[CRTech] Christian Radio Tech [MSG 81059]
[Thread Prev] [-- Thread Index --] [Thread Next] [Date Prev] [-- Date Index --] [Date Next]
Re: Rivendell curiosity
To: CRTech <crtech@crtech.org>, Tom Beattie <tom@tombeattie.com>
Subject: Re: Rivendell curiosity
From: Lamar Owen <lowen@pari.edu>
Date: Thu, 12 Oct 2017 14:14:36 -0400
Content-language: en-US
In-reply-to: <CAL=56PpZmzDAcPFL=21LC49U88QOpXe3p7eguyy3bZiijWL91A@mail.gmail.com>
References: <CAOkhgWozDHU6V-hTiWiV-CCoSG66JewVVvyxLXQtURAFn_vLmg@mail.gmail.com> <E077351015F74F4EAAFFC02321F37A42A186242F@EMP-MB-01.emp.ark.nwc.edu> <E077351015F74F4EAAFFC02321F37A42A1862471@EMP-MB-01.emp.ark.nwc.edu> <CAOkhgWrNpFWiovRGj8r1Anwp32RgGZqnAMNLZsv2yb-PATqSpg@mail.gmail.com> <005301d34119$645d2490$2d176db0$@voiceofchristmedia.org> <CAOkhgWqAMRhentrDvdyzPV_RqwT-KiJB+bTFbT+ZuVPuWtYwyg@mail.gmail.com> <006401d34123$54364560$fca2d020$@voiceofchristmedia.org> <CAOkhgWorYabwH6OnEP-zgVQMT2BbzjEc1thwyX_CDd0rGpEGKw@mail.gmail.com> <83ca06ba-d793-a003-f53a-5db4043a628c@pari.edu> <51E29CA2-FE64-4D04-8BEB-34B8D1F7DDFB@ieee.org> <1ae45548-616d-1f35-c3b5-23cdfea73b7a@pari.edu> <CAL=56PpZmzDAcPFL=21LC49U88QOpXe3p7eguyy3bZiijWL91A@mail.gmail.com>
User-agent: Mozilla/5.0 (X11; Linux x86_64; rv:52.0) Gecko/20100101 Thunderbird/52.3.0
On 10/12/2017 12:26 PM, Tom Beattie wrote:
Kinda off topic but... those Q6600 processors were something special. ...
Perhaps a bit off-topic, but I'll indulge a bit if that's ok.

Indeed; the Q6600 was one of those 'right sized' processors that were a couple of steps down from the top end but still plenty of power for their generation.  What's wild is how good the Q6600 still performs relative to the newest stuff out there; I use PassMark Software's www.cpubenchmark.net/index.php page to do comparisons, and the Q6600 is still present in the High to Mid-Range CPU chart, and scores better than some fifth-generation Core i3's and a few second-gen Core i7's.  The most-revealing comparison of all, however, is the 'Single Thread Performance' ( https://www.cpubenchmark.net/singleThread.html ), which influences audio production tasks more than anything else, in terms of CPU speed at least.  Mixbus can effectively use 4 or more cores for the DSP in the plugins, but some things cannot, and so single-core speed does become a bottleneck.

That's why, when I was first looking for a quad-core laptop to do production on, I picked my older Dell Precision M6500 with an i7-740QM.  Now, just like the Q6600, this is a 'right-sized' processor that is not that much slower than the 820QM, 840QM, or even the 920XM and top-end-of-Nehalem 940XM but for a whole lot less money (the 940XM processor alone costs more than I paid for the whole laptop, even on eBay).  That i7-740QM has just a little less performance than the desktop Q6600, at a lower clock speed, incidentally.  Further, the i7-740QM beats out most of the bargain i3's and even a few i5's up through the 5th generation; see https://www.cpubenchmark.net/laptop.html and look at everything below a score of 3,206 (the 740QM isn't on the 'new' laptop CPU list, but its results are still there linked from the Single Thread list).

When I decided it was time to upgrade (thanks almost entirely to a donation from a listener!), I looked again for a 'right-sized' CPU, and determined that if it wasn't at least 2.5 times faster it wasn't worth the cost of upgrade (being a good steward, of course).  I settled on the i7-3740QM; while again the 38x0QM's and the 39x0XM's are a little faster (10%-20% at most), the i7-3740QM is a whole lot less expensive, much more obtainable, and not that much slower than even the expensive top-of-the-line E3-1535M v6. (Single-thread: i7-3740QM 1,876 ($199); E3-1535M v6 2,303; Full performance: i7-3740QM 8,339; E3-1535M v6 10,797 (a little less than 30% faster multithread; 23% faster single-threaded).  For that matter, look at the laptop CPU chart again, and look at all the chips below the 8,339 mark, such as the seventh-gen i5-7440HQ.  I'd much rather have near the top end of the third-gen than the rock bottom of the latest generation, which would be significantly slower, but more costly.

The M6700 is easily 3x faster on almost everything than the M6500, and that is the jump I will need to be able to justify upgrading again.  May be a while for that performance level to be reached.

How does it stack up in the real world of audio production?  In my workflow, producing a 30-minute (well, 28:00) broadcast, the export times with later Mixbus were approaching 10 minutes or more with the M6500; I cut that to 3 minutes or less with the M6700.  It really did make a big difference.  My original production machine, a Dell Precision M4300 with a Core2Duo T9300 (a reasonably fast CPU, with single-thread performance of 974 (greater than the single-core of the i7-740QM!) and a multithread perfomance of 1,666, which isn't shabby at all) was taking over 15 minutes to export on an older Mixbus (3.something, I think; 3 years ago now) on that machine, and an upgrade was absolutely necessary for me to be able to get my broadcast production done in a timely fashion.

The Dell Optiplex 745 with the Q6600 and a Toshiba 3TB SATA data drive takes roughly the same amount of time to export as the M6500 does for the long-form broadcast, but for the typically smaller multitrack work I do on the box (3-5 minutes, 8-12 tracks), it's fast enough (less than 5 minutes).  The killer on the 745 is the 4GB limit; I can't do as much multitasking on it.  Newest Mixbus 4 is ok with 4GB, but to run the more sophisticated Mixbus32C really requires more than 4GB (Mixbus32C includes component-level (resistors, capacitors, PC traces, etc) emulation of the Harrison 32C console's EQ strip on every channel; Harrison list quad-core in their minimum requirements for Mixbus32C because of that.  Yes, it sounds fantastic; best DSP EQ I've ever used).

I hope that's not considered to be too long or too off-topic; hope it's useful to some.

--
Lamar Owen
Chief Technology Officer
Pisgah Astronomical Research Institute
1 PARI Drive
Rosman, NC  28772
828-862-5554
www.pari.edu

References: Rivendell curiosity
(Willie Barnett <wbradiolists@gmail.com>, 9 Oct 2017 15:11:07 -0000)
RE: Rivendell curiosity
("Gullikson, Brian F" <bfgullikson@unwsp.edu>, 9 Oct 2017 15:16:08 -0000)
RE: Rivendell curiosity
("Gullikson, Brian F" <bfgullikson@unwsp.edu>, 9 Oct 2017 15:18:00 -0000)
Re: Rivendell curiosity
(Willie Barnett <wbradiolists@gmail.com>, 9 Oct 2017 15:33:59 -0000)
RE: Rivendell curiosity
(<tech@voiceofchristmedia.org>, 9 Oct 2017 16:12:29 -0000)
Re: Rivendell curiosity
(Willie Barnett <wbradiolists@gmail.com>, 9 Oct 2017 16:27:05 -0000)
RE: Rivendell curiosity
(<tech@voiceofchristmedia.org>, 9 Oct 2017 17:23:39 -0000)
Re: Rivendell curiosity
(Willie Barnett <wbradiolists@gmail.com>, 9 Oct 2017 17:43:26 -0000)
Re: Rivendell curiosity
(Lamar Owen <lowen@pari.edu>, 11 Oct 2017 21:48:18 -0000)
Re: Rivendell curiosity
(Sherrod Munday <smunday@ieee.org>, 11 Oct 2017 21:57:37 -0000)
Re: Rivendell curiosity
(Lamar Owen <lowen@pari.edu>, 12 Oct 2017 16:04:19 -0000)
Re: Rivendell curiosity
(Tom Beattie <tom@tombeattie.com>, 12 Oct 2017 16:27:00 -0000)
Prev by date: Re: Rivendell curiosity
(Tom Beattie, 12 Oct 2017 16:27:00 -0000)
Next by date: RDS for satellators
(Scott Todd, 13 Oct 2017 00:55:21 -0000)
Prev by thread: Re: Rivendell curiosity
(Tom Beattie, 12 Oct 2017 16:27:00 -0000)
Next by thread: Re: Rivendell curiosity
(Fred Gleason, 9 Oct 2017 17:19:48 -0000)
CRTech.org