OT: RME RayDAT Questions

[William Bushnell]

A question for those who are using the new RME RayDAT card. If you had a PCI card and DigiFace setup before, have you noticed an improvement in performance (lower latency)? If so how much?

Also, where did you purchase the card. I obtained my PCI / digiface products from JJRShop on Robert's suggestion but they appear to not be carrying RME now. (only one RME product listed on their web site)

Thanks.
____________
William Bushnell

[Bob L]

Well the driver does go down to 32 samples... and it seems to work so far.

I am using 2 x 64 for the Golden Nugget system.

Just search Google for RayDat and you should find many dealers selling it.

Bob L

[William Bushnell]

Thanks Bob. $849.00 seems to be the standard price.
____________
William Bushnell

[Mogers]

Hi William

I'm using a RayDat with SAC, and also happily running at 2x64. Computer is a Quad Core Intel Q6700, with XP SP3, 4GB RAM fitted (but not all available to Windows), Asus P5Q Pro motherboard.

I did a series of big shows with all 36 inputs being used (4xADAT, plus a SPDIF and an AES/EBU), EQs and dynamics on pretty much everything, lots of JMS RCDelays, a few instances of the SAW reverb, SAW-link recording everything to an internal Seagate SATA drive, another instance of SAWstudio running to do surround playbacks (routed direct to the speakers using Totalmix and controlled using a Frontier Tranzport), plus 16-fader MIDI control of SAC using a Yamaha 01V96 (with the Behringer BCF2000 MIDI template and a converter program that I wrote), and the whole thing ticked along happily at about the 35-40% mark, and not a single dropped buffer for the whole 2-week run.

I was very impressed, and fortunately for me, so were the clients! Nobody mentioned any latency issues at all... I've relied on RME gear for over 10 years now, and the RayDat seems to be as rock-solid as everything else.

It was a fun series of gigs - an official tribute to Syd Barrett - with an extraordinary lighting rig from Pete Wynne-Wilson, Pink Floyd's original lighting designer. Some cool pics and a review at http://www.brain-damage.co.uk/concerts/syd-barrett-festival-the-city-wakes-cambridge-oct-november.html

cheers
Mark

[Bob L]

Way to go Mark. :)

Bob L

[Trackzilla]

So...just to check my understanding of things...why would one choose 2x64 rather than 1x128? Won't they both be the same latency? and although it certainly isn't an issue in this case, wouldn't 1x128 yield slightly lower CPU consumption?
Am I confused or misunderstanding something?

[Bob L]

Because it gives one extra buffer worth of a safetynet in case Windows steps on the engine threads... the same latency... yes... but split into two separate loads... therefore, if Windows steals the cpu for a time, you still have one pre-cached buffer to help get through that period without glitching the audio.

And yes... you will generally start to see the cpu load drop some as the buffer size gets bigger... so which option you select is totally dependent on your system and setup and which gives you the best results overall.

There is no correct answer.

Bob L

[Trackzilla]

Your logic is, as it is so often, impeccable. I hadn't thought of the 'safety net' aspect.

[905shmick]

For those of you running the RayDAT cards, have you run into any problems with them? I'm wondering if I should scrap my original plan of 3 HDSP 9652 cards and go with 2 RayDAT cards. I know this will be only give me 64 channels vs 72.

Is there any advantage to the RayDAT solution other than 32 ports per card? Will it offload any additional processing from the SAC CPU?

Thanks.

[Bob L]

No cpu benefit... although easy to sync internally... the new RayDAT cards also have a nice feature where the clock sync is established as soon as power is applied to the machine... instead of waiting until the drivers are loaded during the Windows boot cycle... helps with keeping hardware converters from floating and spiking during the boot.

The newer drivers also allow 32 sample size buffers... although I still use 1 or 2 x 64 in my systems.

You can also control WDM device loading which helps eliminate the Windows WDM Midi device limit problems while still allowing you to use all channels in ASIO mode with the latest driver.

Bob L

[Bud Johnson]

No cpu benefit... although easy to sync internally... the new RayDAT cards also have a nice feature where the clock sync is established as soon as power is applied to the machine... instead of waiting until the drivers are loaded during the Windows boot cycle... helps with keeping hardware converters from floating and spiking during the boot.
Bob L
My hammerfall 9652 sends clock (MY converters all lock) as soon as the system beep and vga screen appears. Interestingly, when my power supply was slowly dying, at some point it stopped doing that and wouldn't send clock till the win drivers loaded. Now with the new power supply, clock is sent right away again. Go figure....

[Naturally Digital]

Hi William

I'm using a RayDat with SAC, and also happily running at 2x64. Computer is a Quad Core Intel Q6700, with XP SP3, 4GB RAM fitted (but not all available to Windows), Asus P5Q Pro motherboard.

I did a series of big shows with all 36 inputs being used (4xADAT, plus a SPDIF and an AES/EBU), EQs and dynamics on pretty much everything, lots of JMS RCDelays, a few instances of the SAW reverb, SAW-link recording everything to an internal Seagate SATA drive, another instance of SAWstudio running to do surround playbacks (routed direct to the speakers using Totalmix and controlled using a Frontier Tranzport), plus 16-fader MIDI control of SAC using a Yamaha 01V96 (with the Behringer BCF2000 MIDI template and a converter program that I wrote), and the whole thing ticked along happily at about the 35-40% mark, and not a single dropped buffer for the whole 2-week run.Mark, did you need to force the affinity on the processes to even out the load across the cpu?

[Bob L]

In some cases you may find forcing the cpu affinity can lower overall performance and maybe even stability... definitely experiement... I just use the RealTime Priority option inside SAC and leave everything else alone.

Bob L

[Trackzilla]

definitely experiment
on my test rig force single CPU made things worse, and force realtime crashed the box.

on my current machine, SAC force single CPU/realtime improves low latency stability noticeably...although i MUST leave SAW as dual CPU/normal or the rig gets mad & starts dropping buffers...

so my experience so far is, different boxes will be happiest with different settings ;)

[mycorn]

Mark, did you need to force the affinity on the processes to even out the load across the cpu?

as stated you have to see what your rig does

on the box i built
forcing anything is a bad idea...

FWIW

[Mogers]

Mark, did you need to force the affinity on the processes to even out the load across the cpu?

Nope - I didn't do anything at all other than leave everything at the default 'normal' setting.

I tested the rig for many hours before the show weeks (it was a new-build system) and decided on the "if it ain't broke, don't fix it' principle...

[Bud Johnson]

as stated you have to see what your rig does

on the box i built
forcing anything is a bad idea...

FWIW
As my first boss used to say,
"Don't force it, get a bigger hammer"