Is it just me or......

[BrokenWolf]

I know I have had my systems down for a week but they were all up and running Friday morning. Most of the systems have turned in at least 1WU including one bigWU that finished up so I should have more systems listed as reporting in and more points (I will connect to work again and get the exact numbers later this evening). Is there an issue with the stats server where WU's turned in not being credited or is it something else?

Thanks,

BrokenWolf

[bruce]

The most recent WUs uploaded from many of your clients (Times are Stanford local database times):
2010-06-11 00:06:07 6013 0 151 150
2010-06-11 09:06:09 6051 0 195 28
2010-06-11 20:04:27 10873 0 0 2
2010-06-12 08:06:26 6012 0 421 164
2010-06-12 08:06:28 6051 0 27 64
2010-06-12 08:06:28 6024 0 143 166
2010-06-12 08:06:28 6073 0 141 38
2010-06-12 10:04:30 10400 48 0 37
2010-06-13 00:05:12 6701 29 0 0
2010-06-13 00:05:12 6701 12 1 0
2010-06-13 02:05:09 6701 38 1 0
2010-06-13 02:05:09 6701 26 0 1
2010-06-13 03:04:52 6603 6 200 148
2010-06-13 03:05:06 6701 95 0 0
2010-06-13 04:05:58 6701 68 1 0
2010-06-13 04:05:58 6701 42 1 0
2010-06-13 04:05:58 6701 27 10 0
2010-06-13 04:06:56 2684 0 15 3
2010-06-13 06:05:48 6077 0 19 54
2010-06-13 07:04:52 6701 78 1 0
2010-06-13 10:04:58 6701 49 1 0
2010-06-13 12:03:38 3466 1 58 393

[BrokenWolf]

Thanks Bruce. After our city power comes back up (high winds brought a tree down on a power line and am using a generator right now) I will connect to work and verify the returned WU's.

Thanks,

BrokenWolf

[bruce]

Also noted elsewhere: The WUs from p6701 are not getting bonus points yet. The Pande Group is aware of the problem.

[BrokenWolf]

Ah ha. That would be where the calculations are off as most of my systems (and NoahJ's) are working on crunching those away.

Guess I missed that post.

Thanks again Bruce.

BW

[djchandler]

Thanks Bruce. After our city power comes back up (high winds brought a tree down on a power line and am using a generator right now) I will connect to work and verify the returned WU's.

Thanks,

BrokenWolf

Now that's dedication!

Hope things are getting better for you folks; we had stormy stuff here in KS too, but not as bad here as other places nearby. We had mostly flash flooding, only in areas notorious for such damage, and nothing like what occurred in AR and OK over the past several days.

Having used the -oneunit flag on my last WU, it should upload ASAP. My client has been trying to upload it without pause now for a couple of hours. It's up to 102 failed uploads so far.

[noorman]

Also noted elsewhere: The WUs from p6701 are not getting bonus points yet. The Pande Group is aware of the problem.

.

What 's the status on this lack of bonus points and/or is this getting seen to any soon ?


.

[noorman]

since there seems not to have been word that things have changed (yet), I 'm asking the question ...
( these WU's run much slower than other recent SMP WU's and then on top don't get a bonus ? )


.

[PantherX]

noorman, just to clarify, what do you mean: (you may edit your post depending on your answer)
A) Do you mean that the WUs that were initially submitted while the Bonus Problem existed, and till now, you haven't received the correct Bonus?
B) You are still having problems with the 6701 WUs?

I have been getting Bonus points since the first P 6701 WU I got: (http://foldingforum.org/viewtopic.php?f ... 45#p147756)

[noorman]

noorman, just to clarify, what do you mean: (you may edit your post depending on your answer)
A) Do you mean that the WUs that were initially submitted while the Bonus Problem existed, and till now, you haven't received the correct Bonus?
B) You are still having problems with the 6701 WUs?

I have been getting Bonus points since the first P 6701 WU I got: (http://foldingforum.org/viewtopic.php?f ... 45#p147756)

.

I 'm going on Bruce's statement (which I linked to) and the fact that if you get a P6701 after a few of the other WU's, your output drops significantly, still, even with the new Core active ...
I know of donors who stopped Folding SMP because they had many SMP Clients running and witnessed drops of between 10,000 and 20,000 PPD just because of a flood of P6701 Wu's on their farm.

So, Bruce was wrong or there is still something else wrong.

I also lived in the belief that with newer Clients and Cores, the possibility to assign with more direction would also evolve, but still WU's are assigned to processors (or systems) who are doing a worse job of it than others, the same goes for GPU cards.
As I reported before (without a reaction from PG), f.e. 353 pointers are quickly dealt with by my 9800GTX+ and GTS250 cards, but not so by my GTX275.
On the other hand, the larger WU's (783 and 787 pointers) are handled more quickly by the GTX275 than the older (less performant) cards are making of those.
Why not have better assignment with a quicker return of Results ???


.

[PantherX]

If my understanding is correct, you want:
A) To get higher PPD with this Project when compared to other SMP2 Projects
B) The WU distribution should be more specific rather than being random

If I am correct, you may want to change the Title to something along the lines of "Requesting Change in K-factor/Base Points (P 6701) To Up PPD" as reading the title made me think that you are having problems for the bonus, Also we can hope that V7 Clients will incorporate point B.

[noorman]

.

I just want the credits for the P6701 on par (pro ratio) with the other SMP WU's

and I would have thought they would have been able to sort out Assignment so that WU's can go to that hardware that is most suited to its demands and that can finish it quickest, that 's all ...


.

[bruce]

Hi noorman (team 734),
Your WU (P6701 R83 C18 G5) was added to the stats database on 2010-06-20 19:05:48 for 5040.63 points of credit.

For new WUs, that problem was fixed some time ago,. I'm not sure of the status of the WUs that were submitted before the problem was fixed. Some Wus may still be due a re-credit of the bonus.

[Flathead74]

I thought Dr. Kasson already addressed this issue.
Here's a link to what I believe is a relevant post:
http://foldingforum.org/viewtopic.php?f ... 40#p148140

The deadlines are not related to the point value in any way. If people with capable machines are having trouble making the deadlines, we can certainly consider changing them. But the points values are deadline-independent (we correct for deadline-length in calculating the k-factor).

FYI, we re-ran the benchmark values on P6701 and got essentially the same points (base value would have been 920 instead of 921 on the second run, k-factor identical). It's possible that some of the other projects were overvalued; we could re-run and adjust those if there's a consensus in favor of doing so (but we usually don't like to take points lower unless there's something really wrong). I doubt there would be a change of more than 10% or so.

One thing to note is that most of the initial A3 projects were *very* similar to each other in terms of the molecular system involved. As we branch out to different types of systems and scientific projects, the scaling between our Core i5 benchmark machine and a given other machine may differ somewhat. (Things like cache size can become quite important.)

Personally, even though the points for the p6701 WUs do seem to be low,
I would rather not have the rest of the WUs devalued to an equal level.

It could be the problem we are up against is that in certain situations
the benchmarking machine just happens to be a particularly efficient system.

Fellow folder, Wrish explained it as follows:
viewtopic.php?f=58&t=14235&p=139631#p139631

The benchmark i5-750 has the highest cache-to-thread ratio of any Nehalem (64K/256K/2 MB per thread), and the Lynnfield revision has the lowest latency to memory of any x86 CPU architecture (3-cycle L1, 8-cycle L2, 33 ns RAM @ 1600 Mhz). Therefore, an i5 750 is going to have the greatest advantage doing the most complex simulations.

[noorman]

.

I can only state that these are too large to be Folding 24/7 (on my system) and for those credits; I switch them off during day hours (almost 2x as expensive as night-time costs)
That way I can keep Power costs down, but then they are running for 3 days if they are out of sync with that schedule I need to keep.
( just switched it back on / that will be the last of them / if it had gained the ~15% extra speed from the optimization of the new Core, I might have let it run, but I see less than 1% gain for this model of WU ? )


.