SETI@home News Vol. 112 (08/08/2010)

I've got my 470's clocked @ 700/1400/1700 and the GFLOPS are showing as 1254

oops, my bad.....

Didn't realise it was kolala's original quote.

Just ignore me (everyone else does :D)
 
Looks like you done a fine job on it, daft question but do you plan on using it as in everyday driving, or is it just for the shows and that sort of stuff. Personally i have no problems with my penny farthing now that the roads are tarmac rather than cobbles :D
 
Yay nice one, i dunno i admit i enjoy the parping and such informing people that they're slacking and all but it's really bugging me now as to how we could attempt to "re-activate" the interest in the 1000 odd inactive members.

Just how in the heck can we "sell seti" for want of a better phrase. Must admit the team spirit is top notch the help/advice second to none, but apart from the natural flow of people who "drift in" and then "drift out" ......i don't know.
I'm not naive i know it can be pricey to upgrade and hardware ain't free, but its how to make people believe and know that its not about how much you can output, or how little but just being part of it and doing what you can.
Hopefully as people see how well the team is doing they might come around to see whats cooking so to speak, dont just swing by, stop for a while, give it a try theres free help available just about 24-7.If you decide not to stay or it's not for you fair enough.
I just feel as though i'd like to be doing something more, apart from a few months running around like a headless chicken in an effort to understand it's allmost set and forget now(touch wood :eek:)
Had a re-hash with work and it's going to be better getting a few people to do the work that have more enthusiasm than me and which will free me up to do stuff i am really interested in. Home is sorted, tax man's happy (the ******) and i am ready to go.
Been thinking of lots of different ideas, but looking forward to learning Area51's perl script for starters. It was going to be way to pricey to get those case badges made up and there just didn't seem to be the interest, so how about trying to raise a few squid for seti???
There would be no politics to it and no committment just those that want to get involved can do so.
Am sure all of us have at least one box of bits n bobs that we call on when in need, so why not persuade "those that want to that is" people to pick say 1 or 2 items they have no use for and that might fetch say a minimum of £2-£3 (to justify postage and fees) and sell them by way of mm?
So give me your ideas of what else we might be doing to help a cause that takes up a fair bit of our time and energy and how to actively encourage people back to the team??

Regardless of what you think of me, give the idea some thought and come up with something better, would be interesting to see just what we can do as a team as well as the contribution to science.

It's also a fair and valid point that some people will think they do enough allready, what with running costs time and energy etc and it's a great start but i can see this team doing a whole lot more.

There's allways a chance the next discovery could be made by someones pc cycles from this team....and if you believe there's no chance of that....well you wouldn't be here to begin with would you? :confused:
 
Nice Video Joeyjojo, guys a legend.

I don't know how to get interest in Seti or any other project it's something that people do out of wonder I think, But you will get a few who stay because they are interested in the subject, It's just finding those few who don't mind and want to try and find the answer to the old question "are we alone" :)
 
Last edited:
edgedemon said:
If you have a serious amount of wu's it might be a memory issue
Click to expand...


3 days work by 3 x GTX295's down the toilet

Hard drive failed :(:(:(:(:(:(:(:(:mad:

The ironic thing is my other PC's are made from 2nd hand parts, the drive that failed was a brand new WD Caviar Black.

This is the second drive I have had die on me, the last one was new as well.....go figure :rolleyes:

Not a happy chap at all at the moment.
 
ToxicTBag said:
Been thinking of lots of different ideas, but looking forward to learning Area51's perl script for starters.
Click to expand...

Shouldn't be too long now! It's been running fine on my machine for about a week now, although its only setup initially to sort out -12 and -529697949 errors. For me, I get about 10 of these a week (-12 and -529697949) - but it finds, cleans and re-assigns them to the CPU with no problems. I am looking to add -177 and other types of errors bit by bit. You will be handy, because you have a Fermi setup (and when your setup goes wrong, in the usual 'Toxic way', it doesn't do it on a small scale)! ;) . All I need now is a willing volunteer who is running 2 or more GPUs, 1 Fermi and 1 lnon-Fermi jobbie......

I would also like to find a couple of people who are running Windows XP 32 bit......

To complete the first release ready for testing, I need to implement a section of code that extracts the version numbers of the GPU app and CPU app that are being used. I can't assume someone has an app_info.xml file, or that it is in a clean state - so all information has to come from client_state.xml. At the moment, these values are hard-coded, which is no good good for general distribution or for maintenance-free code. However, I did finally figure out how I can get the required info last night, so I just have to implement that in a reasonably efficient way.
 
Grimgoth said:
edit: and while we are here, why not plan to customise the benchmarking, so that we can get a real taste of the pc; have the same tasks multipled for 2, 4, 6, 8, 12 threads (or evenmore) for the cpu, and 1,2,3,4,5,6,7,8,etc numbers for gpus- and then we just time the whole shebang.
Click to expand...

The task of multiplying out a task for multi-core and multi-GPU setups is in itself not difficult. However, there are a number of issues. Firstly, the tasks would have to be renamed. Not a major problem - but it could be that there are issues with 'file already exists' type errors, resulting in a good work unit being over-written by a benchmark task (an old task may have been re-sent to a client that has been bouncing around the Berkeley servers for a while, and just happens to coincide with the name of a benchmark task). The next problem is the timing. Say you have a batch of 20 CPU tasks, and lets say 50 GPU tasks. When do you declare the test to be complete? Not all of the CPU cores will simultaneously process the same task in the same length of time since the CPU itself will allocate other OS related tasks to certain cores - and not to others. Also, it is likely that GPUs connected to an 8x slot will load an unload the task slower than those using the 16x slots. The result is that 1 or more cores (or GPUs) will finish the benchmark task(s) allocated to them, whilst the others are still processing theirs. The test will therefore become skewed, in the sense that you are still timing the task(s) being finished, but the some of the resources of the machine have gone on to do other things (or not as the case may be).

The original idea was to get a relative (or absolute if possible) indication of GPU performance, assuming that a GPU is only being used for S@H. To my way of thinking, this is best done by running a single task at a time, and noting the result. Yes, the results will be skewed a little and will to a certain extent be influenced by issues such as hard disk performance etc, but the overall effects of such things will be minimised.
 
halz said:
3 days work by 3 x GTX295's down the toilet

Hard drive failed :(:(:(:(:(:(:(:(:mad:

The ironic thing is my other PC's are made from 2nd hand parts, the drive that failed was a brand new WD Caviar Black.

This is the second drive I have had die on me, the last one was new as well.....go figure :rolleyes:

Not a happy chap at all at the moment.
Click to expand...

Feel for you. Is it definitely dead along with the data?
 
You must log in or register to reply here.
Back
Top Bottom