S5R5 Performance Analysis

Mike Hewson
Mike Hewson
Moderator
Joined: 1 Dec 05
Posts: 6591
Credit: 323648764
RAC: 174685

Does anyone

Does anyone know-of/could-tell-me the file format/structure for the skygrid files? :-)

[ ie. skygrid_XXXXHz_S5R5.dat ]

Cheers, Mike.

I have made this letter longer than usual because I lack the time to make it shorter ...

... and my other CPU is a Ryzen 5950X :-) Blaise Pascal

Bikeman (Heinz-Bernd Eggenstein)
Bikeman (Heinz-...
Moderator
Joined: 28 Aug 06
Posts: 3522
Credit: 756207174
RAC: 1148699

RE: Does anyone

Message 90064 in response to message 90063

Quote:

Does anyone know-of/could-tell-me the file format/structure for the skygrid files? :-)

[ ie. skygrid_XXXXHz_S5R5.dat ]

Cheers, Mike.

Hi Mike.

Those are just zipped ASCII files. If you rename them skygrid_XXXXHZ.S5R5.dat.zip , they will unzip to an ASCII file skygrid_XXXXHz_S5R5.dat with two coulumns of ASCII floating point numbers, which are sky coordinates in RA,dec.

CU
H-B

Mike Hewson
Mike Hewson
Moderator
Joined: 1 Dec 05
Posts: 6591
Credit: 323648764
RAC: 174685

RE: Hi Mike. Those are

Message 90065 in response to message 90064

Quote:

Hi Mike.

Those are just zipped ASCII files. If you rename them skygrid_XXXXHZ.S5R5.dat.zip , they will unzip to an ASCII file skygrid_XXXXHz_S5R5.dat with two coulumns of ASCII floating point numbers, which are sky coordinates in RA,dec.


Thanks very much! :-)

Cheers, Mike.

( edit ) Indeed it looks like that dec goes from 'exactly' -PI/2 to ~ +PI/2, and for each such dec RA goes from exactly 0 to ~< 2 * PI

I have made this letter longer than usual because I lack the time to make it shorter ...

... and my other CPU is a Ryzen 5950X :-) Blaise Pascal

Bikeman (Heinz-Bernd Eggenstein)
Bikeman (Heinz-...
Moderator
Joined: 28 Aug 06
Posts: 3522
Credit: 756207174
RAC: 1148699

RE: ( edit ) Indeed it

Message 90066 in response to message 90065

Quote:

( edit ) Indeed it looks like that dec goes from 'exactly' -PI/2 to ~ +PI/2, and for each such dec RA goes from exactly 0 to ~< 2 * PI

Yup, like peeling an orange in one piece. The grid is isotropic so there will be more points per fixed declination near the equator compared to the poles (proportional to the circumference of the circle of latitude at that declination).

CU
H-B

archae86
archae86
Joined: 6 Dec 05
Posts: 3160
Credit: 7261408559
RAC: 1549335

RE: Here is all the 3.01

Message 90067 in response to message 90058

Quote:
Here is all the 3.01 available for my premium host, a Q9550 running stock.


I have another 101 results available from that host, which show what to first eye appears to be a possible minimum boundary execution time near 10600 CPU seconds. However this is not from cycle minimum phase--just interspersed among much longer neighbors.

I also have smaller but similar span collections from a Q6600 and an E6600. I'm reluctant to clutter the thread with more raw data (though I love seeing the graphs Bikeman has made and the commentary from all). So I'll await expression of interest, and possibly send the actual data by Private Message or other back channel to anyone interested in it.

To my eye the other two hosts show a similar overall pattern to the Q9550--fairly tight clustering of times and a pretty orderly slope near the cycle peak, shifting to a very broad scatter for points farther away from peak. However they don't hint at a minimum time cluster, so my eye may have constructed an artifact out of randomness on the Q9550.

None of my three hosts have returned work spanning either a cycle peak or cycle minimum yet.

Mike Hewson
Mike Hewson
Moderator
Joined: 1 Dec 05
Posts: 6591
Credit: 323648764
RAC: 174685

RE: Yup, like peeling an

Message 90068 in response to message 90066

Quote:
Yup, like peeling an orange in one piece. The grid is isotropic so there will be more points per fixed declination near the equator compared to the poles (proportional to the circumference of the circle of latitude at that declination).


Thanks, it's coming back to me now. Isotropic implies areal density ( number of points per [delta_RA x delta_dec] element ) is a constant ( or awfully/sufficiently close to that, within some tolerance ) over the entire celestial sphere.

Time for some spherical trig - I'm trying to establish a de-novo derivation of our runtime formulae, maybe with the wiggles, from the actual skygrid files, per frequency band etc ( as per your very thoughtful suggestions via PM ) ......... wish me luck! :-)

Cheers, Mike.

I have made this letter longer than usual because I lack the time to make it shorter ...

... and my other CPU is a Ryzen 5950X :-) Blaise Pascal

Bikeman (Heinz-Bernd Eggenstein)
Bikeman (Heinz-...
Moderator
Joined: 28 Aug 06
Posts: 3522
Credit: 756207174
RAC: 1148699

Hi! Some more S5R5 results

Hi!

Some more S5R5 results for windows machines, from data that archae86 made available.

1) stock-clocked E6600 (2.4 GHz on a Conroe-class duo)

data file

Min frequency in data : 632.35
Max frequency in data : 632.80
Min predicted runtime : 15567
Min observed runtime : 13207
Max predicted runtime : 23177
Max observed runtime : 23616
Rel. predicted runtime variation: 1.489
Mean predicted runtime : 18103

2) a stock-clocked Q6600 (2.4 GHz on a Conroe-class quad)
data file

Min frequency in data : 730.05
Max frequency in data : 730.40
Min predicted runtime : 18445
Min observed runtime : 14933
Max predicted runtime : 26169
Max observed runtime : 27353
Rel. predicted runtime variation: 1.419
Mean predicted runtime : 21020

3) Q9550 stock-clocked (2.83 GHz on a Penryn-class quad)
data file

Min frequency in data : 748.70
Max frequency in data : 750.35
Min predicted runtime : 12427
Min observed runtime : 10591
Max predicted runtime : 20148
Max observed runtime : 20518
Rel. predicted runtime variation: 1.621
Mean predicted runtime : 15001

Some S5R4 reference data for comparison will follow later
CU

H-B

Elphidieus
Elphidieus
Joined: 20 Feb 05
Posts: 245
Credit: 20603702
RAC: 0

It baffles me to see my

It baffles me to see my 3.0GHz Xeon underperformed by a huge margin compared to a 2.66GHz Core 2 Duo, both on a Mac platform.

WorkUnit ID 47744130

Paul D. Buck
Paul D. Buck
Joined: 17 Jan 05
Posts: 754
Credit: 5385205
RAC: 0

RE: It baffles me to see my

Message 90071 in response to message 90070

Quote:

It baffles me to see my 3.0GHz Xeon underperformed by a huge margin compared to a 2.66GHz Core 2 Duo, both on a Mac platform.

WorkUnit ID 47744130

If you use the Xeon for work, leave Safari open, leave other tasks running ... sure can happen ...

Safari pointing at the IBERCIVIS site sucks 25% or so CPU even when you do nothing ... because it has an active plug-in that is updated even when in the dock ...

Look at performance monitor to see what is stealing CPU ... you might be surprised... all lost CPU equates to longer run times ... that CPU has to come from somewhere ...

Elphidieus
Elphidieus
Joined: 20 Feb 05
Posts: 245
Credit: 20603702
RAC: 0

Nope. It was left overnight

Nope. It was left overnight to crunch, nothing was active except BOINC. Activity Monitor did not register anything out of the ordinary. All 8 cores register around 98% usage.

I've raised this question before on another thread, but it seemed nobody can identify the problem. Even peanut himself realised that his MP Xeon underperformed compared to his Mac Mini.

@peanut

Sorry to put you in the limelight, but your MP Xeon and mine shares an almost identical configuration and OS version with the exception of memory.

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.