I agree with Joe about the hardware being better. Take my P4 and intell Mac core 2 duo. There is 4 years difference between the 2. I run opp aps on the P4 and just stock on the Mac. The P4 has a rac of close to 500 and the mac just over 1000. divide the Mac by 2 and im getting a rac of 500 per core running stock. quite an improvement for just 4 years.

---

[/quote]

Old James

I upgraded from a 2.5 gig athalon, to a 6 core AMD 1055, new motherboard, 4 gig of ram, and a 1 gig video card for about 500.....it kicks ass....it chrunchs SETI so fast, its not funny.....and i can play any game i want, watch movies, music, without a problem....HD movies , full HD.....

consider an upgrade

:)

...
SO: what's up with the lightyear difference in crunching time? The cache? Certainly NOT the 3/4 GB extra RAM. It, then, must be that the Astropulse v505 kernal is multi-threaded; and that pisses me off.

Although the 512 KiB L2 cache on your P III helps, each core of the wingmate's system has access to 2 MiB of L2 cache. That makes a huge difference, and your P III also probably has much slower RAM to handle the cache misses. Beyond that Intel has improved many things in recent years; branch prediction, out of order execution, speculative execution, hardware prefetch, etc.

No, it's not multi-threaded. My 1.4 GHz Pentium-M laptop would have done that task in about 38 hours rather than the 86 hours your P III took. I don't really know how much of that difference is due to its 1 MiB L2 cache and how much is due to architectural improvements. My 1.6 GHz P4 system with 256 KiB L2 which is temporarily out of service has Astropulse run times very similar to your P III though.

                                                                 Joe

L2 Cache memory is the most significant item that speeds up AP tasks.

The figures for my Pent M @ 1.86GHz, with L2 cache of 2 MB are;
Measured floating point speed 1726.98 million ops/sec
Measured integer speed 3436.01 million ops/sec

Does AP tasks in about 25 hrs [90,000s] (that's with low radar blanking).

AP on 5101707

Roger (that).

Its come to my attention that the last SETI WU I crunched, i.e., Astropulse v505 Anonymous platform (CPU), my wingman crunched the same WU 6.9521462065648671395936034627871x faster than me. The difference betwixt our machines (mine / wing):

Measured floating point speed: 1396.14 million / 2584.49 (million ops/sec) +1.851168220952053518988067099288x better
Measured integer speed: 2233.86 / 5383.68 (million ops/sec) +2.410034648545566866321076522253x better

So HOW does wingman do 6.9x faster WU crunch than me? Memory? mine vs. theirs:

Memory: 1279.46 / 2045.58 (MB)
Cache: 0 / 4096 (KB)

Dunno why its indicating ZERO cache for my Intel P-III S 1400MHz (what makes the P-III chip an S version is the 1/2 MB cache). That being said, wingman has:

GenuineIntel Intel(R) Core(TM)2 Quad CPU @ 2.66GHz [Family 6 Model 15 Stepping 7]

SO: what's up with the lightyear difference in crunching time? The cache? Certainly NOT the 3/4 GB extra RAM. It, then, must be that the Astropulse v505 kernal is multi-threaded; and that pisses me off.

Your PIII probably has PC133 SDR single channel memory, your wingman's probably has PC6400 DDR dual channel memory, there's a huge difference in memory speed between the two,

Later CPU's have a better CPUID (which Boinc can detect), with earlier CPU's a lot of them share the same extended family, family code, extended model and model number,
and what they are (PIII, Xeon or Celeron) depends on how much cache they have, (Boinc can't detect this) see page 20:

http://www.intel.com/Assets/PDF/appnote/241618.pdf

I have mentioned this on the boinc_alpha list, it just requires someone with programming knowledge to implement it, are you up to it? ;-)

A C2D processor is also a lot better than an old PIII, my PIII @896MHz with PC133 memory @112MHz takes over 520,000 seconds to do an Astropulse Wu, your PIII compared to mine is already two thirds quicker,
While your wingman did his Wu in 42,000 seconds @2.66GHz, my E8500 @4.1GHz can do Wu's in around 25,000 seconds, and my HD5770 can do them in around 6,500 seconds,
other's have faster CPU's and GPU's that can do them faster still,

Claggy

Professor Ray, is that your only computer? Or just the only run you have which is running seti? I only saw one computer listed on your computers page. Is that the computer you use every day for your daily computing tasks though?

---

-baron_iv
Proud member of:
GPU Users Group

http://setiathome.berkeley.edu/workunit.php?wuid=682343828

The WU needed about same long as it would take on my AthlonXP 2000+. I think that's OK for this CPU.

---

I encourage you to look up the architectural changes in the CPU generations between the Pentium 3 and the Core2. There were SIGNIFICANT changes in everything from branch prediction, to instruction sets, memory speeds, on-die memory controllers, L2/L3 cache increases, the list goes on and on. Even a faster clocked P3 is going to lose to a much lower clocked Core2 due to these changes. Yes, optimized apps will help, but not enough to make up for the difference in architecture. For reference, intel's fastest (clock speed) CPU to date has been the Pentium 4, at 3.8ghz. Even on a single-threaded workload, a modern core2 or core i3/5/7 would leave the P4 in their dust, and they're clocked considerably lower. Clock speeds matter, but there are a lot of other things which matter more.

As the others have said, it might be time to retire your PIII, if you can afford something else. I'm guessing that AP unit took you several days to complete. The new Sandy Bridge architecture from intel is really a marvelous thing to behold and it would be many many times faster than your P3. The prices are quite reasonable too. If you preferred to go with AMD, the choices are just as interesting...you can get a 6-core CPU with unlocked multiplier for less than $250 these days, which is astounding bang-for-your-buck. The difference between either and your P3 would be like night and day. Most people, myself included, probably upgrade when they don't really need it. You, however, DO need it...badly. ;)

---

-baron_iv
Proud member of:
GPU Users Group

Wasn't aware I was snarky... I said I'm no expert (the truth) and seen from the point of view of an non expert, the simple answer is you have an old P3...

Woah. FIVE replies already to my rant?

But I am using P-III optimized app! See? It says:

Astropulse v505 Anonymous platform (CPU)

AND, JohnDK, you can keep your snarky comments to yourself; albeit, you probably meant them to be nicely snarky. And then, as if things aren't bad enough, [seti.international] Sutaru Tsureku Volunteer tester dog-piles on.

I see that he meant that nicely by holding back the coup de gras stating the app is NOT multi-threaded.

Just so all you guys know:

I have a GF II w/64 MB of DDR RAM on-board the video card. ALSO, I have TWO SCSI HDD devices having internal transfer rates (buffer to media) equivalent to ATA133 bandwidth (sustained).

(HANGS HEAD IN SHAME)

I know, I am so LAST YEAR that its not funny.

This is the WU in question?
http://setiathome.berkeley.edu/workunit.php?wuid=682343828

The speed difference is I guess because of the system architectures (old/newer).

After (http://setiathome.berkeley.edu/ap_faq.php) the min requirements would be a 1.6 GHz CPU.

The AP app (stock + opt.) is not multi-threaded. One app/WU use one CPU(-Core).

AP WUs need ~ 10 hours on my Intel Core2 Duo E7600 @ 3.06 GHz with DDR2 800/5-5-5-18 and ~ 15 hours on my AMD Phenom II X4 940 BE @ 3.0 GHz with DDR2 1066/5-5-5-18 - with opt. AP app.

---

It's also recommended that you use at least a 1.6ghz CPU for astropulse work. your 1.4 is pretty slow and should probably only be used for Multibeam work.

---

In a rich man's house there is no place to spit but his face.
Diogenes Of Sinope

Your wingman is using the optimized application. Those are more involved in that they don't update automatically but they certainly increase crunching speed. There are a number of other threads here that talk about the opti apps.
Dan

---

Not being an expert in any sense, isn't the simple answer that your P3 is old? :)

Its come to my attention that the last SETI WU I crunched, i.e., Astropulse v505 Anonymous platform (CPU), my wingman crunched the same WU 6.9521462065648671395936034627871x faster than me. The difference betwixt our machines (mine / wing):

Measured floating point speed: 1396.14 million / 2584.49 (million ops/sec) +1.851168220952053518988067099288x better
Measured integer speed: 2233.86 / 5383.68 (million ops/sec) +2.410034648545566866321076522253x better

So HOW does wingman do 6.9x faster WU crunch than me? Memory? mine vs. theirs:

Memory: 1279.46 / 2045.58 (MB)
Cache: 0 / 4096 (KB)

Dunno why its indicating ZERO cache for my Intel P-III S 1400MHz (what makes the P-III chip an S version is the 1/2 MB cache). That being said, wingman has:

GenuineIntel Intel(R) Core(TM)2 Quad CPU @ 2.66GHz [Family 6 Model 15 Stepping 7]

SO: what's up with the lightyear difference in crunching time? The cache? Certainly NOT the 3/4 GB extra RAM. It, then, must be that the Astropulse v505 kernal is multi-threaded; and that pisses me off.