![[H]ard|Forum](/styles/hardforum/xenforo/logo_dark.png){kind=logo}
New extra large -bigadv on Beta
- Thread starter Nathan_P
- Start date
404 - File not found
Gawd
- Joined
- Apr 2, 2007
- Messages
- 525
HFM says 300k credit for this one. Seems a bit high for this kind of rather "slow" system - or ?
That's total credit for over 3 and 1/2 days of work, so that's fine.
may i be worthy
[H]ard|DCer of the Month - December 2010
- Joined
- Aug 17, 2010
- Messages
- 1,186
I finally got the chance to run 3 x Linux for the start of this coming weekend, and using -bigbeta was lucky enough to score 3 of the new units.
I got excited to see this:
but then I realised i had not updated HFM on that machine to the latest downgraded points:
Oh well, I had a million point a day farm for a few days. But 300k per rig is pretty sweet, I remembered when I cracked 150k and thought that was as good as it got.... and this is all from software upgrades and new units... good times.
I got excited to see this:
but then I realised i had not updated HFM on that machine to the latest downgraded points:
Oh well, I had a million point a day farm for a few days. But 300k per rig is pretty sweet, I remembered when I cracked 150k and thought that was as good as it got.... and this is all from software upgrades and new units... good times.
HFM says 300k credit for this one. Seems a bit high for this kind of rather "slow" system - or ?
Thanks for that info BTW. Was thinking of setting up the same on my end this weekend. Any data points for legacy -bigadv on the dual E5620?
Sure. This was from my last 6901 packet which scored for around 82k points.
Code:
[13:50:32] Project: 6901 (Run 10, Clone 13, Gen 49)
[13:50:32]
[13:50:33] Assembly optimizations on if available.
[13:50:33] Entering M.D.
[13:50:40] Mapping NT from 16 to 16
[13:50:43] Completed 0 out of 250000 steps (0%)
[14:16:12] Completed 2500 out of 250000 steps (1%)
[14:41:42] Completed 5000 out of 250000 steps (2%)
[15:07:10] Completed 7500 out of 250000 steps (3%)
[15:32:40] Completed 10000 out of 250000 steps (4%)
[15:58:08] Completed 12500 out of 250000 steps (5%)
[16:23:39] Completed 15000 out of 250000 steps (6%)
[16:49:08] Completed 17500 out of 250000 steps (7%)
[17:14:38] Completed 20000 out of 250000 steps (8%)
[17:40:06] Completed 22500 out of 250000 steps (9%)
[18:05:37] Completed 25000 out of 250000 steps (10%)posted my first 6903, have 3 in the works now, just switched 4 other boxen over as well, my other boxen are too slow to do it me thinks. my slowest and oldest 16 thread boxen run about 42-43min TPF on a 6901. Not sure if it would be a gain to move them to 6903's
yeah thats about what I came up with as well so I switched all my 16 thread boxen over to bigbeta, guess we'll see how the ppd/ppw avgs out when things settle in
I think there is a time limit on how long it takes to upload a WU. If you don't send it back quick enough the client times out and tries again. Not a big issue if you have a fast connection. However With 2x SR-2 at the same speed and a cruddy connection. 2 times when they have tried to upload at the same time I get timeouts.
I have stopped one client to let the other upload like I did 2 days ago. I will have to restart the other client and hope it held onto the WU.
I have stopped one client to let the other upload like I did 2 days ago. I will have to restart the other client and hope it held onto the WU.
See the dynamic load balancing aspect of the thread here:
http://foldingforum.org/viewtopic.php?f=66&t=18830
Would be interesting to hear if you resume from a checkpoint whether dlb gets engaged, and if so, what gains you see.
My SR-2 is running at 183*23, ddr3-1830 7-8-7 1T currently (dialed back overclock from 187 due to some instability on 6903). 6903 TPF 22:10 -> 21:40 with dlb at 183*23.
http://foldingforum.org/viewtopic.php?f=66&t=18830
Would be interesting to hear if you resume from a checkpoint whether dlb gets engaged, and if so, what gains you see.
My SR-2 is running at 183*23, ddr3-1830 7-8-7 1T currently (dialed back overclock from 187 due to some instability on 6903). 6903 TPF 22:10 -> 21:40 with dlb at 183*23.
I finally got the chance to run 3 x Linux for the start of this coming weekend, and using -bigbeta was lucky enough to score 3 of the new units.
I got excited to see this:
but then I realised i had not updated HFM on that machine to the latest downgraded points:
Oh well, I had a million point a day farm for a few days. But 300k per rig is pretty sweet, I remembered when I cracked 150k and thought that was as good as it got.... and this is all from software upgrades and new units... good times.
Last edited:
Activate: AMD
[H]ard|Gawd
- Joined
- Nov 6, 2004
- Messages
- 1,994
I have yet to pull one of these down... folded a bunch of 6901's but neither hide nor hair of one of these bad boys :\
This is on a 3.6ghz 24 thread SR2:
before:
Code:
[21:56:29] Project: 6903 (Run 11, Clone 15, Gen 1)
[21:56:29]
[21:56:29] Assembly optimizations on if available.
[21:56:29] Entering M.D.
[21:56:36] Mapping NT from 24 to 24
[21:56:39] Completed 0 out of 250000 steps (0%)
[22:24:58] Completed 2500 out of 250000 steps (1%)
[22:53:17] Completed 5000 out of 250000 steps (2%)
[23:21:36] Completed 7500 out of 250000 steps (3%)
[23:49:56] Completed 10000 out of 250000 steps (4%)
[00:18:15] Completed 12500 out of 250000 steps (5%)
[00:46:35] Completed 15000 out of 250000 steps (6%)after:
Code:
[03:59:13] Resuming from checkpoint
[03:59:27] Verified work/wudata_01.log
[03:59:28] Verified work/wudata_01.trr
[03:59:28] Verified work/wudata_01.xtc
[03:59:28] Verified work/wudata_01.edr
[03:59:29] Completed 27580 out of 250000 steps (11%)
NOTE: Turning on dynamic load balancing
[04:26:08] Completed 30000 out of 250000 steps (12%)
[04:57:17] Completed 32500 out of 250000 steps (13%)
[05:26:43] Completed 35000 out of 250000 steps (14%)
[05:52:43] Completed 37500 out of 250000 steps (15%)Just dropped a few -bigbeta and now all 6900's except one 2684... two months ago I would be in heaven but now it is killing my ppd!
At least for the last frame, looks like you went from ~28min -> 26min TPF. How do the later frames with dynamic load balancing on look?
They stayed pretty consistent throught the Wu at ~26
I'm turning in frames at 26 mins TPF right now despite the fact that I restarted the PC between WUs and it doesn't appear DLB has turned on.
I do think there is something causing inconsistent performance, but at least based on what I'm seeing, DLB isn't the cause.
I do think there is something causing inconsistent performance, but at least based on what I'm seeing, DLB isn't the cause.
Are you still getting these? I haven't gotten a new 6903 with -bigbeta flag set since Friday/Saturday. Five capable machines (one I switched to -bigadv just to see if -bigbeta was not working)
Activate: AMD
[H]ard|Gawd
- Joined
- Nov 6, 2004
- Messages
- 1,994
havent pulled any down yet either
I would have been thrilled about a month ago to have all these 6901's coming my way!
SatTech
2[H]4U
- Joined
- Dec 5, 2005
- Messages
- 3,381
Pulled a 6903 this morning with the -bigbeta flag.
Which filesystem are you using? (output of mount command)
I've found getting dlb to engage on the SR-2 is hit-or-miss after restarting units -- looks like the default auto logic triggers around 5% load imbalance threshold-- likelihood of hitting the threshold goes up with a larger number of cores. When dlb does engage, it's consistently ~30 seconds better TPF on the SR2. The AMD system is 30+ seconds better TPF as well. (now getting 11:45 TPF on the AMD system @ 2.7ghz when dlb engaged).
Regarding 6903 availability, I went about 18 hours with no 6903s being delivered. I wonder how often new work units are generated.
I've found getting dlb to engage on the SR-2 is hit-or-miss after restarting units -- looks like the default auto logic triggers around 5% load imbalance threshold-- likelihood of hitting the threshold goes up with a larger number of cores. When dlb does engage, it's consistently ~30 seconds better TPF on the SR2. The AMD system is 30+ seconds better TPF as well. (now getting 11:45 TPF on the AMD system @ 2.7ghz when dlb engaged).
Regarding 6903 availability, I went about 18 hours with no 6903s being delivered. I wonder how often new work units are generated.
Last edited:
I might -oneunit the client and reinstall
There is something amiss. I have been pulling 6903s without fail on my x SR-2s.
I got a 6903 with my 4p 6172 box 15:50 tpf
DSee
2[H]4U
- Joined
- Aug 31, 2007
- Messages
- 2,175
Just check HFM and average was around 14:30 last time I had the 6903 on the 4P Opteron 6174 box.
The quad 6166 box just pulled another 6900...
Happen to know any way/ have thoughts to fix? Strange thing is that it is happening to five boxen. I even have slight variations now like
-smp -bigbeta
-smp 48 -bigbeta
-smp -bigadv (I saw the post saying it went over to bigadv so I have the dual 6164's on this)
Ubuntu 10.10 and 11.04 and etc
Other funny thing is that up until Fri/ Sat the four boxen were all pulling 6903's.
The server that dishes out the 6903 (130.237.232.237) is in accept only mode. No flag combination or number of re-installs can change that.
We are all in the same boat at least they are still bigadv and not SMP this time. Fold on.
We are all in the same boat at least they are still bigadv and not SMP this time. Fold on.
Its got 989 work units its dishing out now... both connect and status accepting...
Activate: AMD
[H]ard|Gawd
- Joined
- Nov 6, 2004
- Messages
- 1,994
pjkenned said:
I can confirm. I just got my first 6903 around 2pm today with -smp -bigadv on
tmi....
but I think I felt similar when I had a 3.3m day.... was a goood day... that I plan on repeating....
3.3m points in a day. I believe that qualifies for:
[ame="http://www.youtube.com/watch?v=WGOohBytKTU"]YouTube - ‪Flight of the Conchords- Business Time‬‏[/ame]