PDA

View Full Version : Video and disks



Tylor
07-31-2001, 01:56 AM
I'm curious about whether video rendering, particularily compression, works with a lot of small pieces of data and reads/writes them to the hard drive a lot, or does it tend to work with larger chunks of data, and/or not requiring as much disk access?

Michael
07-31-2001, 02:08 PM
i assume that your question is motivated by what kind of HD to buy. video, like audio records in large chunks of info. so, an HFS+ volume is no real advantage here. (i don't use it on my audio/video work drives) access speed is definitely a factor however. the quicker the info can be accessed, the smoother your video will look. so get the fastest, largest HD you can afford. LVD scsi drives are still king here though some of the new faster IDE drives will also work.

Tylor
07-31-2001, 03:50 PM
Yes, that's the reason behind this. Initial capture will be to a SCSI RAID. For anything outside of cut and paste though, I prefer to work with the video footage uncompressed, which takes up masses of space. I also plan to store a number of finished files in high quality, which will eat up another large chunk. I'm trying to find out if filters and recompressions would be significantly slowed down by using ATA drives compared to SCSI?

professor
08-06-2001, 10:04 PM
what kind of software/system will you be using? either way you would be happier with scsi drives. go ahead and get a dual channel card to stripe your drives for uncompressed resolution.

usually rendering is written to the drives one frame at a time.

TZ
08-07-2001, 01:46 PM
In another month Acard with have hardwaree RAID ATA/100 cards $199. There are external FireWire RAID solutions also. And at $200 for 60-80GB of IDE it is getting easier and cheaper to justify ATA interface. For serial recording of video, they should be fine. And maybe a scratch SCSI RAID for manipulating.

DHB
08-08-2001, 04:03 PM
Your question is something I posed to my son when he got going on his miniDV-to-the-Mac-burn-a-VCD-and-make-it-work-on-the-DVD-player-in-the-living-room project.

He had to render (correct term?) several MPEG-1s for this and they usually ran all night. I asked if his Mac spent most of its time rendering or thrashing the disks. He said there was little disk activity compared to the amount of time spent rendering.

Sounds like a good candidate for a 1.5Ghz G4...ALTIVEC is goodness here. http://macgurus.com/infopop/emoticons/icon_biggrin.gif

The pieces involved: PT Pro, G3/500mhz (10x), 256M ram, VST ide card with two Maxtors (25G & 40G). Boot spindle an old 2G Seagate 'cuda. I don't recall what he was rendering with but I know he really piled on the memory allocation for it.

As far as ata/66+ vs. SCSI (LVD and up)...
For comparisons:
Last night I benchmarked a IBM 75GXP (ATA/66, 15G) I brought home from the office against an Atlas 10k II 36G U160. The IBM drive was on a VST ide card (ata/66) while the Atlas was purposely hanging from an Adaptec Powerdomain 2940u2w (a LVD card).

Each drive was freshly formatted via FWB's 4.5 HDTK. The Mac using OS 8.5.1, booted off an internal Seagate 'cuda (2G). The system was a PowerCenter Pro with 64M of ram, a 604e/210mhz cpu, builtin video. These were the only two PCI cards installed.

The results were rather surprising:
using Express tools (the benchmark of choice around here)
.........Sustained Reads.......Writes (MB/second)
.......IBM 75gxp....26...........12
.....Atlas 10K II....53...........31

(there, I finally figured out one way to do tables... http://www.macgurus.com/ubb/rolleyes.gif )

I intend to throw one of the sweet 10x G3's I got from the Gurus in it to see if the IBM drive's performance will 'scale up' with the boost in cpu performance (the Atlas may well improve too) as 'ide' drives generally do in the PC world. Others here may be able to shed more light on that scenario.

Note that the IBM (75 or 60gxp flavor) is considered to be noticeably faster than the fastest/biggest dawgs that Maxtor has (like the 40G my son has).

I had tried the Atlas on a 29160N (a U160) Adaptec board earlier this week with the same benchmark results. The theoretical potential is much higher for this board/drive combo. Once I throw the G3 in I'll retest both scsi cards to see if I'm cpu or FSB-speed bound in this benchmark.

The 8M onboard cache of the Atlas may skew the results a bit since the max on the benchmark is 8 meg as well. Quantum quotes a sustained 41 MB/sec performance number for the Atlas 10K II. Thus, the sample size used is suspect.

I'm thinking of pushing the 10x G3 towards 600mhz in the PC Pro just for grins http://www.macgurus.com/ubb/eek.gif with some 50ns 128M dimms I have in hand. Reaching 540mhz+ and building a Raid pile with some Atlas 10K IIs (I have four currently) would be an interesting scenario.

Moving all the good pieces over to a PT Pro and rerunning everything would give some real insight into FSB-speed vs. memory interleaving vs. cpu speed...yes, I see a multi-dimensional spreadsheet with all the results. I also have the SoftRaid drivers that came with the VST card to try out.

One of these days I'll actually start doing some real work with all these toys...errr high tech tools. http://macgurus.com/infopop/emoticons/icon_wink.gif


[This message has been edited by DHB (edited 08 August 2001).]