I got the 375.95 from Geforce Experience, as I have always done. But then I went to the nVidia repository and installed 375.70 and it works.
Tullio

OK, I'll test the Win 7 / driver 375.95 combo on my BOINC v7.7.0 (home build from master, well past v7.6.33 development version), and see what I can find out.

Well, that went smoothly enough - no real change from

26-Nov-2016 22:04:54 [---] Starting BOINC client version 7.7.0 for windows_x86_64
26-Nov-2016 22:04:55 [---] CUDA: NVIDIA GPU 0: GeForce GTX 1050 Ti (driver version 375.70, CUDA version 8.0, compute capability 6.1, 4096MB, 3052MB available, 2138 GFLOPS peak)
26-Nov-2016 22:04:55 [---] OpenCL: NVIDIA GPU 0: GeForce GTX 1050 Ti (driver version 375.70, device version OpenCL 1.2 CUDA, 4096MB, 3052MB available, 2138 GFLOPS peak)

to

27-Nov-2016 16:50:51 [---] Starting BOINC client version 7.7.0 for windows_x86_64
27-Nov-2016 16:50:52 [---] CUDA: NVIDIA GPU 0: GeForce GTX 1050 Ti (driver version 375.95, CUDA version 8.0, compute capability 6.1, 4096MB, 3052MB available, 2138 GFLOPS peak)
27-Nov-2016 16:50:52 [---] OpenCL: NVIDIA GPU 0: GeForce GTX 1050 Ti (driver version 375.95, device version OpenCL 1.2 CUDA, 4096MB, 3052MB available, 2138 GFLOPS peak)

- and the cuda 8.0 task I was running (from GPUGrid) picked right back up where it left off.

I did get the international NVidia driver direct from my local nvidia.co.uk website, rather than GeForce Experience, and I installed it by hand 'driver only' (custom choice) without any of the extras. But I see they gave me a separate installation of Vulkan libraries without asking me...

... But I see they gave me a separate installation of Vulkan libraries without asking me...

The Vulkan Runtime/loader I'd expect since, for gaming, ultimately it's intended to become ubiquitous. [Compute also, but less mature]. LunarG supply the Windows SDK, loader (ICD) and runtime with the latter two meant to be tacked onto drivers and/or application installers just like that. The current GFE and m$ situations being mentioned around seem like a good idea to avoid to me. I see no good coming from those (but that's a topic for another time/thread)

---

"Living by the wisdom of computer science doesn't sound so bad after all. And unlike most advice, it's backed up by proofs." -- Algorithms to live by: The computer science of human decisions.

MB8_win_x86_SSE2_OpenCL_ATi_HD5_r3557.exe
AMD Radeon HD 6570 - Catalyst 11.12
Windows XP

1) RAM problem
Is it normal for ATi_HD5_r3557 to use so much RAM?
It starts at 100 MB, slowly increases and ends in 400-700 MB depending on task (r3500 used to use 100-120 MB)
(Video RAM usage is OK - constant during run 230-270 MB depending on task)

This RAM usage increasing is visible in Process Lasso and is also recorded by BoincTasks:
https://s21.postimg.org/5etbuf8xx/Boinc_Tasks_RAM_r3500_vs_r3557.png


2) Checkpoints problem
On "Very VHAR" tasks this app almost do not checkpoint on my computer:
- I monitored "by eye" tasks with ar=103.05 and ar=12.63
- tasks run to completion in normal time ~33 minutes
- they do not checkpoint until last minute - files state.sah and boinc_task_state.xml are created only @ minute 32 and after another minute the task finishes
In these last-minute files are these lines:
<prog>0.10701535</prog>
<fraction_done>0.926736</fraction_done>

- in init_data.xml is the checkpoint_period set by me (and other non-VHAR tasks obey it):
<checkpoint_period>47.000000</checkpoint_period>

- during run (using BOINC 6.10.58) BOINC Manager and BoincTasks show (after ~8 minutes) fake frozen "Progress" of 86.9 % which stays the same till ~minute 32

- After the startup CPU-only phase (~13 s) the RAM usage jumps (in 1 s) from 100 to 520 MB and stays the same till the end

- if the computation is interrupted (e.g. by Snooze) the tasks restart from beginning as if run for the first time:

Work Unit Info:
...............
Credit multiplier is :  2.85
WU true angle range is :  103.055695
Used GPU device parameters are:
	Number of compute units: 6
	Single buffer allocation size: 128MB
	Total device global memory: 512MB
	max WG size: 256
	local mem type: Real
	LotOfMem path: yes
	LowPerformanceGPU path: no
	HighPerformanceGPU path: no
period_iterations_num=50
Running on device number: 0
Priority of worker thread raised successfully
.....

---

 


- ALF - "Find out what you don't do well ..... then don't do it!" :)
 

On the picture one 8,12 task shows 548MB and another - 447MB RAM usage - seems excessive RAM usage observed before 8.20 too.
What specific for those 2 tasks (versus those with ~100MB RAM usage)? Are they both VHAR?

Described VHAR behavior specific to SoG algorithm.
Maybe for low-performance devices (30 mins for VHAR looks quite slow) worth to add artific sync point to allow checkpointing.

EDIT: and maybe that looked like memory leak issue actually not leak at all but just the room to keep all events for enqueued kernels. If so adding additional sync points will help both issues.

---

SETI apps news
We're not gonna fight them. We're gonna transcend them.

On the picture one 8,12 task shows 548MB and another - 447MB RAM usage

They are old Downloads - I had 2-3 SETI@home tasks when I installed Lunatics_Win32_v0.45_Beta6-for-SoG_setup
(one of them may also be partially done by both apps, I don't remember - but I have stderr.txt archived by BoincMonitor (to 30 days back) if you want me to find some info)

30 mins for VHAR looks quite slow

The last such task:
http://setiathome.berkeley.edu/result.php?result_name=05no09ad.29451.67.3.30.158_0

EDIT: and maybe that looked like memory leak issue actually not leak at all but just the room to keep all events for enqueued kernels. If so adding additional sync points will help both issues.

I forgot to post previously - if I restart a "normal" (non-VHAR) task at e.g. 50% when it may be using 300 MB RAM - the app starts at 100 MB again (and continues from 50% done) and do not "jump" to 300 MB so at the end (e.g. 95%) it uses less RAM than if run without restart.

---

 


- ALF - "Find out what you don't do well ..... then don't do it!" :)
 

I forgot to post previously - if I restart a "normal" (non-VHAR) task at e.g. 50% when it may be using 300 MB RAM - the app starts at 100 MB again (and continues from 50% done) and do not "jump" to 300 MB so at the end (e.g. 95%) it uses less RAM than if run without restart.

this sounds more like memory leak indeed,I'll look where it can hide.

What Cat versio this correspond to?
Driver version: CAL 1.4.1646
Version: OpenCL 1.1 AMD-APP (831.4)

What if Cat version will be chnaged - memory usage differs?

---

SETI apps news
We're not gonna fight them. We're gonna transcend them.

What if Cat version will be changed - memory usage differs?

Catalyst 11.12 (posted earlier)
I don't want to change to Catalyst 12.01 (Windows XP, remember? Those are the only two versions with working OpenCL)

---

 


- ALF - "Find out what you don't do well ..... then don't do it!" :)
 

I am running only one instance. I noticed it is not the GPU board making noise with its fan but the main fan of my HP desktop. The GPU board temperature is around 55 C and its fan is at 34%. For some reason the 1050 board is using more power that the GTX 750 OC it had before.
I had to downgrade the NVidia driver from 375.95 to 375.70 otherwise BOINC would not recognize it.
Tullio

Then you run other instances of other science also on 1050Ti? Or just 1 instance per every board?

Will check the driver ver when I get home...but it's working OK for me...
My questions were 'cause I've got a board that is PCIe 2.x on a 1050Ti, which is a PCIe 3.0 card...so have a problem there & also only Quad-core Xeon to serve to that GPU!

So was wondering how to give best results from the GPU...as it's mostly doing SETi@home...only every other weekend I've got some time for some playing!
But even that was done on 730 before...so 1050 should be like a Space shuttle compared to Appolo!
;)

---

non-profit org. Play4Life in Zagreb, Croatia, EU

this sounds more like memory leak indeed,I'll look where it can hide.

Two notes:
- Seems that BoincTasks records the RAM usage only from the last restart - near the end ~90% I restarted a task which was at >300 MB RAM and it grow only to 120 MB at finish and that 120 MB is shown now in History tab of BoincTasks

- Some tasks seem to not grow or do it very slowly - this task was paused hours ago (so I don't remember its RAM usage) and when I restarted it it run for >20 minutes without much grow:
(it was at ~98 MB for a few minutes when I watched, "usually" the grow I see is 1 MB per few (5-8) seconds)

blc4_2bit_guppi_57424_45953_HIP74190_0027.17319.416.18.27.165.vlar_0	01:26:24 (00:02:16)	0.04C + 1ATI	107.84 MB	115.79 MB	ar=0.008858

http://setiathome.berkeley.edu/result.php?result_name=blc4_2bit_guppi_57424_45953_HIP74190_0027.17319.416.18.27.165.vlar_0

---

 


- ALF - "Find out what you don't do well ..... then don't do it!" :)
 

But even that was done on 730 before...so 1050 should be like a Space shuttle compared to Appolo!

Apollo can go to the moon, Space shuttle can't ...
So I'm not sure which is a greater achievement?

---

 


- ALF - "Find out what you don't do well ..... then don't do it!" :)
 

I've downloaded Geforce driver 376.09 and BOINC recognizes it. I am running 5 projects on the WIndows PC, SETI@home, SETI Beta, Einstein@home, LHC@home. and Atlas@home. Only Einstein and SETI have also GPU tasks, and I am running them one at the time, with 2 CPU tasks running alongside.
Tullio

---

- Some tasks seem to not grow or do it very slowly -

Now I think that (all?) VLAR tasks "grow" very slow in the last 1/3 of processing.
Didn't look very closely to the exact % but if VLAR task is restarted at ~70-75% it starts with ~96 MB RAM usage and slow (10-20 KB/s) "grows".

I now watched blc4_2bit_guppi_57424_45606_HIP73165_OFF_0026.19637.416.17.26.219.vlar   ar=0.011183
- restarted from 73% at 96 MB and go to 99% at 114 MB (% as shown by BoincTasks)

---

 


- ALF - "Find out what you don't do well ..... then don't do it!" :)
 

And if no restarts?

---

SETI apps news
We're not gonna fight them. We're gonna transcend them.

I've downloaded Geforce driver 376.09 and BOINC recognizes it. I am running 5 projects on the WIndows PC, SETI@home, SETI Beta, Einstein@home, LHC@home. and Atlas@home. Only Einstein and SETI have also GPU tasks, and I am running them one at the time, with 2 CPU tasks running alongside.
Tullio

thx for the answer! ;)

But even that was done on 730 before...so 1050 should be like a Space shuttle compared to Appolo!

Apollo can go to the moon, Space shuttle can't ...
So I'm not sure which is a greater achievement?

Moon is not that great achievement in Space...exept a PR stunt! ;)

But anyway, trying to get my GPU to crunch 4 instances of GPU jobs, but only 3 instances of same science (cuda jobs or SoG or sah jobs)...this is my app_config.xml:

<app_config>
    <app>
      <name>setiathome_v8</name>
      <gpu_versions>
          <gpu_usage>0.25</gpu_usage>
      </gpu_versions>
    </app>
	<app_version>
       <app_name>setiathome_v8</app_name>
       <plan_class>cuda60</plan_class>
	   <max_concurrent>3</max_concurrent>
       <ngpus>0.25</ngpus>
       <cmdline></cmdline>
    </app_version>
    <app_version>
       <app_name>setiathome_v8</app_name>
       <plan_class>cuda50</plan_class>
	   <max_concurrent>3</max_concurrent>
       <ngpus>0.25</ngpus>
       <cmdline></cmdline>
    </app_version>
    <app_version>
       <app_name>setiathome_v8</app_name>
       <plan_class>cuda42</plan_class>
	   <max_concurrent>3</max_concurrent>
       <ngpus>0.25</ngpus>
    </app_version>
    <app_version>
       <app_name>setiathome_v8</app_name>
       <plan_class>cuda32</plan_class>
	   <max_concurrent>3</max_concurrent>
       <ngpus>0.25</ngpus>
    </app_version>
	    <app_version>
       <app_name>setiathome_v8</app_name>
       <plan_class>opencl_nvidia_SoG</plan_class>
	   <max_concurrent>3</max_concurrent>
       <ngpus>0.25</ngpus>
    </app_version>
	    <app_version>
       <app_name>setiathome_v8</app_name>
       <plan_class>opencl_nvidia_sah</plan_class>
	   <max_concurrent>3</max_concurrent>
       <ngpus>0.25</ngpus>
    </app_version>
</app_config>

& it doesn't work...can someone please help out?! ;)

---

non-profit org. Play4Life in Zagreb, Croatia, EU

- Some tasks seem to not grow or do it very slowly -

Now I think that (all?) VLAR tasks "grow" very slow in the last 1/3 of processing.
Didn't look very closely to the exact % but if VLAR task is restarted at ~70-75% it starts with ~96 MB RAM usage and slow (10-20 KB/s) "grows".

I now watched blc4_2bit_guppi_57424_45606_HIP73165_OFF_0026.19637.416.17.26.219.vlar   ar=0.011183
- restarted from 73% at 96 MB and go to 99% at 114 MB (% as shown by BoincTasks)

Try to save task with big RAM consumption and run it offline with 2 options set:

1) -tt 5

2) -tt 60

Look for peak RAM consumption - how it differs in these 2 cases?

---

SETI apps news
We're not gonna fight them. We're gonna transcend them.

MB8_win_x86_SSE2_OpenCL_ATi_HD5_r3557.exe
AMD Radeon HD 6570 - Catalyst 11.12
Windows XP

2) Checkpoints problem
On "Very VHAR" tasks this app almost do not checkpoint on my computer:

Try r3578: https://cloud.mail.ru/public/AeMs/SYvE7Eiei
Any improvements in progress reporting/checkpointing?

---

SETI apps news
We're not gonna fight them. We're gonna transcend them.

@Raistmer: Any suggestion for a 1050ti card in a command line?

---

non-profit org. Play4Life in Zagreb, Croatia, EU

I made a short test to train myself how to record RAM usage.
(longer will be when I have more time
- Note/Edit: this post was written ~20 h ago and was on "Preview" when Internet stopped; Now longer test is running and I post this)

For this short test results are almost the same for default 15ms (empty mb_cmdline.txt), -tt 5 and -tt 60

PG0395_v8.wu

-tt 5
Target kernel sequence time set to 5ms

"Process(MB8_win_x86_SSE2_OpenCL_ATi_HD5_r3557)\Working Set"
"12/01/2016 18:30:49.288"	" "
"12/01/2016 18:30:59.288"	" "
"12/01/2016 18:31:09.288"	"96022528"
"12/01/2016 18:31:19.288"	"97103872"
"12/01/2016 18:31:29.288"	"98578432"
"12/01/2016 18:31:39.288"	"100204544"
"12/01/2016 18:31:49.288"	"101810176"
"12/01/2016 18:31:59.288"	"103235584"
"12/01/2016 18:32:09.288"	"104853504"
"12/01/2016 18:32:19.288"	"106450944"
"12/01/2016 18:32:29.288"	"108068864"
"12/01/2016 18:32:39.288"	"109727744"
"12/01/2016 18:32:49.288"	"111333376"
"12/01/2016 18:32:59.288"	"112959488"
"12/01/2016 18:33:09.288"	"114565120"
"12/01/2016 18:33:19.288"	"115978240"
"12/01/2016 18:33:29.288"	"117358592"
"12/01/2016 18:33:39.288"	"118685696"
"12/01/2016 18:33:49.288"	"120037376"
"12/01/2016 18:33:59.288"	"121438208"
"12/01/2016 18:34:09.288"	"122822656"
"12/01/2016 18:34:19.288"	"124170240"
"12/01/2016 18:34:29.288"	"125440000"
"12/01/2016 18:34:39.288"	"126820352"
"12/01/2016 18:34:49.288"	"128233472"
"12/01/2016 18:34:59.288"	" "
"12/01/2016 18:35:09.288"	" "

-tt 60
Target kernel sequence time set to 60ms

"Process(MB8_win_x86_SSE2_OpenCL_ATi_HD5_r3557)\Working Set"
"12/01/2016 19:20:48.382"	" "
"12/01/2016 19:20:58.382"	" "
"12/01/2016 19:21:08.382"	"96006144"
"12/01/2016 19:21:18.382"	"97083392"
"12/01/2016 19:21:28.382"	"98365440"
"12/01/2016 19:21:38.382"	"99975168"
"12/01/2016 19:21:48.382"	"101584896"
"12/01/2016 19:21:58.382"	"103215104"
"12/01/2016 19:22:08.382"	"104824832"
"12/01/2016 19:22:18.382"	"106434560"
"12/01/2016 19:22:28.382"	"107847680"
"12/01/2016 19:22:38.382"	"109490176"
"12/01/2016 19:22:48.382"	"111091712"
"12/01/2016 19:22:58.382"	"112717824"
"12/01/2016 19:23:08.382"	"114323456"
"12/01/2016 19:23:18.382"	"115752960"
"12/01/2016 19:23:28.382"	"117121024"
"12/01/2016 19:23:38.382"	"118390784"
"12/01/2016 19:23:48.382"	"119812096"
"12/01/2016 19:23:58.382"	"121114624"
"12/01/2016 19:24:08.382"	"122503168"
"12/01/2016 19:24:18.382"	"123969536"
"12/01/2016 19:24:28.382"	"125022208"
"12/01/2016 19:24:38.382"	"126447616"
"12/01/2016 19:24:48.382"	"127762432"
"12/01/2016 19:24:58.382"	" "
"12/01/2016 19:25:08.382"	" "

---

 


- ALF - "Find out what you don't do well ..... then don't do it!" :)
 

2) Checkpoints problem
On "Very VHAR" tasks this app almost do not checkpoint on my computer

MB8_win_x86_SSE2_OpenCL_ATi_HD5_r3557

Have NO problem to checkpoint on AR:
1.389965
2.598181
2.725368 (maybe checkpoints every ~1.5 minutes)

Have semi-problem to checkpoint (every ~3 minutes, not at <1 minute as is set) on AR:
4.195846

HAVE problem to checkpoint on AR:
12.638138
103.055695

I will try offline test with r3578 for AR 103.055695 (I have the WU)
and in BOINC for AR 4.195846 (the WU is paused in BOINC)

---

 


- ALF - "Find out what you don't do well ..... then don't do it!" :)