Anybody successfully recording 96 tracks at 96K

[David Spearritt]

Hmmm. Sustained 28 MB per second going to disk for what 3 hours, storing 100GB per hour, bulletproof?

Its a good question.

[d_fu]

Not possible with current DAWs, I'd say, PCI performance being part of the issue, along with disk performance.

[Karl Jackson]

Klark Teknik, in partnership with SADiE, recently announced a box that can do this. Maybe someone using SADiE's MTR software can speak to its stability.

kj

[mpdonahue]

Brian,
We've never had a reason to go to 96 tracks at 96k, but have done 64 tracks at 96k on our Pyramix with a pair of MADI inputs from a pair of SSL Alphalinks. It required a RAID 0 (Striped) but worked without any real brain damage. We use the onboard hardware raid on our Asus P5W DH motherboard and a pair of Seagate 7200.10 500GB drives at SATA-300.
All the best,
Mark

[Remoteness]

Kooster is running 96 tracks at 96K in his new digital mobile RV unit.

He has four Tascam X48s set-up as a dual 96 track system.

Maybe he'll chime in and report on his situation.

[Kooster]

If there is any DAW out there that is bulletproof recording at any sample rate I’d like to see it. All DAW’s are just computers and all computers crash at some time or another (some more than others). The closest anyone can come to bulletproof is to always have a redundant machine running. My Radars have never let me down on a show in almost 5 years but even the Radars have hung on me during non-critical times.

As for my new mobile unit…
The digital chain has been shop tested from the mic-pre thru the fiber to the AES Da, to the X48’s and console all at 96K with all 96 X 2 (192) tracks in record for 3 hours without a hick up.
I have yet to have a client request a recording done at 96K as most of my work is for TV but when it happens I am confident the system will work flawlessly. Again the only reason I am confident is that I have a redundant system running at all times.

Kooster

[Roland]

Originally Posted by d_fu

Not possible with current DAWs, I'd say, PCI performance being part of the issue, along with disk performance.

I'm sorry but this isn't true, Uncompressed HD video runs at 135meg a second sustained, a simple raid system should be able to sustain 28 meg a second. Of course you have to ask yourself how many projects require that track count, it is fairly extreme.

Regards


Roland

[BrianT]

Thanks for all the feedback.

Not to sound like an ingrate, but I have systems that can print the hard disk load without issue. In testing, I've printed 200 tracks at 96K for 4 hours without a problem using Nuendo on hefty Opteron systems using 8 drive RAID, in terms of the comp and hard disk.

Truthfully, I know I can build a machine that will do 96 tracks at 96K rock solid. I have a number of current systems running at 48K with a pair of RME MADI cards for 112 I/O. My burn-in on those machines is 200 tracks at 48K for 6 hours. They have recorded ~100 tracks at 48K for 1.5 hours several times a week for the last 3 years. Failure is just about non-existent, outside of the rare operator error. "Oh, I guess you can even fill up 2TB if you never check the disk". Those systems are now about 3 years old, and I know a lot more than I did then, as well as the state-of-the-art in PCs has moved forward by a big step.

The issue with 96 tracks at 96K becomes how you distribute the data load across the system PCI buses, as somebody above has already pointed out. 96 channels at once in both directions (In and Out) is about 60MB/sec on the PCI bus which is locked down to it's lowest speed at 32 bits/33MHz by the audio cards themselves. Add in any PCI preriphs, like UAD-1, Powecore, etc and you're doomed on a single PCI bus, possibly even spread over 2 PCI buses, depending on the specifics. And you have to make the bandwidth available to both the audio cards and the RAID controller without either impacting the other. To be honest, I've got all that sussed out and am confident I can do the deed.

I'm just looking for anybody who's already there, to compare notes before I build some new systems. I really appreciate you guys taking the time to give some info. I'd love to talk to somebody doing this in the real world on a DAW, but hey, if I turn out to be the first, that's kind of cool too, I guess.

[BrianT]

Originally Posted by Roland

I'm sorry but this isn't true, Uncompressed HD video runs at 135meg a second sustained, a simple raid system should be able to sustain 28 meg a second. Of course you have to ask yourself how many projects require that track count, it is fairly extreme.

Regards


Roland

You're right on the math. My RAID arrays test at about 450MB/sec read or write on a single file, and will do full uncompressed 4:4:4 12 bit HD if necessary (like anybody wants to do that.............).

But it does get much tougher doing audio vs video as the sample rate and track count get higher. Funny enough, I can write more tracks than I can read, when pushing the extreme. Because when writing, the disk controller gets to throw the data anywhere that happens to be open, but during read, the DAW is specifying what data it demands.

The system I'm building as a prototype tests at ~450MB/sec using Blackmagic Speed Test which is a single file test. That same system tops out at writing about 220 tracks at 96K. Which is about 65MB/sec across 220 files. So the penalty, at least on this setup, for writing to 220 files vs 1 is about a 7:1 ratio. Could be worse, I guess.

[dkatz42]

Yep, it's the file count (and thus the head motion) that kills you in an app like this, not the sustained I/O rate. Once we achieve high performance, reasonably-priced solid state media, things are going to rock.

Meantime the adventurous (with a lot of RAM) can create a RAM disk and see what kind of performance they can get.

[BrianT]

Originally Posted by Kooster

As for my new mobile unit…
The digital chain has been shop tested from the mic-pre thru the fiber to the AES Da, to the X48’s and console all at 96K with all 96 X 2 (192) tracks in record for 3 hours without a hick up.

Kooster

You're my kinda guy. I like to beat the heck out of this stuff before anybody needs to use it in actual combat. My rule-of-thumb is I aim for successful testing at 2x the capability that will ever actually be required in use.

[DGaines]

With Pyramix version 6 with MassCore you can record 256 channels at 44.1/48 kHz and 128 at 88.2/96 kHz.
Dennis

[HughH]

Hello,
112 tracks is done regularly at 48k using Nuendo and dual madi systems, sometimes on a single sata drive although most opt for a raid. RME supports 3 madi cards so I suppose at 96k you could accommodate enough I/O for 96 tracks, but you'd better have a backup running.
I agree with BrianT - not really an issue with a good hardware raid on an isolated buss and audio cards on another buss. But again, run with a backup. It's still a computer.
Kooster - good thing you're running those x-48's with a backup. I'd be disappointed in you if you weren't.
H

[mpdonahue]

The next logical topic for this thread is the one of individual files vs. n-channel interleaved files.
RME is supplying it's little recorder application that allows the user to record any number of channels as N-channel interleaves, like Pyramix does with it's PMF file format. (I'm told there is a vicious rumor that there may be support for these files in an upcoming version of Pyramix, any comment Dennis....)
The RME application is really not ready for primetime (Just a little too rudimentary and unstable for my tastes) but offers a look at the future of large track count recording.
With the future enhancements of multi-card recording this could be a very powerful recorder for virtually free.
All the best,
-mark