HostID 1001562 - Richard Haselgrove's Q6600 Quad Core
15 Feb 2008 8:29:12 UTC
Topic 193515
(moderation:
As Richard announced here and here, he is dedicating his Q6600 to a timing run using the 4.32 'power users' app and the new >800 frequency data. He has invited other crunchers to follow his adventures and with the advent of Mike Hewson's really useful ReadyReckoner (RR), it will be very easy to do that.
Just over a day ago, I published a list of the completed tasks for Richard's host ID 1001562 that were recorded on the website at that time, leading up to the transition to the new 4.32 version of the science app. The published list is exactly what is required to paste into the current latest version of RR, which will then show you all you ever wanted to know about the cyclic variation of task runtimes. You can view tabulated data, summary outputs, and very nice plots.
Since there is likely to be some interest in Richard's work and since I intend to follow it myself, I've created this thread where discussion of this particular project of Richard's can take place. Please feel free to post here if you have anything to add to the discussion of the results coming from this particular host. If you wish to discuss your own or another host's results, please start a separate thread.
Below, you will find the latest full set of data for Richard's host. It contains all that was previously published and the new 4.32 runtimes that have accumulated since the app change. I intend to publish each day, just the additional results from the previous 24 hours (if possible). No results will be missed but I may not always be able to publish immediately.
0775.25,116,26752,4.32 0777.50,004,33027,mixed 0777.50,008,35535,4.26 0777.50,010,35235,4.26 0777.50,013,34562,4.26 0777.50,017,34187,4.26 0777.50,032,31903,4.26 0777.50,043,30665,4.26 0777.50,052,29987,4.26 0777.50,059,29705,4.26 0777.50,071,30035,4.26 0777.50,085,30951,4.26 0777.50,091,31901,4.26 0777.50,098,32670,4.26 0777.50,105,33406,4.26 0777.50,113,35474,4.26 0777.50,114,35313,4.26 0777.50,142,33655,4.26 0777.50,145,33431,4.26 0777.50,149,32914,4.26 0777.50,172,30268,4.26 0777.50,180,29833,4.26 0777.50,182,29682,4.26 0777.50,184,29655,4.26 0777.50,185,29640,4.26 0777.50,202,30312,4.26 0777.50,215,31699,4.26 0777.50,225,33078,4.26 0777.50,236,35031,4.26 0777.50,239,35561,4.26 0777.50,240,35643,4.26 0777.50,268,33189,4.26 0777.50,270,33476,4.26 0777.50,271,32981,4.26 0777.50,274,32680,4.26 0777.50,278,32215,4.26 0777.50,281,31710,4.26 0777.50,326,24725,mixed 0909.15,489,25740,4.32 0909.15,490,25849,4.32 0909.15,491,26028,4.32 0909.15,492,25933,4.32 0909.15,493,26242,4.32 0909.15,494,26463,4.32 0909.15,495,26495,4.32 0909.15,496,26472,4.32 0909.15,497,26713,4.32 0909.15,498,26940,4.32 0909.15,499,26882,4.32 0909.15,500,27144,4.32 0909.15,501,27318,4.32 0909.15,502,27352,4.32 0909.25,499,26836,4.32
Note that the data is sorted in ascending order of frequency and then in ascending order of sequence#, irrespective of the chronological order in which the tasks were crunched. The four columns represent Frequency, Sequence#, Runtime, App_Version respectively. The word 'mixed' simply means that more than one version was used on that task.
Cheers,
Gary.
HostID 1001562 - Richard Haselgrove's Q6600 Quad Core
)
It's amazing what a difference a few more data points make.
Yesterday, Richard published this output from the first 10 tasks completed
Today, with just 14 points at frequency 909.15, here is the summary output
It seems that the 4 extra points have made an amazing difference. Richard is in for some really quick runtimes once he gets down to the trough.
The plot of the 909.15/4.32 freq/app combo looks great. Take a look for yourself - it's so easy. Just grab RR_V6A and paste the complete data from above and use the 'Prev'/'Next' buttons as appropriate to pull up the exact plot you wish to see. You can easily examine the the 777.50/4.26 data as well while you are there.
Cheers,
Gary.
Next four results will be
)
Next four results will be cooked in a couple of hours.
In the meantime, here are the RR results for v4.07 for the same machine - well, actually its twin sister, but under the same experimental conditions.
That led to this chart:
(direct link)
which looked very nice, but putting the data into the RR shows it to be quite a way off the theoretical line.
Mike, you might like to have a play with:
464.35,16,38788.09,4.07
464.35,17,38636.52,4.07
464.35,18,38493.59,4.07
464.35,19,38341.72,4.07
464.35,20,38269.75,4.07
464.35,21,38214.47,4.07
464.35,22,38252.78,4.07
464.35,23,38271.33,4.07
464.35,24,38251.06,4.07
464.35,25,38383.02,4.07
464.35,26,38495.83,4.07
464.35,27,38624.69,4.07
464.35,28,38760.09,4.07
464.35,29,38999.58,4.07
464.35,30,39135.77,4.07
464.35,31,39266.78,4.07
464.35,32,39447.55,4.07
464.35,33,40120.73,4.07
464.35,34,40217.83,4.07
464.35,35,40649.02,4.07
464.35,36,40757.98,4.07
464.35,37,41196.08,4.07
464.35,38,41665.31,4.07
464.35,39,41958.31,4.07
464.35,40,42216.77,4.07
464.35,41,42513.52,4.07
464.35,42,43080.69,4.07
464.35,43,43430.56,4.07
464.35,44,43836.02,4.07
464.35,45,43883.66,4.07
464.35,46,43404.98,4.07
464.35,47,43006.2,4.07
464.35,48,42717.13,4.07
464.35,49,42198.92,4.07
464.35,50,42008.16,4.07
464.35,51,41584.66,4.07
464.35,52,41328.98,4.07
464.35,53,40787.58,4.07
464.35,54,40402.31,4.07
464.35,55,40425.52,4.07
464.35,56,39871.63,4.07
464.35,57,39489.52,4.07
464.35,58,39430.14,4.07
464.35,59,39253.53,4.07
464.35,60,39008.66,4.07
464.35,61,38776.59,4.07
464.35,62,38640.3,4.07
464.35,63,38486.91,4.07
464.35,64,38348.38,4.07
464.35,65,38271.42,4.07
464.35,66,38254.17,4.07
464.35,67,38216.27,4.07
464.35,68,38272.89,4.07
464.35,69,38289.89,4.07
464.35,70,38317.14,4.07
464.35,71,38486.77,4.07
464.35,72,38661.41,4.07
464.35,73,38831.41,4.07
464.35,74,39069.33,4.07
464.35,75,39221.84,4.07
464.35,76,39318,4.07
464.35,77,39411.13,4.07
464.35,78,40143,4.07
464.35,79,40157.05,4.07
464.35,80,40687.83,4.07
464.35,81,40808.23,4.07
464.35,82,41228.2,4.07
464.35,83,41710.05,4.07
464.35,84,41980.47,4.07
464.35,85,42274.42,4.07
464.35,86,42520.84,4.07
464.35,87,43044.89,4.07
464.35,88,43379.23,4.07
464.35,89,43713.05,4.07
464.35,90,43946.31,4.07
464.35,91,43417.66,4.07
464.35,92,43018.31,4.07
464.35,93,42657.48,4.07
464.35,94,42282.02,4.07
464.35,95,42038.22,4.07
464.35,96,41619.52,4.07
464.35,97,41430.89,4.07
464.35,98,40902.63,4.07
464.35,99,40514.13,4.07
464.35,100,40499.64,4.07
464.35,101,39929.56,4.07
464.35,102,39555.58,4.07
464.35,103,39469.75,4.07
464.35,104,39258.8,4.07
464.35,105,39044.02,4.07
464.35,106,38829.92,4.07
464.35,107,38648.83,4.07
464.35,108,38542.33,4.07
464.35,109,38372.13,4.07
464.35,110,38336.14,4.07
464.35,111,38269.7,4.07
464.35,112,38234.31,4.07
464.35,113,38217.36,4.07
464.35,114,38121.77,4.07
464.35,115,38327.67,4.07
464.35,116,38353.41,4.07
464.35,117,38608.02,4.07
464.35,118,38736.36,4.07
464.35,119,38910.86,4.07
After 18
)
After 18 results:
I hadn't looked before, but I seem to be sharing this dataset with host 997488 (Peanut's 8-core X5365 Darwin @ 3.00GHz), and host 1094135 (an anonymous 8-core X5355 Linux @ 2.66GHz). No wonder I can't keep up! But it's a nice little cross-platform comparison we've got running here - Gary, you might like to add those to your monitoring list.
RE: Mike, you might like to
)
Indeed! The plot noticeably wanders off line in the higher cycles, with either RR_V6A or RR_V7A ( you can now breathe! ). It spurred my question about the choice of period brought up here. As RR's current choice for that ( 0.000206 ) was derived from ~ 800Hz, and your listing is from ~ 460Hz, it's fortunately a gentle effect. I'm clearly going to need a deeper derivation of that 'grid density constant' in the quadratic for the period. I have a feeling that's non-trivial, but it is certainly important enough to solve as the tasks progress upwards from 900Hz.
Cheers, Mike.
( edit ) Actually, here's some of Peter's from even lower down:
254.1,37,28588
254.1,36,28441
254.1,34,28031
254.1,32,28413
254.1,31,29706
254.1,30,30560
254.1,29,31579
254.1,26,31900
which shows similiar angst.
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
RE: RE: Mike, you might
)
Mike, for various reasons that I'll explain later, if you think it worthwhile, make these two changes and rerun the RH data and see if you like the better fit in the higher cycles.
1. Change the constant 0.000206 to 0.0002042
2. Change the frequency in the data from 464.35 to 470.45
I've done that and to me it looks a lot better. What do you think?
While you are trying it out, I'll work on the explanation :).
Cheers,
Gary.
Here is a little comparison
)
Here is a little comparison of Richard's & my shared work on the 909.15 sequence. The lower sequence numbers (~456) seem to be near the trough so we should be able to see if the RR predicts the trough well. My guess is that the RR trough looks pretty close. The -0 is a quirk of different versions of MySQL on my computers I think. The #'s are freq, seq#, my hostid, my time, Richards Host=1001562, and Richard's time.
909.15 456 997488 20522.6 1001562 -0
909.15 457 997488 21224.2 1001562 -0
909.15 459 997488 20786.3 1001562 -0
909.15 460 997488 21451 1001562 -0
909.15 461 997488 21551.3 1001562 -0
909.15 463 997488 21364 1001562 -0
909.15 464 997488 22073.3 1001562 -0
909.15 465 997488 21551.9 1001562 -0
909.15 466 997488 21521 1001562 -0
909.15 467 997488 21782.9 1001562 -0
909.15 469 997488 22161.7 1001562 -0
909.15 470 997488 21996.1 1001562 -0
909.15 471 997488 22080.5 1001562 -0
909.15 472 997488 22188.7 1001562 -0
909.15 475 997488 22755.8 1001562 24341.3
909.15 476 997488 22843.7 1001562 24424.4
909.15 477 997488 22886.4 1001562 24466.8
909.15 478 997488 22948.2 1001562 24515.6
909.15 481 997488 23195.4 1001562 24760.3
909.15 482 997488 23414.1 1001562 24973.9
909.15 483 997488 23586.3 1001562 25188.8
909.15 484 997488 23711.6 1001562 25318.8
909.15 485 997488 23648.1 1001562 25212.3
909.15 486 997488 23567.1 1001562 25144.8
909.15 487 997488 24128.8 1001562 25724.1
909.15 488 997488 23974.6 1001562 25557.1
909.15 489 997488 24191.7 1001562 25740.9
909.15 490 997488 24239.4 1001562 25849.1
909.15 491 997488 24466.5 1001562 26028.9
909.15 493 997488 24618 1001562 26242.6
909.15 494 997488 24870.1 1001562 26463.2
909.15 495 997488 24889.1 1001562 26495.9
909.15 496 997488 24842.1 1001562 26472.7
909.15 497 997488 25098.3 1001562 26713.8
RE: Mike, for various
)
First, I've made a minor change - with RR_V7C one can enter/vary a sky grid density value.
Second, yes it does look better! Do tell ....
Hmmm ... I think I'm heading to a three parameter least squares curve fitting exercise, one free degree of which is within the sine phase - & that'll have to be expanded as a polynomial now. :-)
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
RE: RE: Mike, for various
)
Just want to mention that it's nice and all to be refining the calculations, but you may want to consider having examples in there like the earlier versions so as to make this a tool for all users, not just people who already know what to do. The input via csv is great, but if I were to try to refer someone to your page, I'd have to provide a tutorial first on what is the frequency, the task number, etc, etc, etc...
Guess it depends on who you view as your primary audience...
RE: Just want to mention
)
Err ... well you'd use whichever version takes your fancy I guess, that's why I wrote them each as standalone versions. :-)
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
Latest report: Period
)
Latest report:
Runtimes are still shortening (Peanut, the RR predicts a trough at 426 - way to go, man!), and BOINC's RDCF is already predicting 6:52 for the cached WUs.
I'm going to have to stop data-watching for the weekend - it should all still be there when I get back - and hand you over to Gary for the next stage. I should still have forum access, though.