Not running on a 7970.

I've got 2x7970 in crossfire, 4 WUs tried, all reset the driver within a couple of seconds, and dont progress from that point.

I'll reboot and try with reduced clocks see what happens.

Moo doesnt run on 7970 either.

Initial run on Collatz seemed to go well, doing those in 525 seconds at full standard settings inside the driver, no voltage applied.

Regards
Zy

Tried bringing down the clocks to 875/1100 (default is 925/1325) still doesnt run. It resets the driver within a couple of seconds as before.

I crashed a few against my 1090T, then aborted, so you have a few to look at.

If you want me to test any different versions, post here or PM me - delighted to help if I can

Regards
Zy

All your tasks say they were aborted so I don't see any error.

Actually I found one that does have something in it. Could you try running it with the argument --gpu-polling-mode -1?

Actually I found one that does have something in it. Could you try running it with the argument --gpu-polling-mode -1?

Or 1. Or anything that isn't 0.

I'll need the app_info format and basic entries - not done one at MW for a while, once I've got that I'll run a few for you

Try this http://milkyway.cs.rpi.edu/milkyway/download/random_stuff/app_info_separation_win64_ati14.xml

got a tag mismatch - back shortly

Didnt run - behaviour the same - driver reset and zero utilisation, I let that run like than for 1 min then aborted. Tried two WUs, same for both

Rebooting - back jn a few

Regards
Zy

Bak .... GPUs set for defaults (925/1375)

If you want to come in on remote and take control fine by me - your choice.

If you want me to run more tests - no problem

Any log files I can set that will help you?

Regards
Zy

It doesn't really matter. It's not worth putting in much effort to find out why it's crashing since the CAL version is discontinued. I'll just make sure the replacement is fine when I get mine in a few days.

I'll get some more Collatz done - but post if you want me to run anything (and/or log files), I'll watch this board for posts.

Regards
Zy

It doesn't really matter. It's not worth putting in much effort to find out why it's crashing since the CAL version is discontinued. I'll just make sure the replacement is fine when I get mine in a few days.

Okie Doke

Regards
Zy

I've made some tests with NV OpenCL app on Obix007's 7970.

input

milkyway_separation_0.82_windows_x86_64__cuda_opencl.exe -np 14 -p 0.3818788877972418000000000 30.0000000000000000000000000 -20.0000000000000000000000000 76.0000000000000000000000000 38.4090067002612400000000000 0.0000000000000000000000000 0.0000000000000000000000000 1.0000000000000000000000000 5.7985829621085900000000000 40.3449848062901640000000000 45.7000000000000000000000000 0.3401773264955795000000000 0.0000000000000000000000000 1.0000000000000000000000000

output

21:33:27 (3956): Can't open init data file - running in standalone mode
Error loading Lua script 'astronomy_parameters.txt': [string "number_parameters: 4..."]:1: '<name>' expected near '4'
Error reading astronomy parameters from file 'astronomy_parameters.txt'
Trying old parameters file
Using SSE3 path
Found 1 platform(s)
Platform 0 information:
Name: AMD Accelerated Parallel Processing
Version: OpenCL 1.1 AMD-APP (851.6)
Vendor: Advanced Micro Devices, Inc.
Extensions: cl_khr_icd cl_amd_event_callback cl_amd_offline_devices cl_khr_d3d10_sharing
Profile: FULL_PROFILE
Didn't find preferred platform
Using device 0 on platform 0
Found 2 CL device(s)
Trying to show unknown MWDoubleExts 6
Device Tahiti (Advanced Micro Devices, Inc.:0x1002) (CL_DEVICE_TYPE_GPU)
Driver version: CAL 1.4.1658 (VM)
Version: OpenCL 1.1 AMD-APP (851.6)
Compute capability: 0.0
Image support: CL_TRUE
Max compute units: 32
Clock frequency: 1125 Mhz
Global mem size: 3221225472
Local mem size: 32768
Max const buf size: 65536
Double extension: Unhandled MWDoubleExts

Compiler flags:
-cl-mad-enable -cl-no-signed-zeros -cl-strict-aliasing -cl-finite-math-only -DDOUBLEPREC=1 -DFAST_H_PROB=1 -DAUX_BG_PROFILE=0 -DNSTREAM=2 -DCONVOLVE=120 -DR0=30.000000000000000 -DSUN_R0=8.500000000000000 -DQ_INV_SQR=6.857229770206179 -DUSE_IMAGES=0

Build status: CL_BUILD_SUCCESS
Build log:
C:\Users\597x2-2\AppData\Local\Temp\OCL35C8.tmp.cl(236): warning: attributes
ignored here
__global const __read_only RPoints* r_pts,
^

C:\Users\597x2-2\AppData\Local\Temp\OCL35C8.tmp.cl(238): warning: attributes
ignored here
__global const __read_only LBTrig* lbts,
^

LOOP UNROLL: pragma unroll (line 302)
Unrolled as requested!
LOOP UNROLL: pragma unroll (line 279)
Unrolled as requested!

Kernel work group info:
Work group size = 256
Kernel local mem size = 0
Compile work group size = { 0, 0, 0 }
Keeping chunk boundaries as multiples of 2048
Range: { nu_steps = 640, mu_steps = 1600, r_steps = 1400 }
Iteration area: 2240000
Chunk estimate: 1
Num chunks: 1
Added area: 512
Effective area: 2240512
Integration time: 47.281927 s. Average time per iteration = 73.878011 ms
Integral 0 time = 48.013559 s
Likelihood time = 2.081959 s
<background_integral> 0.000076204271302 </background_integral>
<stream_integral> 0.000000007248892 19.757344732480210 </stream_integral>
<background_likelihood> -5.610462983778151 </background_likelihood>
<stream_only_likelihood> -238.985842705200840 -188.059457448391440 </stream_only_likelihood>
<search_likelihood> -5.289947178575774 </search_likelihood>
<search_application> milkywayathome_client separation 0.82 Windows x86_64 double OpenCL </search_application>
21:34:18 (3956): called boinc_finish

Sesef

Could you lead me by the nose on dowloading an NVidia one and running it on a 7970, and I'll give it a go. Is it app_info? If so, whats the file settings etc.

Regards
Zy

You can find the applications right here on Arkayn´s page: http://www.arkayn.us/forum/index.php?action=downloads;cat=11

---

You can find the applications right here on Arkayn´s page: http://www.arkayn.us/forum/index.php?action=downloads;cat=11

thanks it works:)
dowload NVidia and edit app_info to

<app_info>
  <app>
  <name>milkyway</name>
  </app>
  <file_info>
    <name>milkyway_separation_0.82_windows_x86_64__cuda_opencl.exe</name>
    <executable />
  </file_info>
  <app_version>
    <app_name>milkyway</app_name>
    <version_num>82</version_num>
    <flops>1.0e11</flops>
    <avg_ncpus>0.05</avg_ncpus>
    <max_ncpus>1</max_ncpus>
    <plan_class>ati14ati</plan_class>
    <coproc>
      <type>ATI</type>
      <count>1</count>
    </coproc>
	<cmdline></cmdline>
    <file_ref>   <file_name>milkyway_separation_0.82_windows_x86_64__cuda_opencl.exe</file_name>
      <main_program/>
    </file_ref>
</app_version>
</app_info>

Fine that it works. Runtime for you is about 66 s, I see. Do you think that there will be improvements with AMD-specific code?

(This question goes also to Matt :))

---

Run Time is 52 secs with GPU set to 1125/1575, the max speed inside the driver, power slider is +5

Validates ok. GPU Temp is max 73 degrees, fan on auto, running 2 x 7970 in crossfire, 1090T CPU @3.6Ghz 16Gb DDR3 RAM @1600, no CPU apps running.

Nice stuff ATI hatdware running NVidia app via OpenCL. Suspect more speed to come with ATI OpenCL app. If additional voltage applied likely to go to 40-45 secs a WU I reckon.

Good start straight off the bat and no tweeking as such

EDIT: Screen a bit laggy so set it to 32 inside MW preferences, much better, WU now runs @53secs. Temp still 73-74, solid, running well

EDIT2: Utilisation is 99% constant on both 7970s, hardly flickers

Regards
Zy

Run Time is 52 secs with GPU set to 1125/1575, the max speed inside the driver, power slider is +5
Validates ok. GPU Temp is max 73 degrees, fan on auto, running 2 x 7970 in crossfire, 1090T CPU @3.6Ghz 16Gb DDR3 RAM @1600, no CPU apps running.
Nice stuff ATI hatdware running NVidia app via OpenCL. Suspect more speed to come with ATI OpenCL app. If additional voltage applied likely to go to 40-45 secs a WU I reckon.
Good start straight off the bat and no tweeking as such
Regards Zy

It's interesting that even at this point the HD 7970 is running the NVidia optimized app at 3.5x faster than the top GTX580 card that I could find in the database. Pretty impressive and the HD 7970 also uses less power.

BTW, the page below needs to be corrected and updated:

http://milkyway.cs.rpi.edu/milkyway/gpu_list.php

The HD 69xx series isn't even listed (unless the top AMD listing of HD5800 is a typo) and no the GTX480 is not faster than the GTX580 according to the WU completion times in "Top Computers".