Radeon pre-4999 Unsupported. (Very few WUs left)

Moderators: mhouston, Site Moderators, PandeGroup

Re: Only 1 WU per day for Radeon 4xxx?

Postby exit151 » Sun Apr 29, 2012 4:43 pm

Thank you for the post snippets and links. I really appreciate it, even though it wasn't the news I hoped to read, lol! I especially liked the poster who explained about the languages and why support has ended. I had no idea there were such differences in the language(s) used, nor did I know there was more than one used between the various ATI cards. I always assumed it would have been the same language just enhanced each time to add support for new code used by newer cards. Guess I'm going to have to petition my wife for a new video card, lol :) I've also learned I'm going to need to do more than just fold blindly and start reading some posts/blogs :) Again, thanks all for the info.
exit151
 
Posts: 8
Joined: Sat Feb 11, 2012 5:06 am

28 hours of unsuccessful attempts to get Core 11 assignment

Postby GreyWhiskers » Tue May 01, 2012 8:21 pm

GreyWhiskers wrote:It would be good to know what the PL strategy is - keep limping along on the 4724 and 4756 projects until all the science those projects need is exhausted and then once an for all, pull the plug; or are there going to be more new projects on the horizon?


I posted on 21 April An Analysis of old Core 11 WU server 171.64.65.103 delay. I've continued to get WUs for the old ATI HD4670 AGP card since then, with delays ranging from a few minutes till many hours from request until the assignment server finally assigns to 171.64.65.103 and a new WU is downloaded.

Is there something different today? It's been 28 hours now of repeated unsuccessful assignment attempts (FAH Control GUI is reporting 55 unsuccessful attempts so far). Below is the latest Server Stat for this server - it appears that some assignments are being made.

Does anyone from the PG who is working these Core 11 WUs have any recent news?

Thanks.

Image
User avatar
GreyWhiskers
 
Posts: 795
Joined: Mon Oct 25, 2010 5:57 am
Location: Saratoga, California USA

Re: Only 1 WU per day for Radeon 4xxx?

Postby Nathan_P » Tue May 01, 2012 9:01 pm

With only 56 WU ready to be sent out you are always playing roulette with regards to getting one, 56 between several thosuand potential cards are not good odds.
Image
Nathan_P
 
Posts: 1704
Joined: Wed Apr 01, 2009 9:22 pm
Location: Jersey, Channel islands

Re: Only 1 WU per day for Radeon 4xxx?

Postby exit151 » Wed May 02, 2012 6:30 am

Hey I gotta say after what I read about the 4xxx series not being run anymore, I was elated beyond belief to see that either last night or today sometime my 4xxx gpu was issued work to do :) w00t!! :)
exit151
 
Posts: 8
Joined: Sat Feb 11, 2012 5:06 am

Re: Only 1 WU per day for Radeon 4xxx?

Postby GreyWhiskers » Thu May 03, 2012 9:07 pm

I'm going to bump this topic again.

It's now been 3 days, 5 hours and 13 minutes since I completed my last completed core 11 WU to 171.64.65.103. There have been 140 unsuccessful attempts (per the FAHControl GUI) to get an assignment for the ATI HD4670 ATI bus GPU.

What's the situation? Is Stanford just silently pulling the plug? The server seems up, and showing 100 available WUs. Am I just unlucky not to have gotten in on any of those 140 unsuccessful assignment attempts?

Thanks.
User avatar
GreyWhiskers
 
Posts: 795
Joined: Mon Oct 25, 2010 5:57 am
Location: Saratoga, California USA

Re: Only 1 WU per day for Radeon 4xxx?

Postby bruce » Thu May 03, 2012 9:18 pm

Bumping won't help. (And is prohibited on our forum. :twisted: )

A server has a minimum number of WUs required before it will issue work. It doesn't know how many people are all trying to get the same WUs, and bad things happen if the server gets 200 requests for the same 100 WUs before the AS finds out there are no more. (I'm using arbitrary numbers.)

Science is the determining factor, not the number of people competing for those WUs. Projects can finish at any time and a researcher can start a new project at any time. If I want to study a protein which can make use of FahCore_11 and finish it before that core is deprecated, given how many idle GPUs are in the "below HD 5xxx" class of machine, I'd be happy to start a new one. If I anticipate it would have to be aborted before it's finished, I'd start it on some other GPU platform.

If FahCore_11 has some specific limitation that applies to my research, the answer is also obvious. Brook and CTM are deprecated, and nobody is going to be changing anything in FahCore_11 for ATI. I saw recently that the spec for OpenCL is now at v1.2. I'm not sure who supports it yet, but there will be some nice improvements over the current version when it is supported.
bruce
 
Posts: 21286
Joined: Thu Nov 29, 2007 10:13 pm
Location: So. Cal.

Re: Only 1 WU per day for Radeon 4xxx?

Postby 7im » Thu May 03, 2012 9:53 pm

GreyWhiskers wrote:
Is Stanford just silently pulling the plug?


No! They loudly pulled the plug 18 months ago with News and Forum posts. It's just taken this long for all the old WUs to drain out. They end when they end... ;)
How to provide enough information to get helpful support
Tell me and I forget. Teach me and I remember. Involve me and I learn.
User avatar
7im
 
Posts: 15237
Joined: Thu Nov 29, 2007 4:30 pm
Location: Arizona

Re: Only 1 WU per day for Radeon 4xxx?

Postby exit151 » Fri May 04, 2012 3:26 am

I'll keep mine online and waiting just in case :)

As an aside (and I understand the situation, sort of with the ATI 4xxx series, for those who don't, see the previous page where there are plenty of explanations, but I find it somewhat sad that despite the fact that my 4xxx card could be still doing some sort of useful work, it sits idle while I have a Fedora linux box which runs 24/7 with dual Xeon 2.4ghz processors (the OLD Xeon!) which has no shortage of work, even in it's really, really old age :( I know you guys have your priorities but for the record, it's a real bummer for the many of us with these cards that nothing can be done with them. :( (my wife didn't like my excuse-attempt at buying a new video card either, rofl), so it'll be awhile before I'm able to put up good numbers here lol :(
exit151
 
Posts: 8
Joined: Sat Feb 11, 2012 5:06 am

Re: Only 1 WU per day for Radeon 4xxx?

Postby GreyWhiskers » Sun May 13, 2012 2:23 pm

Just as I was about to give up and delete the folding slot for the old ATI card, along comes another WU.

In this case, it was some 43 hours between completion of the previous WU, and over 70 unsuccessful attempts to get an assignment from the assignment server, and finally getting a new assignment, Project: 4732 (Run 1, Clone 41, Gen 105).

Well, let's see what happens in a few hours when it completes.
User avatar
GreyWhiskers
 
Posts: 795
Joined: Mon Oct 25, 2010 5:57 am
Location: Saratoga, California USA

Re: Only 1 WU per day for Radeon 4xxx?

Postby p24601 » Sun May 13, 2012 5:28 pm

Buying a mid range card is not that expensive, at a cost of a night out you can buy a Sapphire HD 6570 2GB (I bought this) http://tppu.me/u/10p (Amazon Site).
cross/'kros/ n: a thing they nail people to.
I Am The Way To A Forsaken People...
Image
p24601
 
Posts: 11
Joined: Thu Apr 19, 2012 4:17 pm

Re: Only 1 WU per day for Radeon 4xxx?

Postby Pikachu762 » Sun May 13, 2012 5:45 pm

Hello everyone,

I have been folding for a number of years, though there was a long gap before I jumped back in recently. I have the latest Windows client, and I selected the option during installation for SMP & GPU. My OS is Windows XP, SP3, my CPU is a E8600 running at 3.33 GHz, and my GPU is a Radeon 4870.

The client downloaded a couple GPU work units, but then for quite a few days it was unable to obtain more WUs. I understand the cause of the situation from reading this thread. I just have a general question below that I hope someone can answer.

Since there isn't much / any GPU work forthcoming, I edited the 7.1.52 client control options so that both IDs are SMP. I'm not sure that does any good versus leaving it as one GPU and one SMP channel. :) I thought I'd ask about that too, if it is better to leave the client running as one GPU and one SMP. Right now, ID 00 is running SMP project 7003, and ID 01 is running SMP project 7020.

If anyone has some advice or insight, it would be appreciated. Thank you in advance.
Pikachu762
 
Posts: 1
Joined: Sun May 13, 2012 5:18 pm

Re: Only 1 WU per day for Radeon 4xxx?

Postby iceman1992 » Sun May 13, 2012 5:57 pm

Pikachu762 wrote:Since there isn't much / any GPU work forthcoming, I edited the 7.1.52 client control options so that both IDs are SMP. I'm not sure that does any good versus leaving it as one GPU and one SMP channel. :) I thought I'd ask about that too, if it is better to leave the client running as one GPU and one SMP. Right now, ID 00 is running SMP project 7003, and ID 01 is running SMP project 7020.
:shock: I think that is not advisable. It doesn't do any good, in fact it does some harm. It would be better if you just set up one SMP slot. Running 2 SMP slots at the same time will make them fight for CPU time. Completion will be slow. So your options are to keep SMP+GPU, or just one SMP slot.
iceman1992
 
Posts: 419
Joined: Fri Mar 23, 2012 5:16 pm

Re: Only 1 WU per day for Radeon 4xxx?

Postby De Berghut » Mon May 14, 2012 11:11 am

Pikachu762 wrote:Since there isn't much / any GPU work forthcoming, I edited the 7.1.52 client control options so that both IDs are SMP. I'm not sure that does any good versus leaving it as one GPU and one SMP channel. :) I thought I'd ask about that too, if it is better to leave the client running as one GPU and one SMP. Right now, ID 00 is running SMP project 7003, and ID 01 is running SMP project 7020.
If anyone has some advice or insight, it would be appreciated. Thank you in advance.


I would also strongly recommend not to use 2 SMP clients/slots parallel, it will slow down the results significantly (while SMP is brought to life to get results as fast as possible => that's what the bonus system is for). Secondly, if the two clients have more threads running than the amount of cores available (which I assume is the case), there will be even more "slowing down" (my English sucks at this point of day it seems...), as the scheduler will have a big deal trying to schedule the calculations as good as possible over the limited resources, taking a huge bite out of the actual load of the CPU.
This is the reason why I try to keep one core "free", so if I do some work on the computer, it doesn't mess up the SMP-folding. 8 (virtual) cores are available, one smp client is running with 6 threads, in the meanwhile a uniclient is running to use the seventh core, leaving one core free for me to use (internet, skype, whatever...) Maybe it's not the best optimum on PPD, but practical I think this is the best set-up for me, as the computer doesn't run as a dedicated computer :)
Computers never do what you want them to, only what you tell them to do...
User avatar
De Berghut
 
Posts: 117
Joined: Fri Jun 27, 2008 1:19 am
Location: Belgium

Re: Only 1 WU per day for Radeon 4xxx?

Postby bruce » Mon May 14, 2012 2:49 pm

One SMP client with (say, 4 or 6 or 8 threads) will use all the SSE resources your i7 has on a single WU. Running two SMP clients will spread those resources across two different WU and each one will run at about half the speed it might have run. Both results will be delayed and the bonus awarded to each one will be much lower than if all the resources had been devoted to a single WU.
bruce
 
Posts: 21286
Joined: Thu Nov 29, 2007 10:13 pm
Location: So. Cal.

Re: Only 1 WU per day for Radeon 4xxx?

Postby GreyWhiskers » Mon May 14, 2012 5:00 pm

Well, it's like the zombie that will never die.

Finished previous WU, then 14 hours knocking on the door, and now a Project: 4724 (Run 8, Clone 316, Gen 122) just downloaded.

There do seem to be a few WUs avail. So, science is being done. Which is why we are in this in the first place.
Project 4724
This set of projects are assigned to the ATI GPU2 core. We're studying the the villin headpiece subdomain (pictured), the lambda repressor monomer, and WW domain fip35. Each of these is commonly studied in experiments probing protein folding; simulating these molecules using GPU2 not only helps us to test the ATI core but allows us to better understand published experimental results.
Image



I received the PCI non-express GT430 card - I'll see about installing it this afternoon - maybe in tandem with the ATI HD4670 AGP. It has a massive heat sink covering the entire the bottom side of the board, and no fans.
User avatar
GreyWhiskers
 
Posts: 795
Joined: Mon Oct 25, 2010 5:57 am
Location: Saratoga, California USA

PreviousNext

Return to ATI specific issues

Who is online

Users browsing this forum: No registered users and 1 guest

cron