My old Pentium 75 (SX969) with 128meg of edo ram managed 1 seti work unit in 1,582,828 seconds(cpu time) and was granted a massive 48.54 credits.

I also gave my AMD K5-P75 with 256meg sdram a go at a seti work unit. After 1,163,352 seconds of cpu time the poor thing was aborted with "Maximum CPU time exceeded".
Both these machines are running fedora core 6 linux. The later versions need the cmov instruction.

The thing is why was the workunit aborted well before its end date. Both machines can finish a workunit before the end date.
Is this a boinc version thing??. The p75 is switched off at the moment. Its quicker to type this message in and wait ..... than boot the machine up and check the boinc version. Its about 15 mins to get it up and running and networked.

Anybody else done a workunit with less than a P75?
My 386 with a cyrix "fastmaths" co-processor runs win3.1 fine but only has 8 meg of ram.

Nairb

---

LOL...

Vintage crunching! Always good to see those old timers soldiering on. :-)

Astro ran a PI/60 for a while a ways back.

The benchmarks don't look whacked, so I'm gonna have to think about the time exceeded abort for a bit.

<edit> IIRC, the max CPU time is calculated from the FPOPs estimate the project sends for the task divided by the FP benchmark. I haven't had to deal with a max time abort for quite a while, so I'll have to root around in my documentation and logging catacombs a bit for more info. ;-)

Alinator

My old AMD K7 @ 604 MHz is running now in this moment.. :-D

AR 0.447801 - 82,286.14 sec. - 39.23 cr.

And MMX optimized! ;-D

---

LOL...

But that one is grandson of Nairb's K5 and son of My K6's. ;-)

Almost a puppy by comparison! :-D

Alinator

...
I also gave my AMD K5-P75 with 256meg sdram a go at a seti work unit. After 1,163,352 seconds of cpu time the poor thing was aborted with "Maximum CPU time exceeded"... running fedora core 6 linux.
...
The thing is why was the workunit aborted well before its end date.
...
Is this a boinc version thing??.

It was running 6.4.5, I don't know any reason that should have mattered, though you were certainly out of the range of systems used for alpha testing. The only possibility I can think of is that the transition to doing calculations in wall time rather than CPU time may have begun before 6.4.5 was produced.

It's now showing "Measured floating point speed 21.39 million ops/sec", which is about half what the project settings for speed were intended to allow. That suggests it hadn't benchmarked before the work was sent, the default is 1000 million ops/sec. so is eligible for work. But the Maximum CPU time calculation based on that would have been about 9 days rather than the 13 it was allowed to run.

Did you perhaps set up the system by copying files from a host with fpops benchmark around 700 million ops/sec?

                                                                  Joe

With both machines up and running at last ... the p75 has an old version of boinc 5.8.16 which I suspect is the reason it finished its workunit. The other K5-P75 has a much later version.

An AMD K7 ..... such power. AND has the cmov instruction so you can use fedora core 8 and later releases so these socket 462 cpu's can be used with seti for a few more years.
I hope we will still be able to use the socket 7 cpu's for a good while yet.

On a practical note: It costs about a 2 quid (uk money) to do each workunit (1 about every 20 days). Its not the best bit of kit for finding E.T. They only get used now and then.

Nairb

---

Hi Josef,
Nothing copied from other machines. It has several goes at the benchmark before it gets it right and the work did get sent before it got the speed right.
This is just a bit of fun to see what can be done with old kit. But to get 2 weeks into the w/u and get aborted is a shame.
I just re-set the k5 up with the old version of boinc..... Then it got 2 w/u.
I will need to change its cpu for a p233 to get them both done on time. Not something the quad core boy have to worry about.

Nairb

---

Anybody else done a workunit with less than a P75?
My 386 with a cyrix "fastmaths" co-processor runs win3.1 fine but only has 8 meg of ram.

Nairb

The slowest computer I had running BOINC was a P75 laptop running Windows 98. Back in Seti Classic I ran a 386 SX-2 16mhz WITHOUT a math co-processor. Setispy estimated 3 years to finish one workunit.

Adding an 80387 co-processor (from Ebay for 10 bucks including shipping, when original price was over $300), cut the time down from 3 years to 1 year. Sadly, it never finished that WU before Classic went offline. :(

Since then I've been more concerned with saving electricity so I only run the most efficient computers (watts per credit)...although I admit it was hilariously fun getting old 386's and 486's running Seti in the old days. :)

---

You will be assimilated...bunghole!