RAID Question

[LvR]

I must surely be missing something here:

Write multiplexing is NOT part

How does one then go about doing a "multiplexed read" if you are not allowed to create it in the first place?

I am surely seeing "multiplexed" writing and reading on the raid cards I have.

What do we call this mode of operation then if its not part of the spec?

As far as the realtime compression goes - well thats up to your needs I suppose. If you can live with the quality produced by single pass software encoders, then mpeg1 and 2 is possible and even DivX given the nice fast processors we have nowadays.

The best is possibly Huff (needs around 3 times less space than raw YUY) if you intend to also edit the material.

[Dogbert]

The best is possibly Huff (needs around 3 times less space than raw YUY) if you intend to also edit the material.

Doesn't it also mean you'll need a third of the transfer rate ?

[LvR]

Thats true as well, and just about any disk nowadays should be able to handle that (about 10M/s) datarate without being hooked-up to a raid.

[Gurm]

I'm pretty certain that DVD-Quality MPG2/Divx can't be compressed real-time. If it CAN, it requires something approaching a 2GHZ chip... and that isn't the best quality.

Perhaps I used the wrong word earlier...

What isn't part of the original RAID specs is multiple writes happening across multiple discs "simultaneously". Only multiple reads. The multiple writes is something added in by many manufacturers since RAM and throughput have become plentiful nowadays.

- Gurm

[Indiana]

All available mainstream IDE RAID controllers (be it Promise or Highpoint) are doing multiplexed writes, so the point if it is part of the specs or not is really moot.
Still he is getting too low throughput considering the HDs he has. I'd try to change/update the driver for the RAID and look for IRQ conflicts.

Uncompressed PAL full-res video takes about 22MB/s which is a bit hard on a normal single IDE drive since this is a sustained 22MB/s needed (in contrary to the max. speeds those HD benchmark programs normally are telling you).
With HuffYUV this is cut to somewhat less than half: 7-10MB/s, depending on the video material, so you would not really need a RAID. Still the extra headroom a RAID provides is fine if you want to use your computer for other things while capturing.
At least an oc'ed AthlonXP1700+ to 1900+ speeds is not fully up to the task of DivX'ing in the highest quality setting in realtime, you'd need about a 2GHz (real clockspeed, not AMDs rating) AthlonXP for this.

[Dr Mordrid]

Originally posted by Gurm
I'm pretty certain that DVD-Quality MPG2/Divx can't be compressed real-time. If it CAN, it requires something approaching a 2GHZ chip... and that isn't the best quality.

Perhaps I used the wrong word earlier...

What isn't part of the original RAID specs is multiple writes happening across multiple discs "simultaneously". Only multiple reads. The multiple writes is something added in by many manufacturers since RAM and throughput have become plentiful nowadays.

- Gurm

MPEG capture

Wrong.

Software as lowly as Ulead's VideoStudio 6 can encode video streams in realtime to MPEG-2 using MPEG.now at DVD specs (6000+ kbps) with CPU's as low as 1 ghz. That said with faster CPU's you can definitely use higher quality settings, but don't say it can't be done.

Even before that one could capture MPEG-2 for DVD using VideoStudio 5.x with a card like the Hauppage WinTV Theater. VS5.x at that time used the Ligos realtime MPEG encoder.

There are numerous dedicated video to MPEG capture tools for sale as well.

As far as DivX goes I was capturing it in realtime using my Marvel G400-TV and a PIII/850 up until just a few months ago when I upgraded many of my systems with AthlonXP's.

History of RAID:

Read this article for more info on its development by IBM to the current day;

http://www.robdata.com/data_files/raidintr.htm

Read the portion on Data Striping. Interleaved reads and writes have been part of RAID from the start because most all RAID levels depend on striping and striping is done by interleaving the writes for performance.

Data redundency and if RAID0 is technically a RAID because of its absense is a separate issue. IMO the name RAID has evolved into a generic term for any striped array of disks and is no longer discriptive of just arrays with parity.

This is certainly how it's being used in everyones product literature and in the media, so arguing the contrary is likely an anachronism.

Dr. Mordrid

[Indiana]

Doc, I belive Gurm meant that it's not possible to do on-the fly mpeg2 or DivX compression with highest quality settings and there he is correct.
If you want to use OTF software compression with these two formats you have to take some of the settings to lower quality values.
Although for DivX an AthlonXP2400+ or a 2.4GHz P4 could probably do the job.

[Gurm]

Hmm... I stand corrected on RAID. I may be wrong about DVD compression, but I seriously doubt you could take the raw bitstream for, say, Harry Potter and spit out a DVD stream in realtime at full quality.

- Gurm

[Dr Mordrid]

I agree that the highest quality is obtained in non-realtime, especially when motion compensation is an issue, but things are getting better on this front.

One can capture very good quality MPEG-2 in numerous ways. With my AthlonXP 1500+ system (the slowest one here) one could easily;

1. capture analog using VideoStudio6's MPEG.now by S-Video input (higher quality from the get-go) to its DVD presets using the WinTV Theater or any other analog card that provides a YUV signal (most).

Settings matter, and the Performance one is critical. If you can get 8 or better you get quite good quality. Most of my systems can do 10-12 with a YUV input.

2. capture analog (S-Video) using the RT-2000's MPEG-2 capture mode at any number of bitrates whereupon it can be burned using the included DVDiT.

3. capture analog (S-Video) using the RT.X100's *m2v capture mode using any number of bitrates. These *.m2v's + their associated *.wav are all ready to burn to a DVD after the capture, again using DVDiT or whatever.

4. capture using the Ligos encoder in MSPro 6.5 from the WinTV Theater to MPEG-2 DVD presets at any number of bitrates (Ligos will be a bit lower in quality than the MPEG.now in VS6 though) and then burn it.

5. capture analog (S-Video + YUV card again) to DVD Workshops MPEG.now realtime capture where it can be immediately be authored and burned without editing software. The settings are similar to those in VS6.

and with the AthlonXP 2000+ or faster systems it's even easier.

The latest DIVX isn't a problem either, but does require a bit of a faster system. Still it's not exactly a high $$ affair, but who would want DivX anyhow? To my admittedly jaded eye it's quality isn't any better than *.wmv, which means it sucks, no matter how it's done.

Windows Media Corona will set the standard in the MPEG-4 arena once it's released because it will provide profiles capable of lossless editing as well as streaming.

Dr. Mordrid

[Gurm]

I find high quality DIVX to be comparable to average quality MPG2. That's my unjaded eye.

- Gurm

[Dr Mordrid]

DocM hands Gurm a prescription for an eye test.....

[Gurm]

I dunno what MPG2's you're used to, but if you take your "average" DVD (aka most porn, stuff from smaller studios) it is kinda crappy. Blocky, pixellated in places. DIVX, on the other hand, can be quite nice. My DIVX of Lord of the Rings is indistinguishable from DVD quality on a TV.

- Gurm