NVIDIA goes VST?

medusa13 · Post by **medusa13** » Mon Sep 06, 2004 3:59 pm

I found this today.
I thought you might want to give it a look.
Seems pretty impressive, tough.

http://www.bionicfx.com/

Martin

cleanbluesky · Post by **cleanbluesky** » Mon Sep 06, 2004 4:03 pm

Hmmm. I have an Nvidia, not a great one though. Do I hear the upgrade fairy calling me to a world where I can get 120 frames running on half-life 2 at 1280x1080(or whatever)?
If true it would solve all my problems as far as processing power is concerned...
But I think it was a company called Zeosynch (Zeotech?) who promised lossless compression that would solve all my storage problems...
But they didn't...

medusa13 · Post by **medusa13** » Mon Sep 06, 2004 4:08 pm

Actually the link is about audio and VSTs, not video.
Sorry about your video card.

Martín

AndreD · Post by **AndreD** » Mon Sep 06, 2004 4:18 pm

On 2004-09-06 16:59, medusa13 wrote:
I found this today.
I thought you might want to give it a look.
Seems pretty impressive, tough.

http://www.bionicfx.com/

Martin

You may have a look @ the off topic section

medusa13 · Post by **medusa13** » Mon Sep 06, 2004 4:21 pm

Sorry for the repost:)Didn't know.

Martín

cleanbluesky · Post by **cleanbluesky** » Tue Sep 07, 2004 4:28 am

It's not a question of being in the off topic section, if you look at the website, the gent is offering to use the video card to do audio processing.
This has two advantages.
1) A top audio card is much more powerful than your CPU.
2) There is far greater pressure to extend graphics-card speed and technology than there is on any audio cards (that are within reach of consumer prices) - for example, look at the amount of card produced by Nvidia verses the amount of Creamware cards there are. Nvidia are a larger company and can bang things out much faster (not that I am criticising creamware, I think their cards are best on market)

snoopy4ever · Post by **snoopy4ever** » Tue Sep 07, 2004 11:28 am

It's a very interesting news to me.., the whole filosophical concept of audo processing hardware gets an extrange twist. If CPU's were (and are) used to process audio, why not GPU's?.., but what happens then with DSP's??.. could then GPU's be clasiffied as a special kind of DSPs?... if that's so.. why, in the first place CW and other manufacturers (Euphonix, Fairlight, Studer, etc..) did not choose to use these GPUs instead of Analog Devices Sharcs?, Are these individual GPUs more expensive than Sharcs (or other DSP chips for that matter)? ... is there a conspiracy on the video chips market rising??.. will I be bald when I get 35?..

oops.. I crossed the line!

It's important to note that these people point out to some problems using the whole processing power of these GPU's. They hope to be fixed with the new PCI standards.

[ This Message was edited by: snoopy4ever on 2004-09-07 12:34 ]

AndreD · Post by **AndreD** » Tue Sep 07, 2004 11:57 am

remember, uad1 is using a GPU too

[ This Message was edited by: AndreD on 2004-09-07 12:59 ]

astroman · Post by **astroman** » Tue Sep 07, 2004 2:20 pm

On 2004-09-07 12:28, snoopy4ever wrote:
... why, in the first place CW and other manufacturers (Euphonix, Fairlight, Studer, etc..) did not choose to use these GPUs instead of Analog Devices Sharcs?, ...

SFP was designed around 1997 - those GPUs simply didn't exist back then, at least not in affordable form.
The 3D gaming market brought an explosion in power, literally. Check out the specs of the old SGI graphic engines which sold as add-ons for 6-figure $ values in those years

Finally it's all about math and Analog devices have an outstanding library for signal processing - but even in this context the pro audio market is only a side product (and a niche market).
That stuff was developed for radar, ultrasound and telecommunication purposes in the first place.
It's just that certain processing strategies overlap and that makes the Sharcs extremely useful.

A couple of years ago smart developers recogized that graphic processors have a similiar overlap to process data streams, particularily together with 'huge' amounts of high speed Ram onboard.
Convolution is a buzzword in this context - but it's of course not all and everything

A processor is as powerful as it's programs and a Pentium is a rather uneffective 'tool' with an instruction set far too complex to be effectively programmed for typical realtime signal processing.
It burns a couple of gigacycles per second just to get stuck with the data in the IO lines.

It could be programmed much better if people had the time (and expertise), as can easily be observed if you compare the resource-hogging NI Kontakt engine with Tascam's Gigastuff (for example).

my 2 cents, Tom

[ This Message was edited by: astroman on 2004-09-07 15:22 ]

kensuguro · Post by **kensuguro** » Tue Sep 07, 2004 10:27 pm

interesting concept, as I was just learning to program pixel shaders in directX.. (programming the GPUs) Wouldn't remotely understand HOW they can get audio to work on GPUs. Perhaps they're working in assembly level. But man, modern GPUs are hell fast, and they have huge ram, and the video card's internal bus speed is really fast. There's no reason these things can be used as audio DSP. Essentially, a video card IS a DSP card that processes video.

Did you know that through DirectShow (part of directX) you can write a VJ software, with a video source, filters, a video mixer, and then a video out? That's essentially modular DSP processing, and it's all in realtime.

[ This Message was edited by: kensuguro on 2004-09-07 23:28 ]

LHong · Post by **LHong** » Wed Sep 08, 2004 12:39 am

Well said, Ken.
Regards,
Long

spiderman · Post by **spiderman** » Wed Sep 08, 2004 5:04 am

with the new NVIDIA GeForce 6600 GT you can put two of them on pci express bus in SLI mode ( remember the old voodoo ..)

snoopy4ever · Post by **snoopy4ever** » Wed Sep 08, 2004 9:21 am

On 2004-09-07 15:20, astroman wrote:
On 2004-09-07 12:28, snoopy4ever wrote:
... why, in the first place CW and other manufacturers (Euphonix, Fairlight, Studer, etc..) did not choose to use these GPUs instead of Analog Devices Sharcs?, ...
SFP was designed around 1997 - those GPUs simply didn't exist back then, at least not in affordable form.
The 3D gaming market brought an explosion in power, literally. Check out the specs of the old SGI graphic engines which sold as add-ons for 6-figure $ values in those years

Finally it's all about math and Analog devices have an outstanding library for signal processing - but even in this context the pro audio market is only a side product (and a niche market).
That stuff was developed for radar, ultrasound and telecommunication purposes in the first place.
It's just that certain processing strategies overlap and that makes the Sharcs extremely useful.

A couple of years ago smart developers recogized that graphic processors have a similiar overlap to process data streams, particularily together with 'huge' amounts of high speed Ram onboard.
Convolution is a buzzword in this context - but it's of course not all and everything

A processor is as powerful as it's programs and a Pentium is a rather uneffective 'tool' with an instruction set far too complex to be effectively programmed for typical realtime signal processing.
It burns a couple of gigacycles per second just to get stuck with the data in the IO lines.

It could be programmed much better if people had the time (and expertise), as can easily be observed if you compare the resource-hogging NI Kontakt engine with Tascam's Gigastuff (for example).

my 2 cents, Tom

[ This Message was edited by: astroman on 2004-09-07 15:22 ]

Thanks AstroMan.

All this new hardware application brings interesting new posibilities.

It would be nice to start thinking about a standard library for developers using GPU's for audio. At least one per GPU manufacturer.

kensuguro · Post by **kensuguro** » Wed Sep 08, 2004 11:55 am

pretty tough situation.. Video card manufacturers can't even set an agreement for DirectX. But, it'll be nice if they could agree on something. Or if someone built something like OpenGL.

cleanbluesky · Post by **cleanbluesky** » Wed Sep 08, 2004 12:39 pm

On 2004-09-07 15:20, astroman wrote:
Convolution is a buzzword in this context - but it's of course not all and everything
[ This Message was edited by: astroman on 2004-09-07 15:22 ]

I think the biggest advantage of offloading convolution to a GPU would be that it is the MOST resource hogging thing imaginable.
I would buy a top-of-the-range Nvidia just to handle my reverb (I would never buy a top-end one just for gaming) and then use a combination of native and Pulsar effects for everything else. I would then be in complete heaven...

astroman · Post by **astroman** » Wed Sep 08, 2004 2:54 pm

On 2004-09-08 13:39, cleanbluesky wrote:
... the biggest advantage of offloading convolution to a GPU would be that it is the MOST resource hogging thing imaginable...

not exactly

but it's chips and potatoes in graphic processing. NOTHING printed today goes without convolution.

That it's so resource hungry isn't due to the sophistication of the processing but due to the (obvious) inability of programmers for optimized coding.
In graphic programming you can easily have a performance difference of a factor of 50 (times, not %) between a smart and a stupid algorithm.
The Pentium may not be the worlds greatest CPU ever, but you can bet you NEVER used more than 15% of it's potential in whatever application that comes to your mind

cheers, Tom (couldn't resist)

symbiote · Post by **symbiote** » Fri Sep 10, 2004 1:46 am

Mmm.

POVRay uses more than 15%, also compiling stuff takes more than 15%, encoding all those audiofiles to flac to save some space also kind of takes more than 15%, converting the 1 Gb of scanned drawings I had lying around also took more than 15%. Converting those 100 POVRay renders to a smaller animation again took more than 15%. I did a whole reinstall this week, so I ghosted my partition before/after, again more than 15%. Then I RARed those to backup to CD, again more than 15%. Then after all this I needed a bit of unwinding in Wolfenstein Enemy Territory, still more than 15%!

And that's just this week!

On the other hand, I'm still with a p3-1ghz and am perfectly happy doing almost everything. I kind of wish I had a Quad Xeon for the POVRay stuff, but I survive!

I guess a few odd encoding chips could handle most of this without taxing the CPU too much. I guess most of it is possible on a slower machine too, given a bit more patience heh.

I kind of wish all these encoding thingies would come standard on computer boards, like codec and crypto chips.

I'm perfectly happy with my audio setup tho

R-type · Post by **R-type** » Fri Sep 10, 2004 2:47 am

It's a sweet idea, if I was Creamware I would be looking at recompiling their awesome range of plugins to a new platform.

Maybe creamware just don't have resources to develop hardware anymore, so what. Apple took their time but eventually realized that they were a software company like microsoft, and things have improved.

Why not a GPU market?

Since the UAD 1 is an old graphics card you have to admit that a PCI express Geforce 6800 would be an amazing audio super computer.

We all know it's all about quality and NOT power when it comes to music however moving creamware to a new platform is something that must be considered a year or two from now.

cleanbluesky · Post by **cleanbluesky** » Fri Sep 10, 2004 3:59 am

The problem with the GPU would be that the models and makes change so fast. Everytime a new NVIDIA comes out, the plugin may need to be updated...
What might be a good idea would be an VST-Wrapper that you could use to get all your stuff to work on the GPU...

R-type · Post by **R-type** » Fri Sep 10, 2004 4:52 am

What if creamware recompiled and licensed for a nVidia GPU then produced it as a pulsar from then on?

As time goes by more and more boards will ahve multiple PCI express slots and creamware could get back to selling plugins.

Imagine a card that could run any number of sweet as fcuk creamware plugins if only you had bought the keys...