
Message board moderation
Previous · 1 · 2 · 3 · 4 · 5 · 6 · Next
Author | Message |
---|---|
Send message Joined: 12 Oct 07 Posts: 77 Credit: 404,471,187 RAC: 0 |
The new v0.18 stock app is also significant faster at PPC Macs to: and my T2500 has gone from 47min down to 25min :-) |
Send message Joined: 4 Oct 08 Posts: 1734 Credit: 64,228,409 RAC: 0 |
Any PC that can see a WU crunched in under 1,000 seconds using the new Windows stock client V0.19 is heading for the credit limiter. This means the faster rigs (Penryn types and core i7s) are just as well off using the stock 0.19 client as when using an optimised client (say 0.16) because of the credit limiter ATM. I think older slower rigs (P4s and P3s) are better off, ATM, using an opti client (0.16). The new stock client (0.19) definitely speeds them up, but they still take about twice the time to crunch a WU as when they use an opti client. |
Send message Joined: 6 Apr 08 Posts: 2018 Credit: 100,142,856 RAC: 0 |
I've updated the Linux apps to zslip (thanks speedimic). I suspect the Windows 0.18 would be just as fast as the 0.16 ;)just posted the new v18d Linux-apps in the code forum...
I fully agree. It's fantastic to see what they do, and a pleasure to use it :) |
Send message Joined: 4 Jul 08 Posts: 165 Credit: 364,966 RAC: 0 |
Just started using the 0.19 stock app still not as quick as the 0.16 ssse3 oppti app. 36min for 0.19 and 17min for 0.16 oppti ssse3 app. If we still getting credit for the 0.16ver think I will go back to using that |
Send message Joined: 30 Aug 07 Posts: 2046 Credit: 26,480 RAC: 0 |
Just started using the 0.19 stock app still not as quick as the 0.16 ssse3 oppti You'll still be getting credit for the v0.16 version, so that should be fine. The reason the windows versions are about half as slow as the optimized versions is that Dave P. is compiling them in visual studio, which doesn't support auto-vectorization. We're working on getting him set up to compile the 64 bit windows applications in cygwin/g++ which will let him compile with SSE2 and should speed them up significantly. |
Send message Joined: 4 Oct 08 Posts: 1734 Credit: 64,228,409 RAC: 0 |
With the reduction in the per WU credit and the lifting of the credit limit by Travis, then work output from PCs will start to reflect the actual work done by a PC host. This looks like the project is crediting for the real world and is close to a cross project average of credit given. A few days like this will settle things down well. Good for you Travis. |
Send message Joined: 4 Jul 08 Posts: 165 Credit: 364,966 RAC: 0 |
Thanks Travis still getting through the 0.19 wu's!! Got the ssse3 app ver 0.19 from zslip going to give that a go. I think its based on ver 0.16 so should be fine if it still gets through them in 15 or so min |
Send message Joined: 4 Oct 08 Posts: 1734 Credit: 64,228,409 RAC: 0 |
Got the ssse3 app ver 0.19 from zslip going to give that a go. I think its based on ver 0.16 so should be fine if it still gets through them in 15 or so min I think you are correct Glenn. I am using the same version (SSSE3) and it crunched the 0.16 WUs in my cache without trashing them as usual during a change. |
Send message Joined: 4 Jul 08 Posts: 165 Credit: 364,966 RAC: 0 |
Gday John, thats good then wont be a painfull change back:) It just a bit more than 50% quicker than stock and with the cut in credits have to get through more quicker to get the same credit/day Edit: just reported 2 wu's got 12.1ea credit for them was getting about 30 per wu. |
Send message Joined: 28 Apr 08 Posts: 1415 Credit: 2,716,428 RAC: 0 |
I'm getting less than 12 credits an hour, not finished running version 17 wu's yet! So now instead of 30 credits an hour I get 12!! I am tempted to dump the 17's! But I wont. These wu's should still be awarding 30 credits in my opinion. |
Send message Joined: 4 Jul 08 Posts: 165 Credit: 364,966 RAC: 0 |
Just cleared my cache of the stock wu's now crunching with the ssse3 opp app and theyre processing in 17m40s. Reported 2 got 12.1cr each wu down from 30 |
Send message Joined: 16 Jan 08 Posts: 18 Credit: 4,111,257 RAC: 0 |
Hi Anybody out there know where I can get the old v0.16 optimised app for windows. Since switching to v0.19 the work unit times on most of my machines have got longer (eg. from 390s to 540s, and 20mins to 35mins on the slower machines). So I'd like to try reverting back to v0.16 to see what happens m4rtyn ******************************* ******************************* |
Send message Joined: 12 Nov 07 Posts: 2425 Credit: 524,164 RAC: 0 |
Hi If you read the other topics you would know that is due to the units being longer not the version. Doesn't expecting the unexpected make the unexpected the expected? If it makes sense, DON'T do it. |
Send message Joined: 16 Jan 08 Posts: 18 Credit: 4,111,257 RAC: 0 |
Hi That sounds like way too much hard work. m4rtyn ******************************* ******************************* |
Send message Joined: 6 Apr 08 Posts: 2018 Credit: 100,142,856 RAC: 0 |
Hi I will put them back in zslip. I was thinking of keeping it simple by having only the latest window/linux versions, but then windows 0.16 is still valid and I would support people being able to choose if there is more than one version available. I'll let you know when windows 0.16 is added back to zslip |
Send message Joined: 12 Nov 07 Posts: 2425 Credit: 524,164 RAC: 0 |
Hi There is no difference except more info is given in the task page, no programing difference. Don't see any reason to keep the 16's up. As Cluster posted in another thread, where this was asked. Doesn't expecting the unexpected make the unexpected the expected? If it makes sense, DON'T do it. |
Send message Joined: 16 Jan 08 Posts: 18 Credit: 4,111,257 RAC: 0 |
Hi Thanks Ice much appreciated m4rtyn ******************************* ******************************* |
Send message Joined: 6 Apr 08 Posts: 2018 Credit: 100,142,856 RAC: 0 |
Hi You're very welcome m4rtyn. zslip now contains all the Windows and Linux optimized apps from 0.16 onwards. |
Send message Joined: 2 Sep 07 Posts: 18 Credit: 180,611 RAC: 0 |
On a P3 1233 Mhz with Win2000 the new stock app v0.19 with 55-57 min is faster than the previous optimized win32 app v0.16SSE with 1:15h. Nice work Travis - Thanks ;) As stated before, there are no improvements in the codebase of the relabeled v0.19 opti app beside the additional logging. So SSE optimized app v0.16 and v0.19 for windows should give the same runtimes. Does this mean the stockapp is better for old PCs with only SSE? |
Send message Joined: 26 Jul 08 Posts: 627 Credit: 94,940,203 RAC: 0 |
On a P3 1233 Mhz with Win2000 the new stock app v0.19 with 55-57 min is faster than the previous optimized win32 app v0.16SSE with 1:15h. The x87 and SSE versions are created using different compilers. So it is perfectly possible that the x87 version is faster than the SSE one, as SSE cannot be used for the time consuming stuff here at MW requiring double precision. Also be aware there are WUs with different length floating around. One should compare only the times of similar ones. |
©2024 Astroinformatics Group