My tiny contribution
30 Oct 2005 23:13:53 UTC
Topic 190092
(moderation:
I have 3 Mac Minis crunching for E@H (and 1 P4 crunching for World Community Grid)...
The minis seem quite suited for E@H, low power consumpution (I think), and the G4s seem to zip through a WU in 23-24K seconds... Not bad for 1.5 Ghz processors...
I
My tiny contribution
)
23-24 Seconds LOL
did you check this if this WUs are valid....
btw if your Host are hidden, we can check this
but 23-24 K Seconds sound to me for an Error,
for Info my Athlon64 3.200+ take about 5 Hours for an validated WU...
Greetings from Germany NRW
Ulli
RE: RE: I have 3 Mac
)
I was trying to say 23-24 thousand seconds... Still seems pretty good for a G4, it was dog slow at folding@home...
Tiny or not ... it all adds
)
Tiny or not ... it all adds up ... and that is the point ...
only for Info 24 thousand
)
only for Info
24 thousand seconds = 400 Minutes /60 = 6.6 Hours not bad but what i say my Athlon 64 3.300 + does this in 5 Hours...
btw it is Science so every Time is going into the Database, fast ore faster ore...
Greetings from Germany NRW
Ulli
RE: only for Info 24
)
Yes, very good, and I am glad a lot of people are interested in this project (I have been an Einstein fan since I was 6 years old)...
I was just impressed with the mini's 32 bit G4 running the optimized code so well...
If E@H comes out with 64 bit code (have they?) I might upgrade to a Power Mac G5 (or Athlon 64 dual core)...
i think People are working on
)
i think People are working on an 64 Bit client, but a warning
do not expect that this client is working faster than the 32 Bit Client...
there are many threads about this and the most people think that there will be no improvement...
for Info
Greetings from Germany NRW
Ulli
RE: I was just impressed
)
When Einstein released the Altivec-optimized Mac app, it cut 40% off the crunching time on my Mini (which is the 1.25GHz "low end" version). The alpha Team MacNN SETI apps do even better compared to the 'stock' SETI app, cutting times in half - or better. That little G4 has an incredible vector unit. Yes, my Athlon 64 3700+ is a little faster, but at twice the clock speed (and 1.5x the cost), it should be, and on Einstein it definitely isn't twice as fast...
How well the Macs do on the various projects is heavily dependent on how well the code is written for the Mac. As much as I love the Rosetta project, their Mac code seems to be 8-10X _slower_ than the Windows version, so I suspect it's a straight "port" of something originally written by an x86 programmer. (Warning: this is based on a small number of WUs, I haven't been doing Rosetta very long.) Einstein's Mac code is I think actually better than the Windows version, and for SETI it's a toss-up, as excellent apps exist for both platforms. SZTAKI and Predictor both also seem equivalent between platforms.
And you're right, the Mini is very low on power consumption, noise... it may not be a quad G5, but it's a very nice little box, a good value for the money. Plus I could put 16 of them in the space taken by my PC... so yes, if Einstein is your project, the Mini is just about ideal!
RE: ...Einstein's Mac code
)
Yes, I would say so. I can't help but wonder why code hasn't been written yet to use the SSE2 instruction set on x86s. Maybe the AltiVec engine is just much more accessible. If AV and SSE2 were equally accessible, then why would they spend the time to optimize an app for the six-percenters, when optimized-for-SSE2 would benefit probably 40-60% of Einstein crunchers? I've read that SSE optimization would be of little benefit, speed-wise, so I can't expect any further "bump" on my 32-bit AthlonXP, or on older, pre-P4 Intel silicon, either, but an SSE2-driven app might tip the scales toward a 64-bit upgrade for me.
Michael
microcraft
"The arc of history is long, but it bends toward justice" - MLK
The primary advantage of
)
The primary advantage of "pure" 64-bit extensions is simply a larger virtual address space, for the most part this only helps things like large database programs.
In the case of some of the "microprocessor" chips we use in the PCs, there are other minor advantages that may accrue with no other special programming ...
For example, I bought two Dell dual Xeon computers with 1G RAM about 3 months apart. Both are 3.4 GHz, both have 2 M L2 cache. Yet the 64-bit version has been seemingly slightly faster with BOINC View recording slightly higher CS/hour numbers ... note, just to make my day, they inverted for the first time I can recall ...
27.44 CS/hour for the 64-bit computer
27.86 CS/hour for the 32-bit computer
So, what do you believe? :)
The historical better speed of the 64-bit machine may have been because of the use of the optimized SETI@Home application (which I am not running on any machine to improve my Einstein@Home and Rosetta@Home standings ...).
I will have to watch this when I turn on the SETI@Home application later ...
But, now, I am kinda interested in the G4s as they are pretty low in cost and might be a good way to improve my throughput ... heck, they go on sale at Frys at times ... maybe I can get one and reduce my heat budget but improve my CS/hour ... wife has been complaining of the fan noise ... :)
RE: ... maybe I can get
)
Paul
Just to keep the wife happy, maybe you should experiment with mobile processors. My Pent M in desktop produces virtually no noise, when installing programs etc. I can even hear the HDD clicking as the head moves in and out.
Andy