SETI v3.03 update announcement

[Poof]

Posted this afternoon at alt.sci.seti (from Matt Lebofsky):



<< We will start making the version mandatory starting as early as this week. We need to make some tweaks to our server to make the version warning messages work properly. We had a bug where people downloaded 3.03, finished their work unit and sent it back, and then their version was &quot;obsoleted&quot; by accident. How annoying. This'll be fixed and tested today or tomorrow. Then the mandatory upgrade messages will start, and a couple weeks down the line we're shutting down the old clients by shutting them down at the router (via a DNS switcheroo).

FYI, over 25% of our current results are coming in from 3.03 clients.

- Matt - SETI@home >>

 

[CurtOien]

Poof,
Thanks for the info.

 

[Fardringle]

Bummer (for the stats)! At least they did give us a bit of warning before our fleets suddenly 'die' from boredom...

 

[JWMiddleton]

Hi TeAm! Has anyone determined if there is a hit with low angle-range WU's with the version 3.03?

 

[Poof]

I may be wrong but I have a feeling that the linux and Mac clients are going to be moved over first - mainly because those were moved first with the 2.x -> 3.00 upgrade.

My problem with this new type of conversion is the fact that I have an alpha and a sparc, both running linux clients, and neither of these have 3.x ports yet...

Well... I've sworn to myself that the minute they reject the results from these two machines (inadvertantly or not), because they are still running 2.x clients due to not having a 3.x port, then they'll both be crackin' for TA!

[EDIT: JWMiddleton - just saw your post - yes there IS still a VLAR hit on 3.03... it manifests a little differently, ie. rather than the WU being run at 50% CPU the entire time, it cycles from 80% down to 50% and then back up to 80% again, never going higher in seti CPU processing... ]

 

[Ben98SentraSE]

It seems so silly that they have to slow down processing because of bandwidth issues. SO MANY BIG COMPANIES are using their computers for Seti@Home and providing a lot of output for Seti@Home, like Intel, SGI, IBM, etc. Why don't one of these companies step up and help them out? Or some company that is specifically a bandwidth company? Add a T3 or 7 for them. That would be SUCH good press for the company.

 

[JWMiddleton]

Poof: That means that RC5 will benifit at those times if one runs both on the same box!

 

[ElFenix]

they're not slowing down processing. they're adding more processing. its not like back with the previous clients where there were no chip-specific optimizations. they're still there. just a lot more crunching to do.


maybe i'll benchmark a 486SX/8 to see if it takes more than 3 years.

 

[Poof]

Poof: That means that RC5 will benifit at those times if one runs both on the same box!

Yep. That's how I've started to run mine on one machine so far to see how it does...

 

[Robor]

&quot;FYI, over 25% of our current results are coming in from 3.03 clients.&quot;

Hmmm... That's not very high. The other way to look at that is nearly 75% of our results are still coming in from older clients. I'm still running the 3.0 CL client on my fleet. I do not look forward to the switch. I wish I could just overwrite the .exe but with the client running the file is in use. I've still got to go to each machine, stop the client, overwrite the .exe, and restart the client. Not a big deal for a few machines but doing it 35 times... :| Oh well, I guess that's the price we pay for maintaining a large fleet. I just hope they don't update the 3.03 client soon after the switch (unless it fixes the low-angle bug).

Thanks for the info!

Rob

 

[BadThad]

Arrrgggg....the end is near! At least there is some warning. I'm running 3.0 until the end! ...yes, I'm a wu hog, LOL. Trying hard to get into the 99.9% group before the switch, almost there at 99.897%.

 

[Assimilator1]

Thanks for the info Poof

Well... I've sworn to myself that the minute they reject the results from these two machines (inadvertantly or not), because they are still running 2.x clients due to not having a 3.x port, then they'll both be crackin' for TA!

Cool ,who you aiming for?hehe

 

[Sukhoi]

BadThad, I'm running 2.04 and 3.0 until the end too. Depending on how slow the v3.03 client is on my P133, I may switch it over to RC5. Right now my P133 is about 3.67 times slower than my C433 (P133 is 2.04, C433 is 3.0). I'm intersted to see what the difference is when both are running 3.03.

I just checked, and my P133 is 6.44 times slower than my C433 at RC5. If my SETI difference is less than that, I'll keep SETI on the P133. If not, I'll run RC5.

 

[Sukhoi]

Oh, I'm also kinda confused what they mean about mandatory cut-off date, vs. DNS change.

Will SETIQ still function if it's getting the &quot;obselete version&quot; error messages? I'm not worried about the DNS and SETIQ, since that's fairly easily fixed.

 

[Poof]

Cool ,who you aiming for?hehe

Gosh Assimilator1, I'm so far back in the pack, I'd just be aiming to get my key rate up! I wish that RC5 cracking rate CPU chart had an estimated crack rate listed for an alpha 21164 433Mhz... Of course, I could always put a client on it right now and see for myself... but I'm trying to pile up as many WUs as I can before they either obsolete my alpha linux client or finally upgrade it and switch it over to 3.03.

And Sukhoi - I'm not sure how Setiqueue would even deal with a pile of returns that might be of mixed clientage (which is sortof similar to what you're talking about), eg., some 3.03s in with others... It could be that the old client results just don't get credited. It would be a mess for any switchover to halt a Setiqueue upload mid-way to complain about a new client being available... I know alot of people on alt.sci.seti are posting about this and hoping that the switch doesn't occur before Setiqueue's changes are made...

And the DNS change and client cut-over are sortof different things too, ie., the DNS change can be gotten around by using a hosts file with the new servername mapped to the old IP. The client cutover can't (unless maybe the version.sah file can be changed... LOL Oooo I'm terrible). I doubt it though 'cause I think this time, the version will be built into the binary....

 

[Sukhoi]

Hmm, maybe I'll just install SETIDriver during the transition period.