New PC up and running

Having problems installing that new stick of memory? Found some great software or having issues with something? Or maybe want to chat about your PlayStation, X-Box, Nintendo, Sega, even your old Spectrum 48k....! Or maybe something you want to sell or acquire (computing related of course!). Let us know here...
Post Reply
UBT - Timbo
UBT Forum Admin
Posts: 9680
Joined: Mon Mar 13, 2006 12:00 am
Location: NW Midlands
Contact:

New PC up and running

Post by UBT - Timbo »

Hi all,

It's been a long, long struggle but at last my shiny new PC is up and running....even if it's only to get it "checked out" so I know all is working OK...so, it'll be crunching some WU's for the next day or two, and then a few drivers need updating on it.

For those interested, it's an Intel i7-3820 @ 3.60GHz (not overclocked), with a single (for now) NVidia GTX580, running Win XP Pro (for now).

Should be capable of some pretty good crunching.... :)

regards
Tim
UBT - Timbo
UBT Forum Admin
Posts: 9680
Joined: Mon Mar 13, 2006 12:00 am
Location: NW Midlands
Contact:

Post by UBT - Timbo »

Hi all,

After a couple of days "running in" the new PC, I found it works very well and makes a nice room heater....running a few CUDA apps (as well as CPU WU's) and the GTX580 was getting pretty warm, showing 99% GPU workload and with temps rising to 88-90 deg C.

I've left the fan control set to automatic and it seems pretty stable.

So, this afternoon, I took the plunge and fitted the 2nd GTX580, plus the SLi bridge adaptor.

It was then that I noticed that the rear grille on the first GTX was blocked by a blanking plate :( which was quickly removed.

A quick restart and already it's back in action with 2 GPU tasks being crunched simultaneously, plus up to 8 CPU tasks (as sometimes 1 CPU is used by a GPU tasks, especially on Einstein WU's).

I'll see how it settles down over the next day.

regards
Tim
Zydor
Posts: 437
Joined: Mon Nov 01, 2010 12:00 am

Post by Zydor »

Watch your temperatures carefully ..... 88-90 is high, too high in a generic sense. NVidia cards tend to run hotter, however, 88-90 is high, and could do with taking down a little - an experienced NVidia user could chip in here if I am wrong re temp for NVidia card as I am an AMD user.

I would be cautious until an - experienced - NVidia user checks that out. 88-90 seems too high for long term 7x24 use. Many may say "its ok I been doing that for weeks at that temperature" ..... however too high a temp will over time reduce card life drastically - like down from 10+ years to circa 6-18 months.

Its one to be cautious about..... speak to an - experienced - NVidia user as AMD and NVidia cards do run at different "normal" temperatures.

If all CPU cores automatically kicked in - then its possible the wrong setting was applied to the config. Have a peak at this thread (read from post one 11 Apr):

http://boinc.thesonntags.com/collatz/fo ... hp?id=1151
david
Posts: 569
Joined: Fri Mar 30, 2007 1:00 am

Post by david »

I run all Nvidia cards on 3 different M/C's

 All currently run at core temp of 82-83C,given the summer we have had,i think that is acceptable,it may change as the weather cools,but i wouldn't expect a great deal

  Cards are as follows:
 
  1) EVGA Superclocked Nvida GTX 780Ti

  2) Gigabyte GTX 580

  3) Gigabyte GTX 650

   I would say 88-90C is a little to high for sustained use and as Zydor says will shorten the life of the card if thats not a factor then go for it

  David
UBT - Timbo
UBT Forum Admin
Posts: 9680
Joined: Mon Mar 13, 2006 12:00 am
Location: NW Midlands
Contact:

Post by UBT - Timbo »

Hi Zydor,

Thanks for the warning....

As things stand, I was fairly sure that the 88-90 deg C temp was a bit on the high side, but as this GPU stuff is all new to me, I thought I'd post the above and see what others thought.

So, the next question is: Can the GPU apps be "pegged back" a bit so they are not operating at "full blast" so to speak ?

Is this a job for an "app_config.xml" setting ?

Also: I don't have an extra case fan to remove heat, though there's an enormous CPU fan, plus a 14 cm PSU fan, and both these seem to removing excess heat from the case. (Both CPU and PSI are running cool to the touch BTW).

Alternatively, if the fan speed is in "automatic" mode then if the card was sensing that the temp is too high then surely the fan speed would be increased to reduce it....the fan speed us currently only at 70% and temp is settled down to about 86 deg C.

FWIW: The single GTX560 is running at 69-70 deg C, though unfortunately, with the latest NVidia drivers/Control Panel, there is no way to manually adjust the fan speed, although the same drivers on the GTX580's do allow this.

If any NVidia fans are looking in, please feel free to respond with any tips.

regards
Tim
Zydor
Posts: 437
Joined: Mon Nov 01, 2010 12:00 am

Post by Zydor »

Read the post below carefully .... it will indicate the principles behind proper useage of app_config etc.

Do bare in mind that post is for a different Project, however, the thrust behind it all remains 98% true for all Projects. Whilst the "new" apps will tend to look after you in the sense of kicking in a default if no app_config is specified, most will have the default grab all the card resource they can lay their hands on - that can be irritating, but its understandable from a Project Admin view when trying to push his/her Project.

Many people these days are "spoilt" by the new BOINC system of app_config etc compared to the "bad old days" of config.sys/autoexec.bat in that now all we do is indicate the hardware we have and the Project backend takes it from there. Many just don't bother with app_config etc any more "cos it runs anyway". Whilst the latter is true .... end result is often a Project grabbing all it can get.

So .... always explicitly type out an app_config.xml and cc_config.xml, you are then in control, not the Project Admins - the latter will grab all they can which is not necessarily what you want.

Do look at the post below carefully, you will see I limit the amount of CPU used and stay in control of what the card does. Once the card does use the resource you want it to, and if it still runs hot, then it would be time to turn down the speed a little until you are more used to controlling the fans.

Safety first will lose you a few credits (not that many) but you will still have a card to play will later (!). It will not take long to get used to the detail of controlling GPUs, zillions of posts out there about it all - but whilst you do that, bring it down a little in terms of speed.

That way you will still have a card left to play with once you are up to speed on the fine details rofl :)

http://boinc.thesonntags.com/collatz/fo ... 9045#19045
UBT - Timbo
UBT Forum Admin
Posts: 9680
Joined: Mon Mar 13, 2006 12:00 am
Location: NW Midlands
Contact:

Post by UBT - Timbo »

Hi Zydor,

Thanks for the briefing.....very informative indeed.

In fact I've read this before and despite re-reading it, (and having been a DOS-junkie, when I was able to re-write config.sys and autoexec.bat files blindfolded, as well as using 4DOS as my command interpretor (instead of MS command.com)), I still find it difficult reading the app_config.xml file so unbearably tedious.....

For instance:

<app_config>
<app>
<name>solo_collatz</name>
<max_concurrent>5</max_concurrent>
<gpu_versions>
<gpu_usage>1.0</gpu_usage>
<cpu_usage>0.85</cpu_usage>
</gpu_versions>
</app>
</app_config>

...

Using the above as an example, Collatz will assume two GPU Cores need two CPU to support (at 0.85 CPU per GPU), and will allocate 3 CPU for CPU WU at one per CPU WU, total useage 4.7 CPU


so I take it:
<max_concurrent>5</max_concurrent>
specifies the maximum number of CPU's to use on BOINC

<gpu_usage>1.0</gpu_usage>
So, does this mean use each GPU fully?

<cpu_usage>0.85</cpu_usage>
and does this mean the maximum level of CPU "support" per each GPU task?

So, 5 CPU's, (of which 2 GPU's take 0.85 CPU "power each"), which makes 1.7 "CPU's", plus 3 CPU's = 4.7


Now, where does it say "use (say) 50% of each GPU" ??


Even writing config.sys files to allocate expanded and extended memory (for use with Lotus 1-2-3) is easier than this !!

regards
Tim
Zydor
Posts: 437
Joined: Mon Nov 01, 2010 12:00 am

Post by Zydor »

so I take it:
<max_concurrent>5</max_concurrent>
specifies the maximum number of CPU's to use on BOINC
Close but not quite ... For me .... it shows 2 x GPU and 3 x CPU, because I have two GPU running on BOINC, so it takes the number of GPUs running, deducts that, leaving three - which are therefore 3 x CPU (of course another 2 CPU are automatically taken to run the GPU WUs no need to specify those two CPU, BOINC assumes it). max_concurrent means The maximum number of tasks of this application to run at a given time beit CPU or GPU WUs

Another example with same numbers:
If I was running 1xGPU (I'm not), then it would mean .... 1xGPU leaving four left over, so would therefore be running 4xCPU WU

<gpu_usage>1.0</gpu_usage>
So, does this mean use each GPU fully?
Yes ..... if you had put (for example) 0.5, that would have meant run 2xGPU WUs on each GPU. 0.33 would have meant run 3 on each GPU ..... etc
<cpu_usage>0.85</cpu_usage>
and does this mean the maximum level of CPU "support" per each GPU task?
Correct ..... in the example it means give each GPU dedicated access to 0.85 CPU
So, 5 CPU's, (of which 2 GPU's take 0.85 CPU "power each"), which makes 1.7 "CPU's", plus 3 CPU's = 4.7
Spot on ..... you win todays Jackpot :)
Now, where does it say "use (say) 50% of each GPU" ??
You would put  <gpu_usage>0.5</gpu_usage>

Full syntax is at (do a quick scroll down first to get the idea of whats there - there is lots :) )

https://boinc.berkeley.edu/wiki/Client_configuration
UBT - Timbo
UBT Forum Admin
Posts: 9680
Joined: Mon Mar 13, 2006 12:00 am
Location: NW Midlands
Contact:

Post by UBT - Timbo »

Hi Zydor

OK - I'm really getting a headache with this app_config stuff....and I'm sure that others might be too. I've read the article but it doesn't give any examples, so one is left floundering....and short of editing and resaving the XML file multiple times (stopping and restarting BOINC every time), which is a pain.

And it might even be that what I'm trying to do, can't be done, given the limited number of "options" in this XML file.

So, I'll go through what I want to do and then if anyone can give me some pointers, then I might be able to suss it out.

1) Those projects that have GPU work available, I have set "Use CPU" preference (on the specific project website) to "No".

This way, I can maximise the amount of work I do using GPU and keeping the CPU's glowing doing work where projects don't have a GPU app.

2) One PC is an hyperthreaded 4 core Intel i7, giving me 8 virtual CPU's, plus it is fitted with 2x GTX580's.

The 8 CPU's will only crunch projects with CPU WU's.
And to minimise the heat produced by the GPU's, I want to reduce the work load on them to maybe 50%.

I've checked the GPU temperatures, when only crunching CPU WU's and they are both at around 38-40 deg C.

But, thus far, when trying various settings I cannot get both GPU's to (say) tick over at 50% load on each.

So, for each "app", one assumes that the app_config.xml can be manipulated to do what I want.


Here's the basic format of app_config.xml:

<app>
     <name>uppercase</name>
     <max_concurrent>x</max_concurrent>
     <gpu_versions>
         <gpu_usage>y</gpu_usage>
         <cpu_usage>z</cpu_usage>
     </gpu_versions>
   </app>

max_concurrent
The maximum number of tasks of this application to run at a given time.


So, x would be 2 (as I have 2 GPU's)

gpu_usage
The number of GPU instances (possibly fractional) used by GPU versions of this app.

So, y would be 0.5 (50% workload on each)

cpu_usage
The number of CPU instances (possibly fractional) used by GPU versions of this app.


So, z would be any number, as specified by the project.


I tried the above, having first edited the app name accordingly and saved the file into the project folder, and for PrimeGrid all I got was "waiting to run" even with NO other CPU or GPU tasks active.

regards
Tim
stuey77
Posts: 8
Joined: Mon Oct 29, 2012 12:00 am

Post by stuey77 »

Hi Timbo

I'm not sure you can achieve what you want by using app_config. I see you are trying to limit 1 task to each card and set each card to only run at 50% by setting gpu_usage to 0.5.

I'm sure I've read somewhere that if your gpu_usage is less than 1, but you are only executing one task on the card, the setting is ignored. I may be wrong because I can't find where I have read it now, but I'm sure I've seen it somewhere. From what I understand, setting the gpu_usage to say .5 would let you execute 2 tasks and .33 would be 3 tasks etc... although this would increase stress on the card. However if only one task is running the task will essentially consume all available resources it requires of the card, unlike on a CPU where you can specify the % used.

Maybe this program will help?  http://efmer.eu/boinc/index.html Not tried it myself, but it looks like you can throttle the GPU based on its temperature. The other solution may be to underclock the cards so less heat is produced. These may allow you to ditch app_config if you're happy to run yet another background task instead!

Do the Primegrid tasks run with the app_config removed?
Zydor
Posts: 437
Joined: Mon Nov 01, 2010 12:00 am

Post by Zydor »

Tim

Which Project (s) do you want to crunch with the GPUs ?

I will write the app_info.xml (s) for you, and then you can look at them and ask questions on what still isn't clear.

Its best that way to avoid me writing an all inclusive Tomb, and you fall asleep before the end of it rofl :)

This is one of those topics, that when the "Penny Drops", we all gurgle a strangulated scream  :toothy7:
UBT - Timbo
UBT Forum Admin
Posts: 9680
Joined: Mon Mar 13, 2006 12:00 am
Location: NW Midlands
Contact:

Post by UBT - Timbo »

stuey77 wrote:Hi Timbo

I'm not sure you can achieve what you want by using app_config. I see you are trying to limit 1 task to each card and set each card to only run at 50% by setting gpu_usage to 0.5.

I'm sure I've read somewhere that if your gpu_usage is less than 1, but you are only executing one task on the card, the setting is ignored. I may be wrong because I can't find where I have read it now, but I'm sure I've seen it somewhere. From what I understand, setting the gpu_usage to say .5 would let you execute 2 tasks and .33 would be 3 tasks etc... although this would increase stress on the card. However if only one task is running the task will essentially consume all available resources it requires of the card, unlike on a CPU where you can specify the % used.
OK - understood - and you've pretty much hit the nail on the head in terms of what I want to do.....it just seems that app_config.xml doesn't have enough options though.
stuey77 wrote:Maybe this program will help?  http://efmer.eu/boinc/index.html Not tried it myself, but it looks like you can throttle the GPU based on its temperature. The other solution may be to underclock the cards so less heat is produced. These may allow you to ditch app_config if you're happy to run yet another background task instead!
Now THAT looks really interesting - TThrottle and the companion program BoincTasks looks like just the job for me....giving useful feedback AND temperature control....I'll report back and let you know how I get on :)
stuey77 wrote:Do the Primegrid tasks run with the app_config removed?
Yup - without the specific PrimeGrid app_config.xml file, PG works fine.....but I'm seeing GPU temps in the high 80's even though the GPU's are running at standard clock ratings (ie NOT overclocked)...GPU load is at 99% and it's this I'd like to reduce so that the GPU temp is reduced.....I regard any WU credit from the GPU's as being a bonus, compared to the CPU credits.....so I don't need to run "full tilt" on the GPU's.

regards and thanks for the input.
Tim
UBT - Timbo
UBT Forum Admin
Posts: 9680
Joined: Mon Mar 13, 2006 12:00 am
Location: NW Midlands
Contact:

Post by UBT - Timbo »

Zydor wrote:Tim

Which Project (s) do you want to crunch with the GPUs ?

I will write the app_info.xml (s) for you, and then you can look at them and ask questions on what still isn't clear.

Its best that way to avoid me writing an all inclusive Tome, and you fall asleep before the end of it rofl :)

This is one of those topics, that when the "Penny Drops", we all gurgle a strangulated scream  :toothy7:
Hi Zydor,

Thanks for the kind offer. I appreciate your patience with me on this :)

I've got a number of GPU projects running on this PC.....so, in theory, the same type of settings in app_config.xml would suit them all.....for now I've been experimenting with PrimeGrid and specifically I have one WU that I've got "on the go" which is "genefer_wr"

However this PC is also doing GPU tasks for SETI, Asteroids, MilkyWay etc, so any of these would be OK.

I'm also trying stuey's suggestion and seeing if that works, as that could be a really simple solution.

regards and thanks
Tim
UBT - Timbo
UBT Forum Admin
Posts: 9680
Joined: Mon Mar 13, 2006 12:00 am
Location: NW Midlands
Contact:

Post by UBT - Timbo »

stuey77 wrote:....Maybe this program will help?  http://efmer.eu/boinc/index.html Not tried it myself, but it looks like you can throttle the GPU based on its temperature. The other solution may be to underclock the cards so less heat is produced. These may allow you to ditch app_config if you're happy to run yet another background task instead!.....
Hi Stuey

I downloaded the program TThrottle v7.30 (the latest one) and installed it on my Q6600 based quad-core (which doesn't do anything else), and which has the GTX560, and this was running a bit on the "warm side" at around 70+ deg C.

After a little bit of "configuring", it's now stable running the CPU's at 100% and the GPU has settled down to 65-67% leading to a GPU temp of 65 deg C.

(The latest NVidia driver doesn't allow one to control the GTX560 fan speed strangely, so this program helps a lot).


So, I installed TThrottle on the i7 octo-core/2x GTX580......

"Configuring" this took a bit longer, as no matter what I did to the "Set GPU" temp box, the temperature was always over the threshold....

With further monitoring of the measured temps, I found that it was possible to set the temps higher than I wanted and then TThrottle has a good chance of bringing the temp down to near enough what you specify.

But if you specify too low a temp, it can't cope and the temp rises anyways. :(

As things stand, I've got 2x PrimeGrid Genefer_WR tasks running (as I couldn't get the single WU I had to start running again, until I downloaded another one...and I got given another "WR" WU - which takes 120 hours to complete :( ) - and the lowest I can get the GPU's down to (temp wise) is 85 deg C.

So, I think the jury is still out on the octo-core, but on the quad-core with a single GPU, it works a treat.

regards
Tim
Zydor
Posts: 437
Joined: Mon Nov 01, 2010 12:00 am

Post by Zydor »

Be careful of TThrotle, its a very good Niche Utility, but it has definite downsides...... it is not a catch-all, cure-all speed device.

TThrotle can be abused, and mask problems which should be attacked at the core issue, not attacked by masking symptoms. At a fundamental level, you don't avoid Speeding Tickets by installing heaps of anti radar devices, you slow down a tad .....

There is a subtle clue in the latter statement ..... :)

I'll work on a few configs for you, they will need close scrutiny as its easy to do it wrong in detail when working "from afar", depending on what you personally need doing,  for each Project.
Last edited by Zydor on Sat Apr 19, 2014 5:41 pm, edited 1 time in total.
UBT - Timbo
UBT Forum Admin
Posts: 9680
Joined: Mon Mar 13, 2006 12:00 am
Location: NW Midlands
Contact:

Post by UBT - Timbo »

Zydor wrote:Be careful of TThrotle, its a very good Niche Utility, but it has definite downsides.

TThrotle can be abused, and mask problems which should be attacked at the core issue, not attacked by masking symptoms. At a fundamental level, you don't avoid Speeding Tickets by installing heaps of anti radar devices, you slow down a tad .....

There is a subtle clue in the latter statement ..... :)
Noted - The main "worry" I have is about the 2x GTX580 combo, as left "unhindered", the GPU's are being stressed too much and their temp is just too much. Now, I could invest in some nice water-cooling rigs....but that's not really "my thing". Instead all I want to do is to have a bit of extra "oomph", on top of what the CPU's are doing....so, throttling back the GPU's somewhat, which reduces their temperature is a very useful function to have.
Zydor wrote:I'll work on a few configs for you, they will need close scrutiny as its easy to it wrong in detail depending on what you need doing,  for each Project.
That would be great.

Put simply, I want the CPU's to only do WU's where no GPU tasks are available, and on other projects to be doing GPU tasks.

As such, I've set my "preferences" on various projects to only accept GPU tasks, if they support them.

This then allows me to be able to manually switch projects and to know that each PC is being used efficiently.....I don't really want to be crunching (for example) SETI CPU WU's if I can do the same tasks better with a GPU.

And in the case of the octo-core, I just want to be able to "dial back" the work-load so the GPU's run a bit cooler.....as any credits gained are just a bonus over and above that the CPU's are doing.

regards
Tim
Zydor
Posts: 437
Joined: Mon Nov 01, 2010 12:00 am

Post by Zydor »

...... I just want to be able to "dial back" the work-load so the GPU's run a bit cooler ...
Run - don't walk - and get MSI Afterburner. The trick is get the GPU fans to work for their living, not dial back GPU Core speed (its free - MSI use it as a "come on" to sell their surperb Graphics cards). It will only give 100% of its functionality to MSI Cards, but it will work on 95% of the time with other card makes. Made by a Russian (lives near the Ukraine Boarder, so he is having a "fun time" at present, and limited latest development work on Beta 19 because of that) his Internet Nic is "Unwinder" ....... he doesn't suffer fools gladly .........

.......he eats them for breakfast rofl :)

He just released 3.0.0 Beta 19. You need to "get into bed" with this software, its now the defacto standard for controlling GPU cards.

Takes a little getting used to, Unwinder has put in a few control measures to stop the Dorks of this world abusing it, but the measures are easily reversed by sensible Crunchers.

http://www.guru3d.com/files_details/msi ... nload.html

Don't get conned by the advert top right - go to the bottom of the article page, the download links are there, make sure you click on Beta 19.

His forum (bookmark it - you will get into the terminology et al over time) is at:

http://forums.guru3d.com/forumdisplay.php?f=55
Zydor
Posts: 437
Joined: Mon Nov 01, 2010 12:00 am

Post by Zydor »

Sorry for the delay - "She-who-must-be-Obeyed" got in the way with boring 'ol domestic stuff (!) ..... she'll kill me when she reads that on her PC rofl :)

In a way this will end up a bit of a "Tomb" - and profuse advance apologies if it outlines some "stuff" you are aware of - its difficult to know how to pitch this when trying to change a mindset. My aim is to "reset" your thinking to focus on GPU, and from that position you then move onto (what is now, minor stuff) the control of (the now) multi core CPU.
The main "worry" I have is about the 2x GTX580 combo, as left "unhindered", the GPU's are being stressed too much and their temp is just too much. Now, I could invest in some nice water-cooling rigs....but that's not really "my thing". Instead all I want to do is to have a bit of extra "oomph", on top of what the CPU's are doing....so, throttling back the GPU's somewhat, which reduces their temperature is a very useful function to have.
You need to "recalibrate" the main thrust of thinking. TThrotle first came on the scene about 10 years ago, the author plugging a then gap in PC performance control. This was way, way before the current abilities to control the hardware via the config settings, the massive changes to BOINC post 6.12.0 and the far better direct control that now exists over the hardware - CPU and GPU. Now, you have far better control over hardware with Utilities such as Afterburner, and the improved configuration files with cc_config.xml and app_config.xml.

The advent of multi-core machines aka 4, 8, 12 core - soon to be 8/16 core with next generation IBM 5XXX CPUs due in Q3 this year - mean the best way to control CPU heat et al (from a Cruncher view) is just focus on CPU "Cores" and their core speed, increasing/reducing numbers of CPU Cores in use. Niche utils exist for CPU control in depth, but the start of the journey has to be with MSI Afterburner (GPUs), and app_config.xml (both GPUs and CPU cores).

Just leave the CPU running at default, that will do for now, get used to GPU control first, tackle the CPU when the GPU control is cracked. that way your CPU will not be in danger of over heating while doing the mind-set change. Leave the CPU WUs alone for now, just do the GPU WUs via Afterburner, when that's cracked and settled, move onto CPU WUs - by then that will be easy. Crack Afterburner GPU control first, leave CPU at default with no CPU WU for now.

From that point, if you need more detailed control, niche utilities exist for the CPU, but the output of current GPUs now so far outstrips the CPU, the latter is minute in comparison, and a per-CPU-Core control (plus BIOS control over CPU speed) is usually enough to regulate CPU heat and output, via the app_config.xml. The latter also sets the number of GPUs to be used on BOINC, and Afterburner will take that setting and apply its control magic of the allowed number of GPUs to be used for BOINC, leaving any left over unaffected by BOINC and available for general PC useage

I often set one GPU not available to BOINC via the app_config, when Gaming, BOINC then leaves alone GPU 0 for me to use on Gaming, and (automatically) only makes GPU 1 available to BOINC.

If you wish I can take you through some sessions verbally via "Skype" (free - basic voice et al is free) if you have a headset/mic (if you don't have a headset/mic, dare I suggest you run -   don't walk - to get one, hugely useful these days), verbal is always immeasurably faster than typed word in getting across this stuff. If you want to take that route, PM me and I will send you my Skype Name via PM, and I'll talk you through it.  It will also be far easier to get hold of me via voice whenever you need to as you make the transition, (I always have Skype turned on when I am on line).

Lets start by getting you going on Collatz with the GPUs. Some Projects are better than others for some cards, but ignore that for now. To "Benchmark" yourself watch James Ying's 580's at Collatz (he is a Chinese businessman working in China - nice guy).

There are many twists and turns to using the "new" cc_config.xml and app_config.xml, so for now we'll keep it straightforward and leave off the fancy stuff. Change your cc_config.xml (located in C:\ProgramData\BOINC, if not there, create one with Notepad - must ensure its saved as a .xml file) to:

<cc_config>

<options>
<use_all_gpus>1</use_all_gpus>
<save_stats_days>180</save_stats_days>
<max_file_xfers>48</max_file_xfers>
<max_file_xfers_per_project>36</max_file_xfers_per_project>
<report_results_immediately>1</report_results_immediately>
</options>

</cc_config>

..... and change your app_config.xml (located in C:\ProgramData\BOINC\projects\boinc.thesonntags.com_collatz - again create via notepad - only notepad, no other util - saving as an .xml file)  to:

<app_config>
<app>
<name>solo_collatz</name>
<max_concurrent>3</max_concurrent>
<gpu_versions>
<gpu_usage>1.0</gpu_usage>
<cpu_usage>1.0</cpu_usage>
</gpu_versions>
</app>
</app_config>

The above is a basic app_config which will get both GPUs running at Collatz, together with 1xCPU on Collatz CPU WUs, total assets in use is therefore 5 (2xGPU and 3xCPU - two of the CPUs support the GPUs, and a third runs on its own doing CPU WUs).

Now go get Afterburner, and install it, don't do anything to it, just take it "as is" for now, get it running before loading up BOINC.

Don't run ahead of yourself, do this basic change first and let it settle, until you are comfortable with whats happening. Run the PC CPU and GPU  at default settings/speeds, no fancy stuff, that way nothing will happen to put your cards at risk, so don't have a concern.

Already I suspect you will see why a headset/mic comes in real handy to explain this stuff - my Skype name is ..... "goldiloc" .... I know ... don't ask rofl :)

Yell when you have all that in place, Afterburner is running, and the basic cc_config.xml and app_config.xml are in place, and WUs running.

Please Please, don't run ahead and do other stuff, lets go step by step, its far better that way in the long run.
Zydor
Posts: 437
Joined: Mon Nov 01, 2010 12:00 am

Post by Zydor »

Sorry .... forgot one thing ...... keeping the PC updated, the latter is hugely important to GPU crunching, don't get complicated about this .... just do it (!) ..... rofl :)

Windows update only goes so far - as users of RadarSynch PC Updater are aware (!) - the first time you use that, it takes an average 2-4 hours for the PC to catchup on the stuff it is missing. Do not be deluded in thinking Windows Update will do the job - yes, it does for the main ones Microsoft are concerned with from a business perspective. From a User end, Windows Update has its limitations, and RadarSynch has it nailed......

Utils abound these days, but if you get into GPU computing for any length of time, it is essential that you have a good set of Utilities - its easy to convince yourself freebie ones are good, until you use a paid for one .....

Go get and pay for these three:

CCleaner - https://www.piriform.com/ccleaner

RadarSynch PC UpDater - http://www.radarsync.com/

ARO 2014 - http://www.sammsoft.com/ARO2014.aspx

There are another half a dozen or so others I use, but that "starter Set" of three will do you fine for now. Proof is in the eating as they say - I haven't Blue Screened for years, and there is no doubt the main reason is RadarSynch PC Updater, excellent easy to use software.

I don't want to jam up the thread by interminable discussions of Utilities, anyone wishing to do so, great, please start another thread. :)
UBT - Timbo
UBT Forum Admin
Posts: 9680
Joined: Mon Mar 13, 2006 12:00 am
Location: NW Midlands
Contact:

Post by UBT - Timbo »

Zydor wrote:...Run - don't walk - and get MSI Afterburner. The trick is get the GPU fans to work for their living, not dial back GPU Core speed (its free - MSI use it as a "come on" to sell their surperb Graphics cards). It will only give 100% of its functionality to MSI Cards, but it will work on 95% of the time with other card makes. Made by a Russian (lives near the Ukraine Boarder, so he is having a "fun time" at present, and limited latest development work on Beta 19 because of that) his Internet Nic is "Unwinder" ....... he doesn't suffer fools gladly .........

.......he eats them for breakfast rofl :)

He just released 3.0.0 Beta 19. You need to "get into bed" with this software, its now the defacto standard for controlling GPU cards.

Takes a little getting used to, Unwinder has put in a few control measures to stop the Dorks of this world abusing it, but the measures are easily reversed by sensible Crunchers.

http://www.guru3d.com/files_details/msi ... nload.html

Don't get conned by the advert top right - go to the bottom of the article page, the download links are there, make sure you click on Beta 19.

His forum (bookmark it - you will get into the terminology et al over time) is at:

http://forums.guru3d.com/forumdisplay.php?f=55
Hi Zydor,

Thanks for posting this.

I've installed AfterBurner on the i7/2xGTX580 rig. All it enables me to do is to adjust the Core Clock and Memory Clock and the Fan Speed - the top two items (Core Voltage and Power Limit) are greyed out.

And in the case of Fan Speed, the NVidia Control Panel was doing that anyways....however, by lowering the Core & Memory Clocks, the GPU temperature has now been stabilised to around 70 deg C.

I could have left the Clocks at "standard" and simply increased Fan Speed, but then the noise would have been too much, plus one has to wonder whether it's sensible to run the fans at 4,000+ rpm, when dialling back the Clocks makes more sense....as mentioned before, I'm not after maximum credits per se......but running some GPU apps is simply a bonus, even if they are taking a little longer to complete......my current estimates are that GPU WU's take around 20% longer to finish.

BTW: I do like the AfterBurner graphic charts which show temps etc ;)

As far as the quad-core/GTX560, that's stable with TThrottle with GPU temps at 69 deg C, so I'll keep that for now, as it's doing fine, though I'll put AB on it next week.

regards and thanks
Tim
Last edited by UBT - Timbo on Thu Jul 03, 2014 6:14 pm, edited 1 time in total.
UBT - Timbo
UBT Forum Admin
Posts: 9680
Joined: Mon Mar 13, 2006 12:00 am
Location: NW Midlands
Contact:

Post by UBT - Timbo »

Zydor wrote:Sorry for the delay - "She-who-must-be-Obeyed" got in the way with boring 'ol domestic stuff (!) ..... she'll kill me when she reads that on her PC rofl :)
Hi Zydor

No problems.....in life, one has to make priorities, and SWMBO is most certainly one of those !!
Zydor wrote:...BIG SNIP...

Yell when you have all that in place, Afterburner is running, and the basic cc_config.xml and app_config.xml are in place, and WUs running.

Please Please, don't run ahead and do other stuff, lets go step by step, its far better that way in the long run.
OK, first thing is that I have a number of Docking (CPU) and Asteroid (GPU) WU's I'm running.....Docking as they will be closing the project at the end of this month, so I'm doing a last minute "push"...and Asteroid, as they are nice and short ( ;) ).

So, I'm not going to be able to "experiment" by downloading and crunching some Collatz WU's using the Collatz app_config, until after (say) 30th April.


However, FYI: I did quickly adapt the app_config you posted and I'm using it with Asteroid, just to see what it did.....and I've noted the following:

Normally, Asteroids, takes 0.01 CPU + 1 GPU. Using your app_config for Asteroids, reduces the CPU's available for Docking from 8 to 6.....and for each Asteroid WU, it shows as 1 CPU + 1 GPU.

Maybe Collatz works differently and needs more CPU "power", but I've got less "crunching" ability as I'm only doing 6 Docking WU's at a time.....while Asteroids hasn't shown any noticeable "speed up" even though it has 2 full CPU's at it's disposal.

I've removed that app_config from Asteroids for now, but it was fun seeing it at work.

regards
Tim
Zydor
Posts: 437
Joined: Mon Nov 01, 2010 12:00 am

Post by Zydor »

If you want to go further with this, just yell when your crunching commitments are done (as such).

At present there is a massive amount we can do, that file was just a generalised  stab in the dark, its easily made more relevant to what you do.

It will take you to the "Next Level" re Credits with those cards, but it will take a learning curve doing it. Yell/PM me if you want to go further and get specific and targeted with the various configuration variations and options

Be aware from get go - winding those cards up fully 7x24 will likely double or more your BOINC electricity bill. Going flat out, they will probably add circa £40-60 a month - in total for both - to the electricity bill .....
UBT - Timbo
UBT Forum Admin
Posts: 9680
Joined: Mon Mar 13, 2006 12:00 am
Location: NW Midlands
Contact:

Post by UBT - Timbo »

Zydor wrote:If you want to go further with this, just yell when your crunching commitments are done (as such).

At present there is a massive amount we can do, that file was just a generalised  stab in the dark, its easily made more relevant to what you do.

It will take you to the "Next Level" re Credits with those cards, but it will take a learning curve doing it. Yell/PM me if you want to go further and get specific and targeted with the various configuration variations and options

No problems - will be in touch in due course :)
Zydor wrote:Be aware from get go - winding those cards up fully 7x24 will likely double or more your BOINC electricity bill. Going flat out, they will probably add circa £40-60 a month - in total for both - to the electricity bill .....
This has already been noticed - I have a energy monitor and I can see what the effect of having 3 PC's on 24/7 is - currently, it's showing as approx 1.4 kWh.....and that's without any other (significant) electrical items in use....to a degree, this was also a reason why restricting the "capability" of the GPU's made some sense to me....

I've not mentioned the ambient temperature yet....just wish I'd got these rigs sorted out before last winter as the warm air heating would have been very welcome. ;)

regards
Tim
Zydor
Posts: 437
Joined: Mon Nov 01, 2010 12:00 am

Post by Zydor »

this was also a reason why restricting the "capability" of the GPU's made some sense to me....
Easiest thing to do is just switch off the machine when you go to bed, that way the cards perform as designed when your up and about - life is easier that way with the configuration / performance.

Having chunky cards then crippling them whilst you use them  doesn't make a lot of sense when you can just turn off the beast before bed and have the same effect re electricity consumption.

It also means the heat appears when you need it most - whilst awake :)

Horses for Courses at the end of the day as they say - we all have different routines.
UBT - Timbo
UBT Forum Admin
Posts: 9680
Joined: Mon Mar 13, 2006 12:00 am
Location: NW Midlands
Contact:

Post by UBT - Timbo »

Zydor wrote:Easiest thing to do is just switch off the machine when you go to bed, that way the cards perform as designed when your up and about - life is easier that way with the configuration / performance.

Having chunky cards then crippling them whilst you use them  doesn't make a lot of sense when you can just turn off the beast before bed and have the same effect re electricity consumption.

It also means the heat appears when you need it most - whilst awake :)

Horses for Courses at the end of the day as they say - we all have different routines.
Hi Zydor,

You're quite right that "might" be an option.,.....excepting that the GPU's go from cold (ie PC not powered up) to "hot" in about 3-4 minutes....and once hot, then there's little to prevent them overheating.....especially as I'm not at home during the day to "baby-sit" them.

As previously mentioned, I regard the GPU's as being able to do work that otherwise, I would not be able to do, so by getting them, I've increased my crunching ability.

Having tried 2x GTX580's in one PC, (which might be considered as "overkill"), it might make more sense (in the short term) to remove one and put it in another PC, which them reduces the temperature inside one chassis and I can then better control the one that is left with either TThottle or something else.

(As neither MSI AfterBurner nor TThottle seem to be able to cope with 2 GPU's).

And in the meantime, I need to fully understand what all the settings do in the app_config.xml file.....and I can do that easily enough through experimentation with the quad-core and GTX560.

regards
Tim
Zydor
Posts: 437
Joined: Mon Nov 01, 2010 12:00 am

Post by Zydor »

At this stage you must do what you feel comfortable with ..... later you will move forward with other aspects.

You can use MSI Afterburner to set a graduated response to heat build up inside the GPUs, that will protect you.

Afterburner is a "must have" for GPU crunching, do get into it from get go (shut down CCC you don't need it, afterburner is by far your best bet)

Go at your pace - yell as and when you need help....
Post Reply