@tuning expert enthusiasts

OpenCL APv7 came not only with much improved behavior on blanked tasks. They also contain few new tuning options that would allow to squeeze few more % of performance out of GPU device. Check ReadMe file in packs thoroughly.

I'm gonna' sit it out awhile during all this turmoil with v7 and v6 AP's else I'm liable to get run over by all the big trucks out there on the freeway. Then when the dust clears and someone can explain what this half-demented creature should do, I'll pop back by. In the meantime Milky Way is quiet and comfortable. My heart belongs to seti AP though.

---

merle - vote yes for freedom of speech

I'm gonna' sit it out awhile during all this turmoil with v7 and v6 AP's else I'm liable to get run over by all the big trucks out there on the freeway. Then when the dust clears and someone can explain what this half-demented creature should do, I'll pop back by. In the meantime Milky Way is quiet and comfortable. My heart belongs to seti AP though.

merle, if you want to stick to stock, just enable AP v7 in your preferences on the account page. If you have a modified app_config.xml, add AP v7 there as well. That's what I did two days ago, started receiving some units today.

I just hope the estimated time is wrong, 13 GPU hours compared to 1 for v6.

---

@Zeus,

No, I just switched to ap7_win_x86_sse2_opencl_nv_r2721
Here is portion of app_info:

Just a heads-up that in your example, the second app_version plan_class is the same as the first one. Should be open_cl_nvidia_100 rather than cuda_open_cl_100, if I'm reading Mike's aistub correctly? As it reads right now, I think it's just a dupe ... Regards, Jim ..

No, one matches 32-bit stock allocations, the other matches 64-bit (the <platform> lines are different). Mike's just trying to make sure no-one trashes a cache.

Most of the v7 GPU WUs that have validated for me have been paired with a ati_mac GPU and the guts has been cut out of credits. Not sure if it is v7 or the ati_mac effect ... time will tell.

cheers

---

Most of the v7 GPU WUs that have validated for me have been paired with a ati_mac GPU and the guts has been cut out of credits. Not sure if it is v7 or the ati_mac effect ... time will tell.

cheers

I noticed that too, so far only two tasks with mac's, but for those chasing credits there will have to be a bit of head scratching.

Thanks Michel,

I'll stick my toe in the water.

--edit
event log says no usable version of app_info.xml

I did a repair of boinc rather than a new install since I have a few MW wu's crunching.

Don't I need a new app_config.xml file now? What's the formal title for APv7?



--edit2
Never mind. I'm working it out. Thanks though.

---

merle - vote yes for freedom of speech

Nope.. http://setiathome.berkeley.edu/results.php?hostid=7374749&offset=0&show_names=0&state=4&appid=20

AP7 credit is low as hell. Thats not that good really.

---

_________________________________________________________________________
Addicted to SETI crunching!
Founder of GPU Users Group

Thanks Michel,

I'll stick my toe in the water.

--edit
event log says no usable version of app_info.xml

I did a repair of boinc rather than a new install since I have a few MW wu's crunching.

Don't I need a new app_config.xml file now? What's the formal title for APv7?



--edit2
Never mind. I'm working it out. Thanks though.

You don't need an app_config.xml... but just in case you're using one, you might need to add v7 to it.

---

OK, I just got some units 7.00 and 7.03 (sse). Are these both cpu wu's?

---

merle - vote yes for freedom of speech

I have a question to add to all of this talk of new apps..

When we moved over to AP_v6, we were asked to run stock apps for a while so that the server-side estimates and credit balancing could get a proper baseline before going to anonymous platform.

Is that going to be the same case with v7?

As it turned out credit wasn't ever "balanced" so no really deep sence to continue that practice. Number of active stock hosts is high enough to not force few on anonymous (few regarding number of stock hosts) platform to force go to stock...

Thanks for replying to that question.

I have both of the new CPU apps sitting in a testing folder on my desktop waiting for my cache of v6 to run out. Trying to determine which one (SSE3 or AVX) will be better on my CPU (FX-6100). I copied the last WU in my cache to the testing folder, so it should be easy enough for me to compare run times between the three apps.

I'll be able to run that test in probably about 3 days (I guess this is the one bad thing about being able to hold a 10-day cache) and will report back with the findings.

---

Linux laptop:
record uptime: 1511d 20h 19m (ended due to the power brick giving-up)

Nope.. http://setiathome.berkeley.edu/results.php?hostid=7374749&offset=0&show_names=0&state=4&appid=20

AP7 credit is low as hell. Thats not that good really.

same here, out of the 2 who validated yet I got 177 and 198 (no overflows, other host was a Mac)

---

My AP v7's are all running incl. gpu. Looks OK so far.

---

merle - vote yes for freedom of speech

OK, AP-credits syndrome si cured.
Thanks so much for the Solomonical fix.
Now cruch for MB only, it pays much more - fullstop

This still day one, none of the app versions have reached their 100 validations as far as i can tell, and I seriously doubt any host has their 11 validations for any of their app versions, have patience.

Claggy

Nope.. http://setiathome.berkeley.edu/results.php?hostid=7374749&offset=0&show_names=0&state=4&appid=20

AP7 credit is low as hell. Thats not that good really.

So, CreditScrew strikes back... ROFL... Can't say I did not expect that, really.

OK, AP-credits syndrome si cured.
Thanks so much for the Solomonical fix.
Now cruch for MB only, it pays much more - fullstop

This still day one, none of the app versions have reached their 100 validations as far as i can tell, and I seriously doubt any host has their 11 validations for any of their app versions, have patience.

Claggy

Agreed.
I have seen simalar at beta, after some time it balanced to normal level.

---

With each crime and every kindness we birth our future.

From my lead ATi host:

AstroPulse v6 (anonymous platform, ATI GPU)
Number of tasks completed 8295
Max tasks per day 2436
Number of tasks today 0
Consecutive valid tasks 464
Average processing rate 515.29 GFLOPS
Average turnaround time 2.54 days
AstroPulse v7 (anonymous platform, ATI GPU)
Number of tasks completed 5
Max tasks per day 40
Number of tasks today 41
Consecutive valid tasks 7
Average processing rate 555.44 GFLOPS
Average turnaround time 0.20 days

So, so far system treats AP v7 running faster than v6 on the same host. Lets see kinetics of this.

18 v7.05 done so far on my #1 cruncher but only 4 validate, the others are all pending. CreditNew in effect. About 127 average, looks like. I get a better idea of how long it took on my 780. Looks like 18 minutes average with only a command line added and running 1 work unit per card. Just downloaded my first NV 7.05 for mac so I should have some of those done by tonight.

Under the "NV Titan ..." title it has unroll 16 ...16384 ...

Under the "For NV x80/x70" title it has ... unroll 18 ... 16384

If you read carefully you will notice that one value is without sleep in use the higher value has sleep switch in use.
Thats the difference.

---

With each crime and every kindness we birth our future.