Welcome to MilkyWay@home

Posts by Zydor

1) Message boards : Number crunching : CPU/GPU Comparison (do we need CPU apps when GPU app is available) (Message 61884)
Posted 12 Jun 2014 by Zydor
Post:
If someone doesn't have 100$ on hand. Why should he spend 100$ more each month on electricity?

That's his choice not yours. Each User has their own reasons for doing what they do in BOINC. Some chase the 'ol Ego thing, some don't give a rats whatsit about Credits and just do it for pure fun when they feel like it. Often the truth lay in the middle. The great strength of BOINC is it (by in large) caters for (nearly) all machine types. Its not for BOINC to "Police" Machines, and long should that remain as a principle.

The software is designed by the Authors for their target batch of Crunchers, but its entirely up to individuals what they do - and long should that remain free of artificial boundaries. At the Open end of the scale Collatz is designed to cater for both the incredibly low powered out of date machine that cant even start - let alone crunch - other projects, as well as the up to date latest Behemoth cards. Some Projects you'll be lucky if the card works only being one generation out of date.

Some CPUs should be banned. Cost of upgrade is less than monthly (or at most two month) cost of use.

Never ...... BOINC draws its strength from being capable of attracting a wide variety of people and motivations. From the Utterly Crazy Barking Mad Credits-addicted-at-all-costs Crunchers beavering away with sweating forehead and clenched teeth at 3am, to the casual Cruncher who does it from time to time just for Fun or a Team Based exercise.

BOINC must stay away from "Elitist" labels. Its not for BOINC or the software designers to dictate to their Crunchers (or the other way round). BOINC is a wide spread vastly varied beast that is capable of supporting many many different levels of PC, and User abilities alike - the day it goes elitist in hardware terms, is the day the Team Based culture dies, then BOINC itself dies.
2) Questions and Answers : Windows : Computation Errors on Cataylist 14.4 (Message 61883)
Posted 12 Jun 2014 by Zydor
Post:
The single largest cause of errors in BOINC is running the cards too fast, and failing to regularly (aka once a week) clean out the cache and undate Windows etc etc. That's particularly so at Milkyway where the apps will rarely run at the full speed of current cards without falling over ( a good thing - means we are not tempted to run the things at full stretch 7 x 24).

Clean out the cache, restart and run it at default speed for a while, if you still get the same issue when running at default speed (and have thoroughly cleaned and updated the card) ..... then its time to scratch the 'ol brain. I acknowledge the reluctance in some to reduce speed drastically for a short while, but its by far and away the fastest way to resolve this.

If it works fine at default speed, slowly increase GPU speed until you get errors (give it an hour or so between upping the speed). When it falls over back off 10 and your good to go. Highly likely you will not be anywhere near the top speed the card is capable of, that's fine, its a (very good) software limitation, not hardware.

Regards
Zy
3) Message boards : Number crunching : R9 290X (Message 60778)
Posted 20 Jan 2014 by Zydor
Post:
Probably not worth it even if they end up doing so, as the 290X is only 1 in 8 Double Precision - you'd be way way better off going for a Single Precision Project.

7XXX were 1 in 4 DP, but the new 290X is only 1 in 8 DP.
4) Message boards : Number crunching : Computation errors (Message 59763)
Posted 30 Aug 2013 by Zydor
Post:
.... run it for 30% for a more accurate overview.


The second "trap" is to get sucked into too strong a competitive edge. The latter has two consequences

- Hammering hardware far too strongly, laptops are particularly vulnerable to that. Burning out a Laptop itsnt worth it .... the world will move on and you get left with the Bill :)

- Take part in your Team Events and contribute, BOINC was not founded on speed records, it was Founded on many PCs contributing a little. Its the sum of the whole that matters.

As you move forward, make sure you use software tools to monitor the Laptop performance, its critical with basic hardware else you can strain it far too much. Your Prime indicator for that is Heat. Do NOT take any chances with heat. Keep it well ventilated, learn how to use the fans correctly. Monitor it carefully. Once you have settled in you'll be fine, its no drama, but it can be if you ignore the basics. Burning a laptop is not worth it.

There are a number of heat monitoring tools around, each with their own "thing", you'll come across them over time, together with the other useful tools. For now take safety first, and a basic heat monitoring tool - having safeguarded the laptop that way, you can move forward and investigate what BOINC has to offer .... and there's a lot.

There are many "Tools" - for now safety first, load GPU Temp and Core Temp, that'll watch your back until you get chance to settle in more:

http://www.gputemp.com/

http://www.techpowerup.com/realtemp/

If you don't know the safe temperatures, Post for help, you'll soon get an answer from someone - and doubtless more and different types of tools that do various things. For now, stay safe and use GPU Temp & RealTemp (or another if you found one) .... but DO use one for CPU and GPU.

Last for now - make sure you have security tools loaded - there are nasty folk around. The basic tools from Windows are free - not perfect to say the least, but they do an effective job for the threat level you face. Use them ....
5) Message boards : News : Badges for crunchers. (Message 59671)
Posted 25 Aug 2013 by Zydor
Post:
You may think it's pointless, but badges that are actually achievable for users who can't afford to go out and buy top-of-the-range dedicated hardware are a huge factor.


Pointless .... not in the slightest..... If you read carefully the two suggested schemes cater for both the "well endowed" cruncher and a casual Cruncher.

As to a million+ .... if a Cruncher with dedicated graphics cards wants to play that game, there are many Projects "paying" far higher at the lower equipped cruncher. 2Million+ is not hard to "get" with a modern graphics card on many projects, even without 24 hr crunching. A three year old 7970 would achieve 2Million plus without changing heartbeat - that's the same as 500 to 1,000 casual crunchers with minimal hardware - that's reality.

Its a fools paradise to try and compare one Project "rates" with another as a Bellweather norm - if that was the norm, the stampede out of Milkyway would have occurred a long time ago.

The Challenge is to craft a scheme that acknowledges everyones efforts. Ignore that and try to play down high end hardware, is as silly as trying to claim low end hardware is the solution to all ills.

There is a clear balance to be struck, and its not an easy one as such. Hit way too low and it becomes meaningless, hit it too high and it becomes "elitist". The realities of Modern Hardware make that a difficult balance to strike - but the balance must be achieved - pitch too high and it flops, pitch too low - as many do - and its a pointless exercise.

That's the real world - like it or not.
6) Message boards : Number crunching : Donating to Milkyway@Home (Message 59669)
Posted 25 Aug 2013 by Zydor
Post:
..... my Wildlife@Home project just got some NSF funding ...


Nice .... Congratulations :)
7) Message boards : News : Badges for crunchers. (Message 59664)
Posted 25 Aug 2013 by Zydor
Post:
Badges ..... here we go :)

Two classic traps:

1. Set Badge levels for todays crunch level - rapidly falls apart as Moores Law kicks in and doubles output each 18 months. eg top three hundred already have more than 100million ........ system falls apart as being too easy, and gets ignored.

2. Too many badge levels resulting from over-enthusiasm to give a tangible sign to lower level Crunchers. That becomes silly, as they get handed out with the fish & chips and the Hot-Dogs .... a bit pointless, particularly as the Crunchers actual Credit level is hardly a "secret".

"Hear no Evil, See no Evil, Speak no Evil" principle is no way to run a Badge system. The system has to be Tangible, mean something, and be something that is actually worth striving for pretty silly if it was easy to get //

Other Projects have recently wrestled with this, and most that didn't fall over at the Start Gate kept the Qualifying Totals High with a large spread between Badges, and kept the number of Badges Down - especially the latter as it quickly reaches 10-15 badges which is silly.

The way around round the Classic Traps is a dual Badge - One for the Total AllTime Credit level, and one for current average credit. The AllTime AllUp total needs to be HIGH, so the scheme lasts a few years not just a few months. As a starter for ten .... the following eight badge level awards illustrate the principle:

AllUpTotal: 1m, 100m, 500m, 1Bn, 5Bn, 10Bn, 50Bn, 100Bn
[As Credit inflation and Crunch Power increases that's easy to amend later ... later adding 500Bn, 1Tn would keep it going for a Decade or more ...)

RAC: Crucial to give lower powered crunchers something to shoot for as well as something for New Starters, this badge reflects CURRENT RAC level, not their all-time highest RAC - it plays to the ego :) ......
5K, 25K, 100K, 500K, 1m, 3m, 5m, 10m

8) Message boards : Number crunching : Naming Convention Problem?? (Message 59652)
Posted 23 Aug 2013 by Zydor
Post:
I just noticed a bunch of WUs going through with a slightly changed file name convention. I have no idea if it makes a difference when piecing this lot together ..... but highlighted below just in case the slightly different name standard makes live "exciting" trying to piece the jigsaw together :)

Recent:
ps_separation_81_DR8_rev_3_1_1372784654_16035748_0

Another more normal name convention:
ps_separation_83_DR_8_rev_3_2_1372784654_16032005_0

Note no underscore between the "R" and "8" in the first WU. There are a lot of:

"ps_separation_80_DR8...."
"ps_separation_81_DR8...."
"ps_separation_82_DR8...."

around, so if it does make a difference ... needs nipping in the bud.
9) Message boards : Number crunching : GPU computation errors on one host but non on the other (Message 59649)
Posted 23 Aug 2013 by Zydor
Post:
The "bad" host shows CL_OUT_OF_RESOURCES and CL_MAP_FAILURE errors.


That's the clue ..... basically BOINC ran out of resources on the hardware to use and gave up. 5XXX cards are way way different beasts in their technical architecture and abilities, compared to 6XXX cards. No matter which Capability Variant is used, there is always a finite capacity unique to that variant. Crudely speaking 6XXX are way faster and better than 5XXX cards - as a generalisation - and will have greater capacity and flexibility. Hence the reason the 5XXX bombed out first.

As a generalisation don't run any more than two of a WU type on a GPU. Sometimes, rarely, three will run successfully. All that happens when more than 2 or three are run concurrently, is BOINC starts to run out of resources to cope, and in any case it time shares between the WUs as the full capacity has been reached - usually after two concurrently running WUs - so there is no or miniscule time saved. Especially when it bombs out crashing with too much being thrown at it. Any miniscule time saved by running more than 2 (sometimews three) is way way overrun by time lost whilst the machine is down for you to get it going again.

A good Rule of thumb for BOINC is a max of two WUs per GPU when the WUs are short run is seconds or a min or so. More than that don't bother - in fact you will lose time eventually running too many at once.

10) Message boards : Number crunching : Computation errors (Message 59647)
Posted 23 Aug 2013 by Zydor
Post:
I did however get one with a predicted 2,212 hour completion estimate and was given 10 days to do it in. As this was an obvious mistake on the part of the research team and their quality control I aborted that one too.


Research Team would know nothing about it ...

This one is an Old BOINC chestnut as old as the hills.

That estimate is a BOINC calculation (not the Research Team - they know nothing about this) trying to estimate how long a WU will take. BOINC often has very little to go on ... ie what kind of CPU and/or GPU used, its speed setting, memory on board, cache used, what the program does, what the program is calculating, how long it actually takes to calculate the various parts of the program, how long it actually takes is completely unknown to BOINC - it has to try and guess the result from what it knows, which initially is very little to zero. Its impossible for BOINC to know how long a Project WU will take to run, right out the starting gate

What BOINC will do is basically take a wild stab at it, and over the subsequent WUs it "learns" from your previous results. Overtime the estimates get better and better, usually from one to a few days worth depending on the WU.

When you switch WU types, the whole process starts again, it learns from experience of WUs crunched. There is literally no other way to do it as the variables are far too many.

The best way to get round this is crunch the WU for around 30-50% of its length, then compute in your mind the time taken to get to that percentage - eg 9secs to get to 33% done means 18secs to go - whatever the "time remaining" counter says. As a few more are completed, estimates for time remaining get better through its experience of past WUs.

Until it learns about the WU on your individual PC, that's the only way its possible to do the calculation. You get round that by taking the time completed shown and the percentage completed shown, extrapolating that to 100%.

Usually after a few dozen / few hours worth of WUs from a Project are crunched, BOINC has learnt enough to give good estimates - until then do the calculation in your mind I outlined above.
11) Message boards : Number crunching : Stderr OutPut Entries - Curious .... :) (Message 59375)
Posted 16 Jul 2013 by Zydor
Post:
Have these entries inside Stderr for the separation runs (nothing appears amiss, they are validating, I am just curious) - spurious or something to look at in due course ??

BOINC: parse gpu_opencl_dev_index 0
<search_application> milkyway_separation 1.02 Windows x86_64 double OpenCL </search_application>
Unrecognized XML in project preferences: max_gfx_cpu_pct
Skipping: 0
Skipping: /max_gfx_cpu_pct
Unrecognized XML in project preferences: apps_selected
Skipping: app_id
Skipping: /apps_selected
Unrecognized XML in project preferences: nbody_graphics_poll_period
Skipping: 30
Skipping: /nbody_graphics_poll_period
Unrecognized XML in project preferences: nbody_graphics_float_speed
Skipping: 5
Skipping: /nbody_graphics_float_speed
Unrecognized XML in project preferences: nbody_graphics_textured_point_size
Skipping: 250
Skipping: /nbody_graphics_textured_point_size
Unrecognized XML in project preferences: nbody_graphics_point_point_size
Skipping: 40
Skipping: /nbody_graphics_point_point_size
BOINC GPU type suggests using OpenCL vendor 'Advanced Micro Devices, Inc.'
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

..... and one other a little further down stderr,

--------------------------------------------------------------------------------
Build log:
--------------------------------------------------------------------------------
"C:\Users\David\AppData\Local\Temp\OCLD5F0.tmp.cl", line 27: warning: OpenCL
extension is now part of core
#pragma OPENCL EXTENSION cl_khr_fp64 : enable
12) Message boards : Number crunching : Milkyway@Home is SAD! (Message 59115)
Posted 26 Jun 2013 by Zydor
Post:
app_info ....

See (scroll to the bottom of the page):

http://boinc.berkeley.edu/wiki/Client_configuration

... for format of use. Run don't walk to use it - its a godsend, once set you just leave it forever, it picks up automatically server-side changes such as new versions blah blah.

For 99% of Users it means no fiddling with massive app_info's any more. Once set, that's it, just crunch on; the server will automatically update and supply the files you need when you start a crunch session.
13) Message boards : News : N-Body 1.08 (Message 58282)
Posted 15 May 2013 by Zydor
Post:
Most files of this nature will show bizarre estimates when run for the first few times - the calculations are so complex, they have little to go on to make an estimate the first few times around.

As a (very) crude generalisation - once they get to around 30% complete the estimates are usually pretty good, after 60% done its pretty well nailed. Cant get specific, it varies hugely depending on whats being crunched, but that's a reasonable guideline.
14) Message boards : Number crunching : Request help updating app_info.xml for Linux (Message 58280)
Posted 15 May 2013 by Zydor
Post:
Could you cut & paste your complete cc_config in here to have a peek at?

- I assume both cards are of the same type and version?

- Does Catalyst Control Centre show hardware enabled for both cards (information-hardware tab)....

- Inside the Performance - Overdrive option at the top of the options showing up mid-bottom screen, do both cards appear inside a drop down box above the small illustration pics?
15) Message boards : Number crunching : Request help updating app_info.xml for Linux (Message 58268)
Posted 15 May 2013 by Zydor
Post:
Straw clutch time .......

Are you creating these ones inside Notepad, and not another editor? Notepad is needed to make sure hidden control characters are not inserted aka word processor software - as these can and are misinterpreted by a BOINC program, and it all falls over.

Use of Notepad is essential - its not optional.
16) Message boards : Number crunching : Request help updating app_info.xml for Linux (Message 58242)
Posted 13 May 2013 by Zydor
Post:
Could try two other things if not all ready done so - strap the two together with a Bridge Strap, and/or enable Crossfire (cant remember the CUDA word for it - escapes me at present :) .... ) in the GPU Control Panel (Catalyst Control Centre for AMDs).

If the previous and these two don't work - buy some incense and beg forgiveness to the gods for upsetting them ..... because there is little else that will stop it given a clean uptodate set of drivers :)
17) Message boards : Number crunching : Statistics Date Wrong (Message 58239)
Posted 13 May 2013 by Zydor
Post:
Weird ..... has to be something on the Op Sys side, as BOINC gets its base timing from the Op Sys Clock set inside the Control Panel, so don't rip apart the BOINC installation again.

Stumped to be honest ..... strange one.
18) Message boards : Number crunching : Request help updating app_info.xml for Linux (Message 58227)
Posted 12 May 2013 by Zydor
Post:
On the problem machines, check to see if the line below is pasted into cc_config in the BOINC directory that's below Program Data parent directory.

<use_all_gpus>1</use_all_gpus>

Probably will not fix both, but hopefully fixes one, move onto the last after this is inserted into cc_config

If cc_config.xml does not exist in the BOINC directory, create it with notepad (make sure its saved as .xml not .txt). As an example mine is below:

<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>
</options>
</cc_config>

Regards
Zy
19) Message boards : Number crunching : Statistics Date Wrong (Message 58225)
Posted 12 May 2013 by Zydor
Post:
What date & time shows up on the windows clock ?
20) Message boards : Number crunching : computation error on HD 6970 - drive 13.4 (Message 58219)
Posted 12 May 2013 by Zydor
Post:
Don't forget Arkayn's comment .... you need to change Operating Systems from XP, without that you don't get off the starting block - you will not get past V13.1 to the current 13.4 due to the need to have OpenCL up and running with current versions. Without being on 13.4 you cant get onto current versions and get back on track with this.

The time to make the dreaded change of Core Operating System from XP has arrived :)

Regards
Zy


Next 20

©2024 Astroinformatics Group