does Magma chassis increase the Scope latency?

lagoausente · Post by **lagoausente** » Wed Mar 08, 2006 1:45 pm

Well, the truth? in theory many people may think, it works ok, but have someone tested it?, doesn´t have the cardbus controlers hidden buffers?,
anyone have tested if there is any really differences beteween a laptop or desktop setup, or could anyone make the test?

[ This Message was edited by: lagoausente on 2006-03-08 13:46 ]

dehuszar · Post by **dehuszar** » Wed Mar 08, 2006 5:51 pm

I not only tested it, it was my working setup for over a year.

You will take a latency hit, but as more of a consequence of a more significant caveat. Most laptop's system busses simply aren't wide enough to handle the amount of data being pushed to the system RAM.

What this means is that the amount of tracks you can push through will be seriously ham-strung (think 10-16), and more demanding devices like the P100, uberPlastic, echo3/blackbox, D-Filter, etc. you will stress the bus to the point where the Scope OS will literally error out, forcing you reduce voices, raise ULLI settings, etc. just to get sound to come out of the device again.

Not for the impatient, and would be a tremendous liability to any potential mobile rig.

The Magma itself is more than capable of handling the throughput. There are many mobile ProTools rigs which use the Magma while only having it's performance lightly trimmed.

Why is this not the same with Creamware cards? When the Creamware cards were first conceptualized, it was way too expensive to use SHARCs with enough onboard memory, and impractical to bolt memory on the cards considering where the audio industry was at that time.

ProTools made different choices, ProTools is also tremendously more expensive. The landscape is different now. Hopefully whatever nextgen PCI cards CWA might intend to release will have those capabilities in place.

But until then, for unbridled production, a laptop is not really a suitable environment. Maybe someday there will PCI-Express based cardbus controller for the Magma and port for the laptops with a big fatty pipeline straight to the memory controller, but we're likely to see a new CWA card before that starts happening.

You might be able to get a decent rig together for specific live stuff, or as a synth box. But if you're hoping to open up an ASIO sequencer, run a lot of voices on certain synths, stack several demanding effects or stream a lot of simultaneous audio channels for all the costs involved, you're better off putting the $800 into a custom shock mounted rack case & a workstation tray (LCD and keyboard on a roll out rack-mountable tray)like they use in server farms. You won't be able to throw it in a backpack (which IS nice), but you would be able to knock it around a bit and you can still put a handle or two on it to make carrying it a little less awkward.

I'm not saying it can't be done and I'm all pessimistic about it, but you need to be aware that there are serious considerations and you will want to do a LOT of research to match a laptop chipset currently available to your Magma rig, and know EXACTLY what you hope to get out of the package deal.

Feel free to do a search of the Z to see me and a few others hashing out the details which fed the above comments. At the very least it will give you an idea of what was involved given the where the laptop market was two years ago. Motherboards and chipsets will have changed quite a bit since then for better or worse.

My $1.50,
Sam

husker · Post by **husker** » Wed Mar 08, 2006 11:18 pm

Wow...that's a great $1.50 worth

My new dual core ASUS laptop has ExpressCard54 (no PCMCIA!), FireWire, USB2...

It will be very interesting to see what the future of Creamware will plug into...

lagoausente · Post by **lagoausente** » Sat Apr 01, 2006 9:38 am

Well, I´m coming again, after some exams.

Really I will not use lot of dps, I would need most of time, one STS, and perhaps one reverb for monitoring, and that´s all. But the dubt is still on the air. I not wonder, how much sofware i could run through a Magma, instead I just want to know, if only one STS sampler would have any more latency through a Magma than in a desktop mainboard. I would need a real test, what I think would be interesting for the knowlegde of the forum.
The controller you tell about, can limit your bandwith, the question is if there are buffers that increases the latency, and if are, how much.
If, for example, on a desktop, have 1.5 ms , and on a laptop 1.7, would be unapreciated, but if you have 6 ms? that would be no so pretty, so I would just know it.

Shroomz~> · Post by **Shroomz~>** » Sat Apr 01, 2006 11:23 am

In 'normal' studio use why would you need to use a reverb for monitoring exactly?

GBauwens · Post by **GBauwens** » Sat Apr 01, 2006 4:49 pm

To give you an idea of real figures, I use mine at 4ms flawlessly with Sx3.
I remember computing the troughput between the chassis and the laptops every time I had a new one. It's chipset dependent, cardbus controller dependent and OS dependent. All laptops aren't equal for sure.
I can have up to 64 ins and outs simultaneously (sts stereo voices and asio drivers) at 32 bits between the card and cubase before getting the message "PCI limit reached".

I tried 3ms but it starts to cracke when my cpu reaches 10 to 15%. It's a pentium4 so it hardly handles buffers under 128. Centrino and Athlon64 will do 64 and even lower apparently so I suppose you could use them at 3ms flawlessly.

Obviously there is a lot of tweaking that may have to be done on a laptop to get to what I have accomplished for mine but I don't want to overwhelm you with technical details for now.

[ This Message was edited by: GBauwens on 2006-04-01 17:39 ]

[ This Message was edited by: GBauwens on 2006-04-01 17:40 ]

lagoausente · Post by **lagoausente** » Sun Apr 02, 2006 7:29 am

Are you telling 3 ms about ASIO?,
I think STS sampler can run without ASIO, with an unapreciated latency. That´s what I´m interested on.
About the reverb, I would decide to want a reverb just went I like it, as simply as that. Maybe I want to sing with a reverb at at time, or maybe not.

Mr Arkadin · Post by **Mr Arkadin** » Sun Apr 02, 2006 8:13 am

Really I will not use lot of dps

(sic)

Hee hee, that's what they all say at the beginning...

GBauwens · Post by **GBauwens** » Sun Apr 02, 2006 8:44 am

Yes I am talking about your ULLI settings which is the speed at which your card will communicate to your CPU and vice versa. It affects ASIO, STS voices and any plug-in that needs to access your system memory i.e. reverbs, delays, etc. It's measured in buffer sizes as exponent of 2 X 64 i.e. 2^0 X 64 = 64 then 2^1 X 64 = 128 then 2^2 X 64 = 256 ...etc...etc.
Your ULLI window approximately matches these buffer sizes with the time it takes for your audio data to travel back and forth (in ms).

GBauwens · Post by **GBauwens** » Sun Apr 02, 2006 8:52 am

Another thing about STS. 16 bit stereo voices are interleaved so that in only uses one 32 bits burst (2X16=32)to send 2 voices from your system memory to your dsp card. I believe it's the same for 16 bit asio drivers but if you use 24 bit asio drivers, one 24 bits voice will use one 32 bits burst and the 8 bits left are just unused.
So at 16 bits vs 24 you can double the number of voices on your available PCI bandwidth.

GBauwens · Post by **GBauwens** » Sun Apr 02, 2006 8:56 am

Last, in order to answer your original question

does Magma chassis increase the Scope latency?

The answer is no.

lagoausente · Post by **lagoausente** » Sun Apr 02, 2006 9:02 am

does Magma chassis increase the Scope latency?

The answer is no.

Well, I think I have read on the forum were the most latencies were about 1.5 ms o less. why do you have more?

have a look here: http://www.planetz.com/Pulsar/PerformanceMIDI.html

[ This Message was edited by: lagoausente on 2006-04-02 12:13 ]

GBauwens · Post by **GBauwens** » Sun Apr 02, 2006 12:56 pm

Your link relates to midi jitter which has nothing to do with latency.

The purpose of the test mentionned is to check the midi accuracy when playing notes.

You get confused when they say latency and actally mean deviation or jitter. It's inherent to Microsoft's opeating systems but some audio cards manufacturers handle the problem better than others. Steinberg has supposedly adressed this issue with their proprietary midex interfaces.
As far as midi on Creamware's boards, well I haven't measured it myself but I guess your link shows someone else did it. Always good to know !

lagoausente · Post by **lagoausente** » Sun Apr 02, 2006 2:33 pm

As I understand, it not only measures jitter latency, even not midi latency, instead the "synths" latency.
Till, I know if I have undertood ok, Creamware takes advantage of latency, since you sent the midi data directly from the in, to the synth, and the audio data, directly to the mix output, without going to windows. I suppose that if you route ok, you will never need anything about windows, unless the latency of the midi track on the sequencer, but not for the realtime playing. I think I´m nor wrong, if I´m please correct me.
And I assume, that the STS sampler would work the same, with the only difference that requires to take the audio from ram, but, even here, not windows at all for live playing.
So, having in account that I´ll have a 3 dps card, would I strees the pcmcia bandwith?, and i still have the doubt, that if I´m not stressing it, will come so quick the audio from RAM to my ears as 1.5 ms o less using a Magma? , Perhaps I finally will have to Buy a Magma only to can test it.

astroman · Post by **astroman** » Sun Apr 02, 2006 4:29 pm

it depends... if you lean back comfortably in your chair this will give you (at least) one extra millisecond of latency

(1 m distance equals 3 ms)

imho the ULLI setting affects only ASIO latency - all other sources are (almost) realtime.
I have an ULLI setting of 25 ms (required by the TripleDat plugin) and I don't care because all sources are (time-)aligned BEFORE the monitor and recorded BEHIND it.
I record with VDAT and edit with Triple.

cheers, Tom

GBauwens · Post by **GBauwens** » Sun Apr 02, 2006 4:43 pm

I think you understand pretty much now. If you are playing sound straight from the DSP's (i.e. B2003) "latency" is only the time it takes to produce the sound and output it from the card (between 1ms and 2 ms depending whether you're using digital or analog outputs. D/A conversion takes a little more time). Just like a hardware unit (synth).

If you're using STS you will have to add ULLI latency to that (between 3 and 25 ms at 44.1KHz).

Practically, at 7ms or below you will not notice any delay when playing notes on your keyboard. At 25ms you will.

Now the PCMCIA bandwidth is not ULLI latency/buffers dependent. It's the number of voices you intend to play from ram which is important. With STS 4000 limited to 32 voices you can hardly max it out.

Hope it helps.

By the way I might sell my magma chassis. If you're interested E-mail me at bauwens@eurovision.net

[ This Message was edited by: GBauwens on 2006-04-02 17:44 ]

[ This Message was edited by: GBauwens on 2006-04-02 17:48 ]

alfonso · Post by **alfonso** » Mon Apr 03, 2006 1:34 am

On 2006-04-02 17:43, GBauwens wrote:

If you're using STS you will have to add ULLI latency to that (between 3 and 25 ms at 44.1KHz).

I can't confirm that. I commonly work at 25ms, because i only use Scope stuff, in SFP environment. STS responds exactly the same with all the ULLI settings.

Post by **garyb** » Mon Apr 03, 2006 9:13 am

that's right. there's only latency when the midi signal travels through the sequencer.

lagoausente · Post by **lagoausente** » Mon Apr 03, 2006 10:06 am

On 2006-04-03 02:34, alfonso wrote:
On 2006-04-02 17:43, GBauwens wrote:

If you're using STS you will have to add ULLI latency to that (between 3 and 25 ms at 44.1KHz).

I can't confirm that. I commonly work at 25ms, because i only use Scope stuff, in SFP environment. STS responds exactly the same with all the ULLI settings.

First, seems there is more confusion that real knowledge. I think that Alfonso have reason on that, the ULLI settings for Asio, have nothing to do with STS.
In the other hand, perhaps 1 ms, well is just a great suposition, based on?
I have to think that the bandwith tells about how many bytes per time can run across a bus, so when you tell about the number of voices you can run, or the number of effects, you have reason, so more o less bytes maybe run. But in the same way, a bus who can run more bytes per time, it´s logical to think that each byte would travel quicker, I´m disorientaded here?
As well, the own Magma, or the pcmcia bus may have a buffer, that can increase the time a byte cross the bus, but this buffer don´t need to be a kneck about how much data per time is managing.
About STS, is diferente from a synth, since data must come from ram to the card, while on the synth doen´t occur.
My insistence on latency use, is because I want the STS to live playin drums, where 7 ms is much more detectable that playing the keyboard, and where the "time" is esential, and of course, no quantition I like at all.

Gordon Gekko · Post by **Gordon Gekko** » Mon Apr 03, 2006 11:15 am

At the very least it will give you an idea of what was involved given the where the laptop market was two years ago. Motherboards and chipsets will have changed quite a bit since then for better or worse.

My $1.50,
Sam

hey man, you thread is very interesting as you came up with pretty good advice

maybe you guys were a bit ahead of the time

now there are laptops out there that can do wonders