It is not the processing of 24bits, that degrades the real-life-performance (speaking of: maximum voices to stream, maximum number of samples in RAM). You are right, 32bit samples were even far easier to handle process-wise.
The problem is, that any 24bit-sample needs more RAM (to reduce the relative buffer-size of future GS-versions without making the performance/audio glitchy, the seek-time of the HD had to be VERY fast, the PCI-bus has to really kick it etc.) and every 24bit-voice, that is getting streamed from HD needs a higher transfer-rate. So the maximum polyphony and the maximum number of samples in RAM will decrease. I don't see, that the machines will improve so enormously, that they will allow satisfacory results within the next two years. Everything had to get bumped up massively: PCI-transfer, HD-transfer, voice-handling etc.. Currently the best machine doesn't really give us satisfactory performance IMO - with only 16 Bit. And in my experience the only situation where you might hear a difference is when you reprocess and EQ the samples a GREAT amount. And who want's to do that with so very well engineered sounds [;)] ? Leave alone 96k, which I consider the real heavy nitpicking at the olympiade of secondary priorities. The more the better? I don't think so. But of course you need that for an ambitious high-end product: Skywalker will buy it, Zimmer will buy it, and they will order a bunch of those discount 5GHz-machines with RAID that are able to heat your sibirian holiday-castle.

And please don't take me too serious [:)].

Roman

Most important is, that the library is playable and can be managed by common systems.

At the moment we are thinking about a 16Bit/24 Bit Pro Edition Release.
Best option would be, that the user has the option in his production process to switch to a different Bit rate.

Herb

Part of the answer would be more RAM and size configurable Giga buffers. There are motherboards out there that can be populated with a great deal of RAM. This is especially true for boards designed for server applications. I am taking deliver on a system that will hold 4 GB. I know of servers that can populate up to 32 GB.

I agree with you Roman about HD performance - the real bottleneck. I am not certain that the extra byte read per sample will be a big problem considering it is the seek time that is the real performace killer (haven't done the math though). I recently saw a test of modern IDE HD streaming 1 MB of data and it didn't even take up the bandwidth of ATA-33 (it was close though) let alone ATA-66/100/133 or Serial ATA. I wonder if solid state HDs will become affordable before disk-based HD start getting as fast as their interfaces?

You guys are still basing your performance concerns against Giga 2.0, we still ahve to see what kind of buffering setups Giga 3.0 will allow and what kind of performance it will have. Its also possible that if offered a 16 bit and 24 bit version, a user can just replace the 16 bit files with 24 bit.

Or consider another form of audio compression similar to the one that is in giga 2.0, This may make the whoel HD stream issue a moot point.

Again, since the lib is supposed to be released on 3.0, who's to know how it will perform. If there is no real performance hit, then why not?

and no I dont want to get into arguments about facts regarding HD speeds and such, its stupid, since there's just so many things the guys at giga could be doing to improve performance.

ha - did you ever look at the prices for just 200MB-solid-state-disks (_MB_ - not GB)? even assuming we'll get 15K/min ATA-disks and the density of data will increase four times (which i'm sure will not happen this or next year) the random access to moving medias will keep slow.

btw: even if some boards can take 4GB (eg. tyan), a single thread can only use 2GB in a 32-bit operating-system (as windows currently is) and 32 GB are available with the advanced-server-version only - and you would not want to configure this version to be used as a sampler [;)]

edit: king, agree 100%

christian

@Another User said:

KING:
...and no I dont want to get into arguments about facts regarding HD speeds and such, its stupid, since there's just so many things the guys at giga could be doing to improve performance.

I agree with you, we should not be speculating about 3.0's performance and there might be a compression-scheme that would lower the transfers or other nifty re-inventing-the-wheel-features, but, I don't think it is "stupid" to talk about these parameters when thinking about what bit-depths will or should be offered/used. Some of us HAVE to speculate about these things and I appreciate, that the team thinks about offering a "twin-bit-depth"-concept. And:

1. Lossless compressions have certain limits and they increase the processing power needed. I admit, there's a lot of CPU-power to use left, since GS 2.5 doesn't need much of CPU for maximum polyphony. Nonetheless a losslessly compressed stream will not make the data vanish from sight.

2. Buffer sizes will ever compete with HD-seek-times (and transfer-rates) and if you watched the performance of HD's improve over the years you will realize, that they were able to at least double the capacities with every year, but the performance-gain went on much slower. Seek-times almost stayed the same at top-of-the-pops 7200RpM drives over the last two years. And 15kRpM-drives, you know - many of us still don't have a climated machine room with stabilized power supplies yet. Not to talk about the noise they spread by design.

I don't like to argue about what noone knows for sure either, but at least SOME factors seem estimateable to me - without having to be a prophet. There are physical limits no software will be able to overcome.

Roman

eh.. [:)]

I've been there and done that with HD's vs Software, thats the main reason I say its not worht it.

The truth is I've heard many people tell me I couldn't get 160 poly out of a 5400 RPM drive, yet I was able to do it on an internal Laptop Drive, with Directsound DRIVERS no less...

Obviously large streams of 10s of large instruments form this drive will choke it, but my point is, Most people said it would be impossible to get 160 at all.

Until its physically here and tested, its is just not worth speculating IMO. Nick's getting incredible performance on his Kontakt PC using 24bit/88.2 files, and I haven't heard if the performance has improved greatly at lowering the bit rate/sample rate.

BTW, The last Giga Format Lossless conversion can nearly half the size of a library.

@Another User said:

btw: even if some boards can take 4GB (eg. tyan), a single thread can only use 2GB

Microsoft sell a 64-bit Xp and 2000 (server, technical end users). If Nemesys designed their code well, they should be able to change their standard int to 64 bits, do a quick recompile and walla - 64 bit addressing [[;)]] Now that you mention threading, it would be interesting to consider a design of GigaStudio that launches multiple thread. GigaStudio is actually similar to many telecommunication applications (i.e. receive, process, transmit) which typically allocate input/output/data manipulation capabilities into multiple threads. Of course, these systems use real-time operating systems which is where my knowledge is, not Windows. Luckily, today's GS can run perfectly on inexpensive PCs.

@Another User said:

Nick's getting incredible performance on his Kontakt PC using 24bit/88.2 files, and I haven't heard if the performance has improved greatly at lowering the bit rate/sample rate

Yes, I've heard that. NI told me about and I read it somewhere on northernsounds, but I wonder what he has been setting the buffer size to. I gave Kontakt a whirl recently on one of my well optimized GS-machines and I had to set the buffer-size so ridiculously high to get satisfactory streaming-results, that a single Gigapiano ate up more than 200MB of RAM (!!??!) This can't be it. I will have to check that with the NI forum sometime, I guess...

Roman

Actually the 160 was with GigaPiano and Cubase playing the MIDI stream. Piano's are notorious for choking systems. I honestly haven't pushed the system just because I dont want to clog up my HD with Gigs. But it would be interesting to see how much poly I could get on this thing with just Giga running on it. I bet it would choke if I played lots of 13th chord 16th notes with release trigggers [:)],...but it didn't seem to really choke, just steal notes, which is basic Giga performance.

I've since made this computer a more "internet" computer, so I doubt it will perform all that well, with all teh fragmentaion and apps that steal memory or have memory leaks. I still use it to do Giga editing tho, and audition sounds over the network.

Anyhow, I'm intersted in seeing actual performance in front of me, or detailed documented performance. Thats all that I worry about. Speculation just leads to needless worrying. Leave that to the develoepers. I mean, all my ideas are things I'd love to see in Giga (or soft samplers in general), I have no idea whats going on with the streaming engine. I've heard rumblings of other cool features, but nothing on how it will perform. If I keep spculating, I could just be setting myself up for disappointment. Still not giving them the benefit of the doubt to take on performance issues, is a bit of a defeatest POV.

Anyhow, just because the developers WANT to release in higher sample rates and bit rates, it doesn't mean that they WILL. If performance takes a major hit I doubt any developer would weigh the benefits of higher sample rate higher than perfomance. Or atleast, not make the choice for us, but give us an option.

I sincerely think that no developer is tyranical enough to force us into something different sample rates if the performance hit is strong. Most should realize that many people dont even have Sound Cards that support higher than 48k, so unless Giga has somesort of internal rendering solution that doesn't require the Scard then its a waste in the giga format (VSTi's are a different story, as it can be all internal).

Lastly, as more options like logic's "freeze" show up, performance will stop being realtime. I mean NO ONE works on 100 MIDI channels all at once. You may have 100+ MIDI channels going, but you never work on every one at the same time. Its physically impossible. You usually work on one section of an orchestra at a time, or you work on one section of a cue at a time, the smarter software gets, the less need we have for a truly realtime "solution". This isn't speculation BTW, the concept of softsampling and VSTi's is still relatively new, so the sequencer developers never anticipated the ways users would be using them. They are beginnning to realize it and I'm sure will start to work on "smarter" ways to do things....like "freeze" (which may be over rated since I'vn't seen it in actual use). These types of features will become standard. (Tho I doubt it will be native in giga, since they are so focused on realtime playback all the time)

Anyhow,...point is... Its going to be about whether its worth it at time of release, not whether its worth it "right now", and time of release of course does mean what they know is appearing in the very near future. Whether its custom software or through relationships with other companies.

@KingIdiot said:

Lastly, as more options like logic's "freeze" show up, performance will stop being realtime. I mean NO ONE works on 100 MIDI channels all at once. You may have 100+ MIDI channels going, but you never work on every one at the same time. Its physically impossible.
Anyhow,...point is... Its going to be about whether its worth it at time of release, not whether its worth it "right now", and time of release of course does mean what they know is appearing in the very near future. Whether its custom software or through relationships with other companies.

King,

I think this abolutely makes sense, but Logic can only frezze audio tracks and audio instruments but NOT MIDI tracks. These have to be bounced to audio first.

Iwan

I kept reading that it would freeze VSTi instruments.

if its only audio tracks then thats crap!!!!! Everyone's been telling me it would freeze MIDI tracks. Goes to show you that I cant believe what everyone says, and I NEED TO SEE IT FOR MYSELF, yet again

Anyhow, I still believe that this type of route is being looked at by people. I've been bugging the guys at nemesys on and off for a "render" option.

I'd imagine that with VSTi's I'd be able to write keyboard shortcuts that would render all but whatever track I'm working on off line, while I get coffee. And short cues can be rendered and replaced via cue points, so it wouldn't mean coffee, maybe a rub of the eyes.

I just haven't gotten into an orchestral VSTi solution yet, because there aren't any options for me (EXS isn't since I dont use Logic)

Kontakt looks good, and with QLSO I'll be able to put my noggin banging to the test.

damn these inconvenieces... I mean I'm trying to write music here, who cares if I dont have to pay for one musician, or hear them fart (smell?).... was that sarcasm....who me? NEVER!! [;)]

@KingIdiot said:

I kept reading that it would freeze VSTi instruments.

if its only audio tracks then thats crap!!!!! Everyone's been telling me it would freeze MIDI tracks. Goes to show you that I cant believe what everyone says, and I NEED TO SEE IT FOR MYSELF, yet again

King,
it does freeze VST instruments (in Logic called Audioinstruments), but not MIDI tracks. Believe me, I do have Logic 6 right in front on my monitor right now, and there is no frezee option for MIDI tracks! Also if you do use Logic 6 on OS X. you will have have no VST support at all!!!!!No Waves, no Altiverb etc...

Strange, No?

Iwan

oh wait...

it does freeze EXS? Well thats MIDI. Or am I misunderstanding something?

External stuff liek Gigastudio is impossible, unless its gigastudio that does the "freezing" and for that to happen it would have to be more incoorporated to the software.

@Another User said:

Also if you do use Logic 6 on OS X. you will have have no VST support at all!!!!!No Waves, no Altiverb etc...

Both Altiverb and Waves will be supported in audio-unit format and thus will work in OS X with Logic. Altiverb is scheduled for next month. I don't know the schedule for Waves, but have seen a press release that they are committed to support this format.

Audio-units is a new plug-in format in OS X that is more capable than VST. Emagic has taken the position to only support audio-units in OS X to enhance that format's chances for success. They still support VST in OS 9 so that as the migration to audio-units happens people can maintain compatibility with existing plug-ins.

You can read more about OS X and audio units from Emagic here:
http://www.emagic.de/support/osx/index.php?lang=EN

@KingIdiot said:

oh wait...

it does freeze EXS? Well thats MIDI. Or am I misunderstanding something?

External stuff liek Gigastudio is impossible, unless its gigastudio that does the "freezing" and for that to happen it would have to be more incoorporated to the software.

King,

Hmm, Do not ask me to much......But in Logic their are so called MIDI tracks, which you can not freeze and their are so called Audio Instruments tracks (ESX24/ ES1/EVP88, etc...), and audio tracks, which can be freezed.

Iwan

@Another User said:

The current release version of Altiverb does not yet run in OS X. We hope to put new versions out within two months

I did receive this Email message from Audioease two weeks ago, so I hope it is your info which is rigth!

Iwan

Logic's 'freeze' does work with EXS. but it does not free up the RAM demands of that EXS instruments - those remain until a different instrument is instantiated, which can't be done to a frozen track. in that respect, it does not provide a great deal of help for VSL/EXS use.

cheers

bah crap!

[:)]

@Another User said:

The current release version of Altiverb does not yet run in OS X. We hope to put new versions out within two months

I did receive this Email message from Audioease two weeks ago, so I hope it is your info which is right!

I was trying to be optimistic about the Altiverb schedule. But quite frankly, I've been hearing the "within two months" story since last November. So it seems they haven't made any progress since then if that is what they are telling you.