Hey Bill592!, It is possible that E@H reports invalid information concerning driver version
Good Mornin Titan
The 'computer' info on Einstein doesn't seem to know Shiʿite about what driver we are using (at least for AMD)
It reads out the same thing 1.4.1848 for all AMD Gpu's (
changed from 14.4 whql to 14.8 whql with both of my machines. atleast so far it looks really good. everything after aug 30 is done with 14.8 whql if you want to check times. i have 2 machines running Einstein with gpu only. i7-3930K with 2x7970 and AMD A10-7850K with apu.
Hannu I checked out your machines tasks looks like you are also experiencing a doubling in reported CPU time on the Perseus Tasks after the 30th or with the new 14.8 drivers. Has your RAC on that host improved?
my rac is down because i did run 2 weeks seti@home just for test it again. it looks like cpu times are doubled but possibly gpu time are litle bit lower. i can't say for sure because atleast for me most of my old tasks are dropped from list. 14.4 i had few tasks over 10000s but with 14.8 highest is little bit over 9000s.
Snow Crash thanks for the input, I am investigating the 13.12 driver per your post and found that while running two Perseus tasks a piece on both of my R9 280X's (1150 mhz Core 1600 memory) I experienced an minor increase in task completion times of about 200-300 seconds vs the 14.4. Also these drivers did report lower CPU times per task and about a 1% drop in GPU utilization as report by GPU Z. I am currently trying a 3 tasks per card to see if there is any gains over 14.4.
3 task per cards looks to be ideal on the 13.12 drivers the few tasks completed so far I have observed a 10% shorter task completion time! On the 14.4 drivers 3 tasks of the Perseus Survey per card resulted in run times of 8400 secs vs the 7500-7600 on the 13.12. Snow Crash thanks for the tip looks like it could boost my daily output significantly!
BOINC (7.2.42) doesn't report
)
BOINC (7.2.42) doesn't report proper driver version.
Apparently it only reports "CAL version" which appears to be the same for all (new?) drivers.
Well that is interesting! I
)
Well that is interesting! I didn't know that BOINC only displays the aticalcl.dll version number, thanks for the heads up.
oops double post
)
oops double post
RE: RE: Hey Bill592!, It
)
If you look on this page you can see that could be some of AMD's fault too:
http://www.hal6000.com/seti/boinc_ati_gpu_cheat_sheet.htm It seems they reused the same CAL/BOINC Version number for the last few.
changed from 14.4 whql to
)
changed from 14.4 whql to 14.8 whql with both of my machines. atleast so far it looks really good. everything after aug 30 is done with 14.8 whql if you want to check times. i have 2 machines running Einstein with gpu only. i7-3930K with 2x7970 and AMD A10-7850K with apu.
I saw a substantial
)
I saw a substantial performance decrease when I upgraded from 13.12 to 14.4
tested for 24+ hours: 3x perseus / 7950 / gpu 1125/ mem 1325/ 16x pcie3
13.12 = ~7800 sec
vs.
14.4 = ~8800 sec
I don't have the details at hand but I also saw similar decrease on my 7850.
I reversed the "upgrade" and will stay on 13.12 until the next whql release - unless I see posts claiming a new beta increases performance.
--------------------------
- Crunch, Crunch, Crunch -
--------------------------
Hannu I checked out your
)
Hannu I checked out your machines tasks looks like you are also experiencing a doubling in reported CPU time on the Perseus Tasks after the 30th or with the new 14.8 drivers. Has your RAC on that host improved?
my rac is down because i did
)
my rac is down because i did run 2 weeks seti@home just for test it again. it looks like cpu times are doubled but possibly gpu time are litle bit lower. i can't say for sure because atleast for me most of my old tasks are dropped from list. 14.4 i had few tasks over 10000s but with 14.8 highest is little bit over 9000s.
Snow Crash thanks for the
)
Snow Crash thanks for the input, I am investigating the 13.12 driver per your post and found that while running two Perseus tasks a piece on both of my R9 280X's (1150 mhz Core 1600 memory) I experienced an minor increase in task completion times of about 200-300 seconds vs the 14.4. Also these drivers did report lower CPU times per task and about a 1% drop in GPU utilization as report by GPU Z. I am currently trying a 3 tasks per card to see if there is any gains over 14.4.
3 task per cards looks to be
)
3 task per cards looks to be ideal on the 13.12 drivers the few tasks completed so far I have observed a 10% shorter task completion time! On the 14.4 drivers 3 tasks of the Perseus Survey per card resulted in run times of 8400 secs vs the 7500-7600 on the 13.12. Snow Crash thanks for the tip looks like it could boost my daily output significantly!