Posts by ausymark

21) Message boards : Number crunching : Rescheduling Hosts - Bad Practice (Message 1470236)
Posted 29 Jan 2014 by Profile ausymark
Hi Richard

I glanced over the link and see that its nowhere near as simple as I had assumed... I will have a good look at it tomorrow to see if I get any flashes of brilliance, or just zombie eyes lol

Cheers :)

Mark
22) Message boards : Number crunching : Rescheduling Hosts - Bad Practice (Message 1470206)
Posted 29 Jan 2014 by Profile ausymark
Hi Richard

Well thats the thing, we wouldnt need to calculate the flops at any instant as they are calculated when boinc first starts and runs the benchmarks on the PC. Then all one has to do is work out the % load that the seti program is having on the core(s), which Im guessing can be done fairly easily, and if not use the total core load, less precise but possibly still more accurate than what is currently being used. That load on the CPU to calculate the 'odometer'should be very minimal in the scheme of things these days. (Linux and Windows system monitors do that with negligible load.)

And to reduce manipulation of the system have BOINC run the benchmark program randomly again every 24 to 40 hours.

Perhaps im over simplifying it and dont know the complexities involved, but i guess the question is - would it be a more accurate system than what we currently have assuming it doesnt adversely affect affect crunching ability by more than half of one percent?

Cheers

Mark
23) Message boards : Number crunching : Will BOINC/ SETI run on a Beowulf Cluster ? (Message 1470201)
Posted 29 Jan 2014 by Profile ausymark
I think your main limitation will be how much power the beast sucks in and how much that will cost you.

As a comparison, not including my Intel i7 2600K, the two nVidia video cards I have in my single PC has just a tad over 3000 GFLOPS, so well over 40 times the expected processing power of your proposal. Its total power draw is under 500W, so in theory uses around 12kWh's of power per day if it was running 24 hours a day, or around $5 per day with Australian power prices.

My guess is that the power draw from your proposed system will be at least 1,000W (1kW). So you have double the power draw, which means on a performance per watt basis my rig suddenly becomes 80 to 100 times more powerful than yours.

(BTW Im not saying mines better than yours, just that there are likely better ways of achieving what you want.)

If on the other hand you have deep pockets, and you have room for several video cards on the cluster then that changes the efficiency of your proposal greatly. Just be prepared for a large electricity bill. ;)

Cheers

Mark
24) Message boards : Number crunching : Rescheduling Hosts - Bad Practice (Message 1470198)
Posted 29 Jan 2014 by Profile ausymark
To the developers, just a thought on how to handle credit calculation/allocation that should solve everyones issues - assuming its possible.

Here is my thought. When Boinc starts it knows the processing Power of the CPU and GPU's of its computer. There does not need to be any estimates of how long a work unit will take, what we do need is something like an Odometer within the seti apps(s) that calculates and tallies how much processing power is being used at that point on that work unit (calculated every second for example), then tallies up the total processing power used to fully process that work unit. (Hence my Odometer analogy - or even 'total trip fuel used' meter - whatever analogy works best for you)

Also by doing it this way fixes any issues we have of diminishing credits due to more efficient ways of processing the work units in software.

So for example, lets say we have a work unit that takes 3 seconds to complete and the processor can process at 600 MIPS per core, and we are using one core (for the sake of simplicity in the example).

Second SETI CPU Loading MIPS Total Mips
1 50% 300 300
2 100% 600 900
3 60 360 1260

Total Work Unit Processing: 1260 MI (Million Instructions)
Allocate Credits for 1260 MI (Whatever that conversion rate may be)

So thats my idea, it could be used for all BOINC projects. Thoughts?

Cheers

Mark
25) Message boards : Number crunching : Help needed (Message 1470157)
Posted 29 Jan 2014 by Profile ausymark
I would suggest you get an installable version of the OS that you are using, not an image. I have seen Windows XYZ often insist a driver X is for hardware Y just because of the image used and not having the OS installed fresh. My guess is you have the wrong driver for a piece of hardware - especially since you seem to have changed pretty much everything. Spend the $$ and get an installable version of your OS, or use Linux - its free afterall ;)

Cheers

Mark
26) Message boards : Number crunching : 780 not so impressive - yet (Message 1465983)
Posted 18 Jan 2014 by Profile ausymark
There is much to be said about having good case cooling compared to water cooling. I recently replaced my front and rear case fans from the standard 50cfm and 40cfm to high flow ones with rates of 80cfm and 95cfm respectively. This has reduced the internal temps by around 10c and reduced the 7 internal fan speeds by around 14% (they dont have to spin as hard as internal temps aren't as high).

I have an nVidia 580 and a nVidia 570 in the case, plus an overclocked intel i7 2600K with appropriate oversized air cooler on it. (that in itself can pump 90cfm of hot air inside the case)- so between all three working hard there is substantial heat build up.

Overall cost for the new fans was a 'massive' $35 AUD. So you don't necessarily have to go full water cooling to keep your system both performing well and reliably.

Just my 2c worth.
27) Message boards : Number crunching : AstroPulse for Intel GPUs, open beta2 (Message 1452631)
Posted 11 Dec 2013 by Profile ausymark
Hi Claggy

Thanks for that confirmation. Looks like i wont be crunching anything on that GPU then. Luckily I have an nVidia 570 joining my 580 in the next week to help crunching. Im sure that will perform better than the intel gpu in the i7 2600K cpu ;)

Cheers

Mark
28) Message boards : Number crunching : Hyperthreading VS Not (Message 1452611)
Posted 11 Dec 2013 by Profile ausymark
I leave Hyperthreading on, but, I set seti up to run on 50% of the processors - and hence use the actual number of real cores. This also leaves half the hyperthreads available to handle normal computing tastks so the system is fluid and responsive.

If i am running a GPU (Graphical Processing Unit - Video Card/Processor) I free up an actual CPU core to feed each GPU. Any loss of processing power on that CPU core is made up for by the superior procesing power of the GPU. Doing it this way also ensures your system does not bog itself down from user and/or server based tasks.

Just my 2c worth. :)
29) Message boards : Number crunching : AstroPulse for Intel GPUs, open beta2 (Message 1452603)
Posted 11 Dec 2013 by Profile ausymark
Sandy Bridge GPU

I know the Sandy Bridge GPU's handle Open CL 1.01. So I was wondering why they arent being included in the trial (Unless they are and I missed it). Yes they maybe limited compared to the two recent gens of intel GPU's but my guess is that they could still outperform the cpu cores by a factor of 2.

Cheers

Mark
30) Message boards : Number crunching : AstroPulse for Intel GPUs, open beta2 (Message 1452601)
Posted 11 Dec 2013 by Profile ausymark
Just some general thoughts on intel GPU results.

Firstly, on hyperthreaded systems I always run seti on 50% of the processors, this equates to the actualy number of cores. The extra hyperthreads I keep 'spare' for normal task usage so the computer still is a fluid beast.

Secondly I always free up a core or two for feeding any gpu's running. This assumes that the GPU can process faster than any CPU doing the same task. So freeing up a core or two or three (depending on how many GPU's you are feeding) is essentual to feeding the "GPU Beast".

In the light of the above, with the Intel GPU in this specific case, i would keep at least one real core free. So, with 50% processor usage, (100% actualy cores), drop that to 37.5% or to 25% if you want to free up two actual cores. (maybe you want to run 2 instances on the GPU).

Doing so should free the CPU to feed the GPU and run a smaller amount of CPU tasks while keeping the computer 'fluid and free'. By not bogging the CPU down you maximise GPU use and overall computer crunching rate should increase.

Just my 2c worth.

PS: Waiting for the linux versions of the test to arrive ;)
31) Message boards : Number crunching : CUDA on 64bit Linux with new V7 Seti (Message 1376451)
Posted 4 Jun 2013 by Profile ausymark
Hi Team,

Even though I am downloading nVidia CUDA WU's Seti V7 does not seem to be processing them. Is CUDA on 64bit linux available yet on the latest SETI V7? If not whats the ETA? Any work arounds until then?

Cheers

Mark
32) Message boards : Number crunching : Which is the best CPU configuration? (Message 1275077)
Posted 24 Aug 2012 by Profile ausymark
lol It was late when I wrote that, I forgot that the i3's are only dual core. I was mainly used to demonstrate the point that a less powerful CPU may struggle under workloads that a more powerful CPU would breeze through.

The way I have my i7 2600K setup is that I have 4 physical cores running at 100%, this keeps all cores relatively busy. Using Hyperthreading on the the i7, which simulates 8 cores, leaves 4 virtual cores "free". These are used to run apps and feed my GPU - nVidia 580gtx.

This setup leaves the computer fluid and responsive for general use as we as enabling sufficient data flow to/from the GPU.

If I leave the PC running 24x7 it will reach RAC scores of over 40K, so it seems to work well for me.

Cheers

Mark
33) Message boards : Number crunching : Which is the best CPU configuration? (Message 1274853)
Posted 24 Aug 2012 by Profile ausymark
Depends what your 4 cores are apart of, what your other application loads are, and how much memory you have.

4 cores on an Intel i7 are going to be more powerful than 4 cores on an Intel i3

Your application load may be light in relation to an i7, but heavy on an i3

And RAM wise running 2 copies (2 cores only) will use less RAM than 4 cores (4 seti copies)

Swapping to/from hard disk is also increased using 4 cores - however given a large enough write to disk time this is also negligible.

My recommendation would be to run 3 cores at 100% and see what the computers response is running other apps.

(Just make sure your CPU/Case has adequate cooling - CPU temps should stay under 85C to ensure long term cpu life.)

Just my 2c worth :)

Cheers

Mark

34) Message boards : Number crunching : GTX680 + NVidia Beta 304.79 + BOINC 7.0.28 = "Error while computing" (Message 1260277)
Posted 14 Jul 2012 by Profile ausymark
Just on a side note, what have your found to be the optimal number of parallel SETI GPU tasks on the 680? (I am guessing its somewhere between 4 and 6 before adding more tasks doesn't gain any more throughput - i.e. the existing running tasks start to run longer with no actual increased overall throughput.)

Cheers

Mark
35) Message boards : Number crunching : Linux 32-bit CUDA Client? (Message 1250770)
Posted 24 Jun 2012 by Profile ausymark
You may want to put a request into the Lunatics team, however my guess is that 32 bit cuda on linux is such a small user population that it may not be worth the effort (Like we could be talking a handfull of people worldwide.)

However it cant hurt to ask ;)

Cheers

Mark
36) Message boards : Number crunching : Question about feeding the GPU (Message 1250768)
Posted 24 Jun 2012 by Profile ausymark
My guestamate was worst case scenario, it could be as high as 50K :)

Cheers

Mark
37) Message boards : Number crunching : Question about feeding the GPU (Message 1250406)
Posted 23 Jun 2012 by Profile ausymark
Hi Irok

My setup is similar to yours (i7 2600K overclocked to 4.5Ghz on air with an nvidia 580). On 64bit Ubuntu 12.04 linux.

The i7 is an interesting beast as its 8 virtual cores, 4 real ones. This allows i7 uses to do something quite unique as far as CPU/GPU computing goes.

Core goals for CPU crunching is to use up to 100% of all cores to crunch (assuming the CPU has appropriate cooling) and ....

Core goals of GPU crunching is to crunch as many GPU work units as the GPU memory can handle while allowing the PC system user to use the graphics ability of their machine without degradation to the 'user experience'.

On the GPU front for me that meant running just 2 work units on it at a time. This is primarily because I play games on the system, as well as use it for normal office/web tasks. (Running 3 work units causes runtime issues with graphics games as graphics memory became contested resulting in GPU Seti work unit errors).

Now we come to the i7. The CPU must keep the data communication with the GPU occurring as quickly as possible with as much data as required. This is where the i7 works well ..... I have seti configured to use 50% of my CPU - that is 4 cores. But why just 4 when I have 8 virtual? Simply that by increasing it past 4 results in longer work unit processing times with very little RAC advantage.

The operating system will schedule each CPU work unit on each of the 4 physical cores. Some will now argue that these 4 cores may not be operating at 100% all the time. Well, great! Your operating system needs some space to do other things besides seti - so the system still remains fluid and responsive to all tasks - including seti.

Now this is also where the virtual cores come in. The operating system thinks there are 8 cores and will assign 'GPU data feeding' tasks for the GPU to one, or more, of the virtual cores not running the seti cpu task. This ensures that fluid scheduling is given to the 'quieter' seti cpu core and allows for the GPU to get the workload communication it requires.

This ends up being the best of all 3 worlds:

1) The real CPU cores being worked very hard
2) The GPU running efficiently - with maximum communication to the CPU
3) The PC itself remaining fluid for user interaction (a core goal of 'running seti in the background')

If you look at my stats atm they dont look great at around just 6000RAC, but thats with the pc running for only 3 to 4 hours a day atm. - so if it was run 24/7 it would be around the 35K RAC mark.

Anyway thats just my 2c worth. :)

Cheers

Mark
38) Message boards : Number crunching : Linux 32-bit CUDA Client? (Message 1250371)
Posted 23 Jun 2012 by Profile ausymark
Hi Justin

The 32bit CUDA client was pulled as it was buggy and generated work unit errors. As most crunchers were running 64bit distributions I believe it was deemed a waste of resources to fix and so it was pulled.

Do what I did .... upgrade to the 64 bit version of your distribution (I did it to both Ubuntu and Mandriva) - I haven't looked back since and my seti scores have increased ten-fold from running the CUDA client. (Actually its more than ten-fold lol)

Cheers

Mark
39) Message boards : Number crunching : i7 Hyperthreading + GPU + Most Efficient Seti Workload (Message 1216812)
Posted 11 Apr 2012 by Profile ausymark
Hi Chris

Simply set the number of CPU's to use as 50%, that way it will run on 4 of the 8 cores.

As far as the GPU side of things I am using the optimised CUDA program from lunatics ( http://lunatics.kwsn.net/ )

In the configuration (app_info.xml ) file for that change the setting that says:

---------
<type>CUDA</type>
<count>1.0</count>
--------

to

-------
<type>CUDA</type>
<count>0.5</count>

---------

This will cause 2 seti instances to run on your gpu.

Hope this helps :)
40) Message boards : Number crunching : i7 Hyperthreading + GPU + Most Efficient Seti Workload (Message 1202253)
Posted 4 Mar 2012 by Profile ausymark
Hi Team

Well my computer online time is going to be sporadic over the next several months so for the moment I am abandoning the experiment. I will be running 4 SETI processes on the cpu and 2 on the nvidia 580 graphics card. So for now i will just crunch on.

The server and data supply issues have also made getting any form of consistent baseline pretty much impossible - so not sure if those issues will get better over time or not.

Bottom line is however that this rig is way faster than anything I have used before. I have already crunched 15 times more work than I have in the preceding 11 years of doing seti - thats just crazily amazing :-)

Cheers

Mark.


Previous 20 · Next 20


 
©2016 University of California
 
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.