Advanced search

Message boards : Number crunching : Problem with deadline

Author Message
Vladimir Fuchedjiev
Send message
Joined: 17 Jan 10
Posts: 2
Credit: 0
RAC: 0
Level

Scientific publications
wat
Message 14920 - Posted: 2 Feb 2010 | 11:41:57 UTC

Hi all :)

I'm a new in GPUGrid. I finish one WU, but servers say "Completed, too late to validate".
I start at 18 Jan 2010 23:39:23 UTC and finish at1 Feb 2010 18:46:14 UTC.
My notebook is:
GenuineIntel
Intel(R) Core(TM)2 Duo CPU T5670 @ 1.80GHz [x86 Family 6 Model 15 Stepping 13]
(2 processors)

and GPU is:
NVIDIA GeForce 8400M GS (255MB) driver: 19038

Need of full capacity of my laptop sometimes and stop to work for a while.

My GPU is not the best one, I know that , but I want to share my resources.
Is a possible to get smaller WU or to enlarge deadline?

Thanks in advance! :)

Profile Michael Goetz
Avatar
Send message
Joined: 2 Mar 09
Posts: 124
Credit: 60,073,744
RAC: 531
Level
Thr
Scientific publications
watwatwatwatwatwatwatwat
Message 14923 - Posted: 2 Feb 2010 | 13:38:17 UTC - in response to Message 14920.

My GPU is not the best one, I know that , but I want to share my resources.
Is a possible to get smaller WU or to enlarge deadline?


Short answer: It's not likely to happen.

Long answer:

This is a question that has been asked before. Unfortunately, GPUGRID relies on getting one set of results back in order to generate the next set of results. Because of this, getting the results back quickly is very important, especially as compared to other projects. That's the reason the deadlines are so short.

Making the deadlines longer to allow for slower GPUs to contribute could slow the project down rather than speeding it up. The effect of your contribution, although well-meaning, could actually hurt the project overall rather than aid it.

As for making the WUs shorter, that's something I don't recall seeing discussed. Perhaps someone from the project itself can comment on that.

To put things in perspective, the GPU in your laptop is some 30 times slower than the current top of the line GPUs being used here. It just takes too long to process the WUs.

VERY few mobile GPUs are fast enough for this project. See this thread for a comparison of various GPUs. Amongst all the mobile GPUs, currently only the 8800, 9800, GTS 150, GTS 160, and GTX2xx mobile processors are fast enough. Those GPUs have a minimum of 64 shaders while the 8400M GS has only 16 shaders. Some desktop cards have as many as 240, and once the new FERMI cards are released they'll have up to 512.


____________
Want to find one of the largest known primes? Try PrimeGrid. Or help cure disease at WCG.

Vladimir Fuchedjiev
Send message
Joined: 17 Jan 10
Posts: 2
Credit: 0
RAC: 0
Level

Scientific publications
wat
Message 14929 - Posted: 2 Feb 2010 | 16:54:50 UTC - in response to Message 14923.

Hi Michael :)

Thanks for you answer.
I see that I have only one chance to join really to the project - to make WUs shorter.

I will wait for more info about that.

Tanks again.

Profile skgiven
Volunteer moderator
Volunteer tester
Avatar
Send message
Joined: 23 Apr 09
Posts: 3968
Credit: 1,995,359,260
RAC: 0
Level
His
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 15152 - Posted: 11 Feb 2010 | 19:36:12 UTC - in response to Message 14929.

Try Einstein. I have the same card in my laptop. Although it is capable of completing tasks, it can no longer do them in time. It also caused a few screen glitches, and probably ran very hot, so I took it off the project. Einstein uses both CPU and GPU, but does not stress the GPU too much, so I would say that it is ok to run that project.

Snow Crash
Send message
Joined: 4 Apr 09
Posts: 450
Credit: 539,316,349
RAC: 0
Level
Lys
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 15158 - Posted: 11 Feb 2010 | 22:41:36 UTC - in response to Message 15152.

Have you tried this card with with the new application version? It shortens the time considerably.

Einstein only uses about 20% of your GPU and needs a full CPU thread to process what they consider to be a GPU WU.
____________
Thanks - Steve

Profile Michael Goetz
Avatar
Send message
Joined: 2 Mar 09
Posts: 124
Credit: 60,073,744
RAC: 531
Level
Thr
Scientific publications
watwatwatwatwatwatwatwat
Message 15170 - Posted: 12 Feb 2010 | 12:02:00 UTC - in response to Message 15152.
Last modified: 12 Feb 2010 | 12:05:29 UTC

Try Einstein.


Definitely do NOT try Einstein.

Their GPU application is not worth running; it is a waste of your GPU resources.

It runs primarily on your CPU; your GPU does hardly any work. Depending on the relative speeds of your CPU and GPU, you'll either (at best) see a very small increase in speed running their GPU app and at worst actually see the GPU app run SLOWER than the CPU app! Not all problems can take advantage of the type of computer inside a GPU (a parallel vector processor), and Einstein may be one of those that can not.

Einstein's GPU app ALSO uses a FULL CPU core in addition to the (mostly idle) GPU. Essentially, you're running a normal CPU app that also happens to tie up your GPU, thus preventing your GPU from doing something useful.

I'd recommend SETI@Home. You can also try Milkyway and/or Collatz, although I'm not 100% sure what the GPU requirements are for those applications.

Profile skgiven
Volunteer moderator
Volunteer tester
Avatar
Send message
Joined: 23 Apr 09
Posts: 3968
Credit: 1,995,359,260
RAC: 0
Level
His
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 15181 - Posted: 12 Feb 2010 | 15:36:57 UTC - in response to Message 15170.
Last modified: 12 Feb 2010 | 16:36:05 UTC

The general consensus among the professionals and the administrators of this site is that you dont use a laptop GPU, as it often causes serious hardware problems.

Just so you know, my Ion will now take between 4 and 5 days to complete a task running 24/7 and utilising an overclocked CPU. The GFlops rating is 40.

The GFlops rating of an 8400M is less, about 28 GLops. So the card will not finish in time!

I suggested Einstein for good reasons - I did not want him to fry his GPU.

Profile Michael Goetz
Avatar
Send message
Joined: 2 Mar 09
Posts: 124
Credit: 60,073,744
RAC: 531
Level
Thr
Scientific publications
watwatwatwatwatwatwatwat
Message 15187 - Posted: 12 Feb 2010 | 18:28:59 UTC - in response to Message 15181.

The general consensus among the professionals and the administrators of this site is that you dont use a laptop GPU, as it often causes serious hardware problems.

Just so you know, my Ion will now take between 4 and 5 days to complete a task running 24/7 and utilising an overclocked CPU. The GFlops rating is 40.

The GFlops rating of an 8400M is less, about 28 GLops. So the card will not finish in time!

I suggested Einstein for good reasons - I did not want him to fry his GPU.


Yes I agree 100%. Actually more than 100% since I actually would go further than that. I also consider it unwise to run CPU BOINC tasks on a laptop for the exact same reason. I don't run BOINC at all on my laptops. Most people running BOINC want to get the most out of it, so they'll set the laptop to max power and run the CPU at full speed 24/7. That's not much better for the machine than running the GPU 24/7.

Consider this: Do you clean out your desktops to get the dust out of them? The dust really kills the cooling ability of the heatsinks. You can't even get to the heatsinks in a laptop to clean them out, unless you really want to disassemble it.

Laptops are far more fragile in a lot of ways, including their ability to discharge heat. They are far more likely to fail than desktops, and heat is much more of a problem. Of note is recent problems Nvidia has had with laptops, with a bunch of recalls, BIOS changes to make the fan run continuously, and so forth.

With that in mind, my advice would be that if you want to keep your laptop running, don't run BOINC at all on it. If you don't care if it fails (e.g., it's old an you don't mind replacing it or it's already obsolete so you might as well run it into the ground) then my recommendation to get the most out of the GPU would be SETI. But my real recommendation is to not run BOINC.

Profile skgiven
Volunteer moderator
Volunteer tester
Avatar
Send message
Joined: 23 Apr 09
Posts: 3968
Credit: 1,995,359,260
RAC: 0
Level
His
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 15192 - Posted: 12 Feb 2010 | 19:48:18 UTC - in response to Message 15187.
Last modified: 12 Feb 2010 | 19:48:56 UTC

Well said.
There are however a few things you can do on a laptop to reduce the heat:
You can change the power options so that it runs underclocked (Vista and Win7).
You can use throttling software.
You can set the system to use the CPU 50% of the time from Boinc - Advanced preferences, use at most 50% CPU time.
You could even run Einstein on one core (of a dual core system) and the GPU (as it only uses about 20%).

You can also get a USB fan tray that the laptop can sit on – it blows the air up from underneath it.

As for dust, I use a Vacuum cleaner to suck any dust from the vents of all my systems, laptops or desktops.

Profile Michael Goetz
Avatar
Send message
Joined: 2 Mar 09
Posts: 124
Credit: 60,073,744
RAC: 531
Level
Thr
Scientific publications
watwatwatwatwatwatwatwat
Message 15195 - Posted: 12 Feb 2010 | 20:34:16 UTC - in response to Message 15192.

As for dust, I use a Vacuum cleaner to suck any dust from the vents of all my systems, laptops or desktops.



You know, I never thought of using a vacuum on a laptop. That's not a bad idea. Thanks!

svetimas
Send message
Joined: 17 Feb 10
Posts: 7
Credit: 163,113
RAC: 0
Level

Scientific publications
watwat
Message 15306 - Posted: 18 Feb 2010 | 17:26:55 UTC
Last modified: 18 Feb 2010 | 17:42:00 UTC

My machine needs 10 hours to finish a task from GPUgrid. I agree with Vladimir, such tasks are too big for 5 days' deadline (and such deadline is too short at all, too short for supporting every project). My machine is "at work", it is NOT always running 24/7 and may be turned off for weekend (and weekends sometimes are longer than 2 days). Unfortunately, my attempt to support this project seems to be a headache...
GPUgrid is not for all, not for simple mortals like me. It is for computer-bourgeoisie :-) only...
____________

Snow Crash
Send message
Joined: 4 Apr 09
Posts: 450
Credit: 539,316,349
RAC: 0
Level
Lys
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 15309 - Posted: 18 Feb 2010 | 17:51:05 UTC

Each project has requirements and it is a delicate balance between the science of the project, the servers, the scientists, and the crunchers. The GPUGrid team is doing excellent work with the time, money, and resources available to them but ultimately they have to make decisions that are best for the project which are not necessarily what we as crunchers want. It is a privilege to participate, it is not a right and we all have to the freedom to choose. If your card can not crunch for two hours a day, you are correct, this may not be the best project for you to participate in. No hard feelings? it is just the technical requirements of the project.
____________
Thanks - Steve

svetimas
Send message
Joined: 17 Feb 10
Posts: 7
Credit: 163,113
RAC: 0
Level

Scientific publications
watwat
Message 15311 - Posted: 18 Feb 2010 | 18:15:22 UTC - in response to Message 15309.

Each project has requirements and it is a delicate balance between the science of the project, the servers, the scientists, and the crunchers.

YES!!! A DELICATE... :-) For example, few days ago the most successful GPU-project MilkyWay had absolutely unacceptable deadline, but overnight... They found a reserve or possibility to raise it almost 3 times. :-)

____________

Snow Crash
Send message
Joined: 4 Apr 09
Posts: 450
Credit: 539,316,349
RAC: 0
Level
Lys
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 15314 - Posted: 18 Feb 2010 | 18:57:48 UTC
Last modified: 18 Feb 2010 | 19:00:21 UTC

What GPUGrid does (over simplified) is to look at atomic interactions in very very very samll fractions of seconds (femto to nano). Because the equation is very difficult to process it simply takes a lot of power and a lot of time. The GPUGrid project builds WU based on the results of previously executed WUs. If they increase the deadline too much it would be an very long time before they had any usable results at all. Once you start to stretch out the time between usable results you have to add more task streams which itself takes a lot more effort in every aspect including the number of WUs necessary to keep all streams populated. The increase of WU streams in turn will slow down the whole process. I am sure if it was as easy as adding another hard drive they would have done that already.

Maybe you could complete one or two WUs during the week and then set the PC to No New Work so you don't have any WUs left over for the weekends. If you start again on Monday you should be able to get a couple more finished by the end of the week. If this is too much effort I wish you the best in whatever project fits your needs.

Milkyway is solving a completly different equation where none of the results are based on previous WUs so comparing the two projects is not fair to either. Do you think MW really figured out how to increase their deadline and executed that plan overnight? I am sure they worked on it for a lot longer than that, it is just that you don't see any change in a project until it is complete.
____________
Thanks - Steve

svetimas
Send message
Joined: 17 Feb 10
Posts: 7
Credit: 163,113
RAC: 0
Level

Scientific publications
watwat
Message 15315 - Posted: 18 Feb 2010 | 19:15:55 UTC - in response to Message 15314.
Last modified: 18 Feb 2010 | 19:16:18 UTC

Thank You, Steve! I understand, what is "immediate science goals", so my 1st reply here was a small frustration of novice, nothing more. Your advice is very good too. I think, I will use GPUgrid (this project is really worth to participate) for filling the gaps between SETI@home tasks.
Good luck!

Profile skgiven
Volunteer moderator
Volunteer tester
Avatar
Send message
Joined: 23 Apr 09
Posts: 3968
Credit: 1,995,359,260
RAC: 0
Level
His
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 15337 - Posted: 19 Feb 2010 | 19:41:51 UTC - in response to Message 15315.

It might help if you configure Boinc to run according to a timetable, and set the additional work buffer to Zero. That way Boinc will try to accommodate you by downloading and running tasks that will finish in time (according to your timetable).
Snow Crash's suggestion that you select no new tasks towards the end of the week is a good idea (perhaps on Thursday afternoons).
Alternatively you could check the box for Confirm before connecting to Internet - just in case it starts downloading new tasks when you dont want it to. Dont do this if you are running CPU tasks as well; it would get annoying!

To Configure a timetable,
Open Boinc Manager in Advanced View,
Click Advanced, Preferences,
and under the network usage tab set the hours to mirror your work hours and click the days of the week you work.

svetimas
Send message
Joined: 17 Feb 10
Posts: 7
Credit: 163,113
RAC: 0
Level

Scientific publications
watwat
Message 15654 - Posted: 9 Mar 2010 | 6:43:32 UTC

skgiven
Thank You!
____________

DataC
Avatar
Send message
Joined: 15 Feb 10
Posts: 9
Credit: 16,891,220
RAC: 0
Level
Pro
Scientific publications
watwatwatwatwat
Message 16146 - Posted: 3 Apr 2010 | 14:13:57 UTC - in response to Message 14923.

Seriously though? my project said the same thing when I turned it in after 5 days. We aren't talking weeks, but days. How is that too long to wait? And what happens to the WUs we finish "too late" to get credit for? Our machines run for 5 days doing research for you and get zippo in return because 5 days is too long?

No offense, but this project seems to do everything possible to discourage contribution. From problems like this to links on the MAIN PAGE to inacurate posts which recomend specific video cards that are supposed to work but don't...
How many more people could be among the productive if they did not get frustrated and move on to a new project?

It would make more sense to have smaller work units so everyone could contribute. And they do exist--I have gotten jobs which say they will take 6 hours or 15 hours. And then I get one that says something ridiculous like 88 hours. Wouldn't it make more sense to send that kind of WU to someone with a $500 video card that can get it done in 5 hours? Or do the servers not know the specs/speed of the GPUs they are using?

Just some thoughts...

Profile liveonc
Avatar
Send message
Joined: 1 Jan 10
Posts: 292
Credit: 41,567,650
RAC: 0
Level
Val
Scientific publications
watwatwatwatwatwat
Message 16147 - Posted: 3 Apr 2010 | 15:47:31 UTC - in response to Message 16146.

I must agree. It's getting harder & harder to participate in GPUGRID.net is this a rich people only club? I had an 8800GT that I could use last week, now I get the message GPU doesn't have enough memory. I know that 256MB isn't much, but I have a weird mainboard. It has a Geforce 9400 IGP. If I set it at 512MB I can only use the IGP. If I set it to 256MB it'll register the 9400IGP & the 8800GT, but just recently, GPUGRID.net reports that I don't have enough memory on my GPU, last week it was enough. It was really depressing, because I spent several days finding & installing the required drivers, finding the BOINC version that could detect both GPU's, find the correct setting in BOIS & Nvidia Control Panel, & now it's not good enough for GPUGRID.net because everybody has to have a GF100 or G200 to participate.
____________

Snow Crash
Send message
Joined: 4 Apr 09
Posts: 450
Credit: 539,316,349
RAC: 0
Level
Lys
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 16156 - Posted: 3 Apr 2010 | 23:08:11 UTC

The methodology this projects' scientists invented and the software they wrote to execute is very computationally intense. As mentioned in prior posts in this thread, this project builds WUs from the results of a previous WU so quick turnaround is paramount.

If the requirements for this project do not meet the hardware that you have, please, contribute what you can here and use your other hardware to further scienctific discoveries on other projects.

Best regards,
____________
Thanks - Steve

svetimas
Send message
Joined: 17 Feb 10
Posts: 7
Credit: 163,113
RAC: 0
Level

Scientific publications
watwat
Message 19046 - Posted: 26 Oct 2010 | 6:06:10 UTC
Last modified: 26 Oct 2010 | 6:06:31 UTC

My next attempt to participate.
Q. What I found as NEW in the project?
A. New and heavy ACEMD2 tasks are at least 2 times bigger as previous tasks by the same deadline for only 5 days.
So, my GTS 250 is now too weak. Hopeless...
Bye... :-)
____________

svetimas
Send message
Joined: 17 Feb 10
Posts: 7
Credit: 163,113
RAC: 0
Level

Scientific publications
watwat
Message 19048 - Posted: 26 Oct 2010 | 9:48:42 UTC

Yes, ultra-heavy tasks and computation errors after hours of processing on my GPU.
See You later (maybe)... :-)
Good luck!

Profile skgiven
Volunteer moderator
Volunteer tester
Avatar
Send message
Joined: 23 Apr 09
Posts: 3968
Credit: 1,995,359,260
RAC: 0
Level
His
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 19052 - Posted: 26 Oct 2010 | 15:24:05 UTC - in response to Message 19048.

You ran one task.

The GTS250 is not a great card for crunching here.

svetimas
Send message
Joined: 17 Feb 10
Posts: 7
Credit: 163,113
RAC: 0
Level

Scientific publications
watwat
Message 19057 - Posted: 26 Oct 2010 | 20:42:14 UTC - in response to Message 19052.

You ran one task.

Yes, of course - I ran one task and got more than 104K points for my one result. ;-)

The GTS250 is not a great card for crunching here.

I said exactly the same already in my previous reply. Few weeks ago my GPU was "great" enough - it was possible to process 2 tasks pro 1 day. But today this "welcome to project, GTS250" stage is gone.
I agree with liveonc - "this is a rich people only club".
Maybe, server status explains a lot? 1K tasks ready to send, 4K - in progress...
Not very much... :-)

Profile skgiven
Volunteer moderator
Volunteer tester
Avatar
Send message
Joined: 23 Apr 09
Posts: 3968
Credit: 1,995,359,260
RAC: 0
Level
His
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 19059 - Posted: 27 Oct 2010 | 7:59:15 UTC - in response to Message 19057.
Last modified: 27 Oct 2010 | 8:19:08 UTC

Few weeks ago my GPU was "great" enough - it was possible to process 2 tasks pro 1 day. But today this "welcome to project, GTS250" stage is gone.
I agree with liveonc - "this is a rich people only club".
Maybe, server status explains a lot? 1K tasks ready to send, 4K - in progress...
Not very much... :-)


Things change... New projects are introduced, new applications built, new drivers used, new cards get supported, and new crunchers come along with their new cards to have a go.

One of my GT240's presently takes about 18h to run a TONI_KKi4 task, 19h for an IBUCH task and 15h for a KASHIF_HIVPR.
The GT240 is a more reliable card for crunching than a GTS250 and it is faster, Much Cheaper and uses less electric, so I dont go along with rich club crunchers. Further, a £50 GT240 can do more work here than an i7-980X in many CPU only projects. Yes the top end GTX200 range of card and the Fermi's are more expensive, but they also do more work. You could buy 4 GTX470's for the price of an i7-980X and get a RAC of 250K with that lot. If you want to crunch get a GPU you can afford, and dont spend a fortune on the CPU.

So we recommend a GT240 as an entry level crunching card. We do not recommend buying Compute Capable 1.1 cards to crunch here, but if you have one and it runs reasonably well by all means use it. Getting it to work well will be the difficult part, especially saying as it is 2 generations old now. If you want to spend more then there are other cards to look at.

The server status?
On this project each running tasks is generated from a previous task. When it is returned it will be used to generate the next task, and so on. Having tens of thousands of tasks would slow the research down. So the present ratio is excellent.

Richard Haselgrove
Send message
Joined: 11 Jul 09
Posts: 1589
Credit: 6,560,769,351
RAC: 5,635,230
Level
Tyr
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 19060 - Posted: 27 Oct 2010 | 8:44:53 UTC

Not only do computation times change, so does the data volume.

I'm just finishing off idc4-IBUCH_GCY_101021-4-200-RND2150, and I see that the main uploaded result file (_4) is 35,919 KB.

That doesn't matter to me - I'm on uncapped broadband, and it can take as long as it likes - but some people on expensive data tariffs might like to be warned in advance of changes like this.

Profile skgiven
Volunteer moderator
Volunteer tester
Avatar
Send message
Joined: 23 Apr 09
Posts: 3968
Credit: 1,995,359,260
RAC: 0
Level
His
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 19061 - Posted: 27 Oct 2010 | 9:32:17 UTC - in response to Message 19060.

I would agree that a bit more info might have been posted when the team announced these new tasks, but I know the team are presently very busy. Perhaps a list of running apps with upload & download requirements would be useful on the site. A list of some tasks is on the site but it is out of date and does not contain any info on task size. I'm sure regularly updating the web pages is low priority at the minute, so if people PM me their observed task upload & download sizes I can maintain a list in the Forum.

While these IBUCH tasks do not utilize the GPU very well and do use a lot of bandwidth, even more than most other tasks, the system requirements are similar to other GPUGrid tasks, and within the size of the longer WU's. GPUGrid task upload sizes generally vary between about 20MB and 60MB. So there is some change but all GPUGrid tasks are heavy when it comes to bandwidth. Fortunately no bandwidth requirement restrictions are applied here, so many users can crunch. For those only uploading one or two tasks a day there is not too much to worry about, at least compared to some CPU projects; you could easily be uploading 40MB every 6h per core/thread attached to some projects. That's about 1.3GB daily on one i7-920.

Profile Beyond
Avatar
Send message
Joined: 23 Nov 08
Posts: 1112
Credit: 6,162,416,256
RAC: 0
Level
Tyr
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 19069 - Posted: 27 Oct 2010 | 14:17:46 UTC - in response to Message 19061.

Since the average WU time has increased greatly (close to double) I think the time allowed for the bonuses should be increased too. With a 4 hour queue a GT 240 can barely make the bonus deadline and the older cards are out of luck. Who thinks a bonus deadline increase is in order?

Profile GDF
Volunteer moderator
Project administrator
Project developer
Project tester
Volunteer developer
Volunteer tester
Project scientist
Send message
Joined: 14 Mar 07
Posts: 1957
Credit: 629,356
RAC: 0
Level
Gly
Scientific publications
watwatwatwatwat
Message 19070 - Posted: 27 Oct 2010 | 15:14:30 UTC - in response to Message 19069.

We can do smaller workunits, but then the deadline must be smaller too to keep the same pace, so it would not be of much help.

There is a lot of space to improve things though. Shorter workunits have several advantages and we are trying to figure out how to use them in a useful way.

gdf



Profile Beyond
Avatar
Send message
Joined: 23 Nov 08
Posts: 1112
Credit: 6,162,416,256
RAC: 0
Level
Tyr
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 19071 - Posted: 27 Oct 2010 | 15:35:41 UTC - in response to Message 19070.

We can do smaller workunits, but then the deadline must be smaller too to keep the same pace, so it would not be of much help.

It's not like we're going to slow down so the pace should be the same. It's just that the longer WUs are pushing a lot of cards past the 1 day bonus limit. The average WU has gotten a LOT longer while the bonus time has not. How's that fair? The real effect is that as cards stop getting the bonuses people pull them off, so a net project slowdown. For instance, I used to run a number of 9600GSO and 8800GT (96 - 112 shader) cards here but pulled them a while ago. Now the GT 240 cards are starting to approach that point.


Post to thread

Message boards : Number crunching : Problem with deadline

//