Posts by Skip Da Shu
log in
1) Message boards : Number crunching : **UPDATE: ATI 58x issue resolved! (Message 44697)
Posted 5 Dec 2010 by Skip Da Shu
I converted one of my XP64 machines over to Ubuntu v10.04 to try DNETC on the ATI card and it works... but I had to use the Berkeley variant install instead of the Debian/Ubuntu package due to a permissions bug in x64 package with ATI driver... it's an open bug on launchpad I was trying to narrow down but it's beyond my ability to pursue further till I get some help from the package devs.


Oh yea, a good workaround was found to this via xhost command in the /etc/init.d/boinc-client script. Not on that machine at the moment or I'd included it but if somebody still needs it reply here.


2) Message boards : Number crunching : **UPDATE: ATI 58x issue resolved! (Message 41648)
Posted 21 Aug 2010 by Skip Da Shu
Is there a possibility to test a linux 32 or 64 bit gpu client app for this project ?
I have an Ati Radeon HD5870 which crunch on Dnetc.
I would prefer to crunch on Milkyway@Home if a linux gpu app has been provided with.

I converted one of my XP64 machines over to Ubuntu v10.04 to try DNETC on the ATI card and it works... but I had to use the Berkeley variant install instead of the Debian/Ubuntu package due to a permissions bug in x64 package with ATI driver... it's an open bug on launchpad I was trying to narrow down but it's beyond my ability to pursue further till I get some help from the package devs.
3) Message boards : Number crunching : **UPDATE: ATI 58x issue resolved! (Message 40858)
Posted 9 Jul 2010 by Skip Da Shu
Boy do I wish...
4) Message boards : Number crunching : **UPDATE: ATI 58x issue resolved! (Message 40670)
Posted 25 Jun 2010 by Skip Da Shu
Thanx much... this explains what I'd thought to be an oddity between two 4850 equipped machines (one has ram clock that will not go below 993 in ccc, other has been running with VRAM clock at 650, both with GPU clock at ~650).

UPDATE: I guess I expected a couple seconds difference not a couple MINUTES!

On a AMD 9850, windozeXP machine w/ 2GB RAM with 4850 (1GB) I took the timings from it running last night at the lower vram clock (forgot I'd set it down from 650 to 475 yesterday). I then ran about 5 MW WUs and 5 Collatz WUs with the higher vram clock setting.

850 core/475 vram vs 850/950.

MW no measurable difference.

Collatz avg 19min 44secs vs 16min 52secs!!

Yet another update:

On a 5830 that was clocked 855 / 855 changing it to 855/1000 and running some Colatz WUs resulted in a decrease in average runtime of 38.5 seconds / WU. And again no change in MW WUs... well with all that proven well 'nuff for me... it's back to low vrams and MW workunits.
5) Message boards : Number crunching : **UPDATE: ATI 58x issue resolved! (Message 40668)
Posted 25 Jun 2010 by Skip Da Shu
Thanx much... this explains what I'd thought to be an oddity between two 4850 equipped machines (one has ram clock that will not go below 993 in ccc, other has been running with VRAM clock at 650, both with GPU clock at ~650).

UPDATE: I guess I expected a couple seconds difference not a couple MINUTES!

On a AMD 9850, windozeXP machine w/ 2GB RAM with 4850 (1GB) I took the timings from it running last night at the lower vram clock (forgot I'd set it down from 650 to 475 yesterday). I then ran about 5 MW WUs and 5 Collatz WUs with the higher vram clock setting.

850 core/475 vram vs 850/950.

MW no measurable difference.

Collatz avg 19min 44secs vs 16min 52secs!!
6) Message boards : Number crunching : **UPDATE: ATI 58x issue resolved! (Message 40663)
Posted 25 Jun 2010 by Skip Da Shu
Thanx much... this explains what I'd thought to be an oddity between two 4850 equipped machines (one has ram clock that will not go below 993 in ccc, other has been running with VRAM clock at 650, both with GPU clock at ~650).
7) Message boards : Number crunching : **UPDATE: ATI 58x issue resolved! (Message 40648)
Posted 25 Jun 2010 by Skip Da Shu
If it does, reduce the memory down to as low as you can get it - memory speed is irrelevant at MW (I am running a 5970 on 300 memory with no issues, others I have seen running 4xxx below 200 with no issues). A low memory setting will save you a bunch on power, run cooler and/or allow a higher gpu clocks without overvolting.

If all that holds true ..... a check on PSU capacity/output/power draw maybe a further check once you get going.

Regards
Zy


Is memory speed also irrelevant at Collatz? In other words, will an ATI card set up for MW work well for Collatz also or does it require more VRAM clock?
8) Message boards : Number crunching : GPU Requirements [OLD] (Message 37111)
Posted 9 Mar 2010 by Skip Da Shu
Anybody got WU run times for the new HD5830 (s/b on list I believe also)?

Trying to decide if I wanna upgrade a HD4850 with one of these.
9) Message boards : News : Visualization/Screensaver Work (Message 37110)
Posted 9 Mar 2010 by Skip Da Shu
No personal vested interest here since I run Linux and I must assume u r talking about a Windoze screensaver.

What's the purpose of a screen saver... eye candy. Maybe actually causing someone to ask what it is and get some interest in crunching. It seems to me the option that fulfills that best is #3.

So, to fulfill it's function #3
10) Message boards : Number crunching : I'm thinking about adding a 4770 to my gtx 260; is this going to work? (Message 36604)
Posted 17 Feb 2010 by Skip Da Shu
You will need a separate dummy plug or monitor ( I use the spare VGA cable) for the second monitor.


FYI, Only in Win...My Linux 64b works fine w/o anything on 2nd card.
11) Message boards : Number crunching : Cruncher's MW Concerns (Message 33962)
Posted 29 Nov 2009 by Skip Da Shu
MW is still a CPU project.

The project has not come out with a ATI/CAL app that I know of.
They have come out with a CUDA app, and from what I remember, it is near baseline with the other CUDA projects.. I.e. GPUGrid.

Being such, the ATI app is optimized.

Run the baseline benchmarks for comparisons.

Stock CPU app vs Stock CPU app. Any third party optimizations(unless adopted by the project at large) remain outside the baseline credit comparison charts.

Given that, MW is not the highest paying project. That ribbon currently goes to Aqua or Primgrid/64 bit apps


Good point, if you cut out the app that is producing XX% (insert unknown but suspected high number) of the credits on the project you are quite right.


But again, any cross project parity is a myth, a dream for DA, and a nigtmare for project admins. And a socialists focus.

Free societies determine the amount of pay/credit given, not the king.


All hail the king... ;-)
12) Message boards : Number crunching : Cruncher's MW Concerns (Message 33853)
Posted 27 Nov 2009 by Skip Da Shu
The projects I mentioned pay better for ME. Your reply to Sailor's post, validates my point. With GPU/CPUs performances all over the map, everybody's facts will be different.

Personally I don't see how anybody could complain about the crediting on this project. Cut it half... still be more than any other project.

My original post was to address the above statement. Nothing else. It has nothing to do with GPUs,CPUs, percentages, power supplies, etc. I simply pointed out that I am complaining about the credit system on this project. MW has set the requirements for GPU computing outside the parameters of my card. Just like Einstein. Fine. But I dont like it. So Collatz and Seti Beta pay me more.

Your statements assumes everyone else in the world is just like YOU. Ridiculous, yes?

So you are comparing GPU on collatz/gpugrid to cpu on MW?

I think it's you that's gone down 'ridiculous alley' my friend. The point is that, like someone else said oranges to oranges... or even oranges to tangerines, MW is the top payor.
13) Message boards : Number crunching : Cruncher's MW Concerns (Message 33812)
Posted 27 Nov 2009 by Skip Da Shu
If you take the best credits for initial expenditure i.e. purchase price your right.

If you factor the electricity bills into the equation I'm not so sure :)
4770's still may hold. Anything with a 48xx or 5xxx well thats a whole other story.


Yup.

I had bought a 4770 when they first came out but never did get it to crunch MW. I thought I'd read that somebody here did get them working but it was after I'd RMA'd mine. MW & a HD4770, that would be the ticket w/o making as many offerings to the electricity gods.
14) Message boards : Number crunching : Cruncher's MW Concerns (Message 33798)
Posted 27 Nov 2009 by Skip Da Shu
You cant compare different kind of GPUs... its like stating "my Phenom x4 955 in XXX makes more credits then my Athlon 1400 in YYY, so XXX is the best paying project..."


Actually I can. Let's say for a moment that your T-bird produces x credits on project YYY and your X4 produces 2x credits on project XXX. Who pays better credits? How did you reach that conclusion?

Anyway, where I was headed was...

Whilst out credit whoring around looking for the cheapest way to just add credits to the team output, I have come to the conclusion that you can do no better than throwing an ATI HDxxxx card into a Windoze machine and pointing it to MW (let's keep the cards to say, sub $200 or perhaps even sub $100).

Acceptable? Is there another answer that I missed?

If not, then I still believe my original statement. MW pays the highest credits for effort/money expended.

Ever since I convinced the guy at the corner store that BOINC credits are valuable and he lets me traded them for merchandise this became important.
15) Message boards : Number crunching : Cruncher's MW Concerns (Message 33794)
Posted 27 Nov 2009 by Skip Da Shu
I don't really understand how two projects can use 100% of the CPU but have such different effects on the core temps.


I see many differences between projects some do not even heat up my cpu's others i see which are memory intensive tasks heat up my cpu more.
Ofcourse it also depends on the optimisations used so in fact to many factors to mention.
Its like running collatz on my 4770 gpu card which make it 64C but when it runs MW it becomes 79 or even hotter.


Could it be because MW does FFTs and Collatz doesn't? I don't know, just throwing that out as a possibility.
16) Message boards : Number crunching : Cruncher's MW Concerns (Message 33793)
Posted 27 Nov 2009 by Skip Da Shu
Personally I don't see how anybody could complain about the crediting on this project. Cut it half... still be more than any other project.


What a ridiculous statement. There are many projects that pay a lot better than Milkyway. Collatz, GPUGRID, Seti BETA just to name a few. Check out the facts before you make such blanket statements.


I don't know in what framework you mean this but I'll assume we are still talking about GPU apps.

I run 3 of these 4 (no "Seti BETA"). I'll take numbers from FREE DC as of 10pm central today in the order you listed:

THE FACTS:

Collatz, 1 GTX-260 card + ~25% of the HD4850: 63.7K (avg 52K)
GPUGRID, 9 GPUs, from GTX-275 down to 9600GSOs, mostly GTS-250s: 80K
MW, ~75% of 1 ATI HD4850 card: 55.6K

So let's evaluate/rank these.

MW - 75% of 1 ATI card produced 55.6K so far today or 74.1K/gpu
Collatz - A GTX-260 + 25% of the 4850 produced 63.7K (but averages 52K) or 51K/gpu
GPUGRID - 9 cards produced 80K so far today or 8.9K/gpu

Now it may seem "What a ridiculous statement" to you but I'm still waiting for you to name one project that produces more credit per GPU than MW.

Suggest you take your own advice "Check out the facts before you make such blanket statements."
17) Message boards : Number crunching : Cruncher's MW Concerns (Message 33546)
Posted 22 Nov 2009 by Skip Da Shu
cuts my electricity bills in half


Yea, now that would be sweet.
18) Message boards : Number crunching : Cruncher's MW Concerns (Message 33519)
Posted 22 Nov 2009 by Skip Da Shu
Personally I don't see how anybody could complain about the crediting on this project. Cut it half... still be more than any other project.
19) Message boards : Number crunching : Using the MW ATi application, Q & A (Message 32638)
Posted 22 Oct 2009 by Skip Da Shu
Hi, I'm getting about a 28% error rate running the v.20(b) MW ATI app.

Stream Allocation : Failed to create Buffer Kernel Execution : Uninitialized or Allocation failed Input streams. Stream Allocation : Failed to create Buffer


This buffer alloc error is occurring on two different machines.
[..]
Both machines are running 4 WUs at a time.

I am wondering if the 4 WUs in parallel is the root cause but would like the developer's opinion on that.

Don't wonder anymore, it is the reason.

Actually I wonder how the error messages can be made clearer than they are. Do you have a suggestion? I would think that if an allocation fails (what reasons for this come to your mind?) the cause is quite clear. But that's only me I guess ;).


I assume that at the point in time the app has ask for some memory for the buffer and the 'alloc' just returns a "didn't happen" message so that you don't know for sure if it's truly an "out of memory" condition or an "out of handles" or something more obscure like that? If so, I think the message is fine.

Purely guessing, I am thinking it tried to start before a completing WU had finished releasing it's memory. Or do you really think 4 WUs ate up 512MB?

If that were the case I would've expected the error to occur more than it did. I noticed on at least the 8 or 9 of the errors from when it was running 4 at a time, it was always the 4th one starting that got the error. Makes sense. However, once it errors out another starts and runs fine (4 concurrent). In fact I can run 5 concurrent and the error rate doesn't go up much. Does this suggest that if the "getmain" or "alloc" or whatever is done were to be re-done that it might just succeed because some cleanup had finished?

Also, does this imply that running 6 concurrent WUs might be fine on a 1GB card or are there other limitations in play here?

Oh well, guess it doesn't really matter at this point.
20) Message boards : Number crunching : Using the MW ATi application, Q & A (Message 32628)
Posted 22 Oct 2009 by Skip Da Shu
Hi, I'm getting about a 28% error rate running the v.20(b) MW ATI app.

Stream Allocation : Failed to create Buffer Kernel Execution : Uninitialized or Allocation failed Input streams. Stream Allocation : Failed to create Buffer


This buffer alloc error is occurring on two different machines.

Both machines have HD4850 cards. One clocked stock 625/993 the other one at 650/975MHz.

The stock clocked one is running WindozeXP 32b, SP3 has ~29 out of the last 100 WUs get the error below. The slightly OC'd one running WindozeXP 64b, had ~27 out of the last 100WUs get the error below.

Both are running the same app version, both using driver v8.12 and BOINC v6.10.13. Both machines are running 4 WUs at a time.

I am wondering if the 4 WUs in parallel is the root cause but would like the developer's opinion on that.

I'm currently changing the Win 64b machine to run 3 at a time in an attempt to see if that improves the error rate at all.

The error in all cases looks like THIS ONE


So far running 3 concurrent WUs seems to give much relief. After starting BOINC back up it ran 31 in a row w/o error, then 2 errors, 5 good, 1 error, 2 good, 1 error, 22 good, more good... etc. etc.

So last count it looks like about 4 errors in 99 WUs or about a 4% error rate.

After seeing the 1st 30 go thru w/o errors I changed the other machine (32b, stock clocks) to only do 3 at a time and it hasn't had an error'd WU since (it's done almost 90).

I then reset the clocks to 640/980MHz on the 64b OS'd machine, left it running 3 WUs concurrent and haven't had any errors since but it'll take a bit more time to get to thru enough WUs to declare it under clean.

Hope somebody finds this helpful.


PS: Now I need an English PUB to celebrate ;-)


Next 20

Main page · Your account · Message boards


Copyright © 2017 AstroInformatics Group