new VJ blog post: about -bigadv16

Kendrak

[H]ard|DCer of the Year 2009
Joined
Aug 29, 2001
Messages
21,141
http://folding.typepad.com/

Very informative and brings things up to date on what is happening in the -bigadv world.

Points:
-new bigadv16 are here and more are on their way
-older -bigadv (8 and 12) are on the way out
-deadlines shortened
-core count is crude and they know that, working on better metrics, but a ways off
 
How are they determining the difference between a core and a thread? Or are they?
 
As before with the BA-8 QRB, the core count (real, virtual, hacked) doesn't really matter. Same applies to the new BA-16. Making the deadline is all that matters.
 
As before with the BA-8 QRB, the core count (real, virtual, hacked) doesn't really matter. Same applies to the new BA-16. Making the deadline is all that matters.

Basically, but a bit of a over simplification.

-New 4 and 8 thread systems where as fast or faster than systems out a few years ago with 8 or 16 threads.
-People used thread hack to get -bigadv WU and said they are coming in under the deadline so we are not hurting anything.
(deadline had a buffer built in for slower 16 thread systems so they could still get the WU returned in time)
-So the deadline is shortened to reflect the actual time PG wants the WU returned and thread count is now secondary.
 
Sounds too like theyre bringing out 12 core units too, which will be nice.
 
Sounds too like theyre bringing out 12 core units too, which will be nice.

Then, after the new bigadv-16 projects stabilize, we will bring the bigadv-12 projects into line (points, deadlines) with the bigadv-16 projects and convert all projects to bigadv-16.

Sounds more like -bigadv12 will be turning into -bigadv16
 
Sounds more like -bigadv12 will be turning into -bigadv16

Yes it does but ViJay has also posted this:-

"We are bringing a new server online for bigadv-16. It will start by offering a new class of bigadv projects, but we will soon add in a number of projects on the same server that are more similar to bigadv projects donors have already seen. "

Sounds to me like there will be a much broader mix of projects that the current ones
 
My guess is they will all have short deadlines on them as well.
 
hmmm I suppose my dual 6128 system isn't going to do so hot anymore QQ. Has 16 cores but doubt it's going to meet deadlines if they are trying to keep oced 3930ks and the like out.
 
hmmm I suppose my dual 6128 system isn't going to do so hot anymore QQ. Has 16 cores but doubt it's going to meet deadlines if they are trying to keep oced 3930ks and the like out.

I'm thinking the same thing about mine.
 
What about my SR2? I have dual hexes.

And I have a dual-quad Xeon boxen as well. Hmmm.
 
What about my SR2? I have dual hexes.

And I have a dual-quad Xeon boxen as well. Hmmm.

You SR-2 with hexes I would think will be fine with a modest OC.

The dual quad..... that might not make it depending on the WU.

A lot of this is speculation until we see hard #'s
 
hmmm I suppose my dual 6128 system isn't going to do so hot anymore QQ. Has 16 cores but doubt it's going to meet deadlines if they are trying to keep oced 3930ks and the like out
If you do have to drop it out of receiving -bgadv, you can rest assured it will kick the snot out of standard SMP work units.
 
Thats the upside.

But it'd be nice if you could specify which level you wanted... My 2 l5639's aren't to hot on 6903's, and honestly running more than 2 days a wu just doesn't sit well with my ideal wu turn-around times.. So if you could dumb it down, to smaller bigs... that'd be nice.
 
hmmm I suppose my dual 6128 system isn't going to do so hot anymore QQ. Has 16 cores but doubt it's going to meet deadlines if they are trying to keep oced 3930ks and the like out.

Just a guess, but I would say any highley OCld hex 6 core / 12 thread will still be able to complete them on time. Most if not all of the frame times from 16 core rigs I have seen have been slower than the higher OCed hexes. But this is just a guess.;)
 
Initial performance numbers --

P8101 is about as "fast" as P6903.
P8101 gets half the credit (per day) of P6901.

So, disappointing and discouraging (unless I've missed something;
would appreciate a cross-check).

4p6174 @ 2.475GHz

P8101:
Code:
[23:05:57] Completed 12500 out of 250000 steps  (5%)
[23:18:46] Completed 15000 out of 250000 steps  (6%)
[23:31:33] Completed 17500 out of 250000 steps  (7%)
[23:44:17] Completed 20000 out of 250000 steps  (8%)

P6903:
Code:
[23:53:01] Completed 50000 out of 250000 steps  (20%)
[00:05:51] Completed 52500 out of 250000 steps  (21%)
[00:18:35] Completed 55000 out of 250000 steps  (22%)
[00:31:24] Completed 57500 out of 250000 steps  (23%)


Kendrak, whenever you feel it's appropriate, can you ask the project
to comment, please?
 
You SR-2 with hexes I would think will be fine with a modest OC.

The dual quad..... that might not make it depending on the WU.

A lot of this is speculation until we see hard #'s


Oh cool .. I hope someone has x5650 XtraSpicy laying around and sell it for "cheap" price .. I'll get my 2nd SR-2 up for this :D
 
Initial performance numbers --

P8101 is about as "fast" as P6903.
P8101 gets half the credit (per day) of P6901.

So, disappointing and discouraging (unless I've missed something;
would appreciate a cross-check).

4p6174 @ 2.475GHz

P8101:
Code:
[23:05:57] Completed 12500 out of 250000 steps  (5%)
[23:18:46] Completed 15000 out of 250000 steps  (6%)
[23:31:33] Completed 17500 out of 250000 steps  (7%)
[23:44:17] Completed 20000 out of 250000 steps  (8%)

P6903:
Code:
[23:53:01] Completed 50000 out of 250000 steps  (20%)
[00:05:51] Completed 52500 out of 250000 steps  (21%)
[00:18:35] Completed 55000 out of 250000 steps  (22%)
[00:31:24] Completed 57500 out of 250000 steps  (23%)


Kendrak, whenever you feel it's appropriate, can you ask the project
to comment, please?

Yeh I'm pulling about 120k ppd on mine..... (just uploaded a 6094 and got a 8101)

Don't our rig pull about that on normal SMP?
 
Don't quite remember. Will let it finish, then run regular SMP.
 
Don't quite remember. Will let it finish, then run regular SMP.

Do you happen to have a log of one that slipped in between -bigadv?
I know I did, but recently did a reinstall.
 
I was lucky enough not to receive a regular unit during the outage/maintenance.

You know what, regular SMP takes about an hour. TPF of 39-40 seconds.
I can run one here real quick. Give me few.
 
Same system,

P6097:
Code:
[01:06:10] Completed 10000 out of 500000 steps  (2%)
[01:08:02] Completed 15000 out of 500000 steps  (3%)
[01:09:53] Completed 20000 out of 500000 steps  (4%)
[01:11:45] Completed 25000 out of 500000 steps  (5%)

211k PPD

50k more than P8101. Something went wrong...
 
Same system,

P6097:
Code:
[01:06:10] Completed 10000 out of 500000 steps  (2%)
[01:08:02] Completed 15000 out of 500000 steps  (3%)
[01:09:53] Completed 20000 out of 500000 steps  (4%)
[01:11:45] Completed 25000 out of 500000 steps  (5%)

211k PPD

50k more than P8101. Something went wrong...
That's what I thought.

Thanks.... I've got a post in the DAB and to Kasson.
 
calculations are putting my SR2, dual hex, @ 3.3 Ghz about 20k ppd less than SMP(probably more), and only 20k ppd above a 2600k @ 4.5
 
tear did you notice your cpu usage bouncing all over the place mine is bouncing between 96% and 100%
 
I already posted in the beta forum...
But as a beta tester I can't talk about it even though you can see it...
per standard Stanford logic :rolleyes:
 
Yes it has and I ran top it says the Kracken is running maybe I need to reboot.
 
It may be accounting issue as well (try running "top" for comparison). Regardless, I'd
recommend against sudden movements :)

If the unit is as "fast" as P6903, I'd leave it...
 
The TPF was bouncing between 12:30 and 18:00 so I rebooted and it is now holding at 99% to 100% so all is OK for now other than the low PPD.
 
P8101:
2P X5570 8C 16T 3.2Ghz 39:50 TPF 3553 PPD Completion Time 2.77 days Preferred Deadline 2.4 days

8P 8431 48C 48T 2.4Ghz 13:10 - 13:30 TPF 152K PPD Completion Time 0.94 days Preferred Deadline 2.4 days

P6903
2P 40:30 TPF 103K PPD Completion Time 2.81 days Preferred Deadline 5 days

8P 13:51 TPF 514K PPD Completion Time 0.96 days Preferred Deadline 5 days
 
Back
Top