Welcome to MilkyWay@home

Server Updates and Status

Message boards : Number crunching : Server Updates and Status
Message board moderation

To post messages, you must log in.

Previous · 1 · 2 · 3 · 4 · 5 · 6 . . . 7 · Next

AuthorMessage
Brian Silvers

Send message
Joined: 21 Aug 08
Posts: 625
Credit: 558,425
RAC: 0
Message 17004 - Posted: 27 Mar 2009, 4:02:19 UTC - in response to Message 16996.  
Last modified: 27 Mar 2009, 4:04:27 UTC

So whats fair about putting all the people that have spent good money on GPU cards to get an advantage(Which is availble to everyone)into a league of their own? What about one for people that don't run the optimised client? Not everyone knows how to install it.


I think before working oneself into a tizzy, one might consider seeking clarification about what Travis meant by this from the home page:

"Also, before a credit firestorm erupts again, we have no plans to award less credit/work to the GPU applications, so you don't need to worry about that."

Upon reading Travis' post I figure he will use the calculation of cerdit that the current wu's are based on only multiplying it to work out for the amount of work do with the gpu wu's. So equal credit per amount of work done.


That's how I read it too, so I was throwing that out there to try to quell the uprising. If we're right about that, if the app does 100x more work, then tasks will get 2800 credit (approx) rather than 28, or 1800 instead of 18. Both are just examples and should not be taken to be any kind of official word on it until things are made more clear, meaning someone asks Travis if that's the plan.
ID: 17004 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Lord Tedric
Avatar

Send message
Joined: 9 Nov 07
Posts: 151
Credit: 8,391,608
RAC: 0
Message 17012 - Posted: 27 Mar 2009, 8:25:01 UTC - in response to Message 17004.  
Last modified: 27 Mar 2009, 8:35:09 UTC

That's how I read it too, so I was throwing that out there to try to quell the uprising.


OK, now that the alcohol has worn off and I've had some sleep.......

This is what I really (really) like about the message-boards, something can be read/interpreted in two or more ways, yet as soon as it doesn't fit into someone else's idea, it's immediately despised and rejected!

An OPINION cannot be wrong - it's an opinion.

Let's wait and see!

Bye Bye from this thread, and I'll keep on crunchin'
ID: 17012 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile borandi
Avatar

Send message
Joined: 21 Feb 09
Posts: 180
Credit: 27,806,824
RAC: 0
Message 17033 - Posted: 27 Mar 2009, 16:57:37 UTC
Last modified: 27 Mar 2009, 17:08:09 UTC

I have a few machines running - Q6600, E6400, a P4 2.8Ghz, and today I got a 4850 and a PSU to stick into an X2-4400. I'd prefer the credits I garnish from the GPU client to go towards my total credit score for MilkyWay, not a separate GPU table. On a good day my current output is 12k ish without the GPU. Technology moves on, and as such we haven't seen splits in 'Top Participant' tables yet. Just because the GPU is such a bonus to this project. It's not a new concept, it's just Moore's Law getting a bigger bump than normal.

Also, take some of the other projects in the BOINCsphere. There are those that run 6-8 different apps, depending on which research group has stuff that needs crunching. We don't see different tables for those.

These WUs, whether CPU or GPU, are all under the same umbrella - MilkyWay@Home. We are all here, crunching for Milkyway@Home. As a result, we all have a MilkyWay@Home score - be it P2 300Mhz, E6400, 3850s, etc.

And I'm all for more work and bigger WUs, Travis. I'm happy to donate my CPU (and now GPU) time to this project.
ID: 17033 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile speedimic
Avatar

Send message
Joined: 22 Feb 08
Posts: 260
Credit: 57,387,048
RAC: 0
Message 17037 - Posted: 27 Mar 2009, 18:23:57 UTC - in response to Message 16862.  

I'm not sure if we can do this having GPU milkyway as just a separate application or not.


Your plan sounds good and maybe the answer. One thing though, will you have any controls in place to avoid those using GPU's to only do those particular workunits? With the way things are now its a free-for-all trying to get work.

I guess what I am trying to say is that will your backend be able to tell that the client is running the GPU application and only send work appropriate as well as knowing that the CPU applications are requesting their work specific loads?
...


That's good question, as BOINC doesn't recognise ATI cards as coprocessors.
For the backend a ATI-gpu-cruched Wu must be looking like a cpu-crunched WU...

mic.


ID: 17037 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Temujin

Send message
Joined: 12 Oct 07
Posts: 77
Credit: 404,471,187
RAC: 0
Message 17038 - Posted: 27 Mar 2009, 19:03:06 UTC - in response to Message 17037.  

That's good question, as BOINC doesn't recognise ATI cards as coprocessors.
For the backend a ATI-gpu-cruched Wu must be looking like a cpu-crunched WU...

That won't matter if they split to 2 different projects.
All they need to do is reduce the daily WU/CPU limit down to 500 at both projects.

For GPU milkyway there'll be 500 WUs that are 1000 times bigger than current that will keep GPUs very happy but will probably make running them on a CPU miss the deadline. Good for GPUs, bad for CPUs

And for CPU milkyway you'll have 500 WUs which will keep CPU machines busy all day but anyone running a GPU will run out PDQ. Good for CPUs, bad for GPUs.

That seems to work both ways.


ID: 17038 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Zanth
Avatar

Send message
Joined: 18 Feb 09
Posts: 158
Credit: 110,699,054
RAC: 0
Message 17083 - Posted: 28 Mar 2009, 14:22:01 UTC - in response to Message 17038.  

That's good question, as BOINC doesn't recognise ATI cards as coprocessors.
For the backend a ATI-gpu-cruched Wu must be looking like a cpu-crunched WU...

That won't matter if they split to 2 different projects.
All they need to do is reduce the daily WU/CPU limit down to 500 at both projects.

For GPU milkyway there'll be 500 WUs that are 1000 times bigger than current that will keep GPUs very happy but will probably make running them on a CPU miss the deadline. Good for GPUs, bad for CPUs

And for CPU milkyway you'll have 500 WUs which will keep CPU machines busy all day but anyone running a GPU will run out PDQ. Good for CPUs, bad for GPUs.

That seems to work both ways.




My i7 can crunch more than 500 MW WUs in 24 hours...

ID: 17083 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Bruce
Avatar

Send message
Joined: 28 Apr 08
Posts: 1415
Credit: 2,716,428
RAC: 0
Message 17086 - Posted: 28 Mar 2009, 15:26:56 UTC - in response to Message 17083.  

That machine really goes eh! I had a peek at the i7 it does a WU in about 13 mins COOL thats about 900 a day! thats a GOOD machine. Wish I had the $$$ to get my own.
ID: 17086 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Temujin

Send message
Joined: 12 Oct 07
Posts: 77
Credit: 404,471,187
RAC: 0
Message 17087 - Posted: 28 Mar 2009, 15:38:05 UTC - in response to Message 17083.  

My i7 can crunch more than 500 MW WUs in 24 hours...

Per core?
ID: 17087 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Bruce
Avatar

Send message
Joined: 28 Apr 08
Posts: 1415
Credit: 2,716,428
RAC: 0
Message 17088 - Posted: 28 Mar 2009, 16:20:35 UTC - in response to Message 17087.  

My i7 can crunch more than 500 MW WUs in 24 hours...

Per core?

no thats total from 8 cores aprox 110 wu's a day per core.
ID: 17088 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
John Clark

Send message
Joined: 4 Oct 08
Posts: 1734
Credit: 64,228,409
RAC: 0
Message 17089 - Posted: 28 Mar 2009, 17:00:09 UTC - in response to Message 17088.  
Last modified: 28 Mar 2009, 17:01:37 UTC

My i7 can crunch more than 500 MW WUs in 24 hours...

Per core?

no thats total from 8 cores aprox 110 wu's a day per core.


My Penny (using CPUs only) takes 10.6 minutes for a 29.5 CS for a ps_s22_ WUs. That is 135 WUs per core per day, or 540 for my quad. I just need a few more cores and a couple of ATI Radeon HD4890 2 crossfire graphics cards, and I am sure the settled RAC after 6 weeks would be about 215K
Go away, I was asleep


ID: 17089 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Lloyd M.

Send message
Joined: 1 Dec 08
Posts: 139
Credit: 8,721,208
RAC: 0
Message 17092 - Posted: 28 Mar 2009, 17:43:07 UTC - in response to Message 16962.  

Renata and Neal Chantrill wrote:
I'll pop over to seti and tell them that they need to split their tables as they do different calculations too.


Excellent point. SETI has two types of Astropulse, plus "Enhanced", yet all the credits are earned on one project.

ID: 17092 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Lloyd M.

Send message
Joined: 1 Dec 08
Posts: 139
Credit: 8,721,208
RAC: 0
Message 17095 - Posted: 28 Mar 2009, 18:15:55 UTC - in response to Message 16967.  

Lord Tedric wrote:
I'm not really bothered either way! The majority of users want fairness and I think some sort of level playing field to see how they compare!

If your not 'doing the same work' how can you compare?


Well, I think it's basically not possible across projects, and quite possible within the same project. Even if the work is somewhat different, the people developing the app can equalize the credits based on difficulty

As for a "level playing field", how can there be such a thing? For instance, on this project, my Opteron 170 outperforms my P4s, even the Xeons (which have a higher clock speed, to boot).

When C2Ds came out, they absolutely annilated my Opty in terms of performance, especially on the KWSN SETI optimized apps, which take advantage of their large L2 cache.

There might be some projects where AMDs can still keep up with C2Qs (though I don't know of any), and I doubt AMD has anything that can run with an i7.

My approach (which suits other things I do with my computers), is to have more machines rather than faster ones. Where's the fairness in that? Though it took a while to accumulate enough of them to be able to outrun a single C2D, I did finally do so. Note that while my rank in this project, measured by TC or RAC is what I consider respectable, it is quite a bit lower when compared to people with the same number of machines.

ID: 17095 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile caferace
Avatar

Send message
Joined: 4 Aug 08
Posts: 46
Credit: 8,255,900
RAC: 0
Message 17096 - Posted: 29 Mar 2009, 19:40:09 UTC
Last modified: 29 Mar 2009, 19:40:19 UTC

Testing, testing. 1...2....3.. Is this thing on? :)

-jim
ID: 17096 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Bruce
Avatar

Send message
Joined: 28 Apr 08
Posts: 1415
Credit: 2,716,428
RAC: 0
Message 17098 - Posted: 29 Mar 2009, 19:54:55 UTC

Server broken! Red stuff all over the place! :-(
ID: 17098 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile borandi
Avatar

Send message
Joined: 21 Feb 09
Posts: 180
Credit: 27,806,824
RAC: 0
Message 17099 - Posted: 29 Mar 2009, 19:56:22 UTC

At least the web pages are up :) Progress is progress.

From the server page, looks like 5 of the services aren't running.
ID: 17099 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
WimTea

Send message
Joined: 16 Nov 07
Posts: 23
Credit: 4,774,710
RAC: 0
Message 17100 - Posted: 29 Mar 2009, 19:56:22 UTC

Looks like MW is almost back again... Just the server that "can't attach shared memory"... 5 more minutes maybe?!?!
ID: 17100 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
WimTea

Send message
Joined: 16 Nov 07
Posts: 23
Credit: 4,774,710
RAC: 0
Message 17101 - Posted: 29 Mar 2009, 19:57:17 UTC - in response to Message 17099.  

At least the web pages are up :) Progress is progress.

From the server page, looks like 5 of the services aren't running.

How 'bout that, 2 posts in the same second. Nice...
ID: 17101 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
John Clark

Send message
Joined: 4 Oct 08
Posts: 1734
Credit: 64,228,409
RAC: 0
Message 17126 - Posted: 29 Mar 2009, 22:43:00 UTC

I wonder what caused the server outrage. It must have been bad for Travis and Dave to need to loose another week end to get things online.

Looks like it's all get up and go. We just need the splitters to give us some work to do. But, the demand will suck the moisture from the WUs because of the length of time the servers went away for.

A lot of us used FreeHAL as a fill in, as Einstein is down as well, and, in my case, Malaria is still formulating their next stage.

As Paul says - live long and crunch!
Go away, I was asleep


ID: 17126 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile banditwolf
Avatar

Send message
Joined: 12 Nov 07
Posts: 2425
Credit: 524,164
RAC: 0
Message 17132 - Posted: 29 Mar 2009, 23:03:23 UTC - in response to Message 17126.  

Well at least this time my other project, Rosetta stayed up with no problems! I kept on crunching. :D
Doesn't expecting the unexpected make the unexpected the expected?
If it makes sense, DON'T do it.
ID: 17132 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
BarryAZ

Send message
Joined: 1 Sep 08
Posts: 520
Credit: 302,524,931
RAC: 15
Message 17154 - Posted: 30 Mar 2009, 7:42:20 UTC - in response to Message 16868.  
Last modified: 30 Mar 2009, 7:43:10 UTC

If RPI is concerned about bandwidth, then additional funding might not solve their concern. The approach of creating much larger GPU work units should help a bit with the bandwidth issue though.

The thing is, the combination of optimized CPU applications (and your implementation of much of the optimization in the regular application), plus the hyper efficient optimized GPU applications PLUS the doubling of your user population in the past three months suggests to me that the real issue is a case of being *too* successful. That is sure to have major bandwidth and general server load issues attached to it.




PS.. While you were away there were some suggestions that some of the participants would like to donate necessary equpiment, etc to help the project out. Maybe you can look into this and see whether or not it is feasible.


The project PIs are looking into this, funding things are really not in my job description :)

ID: 17154 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Previous · 1 · 2 · 3 · 4 · 5 · 6 . . . 7 · Next

Message boards : Number crunching : Server Updates and Status

©2024 Astroinformatics Group