486DX-66 MHz; how many WU's? Update: WU took just 3 days!

[Elledan]

Okay, how many years would it take for this little beast to finish one WU?

Also, would this be worth it?

[edit]: update added in new post.

 

[Migroo]

How long - no idea... 3 days?

Worth it - it depends upon how valuable a WU is to you and wether there is something better for that PC to be doing in accordance to your priorities... There are lots of uses for old machines like that - Seti is but one.

 

[Elledan]

WU's aren't that valuable to me (I've already two other PC's crunching data (600+400 MHz)), but this machine will become a node in my soon-to-be-build Beowulf cluster, so I intend to test it thoroughly before using it.

If it's worth it to let it crunch for Seti, I'll probably leave it running for a couple of days.

3 days sounds reasonable

 

[Tarca]

More than 3 days

I have a P/75 and it takes 4.5 days.

I say run it for a week and see how long it takes. Your not the first person to ask this question. Let us know how long it will take in a few days.

-Tarca

 

[Elledan]

I'll do so

 

[Hellburner]

From HERE you're looking at something in the 300 - 400 hour range for the 486DX-66.

 

[Migroo]

Elledan, just being a complete newbie... what is Beowulf? This cluster sounds intriguing

 

[Elledan]

<< Elledan, just being a complete newbie... what is Beowulf? This cluster sounds intriguing >>


Check my thead in General Hardware regarding linux-clusters

Basically, a cluster is a collection (2+) of PC's (can be x86, Alpha, Macintosh etc.) which are all linked together, usually via Ethernet (because it's fast and cheap). These nodes (every system in a cluster is called a node) all run a copy of the same OS (usually Linux, because it's fast and cheap). Together these nodes function much like a parallel system, with multiple CPU's, and an (possibly) insane amount of HD space and memory (every node has its own RAM and HD).

It's actually much like a supercomputer (except for a few critical differences), but for a much lower price.

 

[Elledan]

<< From HERE you're looking at something in the 300 - 400 hour range for the 486DX-66. >>


Let's see... 395 / 24 = 16.458333 days. Say 17 days.

That's... slow... :Q

 

[Tarca]

You might rethink running that. I forgot about Arstechnica's spread sheet.

 

[Elledan]

<< You might rethink running that. I forgot about Arstechnica's spread sheet. >>


Hehe... I'll just use Seti as a way to test the stability of the system when stressed 'a bit'. Perhaps I'll let it run for 17 days, but I most certainly won't use it as a 'crack-unit' for Seti

 

[Migroo]

Thanks Elledan - wow, sounds really cool I'll look into it.

 

[Elledan]

<< Thanks Elledan - wow, sounds really cool I'll look into it. >>


You're most welcome

BTW, the thread I referred to has been moved from GH to HT.

 

[TwilightZone77]

I have a 486SX-25mhz. WUs are resubmitted by Berekely every 4 weeks, correct? Would this machine be able to crunch a WU before it gets resubmitted? Probably not, I'm thinking. Just curious.

 

[Elledan]

<< I have a 486SX-25mhz. WUs are resubmitted by Berekely every 4 weeks, correct? Would this machine be able to crunch a WU before it gets resubmitted? Probably not, I'm thinking. Just curious. >>


A 486SX (and any 386 and lower, for that matter) has no FPU. Seti needs the FPU to work quickly, otherwise the CPU has to emulate an FPU, which is very, very slow.

I doubt that you'll finish one WU within four weeks.

 

[TwilightZone77]

<<

<< I have a 486SX-25mhz. WUs are resubmitted by Berekely every 4 weeks, correct? Would this machine be able to crunch a WU before it gets resubmitted? Probably not, I'm thinking. Just curious. >>


A 486SX (and any 386 and lower, for that matter) has no FPU. Seti needs the FPU to work quickly, otherwise the CPU has to emulate an FPU, which is very, very slow.

I doubt that you'll finish one WU within four weeks. >>



Oh, that's right, I forgot it doesn't have an FPU. Thanks.

 

[ScottMoDSLR]

<< I have a 486SX-25mhz. WUs are resubmitted by Berekely every 4 weeks, correct? >>



Actually, Berkeley stopped doing that several months ago. From the Seti FAQ:

If you don't get a result back from one PC how long does the server wait until sending the same data out again?

For quite a while now we have been sending the data out multiple times over the course of a day or so without waiting for a result at all. That allows us to delete the file from our local disks sooner than we otherwise would (as soon as we've received at least two results) and keep our splitters generating data files at a high rate.


http://setiathome.ssl.berkeley.edu/faq.html#q1.17

Berkeley sends the WU out to numerous crunchers in a few days & waits for the results to come in. The splitters are the current bottleneck in the Seti WU chain, so they have been trying to get them to work as fast as possible

 

[Sukhoi]

IMHO it's a waste of power to run SETI on any 486.

 

[Elledan]

<< IMHO it's a waste of power to run SETI on any 486. >>


Agreed. Unless you find 1 WU every 17 days worth it

 

[Poof]

Unless you find 1 WU every 17 days worth it

:Q :Q :Q

Hey! My sparc IPX takes 16 days per WU with the 3.03 sparc linux client! That little guy is a valid contributor towards the search for ETI and has been crunching for almost 1 1/2 years!!!!!!!!!!!!

And as an FYI, it was taking well over 400hrs/WU on my 486 DX4/100 running Mandrake 6.5 with the 3.0 linux client. It would probably be double for 3.03, although perhaps a bit faster with the win CLI.

 

[Elledan]

OMFG!!! I just can't believe this

Today I turned on the monitor of this little beast and saw to my surprise that it had finished its first WU. Just 3 days....

Not bad for a PC which is almost 10 times slower than my fastest (600 MHz) PC (which takes less than half a day per WU).

They don't make PC's like they used to....

[edit]: it appears that Migroo's 'guess' was right

 

[cirrus1]

wow :Q

My guess would have been a few weeks...

How much ram does that thing have in it? The 3.03 CLI eats about 12-13 mb of RAM AFAIK, and on some lower end machines that is enough for making them slow...

 

[Elledan]

This beast has a friggin' 16 MB of RAM in it. It runs SuSe Linux 6.3 (minimal install) w/ the 3.03 version of the Seti client.

 

[Poof]

Hmmm.... wonder if that was a high angle range WU. Those go much faster. And since you're running SuSE and the linux client, the VLARs would run faster on that as well.

One thing that I think impacts my 486 is the fact that it originally was a DX33 and I replaced the CPU with one of those Cyrix :Q DX4-100 jobbies. I think right there, I answered my question as to why it performs the way it does!

[EDIT: Oh and BTW, my little sparc uploaded a WU last night. Took it 16d 9:57 hours. LOL]

 

[Elledan]

<< Hmmm.... wonder if that was a high angle range WU. Those go much faster. And since you're running SuSE and the linux client, the VLARs would run faster on that as well. >>

What exactly is a 'high angle' WU?



<< One thing that I think impacts my 486 is the fact that it originally was a DX33 and I replaced the CPU with one of those Cyrix :Q DX4-100 jobbies. I think right there, I answered my question as to why it performs the way it does! >>



Actually, my 486 has an original Advanced Micro Devices CPU. Maybe that's why it's so fast :Q

*pokes the Intel-guys on the forum*





<< [EDIT: Oh and BTW, my little sparc uploaded a WU last night. Took it 16d 9:57 hours. LOL] >>

ROFL... I would have retired the little critter a long time ago