Dropping Seti @ Home & possibly Boinc

richard smith said:
Cant run boinc or anything at all at work these days tho. Group wide auditing software you see :(
Click to expand...
All our work machines have that as well, at boot every installed program is checked and removed if not on the list, so no DC at work :(

@Bulldog66, you say Boinc is killing the machines by taking all CPU/RAM but do you mean by that that you have multiple core machines running Boinc using 256Mb RAM when there isn't 256Mb RAM free?
 
Biggles 266 said:
Were you using optimised SETI science apps though? If so, that's why your RAC is down. Enhanced should give about the same credit as the old SETI, but only with standard clients. An optimised science app meant that you were getting through work quicker than standard and that boosted your RAC. If/when there are optimised Enhanced apps, then your RAC will level out again.
Click to expand...
Yep i was using the optimized 4.18 client and now im using the optimized 5.12 client.

Once every one is crunching enhanced work units, everything will level out because you just plain and simple get less credits for an enhanced work unit.
 
Gotta say tho, some of the enhanced units I have crunched are giving me double the credits of the old WU's, e.g. some are at 60-70 credits per wu. Yes some really noisy fast ones are coming in at sub 20 per wu, but if I look at my completed WU's on b0rkelys site there are defo some bigger ones.

I think it will take a few weeks to settle down. And the memory footprint per client is still huge compared to the old version, which is I feel going to be the main factor. Because as we know, especially on windows boxes its not hard to get them to start swapping to disk.
 
Bulldog66 said:
If I find this is not the case I will then put seti down to 1%, put predictor to 98%.

I need a non invasive boinc project that is neither HD hungry or greedy with resource allocation.

Any ideas..
Click to expand...

Screw Predictor. Their admins never reply, issues go unresolved for ages and there is no indication that we're even doing useful work.

If you want a BOINC project that is easy on the RAM and hard disk space then look no further than Leiden Classical and Einstein@Home. Leiden Classical is however shut to new accounts just now, and whilst I could get you an account, it could only be as a member of the Ars Technica team... which leaves us with Einstein.

It uses about 8 MB of RAM and there are a whole bunch of optimised clients available for it, which are at least 4 times quicker than the normal ones. It doesn't use that much in the way of hard disk space either. It's the best way forward for credits.
 
Now this is just stupid, looked at the work in progress on seti enhanced and got 1 WU been running for 8 hours with 8 1/2 hours still to go, how the hell can Borkley expect to keep people crunching for them when they produce WU like that,

Image2.jpg


That was the good thing about Seti, most WU only used to take about 2 hours not 16!
 
Yes its the dual Xeon, checked the result for 13hours and 62 points, all WU's now aborted and stopping seti :mad: will have a break from DC and think about a new project maybe :confused:

Just about to turn o** for the 1st time 2 computers running 24/7 for the last 6 months
 
Last edited:
I'm seeing about double the WU times on my old P3 (11 hours to 24) and about 50% longer on my (Intel) Mac Mini (~2 hours to ~3.25). I've not tried a P4 yet, but I found that the one thing my old Prestonia Xeon was good at was SETI. It's not keen on Folding (down to two clients throttled to 90% CPU right now), and I guess it'll not like SETI Enhanced much either given the above. :(

An optimised SETI Enhanced client might help I suppose (they do already exist), although I'm not sure what affect that will have on the credit scoring (in theory I'd expect to score the same for faster completion, which is pretty much the exact opposite of what happened previously). I get the impression this new scoring is contrary to the original Cobblestone concept, but whatever.

Sounds like Einstein might be the way to go for me. Predictor and LHC just plain suck, and folding seems to have lost it's interest since the demise of the 600-pointers and QMDs. :(

I really don't want to see people leave DC (hell, we've lost a hell of a lot of good crunchers since BOINC happened), but with what's going on in the DC world these days, I'm not entirely surprised. Never thought I'd be considering quitting SETI for good, that's for sure. Mind you, I suspect I might have said the same six months or even a year ago, and I'm still here, just about.
 
I just did some real quick sums from the Mac Mini.

Old = 9,300 - 9,600 seconds, ~20 points.
New = 11,700 seconds, ~53 points.

Looks like either the Windows Enhanced client doesn't like P4s, or someone did a great job of optimising for the Intel Macs. Either way, I suspect that unless the optimised Windows client does much better than the standard one, running something other than SETI on P4-based systems (and that includes Xeons) would be a very good move.

Sure shows the potential of the Core Duo though - and therefore also the Core 2 Duo (a.k.a. Conroe).
 
I know what you mean h2o, seems the guys at seti just can't stop tinkering with the credits depending on which camp is shouting "its unfair".

Never thought i'd even think about leaving seti but my origional idea was that I would do seti / astropulse - even remember doing some beta astropulse wu's about a year ago as a mix. but as always Borkerley can't seem to get there finger out and get it sorted. As for credits, seems we have gone full circle to all wu's grant the same credit per wu for each of the 4 people who crunched it, which is in my mind not to far off a 1 point per wu ala Seti classic.
Hope you find another DC project thats more to your liking.

Babyface UK
 
Yes, at least it's counting actual work done now, rather than some effort-based estimate tnat never seemed to work quite right. At least people can understand it a little better I suppose.

One thing to bear in mind with Enhanced is that there's a much bigger variation in WU times - way bigger than with the old client. In fact, some WUs complete quicker with the new. I guess there's only a certain amount of science they can do with high AR workunits.

As for Astropulse, well, seeing they only got 1/3rd of their 'funding drive', I guess we'll see, but with BOINC stable and enhanced out, maybe they'll turn their attention back to it.
 
I think that the problem with HT'ed P4's and Xeons is that the larger memory footprint of the enhanced WU's is highlighting the limitations of the architecture. Which is probably why the Intel core duo and the AMD64 machines don't seem to be getting trashed as much by it.
 
Seems like a fair assessment. My Xeon box certainly gets trashed by big F@H WUs, on the occasions it actually gets them. Looks like Einstein might be the best choice for overworked P4s then. At least Intel finally saw the light and killed the damn things off. I guess we have Banias et al. to thank for that. :)
 
While I understand how frustrating it is for the rest of you, it's so cool to see my ikkle Mac Mini beating the living daylights out of P4s (inc. Xeons). I don't have enough numbers to go on yet with Athlon 64s, but I did find one A64 3500+ that had been well stomped. :D :cool:
 
You must log in or register to reply here.
Back
Top Bottom