Classic Team Lambchop SETI workunit benchmarked on various CPUs

[craiggloyd]

Something near and dear to me, Seti@home Classic before the BOINC days. ArsTechnica was a huge contributor to SETI classic with their 'Team LambChop' website complete with benchmarking stats. I found the site via the WayBackMachine. Not to mention contribution to early Work Unit caching development via SETI Driver and SETI Spy. This is where to find the official work unit. Genuine 3.03 win32 GUI and 3.08 wiin32 CLI executables can be found in my Mega cloud link within my Youtube video description or probably with Wayback as well. https://web.archive.org/web/20080513235717/http://www.teamlambchop.com/bench/index.htm

3.03 GUI uses more than 1 core at once, despite being made in 2002.. just shy of 2 cores of my 6-core i5-8500. 1 full core for crunching and then a highly inefficient GUI run on another core. The commandline version of Seti 3.08 is more efficient and just uses 1 core, while, for some reason having a more efficient crunching engine. Why couldn't the GUI have the same crunching engine?
Let's bring this back (as a single core bench). How fast does a 14th gen Intel run this on a P-core or a Zen 4 AMD? Probably 10 minutes per core.

Results from official Lambchop WU, recorded at Arecibo (RIP) on Sunday April 11th 1999 at 21:43:18 ;
i5-8500 core in question at 3.8GHZ;
v3.08 Commandline SETI: 34 minutes
v3.03 GUI: 44M

3.2GHZ i5-2400 Sandy Bridge
v3.08 CLI: 40 minutes, wow SB u rock!
v3.03 GUI: 56 minutes

2.8GHZ Pentium D 820;
v3.08 CLI: 90 minutes

1.6GHZ Pentium M 725
v3.08 CLI: 94 minutes , but only 7.5w TDP yo, and undervolted in Windows XP!

1.9GHZ AMD Athlon 'Thoroughbred
v3.08 CLI: 187 minutes

Another WU I found took longer, at 45 minutes on 3.08CLI and 62 minutes on 3.03 GUI but that is not genuine TeamLambchop bench WU.

I have compared this Tomshardware 2003 article with my Pentium D and scores nearly identically, so they probably used our work unit. But unfortunately for no other articles. https://www.tomshardware.com/reviews/prometeia-mach-2,643-23.html

*I don't know why after editing my thread it goes back into moderation mode then gets deleted.

 

[crashtech]

Thanks for posting. The subject far predates my involvement with DC other than Folding, but it's interesting historical data nonetheless, and I can't imagine why it would be deleted except by some kind of glitch...

 

[craiggloyd]

Around the same time SETI@home was released, there was Parabon! Biological science Work Units long before Folding Or World Community Grid was a thing. At first they offered financial compensation but then stopped that. Did not stay active for very long but another piece of history.

D2OL project was interesting as well, ran on JAVA, ended in 2009. https://www.wired.com/2003/04/grid-computing-spreads-to-sars/


Parabon to Deliver Power To Cancer Research
July 28, 2000

NEWS BRIEFS

Fairfax, VA. — Parabon Computation, Inc., a pioneer in Internet computing, will deliver computing resources to the National Cancer Institute (NCI). These computing resources will help researchers understand how certain drugs affect cancer cells.

Parabon’s Frontier computing platform will be used by the Genomic & Bioinformatics Group (GBG) of NCI’s Laboratory of Molecular Pharmacology for research that could help answer complex questions that arise during the analysis of cancer-related data. These computational studies will assist in the analysis of new anti-cancer drugs. Parabon’s computational resources may enable the NCI research team to accelerate cancer research by solving complex computational problems in a matter of days rather than the many months currently required. This partnership may also lead to new ways for researchers to answer questions that extend beyond the bounds of traditional supercomputing.

“Cancer research is just one example of scientific and business research that has always been limited by traditional computing resources. Now with Parabon’s Internet computing solution, these limitations are gone,” said Steven L. Armentrout, Ph.D., president and CEO of Parabon. “We are very excited about Parabon’s support of the NCI project-it allows individuals to directly contribute to cancer research and may lead to important discoveries.”

Parabon is providing supercomputation resources to NCI in a brand new way: by harnessing the idle power of computers across the Internet. Parabon’s Frontier platform provides flexible and variable power and delivers it on demand to researchers and businesses worldwide. Complex calculations can be greatly accelerated, significantly reducing research and development time across a variety of industries.

Within the Frontier platform, large computational problems are broken up into smaller tasks and distributed via the Internet to individual computers on which the Pioneer compute engine software has been installed. Pioneer processes a task only when the software senses that the computer is idle. Once a task is completed, the result is sent back via the Internet to Parabon’s server. Parabon’s system is thoroughly protected with multiple layers of security, protecting an individual’s computer and access by third parties.

As part of its Compute Against Cancer(SM) program, Parabon is donating computing resources to assist additional cancer research projects. By providing less costly and more efficient supercomputing service, Parabon’s technology enables researchers to conquer complex problems they would otherwise not be able to solve. Parabon is dedicating substantial computational resources to ongoing research and treatment projects in support of cancer patients and their families. Additional projects and partnerships will be announced throughout the year.

The National Cancer Institute (NCI) leads research efforts to conquer cancer in all its forms. NCI is one of 25 institutes in the National Institutes of Health, which is part of the U.S. Department of Health and Human Services. NCI conducts, coordinates, and funds cancer research, and provides vision and leadership for the cancer research community. NCI conducts and supports programs to understand the causes of cancer; prevent, detect, diagnose, treat, and control cancer; and disseminate information to the practitioner, patient, and public. More information can be found at http://www.nci.nih.gov .

Parabon delivers virtually unlimited computing capacity by harnessing the power of millions of individual computers via the Internet. This flexible, cost-effective resource will make it possible for businesses to accelerate time to market, researchers to discover new cures and scientists to advance discovery. For more information, visit http://www.parabon.com .

 

[TennesseeTony]

I miss Oxford's United Devices (Grid.org).

grid.org - Wikipedia

en.wikipedia.org

 

[mmonnin03]

I ran United Devices prior to switching to FAH somewhere in 2001/02

 

[Assimilator1]

Something near and dear to me, Seti@home Classic before the BOINC days. ArsTechnica was a huge contributor to SETI classic with their 'Team LambChop' website complete with benchmarking stats. I found the site via the WayBackMachine. Not to mention contribution to early Work Unit caching development via SETI Driver and SETI Spy. This is where to find the official work unit. Genuine 3.03 win32 GUI and 3.08 wiin32 CLI executables can be found in my Mega cloud link within my Youtube video description or probably with Wayback as well. https://web.archive.org/web/20080513235717/http://www.teamlambchop.com/bench/index.htm

3.03 GUI uses more than 1 core at once, despite being made in 2002.. just shy of 2 cores of my 6-core i5-8500. 1 full core for crunching and then a highly inefficient GUI run on another core. The commandline version of Seti 3.08 is more efficient and just uses 1 core, while, for some reason having a more efficient crunching engine. Why couldn't the GUI have the same crunching engine?
Let's bring this back (as a single core bench). How fast does a 14th gen Intel run this on a P-core or a Zen 4 AMD? Probably 10 minutes per core.

Results from official Lambchop WU, recorded at Arecibo (RIP) on Sunday April 11th 1999 at 21:43:18 ;
i5-8500 core in question at 3.8GHZ;
v3.08 Commandline SETI: 34 minutes
v3.03 GUI: 44M

3.2GHZ i5-2400 Sandy Bridge
v3.08 CLI: 40 minutes, wow SB u rock!
v3.03 GUI: 56 minutes

2.8GHZ Pentium D 820;
v3.08 CLI: 90 minutes

1.6GHZ Pentium M 725
v3.08 CLI: 94 minutes , but only 7.5w TDP yo, and undervolted in Windows XP!

1.9GHZ AMD Athlon 'Thoroughbred
v3.08 CLI: 187 minutes

Another WU I found took longer, at 45 minutes on 3.08CLI and 62 minutes on 3.03 GUI but that is not genuine TeamLambchop bench WU.

I have compared this Tomshardware 2003 article with my Pentium D and scores nearly identically, so they probably used our work unit. But unfortunately for no other articles. https://www.tomshardware.com/reviews/prometeia-mach-2,643-23.html

*I don't know why after editing my thread it goes back into moderation mode then gets deleted.

Cool! I haven't run that SETI benchmark in a very long time!
Thanks for bringing this back , I'm pretty sure we used this benchmark WU here. And I tried to do something like this with BOINC, but it just proved too complicated for me, and their wasn't much demand for it either.

I'll run it on my Ryzen rig in a minute, I could run it on my 6 core Ivybridge o/c to 4 GHz, but you've kind of got that covered with your SB.
I didn't know (or had forgotten) that the GUI could use 2 cores, I guess it was meant for dual socket rigs, although it's a total waste using a 2nd core/CPU for the GUI! lol.

Couldn't find the TLC bench WU on the wayback site, but it turns out that I still have it in my HDDs storage! lol.
Not sure if it's the right one though, file properties for the SAH file show a date 21-12-2000 (UK date format), is that the right one?

Btw, I have vague memories of CLi v3.03 being faster than v3.08, does that ring any bells?

 

[craiggloyd]

Hi! It should be on wayback here (https://web.archive.org/web/20020210123651/http://www.teamlambchop.com/bench/benchfile.htm) I also have it on my Mega cloud drive in youtube description.
and I have taken a screenshot of what it shows in the GUI.
Yes commandline version runs faster, more efficient algorithm.

 

[Assimilator1]

Hi! It should be on wayback here (https://web.archive.org/web/20020210123651/http://www.teamlambchop.com/bench/benchfile.htm) I also have it on my Mega cloud drive in youtube description.
and I have taken a screenshot of what it shows in the GUI.
Yes commandline version runs faster, more efficient algorithm.

Err, as I mentioned I already have it on my HDD , although I do want it confirmed that it's the same version.

Also I was talking about the thinking that the CLi v3.03 is faster than v3.08 CLi (because you only tested with GUI 3.03, for some reason I didn't think there was a GUI 3.08, lol).

Anyway, with CLi v3.08 I get "cpu_time=3978.937500" with my Ryzen rig (specs below). Which is 66 minutes, hmm, that's not great!
I've limited my CPUs power draw, so that probably puts its multi core clock speed near default.
I'm going to run it again with CLi v3.03.... (the CPU is currently clocked at 3.6 GHz with ~94% load, SETI is running along side BOINC, as it was yesterday. Although then I had BOINC set to only use 50% of the CPU).

 

[craiggloyd]

Check CPU time from Task Manager to see if they come close to matching because I do not know how SETI calculates time. Check if priority is set above 'low' for SETI. You have hyperthreading? You need a full idle core to get good data, perhaps that is the reason despite BOINC set to 50%. Hey.. can i get cli 3.03 from you? 3.6GHZ is fine.
Yes the radio telescope header GUI screenshot, that is genuine Lambchop

 

[Assimilator1]

BOINC was at 50% for the 1st run and then at 88% for the 2nd, my CPU is 6 core +HT, so 12 threads all in. In both runs SETI had a full core available and was showing at about 8% in TM, which corresponds to a whole thread (with one HT thread being unused in the 2nd run).
I can't remember how you set priority for SETI, lol. But whenever I looked at it in TM, it was running at full bore.

v3.03 CLi time was 43 mins, so yep, as I remembered, v3.03 is quicker, but I think v3.08 did more science?
Btw, the last benchmark I did with this WU appears to be on 25-03-2003 on my XP o/ced to 2.14 GHz (no other system details I'm afraid), it managed a time of ~145 mins (2hrs 25 mins) on v3.03.

I don't understand your comment about the GUI screenshot. Is that how you identify the WU?

I'll see if I can attach CLi v3.03 to my post , nope, damn it, not an allowed file type. I'll PM you with my email address.
(btw, I've got CLi clients all the way back to v1.2, not windows that one, 1.3 for windows. And I have GUI clients back to v1!).