What is up with this new applciation.

my cpu times have gone up, this make the gpu elapsed time and ms/step also have gone up.

I see similar results across all three hosts.

Example, top three results are from 6.53, you can see an increase in cpu time, gpu time and ms/step. All the previous results are very cconsistant. No other changes were made in clients it shows up as soon as 6.53 was started. Does not matter what version clinet i was running on each host.

cputime - credit claim - granted - version - driver - cpus - client - gpu time - ms/step
51,765.08	3,232.06	3,232.06	6.53	178.24	1.90	6.4.02	62,292.532	73.285
51,160.42	3,232.06	3,232.06	6.53	178.24	1.90	6.4.02	62,612.755	73.662
52,259.07	3,232.06	3,232.06	6.53	178.24	1.90	6.4.02	63,227.636	74.385
49,119.95	3,232.06	3,232.06	6.52	178.24	1.90	6.4.02	59,451.948	69.943
49,250.64	3,232.06	3,232.06	6.52	178.24	1.90	6.4.02	59,452.824	69.944
48,724.11	3,232.06	3,232.06	6.52	178.24	1.90	6.4.02	59,492.902	69.992
48,458.51	3,232.06	3,232.06	6.52	178.24	1.90	6.4.02	59,268.219	69.727
48,364.13	3,232.06	3,232.06	6.52	178.24	1.90	6.4.02	59,162.575	69.603
49,230.23	3,232.06	3,232.06	6.52	178.24	1.90	6.4.02	59,228.002	69.680
49,079.31	3,232.06	3,232.06	6.52	178.24	1.90	6.4.02	59,123.869	69.557
49,090.84	3,232.06	3,232.06	6.52	178.24	1.90	6.4.02	59,325.965	69.795
48,103.39	3,232.06	3,232.06	6.52	178.24	1.90	6.4.02	59,306.325	69.772
49,156.69	3,232.06	3,232.06	6.52	178.24	1.90	6.4.02	59,354.071	69.828
49,170.42	3,232.06	3,232.06	6.52	178.24	1.90	6.4.02	59,486.919	69.985
49,379.95	3,232.06	3,232.06	6.52	178.24	1.90	6.4.02	59,273.031	69.733

---

Alpha Tester ~~ BOINCin since 10-Apr-2004 (2.28) ~~~ Join team USA

My last 2 workunits (shorter WU) are ca. 30% shorter and I get only 2435.94444444444 credits for it. Now I have oftener to make a manually update to get new work (24 hour recall problem).

---

You have to look at the workunit name. Otherwise it is normal that the workunit time changes. Different workunits can have different time per step, different total times and therefore different credits.

Usually, we submit a small of set of WU to test and then more and more. We are preparing several molecular experiments to run concurrently. We are actually quite excited about it. Movies will be also published soon.

GDF

We will check. There should be only minor differences with previous application.

gdf

My last 2 workunits (shorter WU) are ca. 30% shorter and I get only 2435.94444444444 credits for it. Now I have oftener to make a manually update to get new work (24 hour recall problem).

1. I had exactly the same (lenght of WU). But for me it`s not a problem.

2. The problem is that app 6.53 said that my WU`s will take over 78h (sure it`s not true - GTX280 OC). It means what`s in follow: on the maschine with 2x280GTX I have 3 WU`s (2 in use) and BM will not request more (the 4th one WU). BM will not request more cause it means for BM that I`ll can`t finish in time. It means that after finishing 2 of 3 WU`s one of my 280 will be idle.

Just for now I`ve changed `Resource share` on every maschines up to 1000 to get work. Hope it will work until problem will be solved (6.54?)



As always - sorry for bad english ;)

---

You have to look at the workunit name. Otherwise it is normal that the workunit time changes. Different workunits can have different time per step, different total times and therefore different credits.

Usually, we submit a small of set of WU to test and then more and more. We are preparing several molecular experiments to run concurrently. We are actually quite excited about it. Movies will be also published soon.

GDF

Looking at the names, they were the same group TEST5 except one TEST6. The INCREASE in ms/step was when the app version changed from 6.52 to 6.53, same task group names. Only the last one on the list is a TEST6 and that shows the same ms/step under 6.53 as the TEST5 under 6.53 (the top three on my list). With 6.53 these are showing 73-74 ms/step. Prior to that I had over 30 results with 6.52 all in the 69 ms/step range.

And the last one completed is same way, increase ms/step.

161682
Name WLX3978-GPUTEST6-1-20-acemd_0
Workunit 120384
Created 13 Dec 2008 13:57:38 UTC
Sent 13 Dec 2008 19:02:36 UTC
Received 14 Dec 2008 14:07:56 UTC
Server state Over
Outcome Success
Client state Done
Exit status 0 (0x0)
Computer ID 6577
Report deadline 17 Dec 2008 19:02:36 UTC
CPU time 51465.75
stderr out

<core_client_version>6.4.2</core_client_version>
<![CDATA[
<stderr_txt>
# Using CUDA device 0
# Device 0: "GeForce 8800 GT"
# Clock rate: 1512000 kilohertz
# Number of multiprocessors: 14
# Number of cores: 112
MDIO ERROR: cannot open file "restart.coor"
# Time per step: 73.314 ms
# Approximate elapsed time for entire WU: 62317.270 s
called boinc_finish

</stderr_txt>
]]>

Validate state Valid
Claimed credit 3232.06365740741
Granted credit 3232.06365740741
application version 6.53

As always - sorry for bad english ;)

Bad english is the language of the science ... ;)

But your english is much better then mine ... :D

---

6.53 - time per step = 131 ms, overall 65550 sec per wu, credit 2435.
6.52 - time per step = 96 ms, overall 82022 sec, credit 3232.

All up getting slightly less credit per day with the new wu's.

We will soon upload new Linux and Windows applications.


gdf

We will soon upload new Linux and Windows applications.

gdf

I hope it will help :) because:

2. The problem is that app 6.53 said that my WU`s will take over 78h (sure it`s not true - GTX280 OC). It means what`s in follow: on the maschine with 2x280GTX I have 3 WU`s (2 in use) and BM will not request more (the 4th one WU). BM will not request more cause it means for BM that I`ll can`t finish in time. It means that after finishing 2 of 3 WU`s one of my 280 will be idle.

Just for now I`ve changed `Resource share` on every maschines up to 1000 to get work. Hope it will work until problem will be solved (6.54?)

It doesn`t help 100%. After rising up `Resource share` to 1000 I get manually 1 more task (the 4th one) and I went sleep. When I awake I checked my box - 2x280 has finished those 4 WU`s and didn`t request more... Box was idle. Nice 24C ;)

---

This is not an application problem. We are trying to normalize the estimated time. A project reset should solve the problem.

gdf

Hi GDF - ran a project reset on one of my boxes. Doesn't solve the (my) problem - in fact it makes it worse. Estimated run times had come down to the 100's, but after a rest WUs were showing 299 hours to completion. I'm still managing the work manually - and these shorter wWUs in fact make things harder to stop my hosts sitting idle.
I'm reluctant to reset the others as they are makig progress, albeit slowly, on reducing the time to completion at download.

P.

I try to reset the project, correct the DCF, but nothing correct the problem.

On computer with 2 GPU cards, when the 2 WU are finished, i can't get more WU if i don't updated manually. And i have only 2 WU, not like before.

Until the 6.4.3, i didn't have any issues, with this project, except some memories problems like any others users.

So you said that it's not a problem with the application or the project, so it's a problem with the new version of BOINC.
But do you know if this problem will be correct?

Jim PROFIT

I reverted to 6.4.3 on the main page.

gdf.

It doesn`t work. Reseting project is making the situation worse as Burdett have written. Estimated time 201 hours. One of my computers (with 280GTX) downloaded just one WU and will not take more. Because of `ETA`.

I`ve stopped all other tasks and I`ve changed DCF to 0.000001 and `Resource share` to 1000. For 1 of my boxes it`s not enough.

I reverted to 6.4.3 on the main page.

gdf.

Just for ur knowledge. I`m using 6.3.19 and 6.3.21 (I don`t trust this new s***). One is x32 and another x64. Going back to 6.4.3 could not solve problem.

---

I further on use the 6.4.5. When I get a WU of the bad lot (this with another name that GPUTEST5 or GPUTEST6), I know, my DCF will be messed up. So I edit the DCF back to 1.00000 and then I get new work. I have on every box with this method 3 or 4 WUs.

---

@GDF (here we started, here I`ll continue):

Still big problems. Here you have some specs and done steps:

XP32
6.3.19 (x32)
Tried both 178.24 and newest 180.48
DCF normal
Resurce share 1000
All other tasks stopped

Manual project update = request 0

Comp is with 2x280GTX

I had 3 WU`s (can`t reach the 4th one...). I`ve finished 2 of them. Crunching last one (using just one 280GTX) with ETA 60h!

1. Manual project update = request 0

2. Resetting project.

3. Project reseted.

4. Manual request.

5. Correct request!

6. Receiving info:
`Message from server: No work sent.
Message from server: Full-atom molecular dynamics on Cell processor is not available for your type of computer.
Message from server: Full-atom molecular dynamics for Cell processor is not available for your type of computer.`

7. Status: No tasks and:
`Next request in 24h`



Tried `Manual project update` few times. Request correct. But always `Message from server: Full-atom molecular dynamics for Cell processor is not available for your type of computer.`



You don`t even have to answer :). It`s just to inform you.

---

Something's screwy - GPUGrid no longer functions as a project that works without a biological coprocessor (me).

I don't understand the logic behind the 'wrong type of processor message' - it appears on my PCs when the DCF issues stop download of a WU .... not meaningful or applicable, but then it causes the 24hour back off which is a bigger problem.

I'm still getting issues with extended times to completion - I am currently showing ranges fron minutes to hundreds of hours. Sure I can manually reset the DCF - but it changes again quickly and over a very wide range. None of the recently available client versions fundamentally improve the situation for me.

Project resets are an irritation that change little for any length of time.

P.

HI,
we are constantly looking at it. We are reporting these problems to BOINC.
No solution yet. The odd part is that we run without problems and most people do the same, but still there is a part of users as you who have problems.

Hope this gets fixed soon.

gdf

Try to update the BOINC client to 6.4.3 please.


gdf