I'm looking to move my machines around with the end of FAD. One of my borged machines is a 1.2 GHz Celeron, based on the Tualatin core Pentium 3. I don't know of anyone running something like this, so I don't know what sort of benchmark it would get, assuming a standard client and Windows.

Anyone got one for me to compare against? Or know somebody who does?

Cheers

---

Hi, NB. An Intel P4 @ 2.0 GHz reaches about 140 and I saw some of the celly 1.2's rated higher than my P4 at Find-a-Drug. Your FPU should do well here.

I'm looking to move my machines around with the end of FAD. One of my borged machines is a 1.2 GHz Celeron, based on the Tualatin core Pentium 3. I don't know of anyone running something like this, so I don't know what sort of benchmark it would get, assuming a standard client and Windows.

Anyone got one for me to compare against? Or know somebody who does?

Cheers

This system of mine is a 750 MHz AMD, it has a RAC is the 80 range. Not sure how a 1.2 GHz Celeron will stack against it, but I'd have to say they ought to be in the same area.

If one of the more knowledgable gurus comes by (Bill Michael or Paul Buck) they might be able to point us at a page that shows systems ordered by CPU type / speed. I know it exists, just not 100% sure where.

[edit]
Close your darn url tags :)
[/edit]

---

point us at a page that shows systems ordered by CPU type / speed

Like this one?

---

point us at a page that shows systems ordered by CPU type / speed

Like this one?

Exactly like that, Oh Great Knowledgable Guru of all things Boinc. :)

Thanks!

---

Like this one?

They seem to be meaningless figures though, if you look at the average credit per CPU. It's probably because the figures include inactive hosts, do not take their resource share into account, number of hours per day they're running, etc.

For example, the RAC shown on the stats site for an Athlon 64 3700+ is a little under 100. For that type of CPU, even if it's only running Rosetta 8 hours a day it would likely get an RAC well over 100 (assuming all runs well).

Similarly, an Athlon X2 4400+ (which is essentially two of the above) should be able to get well over 200 per 8 hour day, yet the RAC shown is just 160. If it runs 24/7 doing nothing but crunch Rosetta, I'd expect it to come in at well over 600 a day.

To get back to the original post...

Housing and Food Services has some 1200MHz Celerons you could compare against. See host 64479 and 38143 for example

---

*** Join BOINC@Australia today ***

For example, the RAC shown on the stats site for an Athlon 64 3700+ is a little under 100. For that type of CPU, even if it's only running Rosetta 8 hours a day it would likely get an RAC well over 100 (assuming all runs well).

I've never had my 3700+ "Rosetta only", but for a while it _was_ SETI only, and the RAC was 700. At 25% share for Rosetta (equivalent of 6 hrs/day), the Rosetta RAC is 116. Of course, it's overclocked a bit.

---

I've never had my 3700+ "Rosetta only", but for a while it _was_ SETI only, and the RAC was 700. At 25% share for Rosetta (equivalent of 6 hrs/day), the Rosetta RAC is 116. Of course, it's overclocked a bit.

Which kinda proves my point :-) With yours, if it ran 24 hours a day, you'd be looking at approx. 450 credits a day, rather than the average of 160 shown on BOINCstats. 700 on SETI sounds like you used one of the optimised apps and were consistently granted more credit than claimed.

Just for interest's sake, I did a comparison of Athlon 64 3700+ hosts (using a hosts xml file from Rosetta). Credits per hour (CPH) based on the standard BOINC formula ranged from 4.1 to 29.27 in the sample I looked at (114 hosts). Mean was 12.83 and median was 12.15

For the 1200MHz Celeron, the range was 3.25 to 6.69 CPH based on 84 hosts. The mean was 5.61 and the median 5.78.

Of course, comparing seemingly similar CPU's doesn't tell the full story - Socket 754 or Socket 939? (Socket 478 or 775 in the case of Intel?) FSB speed? Overclocked? Memory speed and configuration? Cool & quiet enabled? Operating System? BOINC Client? etc etc.

I could publish a table of median/mean for each procesor type (where there's more than 10 in the sample) sometime, with a disclaimer that "your mileage may vary".

---

*** Join BOINC@Australia today ***

I don't believe the data will accurately determine the RAC of a computer dedicated to just one project. If my scores were as low as the ones shown in the listing to which Bill linked, I would find a project more suitable for my computers.

Biggles, let us know what your RAC is after a month of dedicated crunching. Don't be discouraged if it starts at 40 to 60. It takes weeks to reach maximum...and only if it runs 24/7. I'll bet your RAC will be over 120.

Ah, the RAC discussion... :)

RAC is only really a good metric when the computer is single project and is running on a very consistent schedule. The RAC on the project web sites will be very dependent on the timing of the return of the work because it is calculated when the quorum of results is formed and credit is awarded.

The RAC calculations on the stat sites tends to be "better" because it is calculated at the same time each day, or on the time stamp of the XML file. In effect, removing a lot of the impact of the timing of the credit awards.

When you go multi-project all this goes out the window, even more so with multiple CPUs in the system because of the way the CPU Scheduler does its assignments. So, you have a lot more randomness in the timing of work completion.

Additional factors that can come into play are the instability of the benchmark which affects the credit calculations, thus awards; etc.

Or you can ignore all of what I said and just use the numbers Webmaster Yoda suggested ... :)

---

Upon consideration, I guess the question was less to do with RAC, and more about what it would benchmark at. Many thanks to Webmaster Yoda for finding the Housing and Food Services hosts to compare to. How did you do it? I can expect a measured floating point speed in the range of 1050-1100 million ops/s, which clock for clock, is similar to an Athlon XP. I'm happy with that.

However, my 2.93 GHz Celeron D that gets a measured floating point speed of 1318 million ops/s is getting moved off of Rosetta. 244% of the clockspeed of the Tualatin core based machine, only 122% of the floating point speed. Clock for clock, it's doing half the work. I think it's going to get stuck on GIMPS.

Thanks for all the comments folks.

---

Many thanks to Webmaster Yoda for finding the Housing and Food Services hosts to compare to. How did you do it?

Nothing complicated. Basically I just assumed they'd have one, given the number of computers they have on the project :-)

So I just looked them up on the Rosetta site and kept paging through until I saw a 1200MHz Celeron.

There are plenty more though - the xml stats file I downloaded shows 83 hosts with a 1200MHz Celeron (excluding Celeron M)

CPU type:      Intel(R) Celeron(tm) CPU 1200Mhz
Mean MFlOps:    961.63
Mean MIOps:    1726.02

---

*** Join BOINC@Australia today ***

Upon consideration, I guess the question was less to do with RAC, and more about what it would benchmark at. Many thanks to Webmaster Yoda for finding the Housing and Food Services hosts to compare to. How did you do it? I can expect a measured floating point speed in the range of 1050-1100 million ops/s, which clock for clock, is similar to an Athlon XP. I'm happy with that.

I'm running a 1.3 GHz Celeron (with an adapter to the Slot 1 interface) on an Intel 440BX2 motherboard with 640MB memory. It benchmarks around 1.2 GFlops for Windows and about 600 MFlops for Linux using the standard client. I use an optimized client for Linux that brings the benchmark back to 1.2 GFlops.

I would have replied earlier, but the discussion went off on RAC. I didn't see any way to give a meaningful reply about RAC since Rosetta has a 32.2% share of 4 projects running on this computer. An "eyeball" average Rosetta WU completion time (CPU time) is about 4 hours, but it varies from 1 to 7 hours depending upon the WU.

---

the xml stats file I downloaded

Yoda, not to hijack the thread, if this xml file breaks down "Darwin" users to the "7.9" or "8.2" point, can you email me? firstname dot lastname at comcast dot net. Boincstats and others lump all the Darwin together. Thanks!

---