And now my NV host gets 0 tasks on each request.
Quotas not reached, 0 tasks allocated today for each of 3 apps.
other 2 not possible for this host due driver constrains.
Can this be reason? Maybe BOINC attempts to get statistics on cuda42 and cuda50 and refuse to allocate tasks for already good known cuda22, 23 and 32 ?...

Ask for CPU work, you'll find all work received are VLARs, and VLARs aren't sent to GPUs, hence you get no GPU work:

All tasks for computer 45274

Claggy

Has .Vlar tasks been tested on ATI 79xx cards?
They might be able to handle the "stress" better then old weak cards.

I threw a new tape on in hopes of getting some non-VLAR results out.

---

SETI@home v7 7.00 windows_intelx86 (cuda22)
Number of tasks completed 117
Max tasks per day 147
Number of tasks today 0
Consecutive valid tasks 114
Average processing rate 48.943858550476
Среднее оборотное время 1.56 days
SETI@home v7 7.00 windows_intelx86 (cuda23)
Number of tasks completed 80
Max tasks per day 110
Number of tasks today 70
Consecutive valid tasks 77
Average processing rate 106.69005525163
Среднее оборотное время 1.30 days
SETI@home v7 7.00 windows_intelx86 (cuda32)
Number of tasks completed 153
Max tasks per day 173
Number of tasks today 0
Consecutive valid tasks 140
Average processing rate 77.402843134318
Среднее оборотное время 1.47 days

So, this host requests only fastest work now, as should be. Will allow number of completed fastest tasks become twice as big as other tasks and consider testing finished on this host.

Application details for host 63368

Raistmer · log out
SETI@home v7 7.00 windows_intelx86 (cuda22)
Number of tasks completed 49
Max tasks per day 56
Number of tasks today 5
Consecutive valid tasks 23
Average processing rate 26.044266543082
Среднее оборотное время 1.78 days
SETI@home v7 7.00 windows_intelx86 (cuda23)
Number of tasks completed 39
Max tasks per day 72
Number of tasks today 0
Consecutive valid tasks 39
Average processing rate 42.440671446687
Среднее оборотное время 2.71 days
SETI@home v7 7.00 windows_intelx86 (cuda32)
Number of tasks completed 40
Max tasks per day 65
Number of tasks today 0
Consecutive valid tasks 40
Average processing rate 46.412524118551
Среднее оборотное время 0.78 days

Slowest app got new tasks for unknown reason. Random came in place ?

AstroPulse v6 6.05 windows_intelx86 (cal_ati)
Number of tasks completed 0
Max tasks per day 45
Number of tasks today 0
Consecutive valid tasks 12
Среднее оборотное время 0.00 days
AstroPulse v6 6.06 windows_intelx86 (opencl_ati_100)
Number of tasks completed 0
Max tasks per day 38
Number of tasks today 0
Consecutive valid tasks 5
Среднее оборотное время 0.00 days
AstroPulse v6 6.06 windows_intelx86 (ati_opencl_100)
Number of tasks completed 0
Max tasks per day 37
Number of tasks today 4
Consecutive valid tasks 4
Среднее оборотное время 0.00 days

Eric, did you replace tape for AP tasks ?
Current tasks have no meaning for such testing.

No work for MB GPU tasks again.

No work for MB GPU tasks again.

Especially not for us Boinc 6 users :-) We're left in the desert, thirsting for ATI MB task, while only getting ATI AP 6.06 tasks, and a few Brook+ tasks even for my HD7870.

Discrimination it is, I tell ya :-)

---

WARNING!! "THIS IS A SIGNATURE", of the "IT MAY CHANGE AT ANY MOMENT" type. It may, or may not be considered insulting, all depending upon HOW SENSITIVE THE VIEWER IS, to certain inputs to/from the nervous system.

Same here.
I try since Eric put the tapes in.
Only a few rare AP`s.

---

With each crime and every kindness we birth our future.

I'm going digging in hopes of finding a tape that can provide both non-outlier astropulse and non-VLAR S@H. Wish me luck.

---

Slowest app got new tasks for unknown reason. Random came in place ?

This is one of those edge cases. There are two ways for the server to measure performance. One is average elapsed run time versus average estimated run time the second is to average the peak flops count vs run time. Most of the time they are about the same, but in your case the second method reports much higher cuda22 speed than the first. Right now the server is using the second, but the web site only reports the first. Since there is little difference most of the time, I will probably switch back to using the first method.

2013-05-20 04:10:47.8199 [PID=22693] [version] [AV#364] (cuda22) setting projected flops based on host_app_version pfc: 53.94G
2013-05-20 04:10:47.8199 [PID=22693] [version] [AV#364] (cuda22) comparison pfc: 53.94G et: 25.98G
2013-05-20 04:10:47.8200 [PID=22693] [version] Best app version is now AV364 (53.25 GFLOP)
2013-05-20 04:10:47.8200 [PID=22693] [version] Checking plan class 'cuda23'
2013-05-20 04:10:47.8201 [PID=22693] [version] plan_class_spec: driver version min: 19038, max: 0, supplied: 26306
2013-05-20 04:10:47.8201 [PID=22693] [quota] [AV#365] scaled max jobs per day: 99
2013-05-20 04:10:47.8201 [PID=22693] [version] [AV#365] (cuda23) setting projected flops based on host_app_version pfc: 44.69G
2013-05-20 04:10:47.8201 [PID=22693] [version] [AV#365] (cuda23) comparison pfc: 44.69G et: 44.69G

---

What criteria for non-outlier task for AstroPulse ?
Only non-overflowed or blanking % should be less than xx% ?(how many if so? )

Non outlier for astropulse means not-overflowed and blanking less than 5%. I went to the main project and found a tape with average blanking of 2%, and 8% of S@H results being vlar.

Jeff is retrieving it now. Of course we'll need to get the existing VLARs out of the way.

---

Non outlier for astropulse means not-overflowed and blanking less than 5%. I went to the main project and found a tape with average blanking of 2%, and 8% of S@H results being vlar.

Jeff is retrieving it now. Of course we'll need to get the existing VLARs out of the way.

fine, I think I'll abort all current AP tasks on ATi host.
Too many cal ones. 8 hours each...and all outliers. Greenhouse effect inevitable, but at least trade it for something good, not just waste...

The new tape "09jl12ad" is online. It should start splitting once the current crop of 3267 S@H results is out in the field.

---

Hosts start to get work... but in weird way, look:

SETI@home v7 7.00 windows_intelx86 (cuda22)
Number of tasks completed 52
Max tasks per day 59
Number of tasks today 31
Consecutive valid tasks 26
Average processing rate 26.176457693184
Среднее оборотное время 1.81 days
SETI@home v7 7.00 windows_intelx86 (cuda23)
Number of tasks completed 71
Max tasks per day 104
Number of tasks today 0
Consecutive valid tasks 71
Average processing rate 45.110189611676
Среднее оборотное время 2.90 days
SETI@home v7 7.00 windows_intelx86 (cuda32)
Number of tasks completed 42
Max tasks per day 68
Number of tasks today 40
Consecutive valid tasks 43
Average processing rate 46.411875624665
Среднее оборотное время 0.79 days

In short: fastest recived nothing, slowest - quite a lot. No quota hit anywhere.

We've still got 2400 VLARs to get rid of before the new ones will show up. An occasional non-VLAR is sneaking through, but mostly we're still up to our eyeballs in VLARs.

---

We've still got 2400 VLARs to get rid of before the new ones will show up. An occasional non-VLAR is sneaking through, but mostly we're still up to our eyeballs in VLARs.

Send the Vlar's to newer ATI cards. I've tried them on main with rescheduling, and my HD7870 doesn't react bad even when running 2 of them at the same time. No screen lags, or nothing. On the contrary, it chews through vlar's as if it just loves them.

---

WARNING!! "THIS IS A SIGNATURE", of the "IT MAY CHANGE AT ANY MOMENT" type. It may, or may not be considered insulting, all depending upon HOW SENSITIVE THE VIEWER IS, to certain inputs to/from the nervous system.

We've still got 2400 VLARs to get rid of before the new ones will show up. An occasional non-VLAR is sneaking through, but mostly we're still up to our eyeballs in VLARs.

Send the Vlar's to newer ATI cards. I've tried them on main with rescheduling, and my HD7870 doesn't react bad even when running 2 of them at the same time. No screen lags, or nothing. On the contrary, it chews through vlar's as if it just loves them.

Good work!!

+1

I'll code that up tomorrow!

Anyone want to try the same test on recent NVIDIA cards?

---

I'll code that up tomorrow!

<edit>Anyone want to try the same test on recent NVIDIA cards?</edit>

Experiences benching VLAR here, over a fair while back, didn't see issues with any Fermi class or newer (compute cap 2.0+). Cards run here were relatively mid to high end though, GTX480 GTX680 and 560tiSOC. Technically speaking at least all Cuda compute cap 2.0 or above should be fine with default settings, due to the way I striped & scaled the long pulsefinds. Maybe more earlier models will be fine too, though a dry run to figure out any credit or runtime balance issues wouldn't hurt IMO.

We've still got 2400 VLARs to get rid of before the new ones will show up. An occasional non-VLAR is sneaking through, but mostly we're still up to our eyeballs in VLARs.

Can this explain why remaining non-VLARs were sent as cuda22 instead of cuda23 ?