Doesn't SGI make their own graphics boards for their high end visualization systems, called Reality Engine and Infinite Reality? My Onyx has a Reality Engine 2 chipset inside of it, and if you look at the specs on the boards it is quite impressive.
Maybe Apple is planning on doing something like this, taking a page from SGI's book. After all, that's part of the reason SGI systems are used so much for visualization...
The library you're talking about provided extra speed by essentially storing a table of values for common math procedures (e.g. sin, cos). Rather than having to compute the value, the computer could just look it up in the table. I'm not sure if it interpolated or used the closest value when there wasn't an exact match available; either way, there'd be some inaccuracies. Furthermore, this led to other problems; for example, I think tan wasn't implemented, so there were two options: look up sin, look up cos, and divide, or calculate tan the hard way. You got different results depending... That's why the library made network games go all screwy; you'd have one value for tan on your computer, the server'd have a different one, and you'd be out of sync.
The library did improve speed, though, and for many purposes the inaccuracies weren't a big deal.
<strong>The library you're talking about provided extra speed by essentially storing a table of values for common math procedures (e.g. sin, cos). Rather than having to compute the value, the computer could just look it up in the table. I'm not sure if it interpolated or used the closest value when there wasn't an exact match available; either way, there'd be some inaccuracies. Furthermore, this led to other problems; for example, I think tan wasn't implemented, so there were two options: look up sin, look up cos, and divide, or calculate tan the hard way. You got different results depending... That's why the library made network games go all screwy; you'd have one value for tan on your computer, the server'd have a different one, and you'd be out of sync.
The library did improve speed, though, and for many purposes the inaccuracies weren't a big deal.</strong><hr></blockquote>
This is how trig functions used to be done, back in the days when memory was fast and processors were slow. In the last 8 years or so, however, the move has been away from table driven algorithms to computational algorithms... to the point where AltiVec algorithms sometimes use several instructions just to build a constant in order to avoid loading it from memory. Fast math libraries these days will use a Taylor sequence or some similar method to compute the trig functions because loading the table would cause a cache miss which is equivalent to 20-30 instructions.
<strong>Do they still teach kids in high school how to do all the trig table/interpolation stuff? Yuk. What a waste of time that was. They can pry my cold dead fingers off my TI-89...
And Taylor series suck, they're responsible for the single blip on my GPA. First damn mid-term in 15 years, Grrrrrr.</strong><hr></blockquote>
are you telling me that th TI-89 does all that stuff for you?!?!?!
I should go get one ....but we stop doing that stuff, we're doing matrices now(and my Ti-83 supports that
Taylor series are rather useful (even if they did mar your grade ). For many functions they converge to many place accuracy within the first few iterations.
And yes, you can take my 89 from me only if I get to use Maple VII on the test.
Other than a normal working relationship w/ Apple why would this rumor develop concerning Nvidia/Apple?
[quote]from the Archintosh article
"With the Raycer buy well over a year and half ago, God only knows what Apple had originally intended with buying a company which specialized in making advanced workstation level graphic engines (GPUs). One would hope that Raycer's technology may actually amount to something tangible for Mac users and indeed if these rumors amount to the truth it would not surprise us to see an announcement by Apple to that effect".<hr></blockquote>
As far as I know, Raycer never produced a single product. Their patents concerned algorithms to speed up 3D rendering? Maybe Nvidia saw the potential and began a concerted effort to obtain the tech?
Just musing, I have no knowledge or expertise in this area at all.
<strong>Nvidia doesnt manufacture anything... they just design boards and outsource the heavy lifting to other companies--like apple</strong><hr></blockquote>
nvidia make the chips, the boards are assembled by other companies, though nvidia probably had the ability to build the boards after they bought 3dfx (which had previously bought STB)
nvidia make the chips, the boards are assembled by other companies, though nvidia probably had the ability to build the boards after they bought 3dfx (which had previously bought STB)</strong><hr></blockquote>
Well if we are being picky Nvidia doesn't actually make the chips, they design them and they are manufactured by TSMC just like ATI's GPUs. Nvidia didn't purchase the assets of 3dfx, simply it's intellectual property. Thus the STB facilities (in Mexico i believe) were sold off to another company during 3dfx's liquidation. Not sure exactly who got it, maybe Jaton?
Comments
<strong>How about a quad engine GPU? 512 megs of pure VRAM
somehow the image of having so much power to run a game that the games code burned up came to my head when you wrote that
Maybe Apple is planning on doing something like this, taking a page from SGI's book. After all, that's part of the reason SGI systems are used so much for visualization...
The library did improve speed, though, and for many purposes the inaccuracies weren't a big deal.
<strong>The library you're talking about provided extra speed by essentially storing a table of values for common math procedures (e.g. sin, cos). Rather than having to compute the value, the computer could just look it up in the table. I'm not sure if it interpolated or used the closest value when there wasn't an exact match available; either way, there'd be some inaccuracies. Furthermore, this led to other problems; for example, I think tan wasn't implemented, so there were two options: look up sin, look up cos, and divide, or calculate tan the hard way. You got different results depending... That's why the library made network games go all screwy; you'd have one value for tan on your computer, the server'd have a different one, and you'd be out of sync.
The library did improve speed, though, and for many purposes the inaccuracies weren't a big deal.</strong><hr></blockquote>
This is how trig functions used to be done, back in the days when memory was fast and processors were slow. In the last 8 years or so, however, the move has been away from table driven algorithms to computational algorithms... to the point where AltiVec algorithms sometimes use several instructions just to build a constant in order to avoid loading it from memory. Fast math libraries these days will use a Taylor sequence or some similar method to compute the trig functions because loading the table would cause a cache miss which is equivalent to 20-30 instructions.
<strong>Do they still teach kids in high school how to do all the trig table/interpolation stuff? Yuk. What a waste of time that was. They can pry my cold dead fingers off my TI-89...
And Taylor series suck, they're responsible for the single blip on my GPA. First damn mid-term in 15 years, Grrrrrr.</strong><hr></blockquote>
are you telling me that th TI-89 does all that stuff for you?!?!?!
I should go get one
And yes, you can take my 89 from me only if I get to use Maple VII on the test.
-Ender
[quote]from the Archintosh article
"With the Raycer buy well over a year and half ago, God only knows what Apple had originally intended with buying a company which specialized in making advanced workstation level graphic engines (GPUs). One would hope that Raycer's technology may actually amount to something tangible for Mac users and indeed if these rumors amount to the truth it would not surprise us to see an announcement by Apple to that effect".<hr></blockquote>
As far as I know, Raycer never produced a single product. Their patents concerned algorithms to speed up 3D rendering? Maybe Nvidia saw the potential and began a concerted effort to obtain the tech?
Just musing, I have no knowledge or expertise in this area at all.
[ 04-19-2002: Message edited by: MaCommentary ]</p>
<strong>Nvidia doesnt manufacture anything... they just design boards and outsource the heavy lifting to other companies--like apple</strong><hr></blockquote>
nvidia make the chips, the boards are assembled by other companies, though nvidia probably had the ability to build the boards after they bought 3dfx (which had previously bought STB)
<strong>
nvidia make the chips, the boards are assembled by other companies, though nvidia probably had the ability to build the boards after they bought 3dfx (which had previously bought STB)</strong><hr></blockquote>
Well if we are being picky Nvidia doesn't actually make the chips, they design them and they are manufactured by TSMC just like ATI's GPUs. Nvidia didn't purchase the assets of 3dfx, simply it's intellectual property. Thus the STB facilities (in Mexico i believe) were sold off to another company during 3dfx's liquidation. Not sure exactly who got it, maybe Jaton?
<strong>
are you telling me that th TI-89 does all that stuff for you?!?!?!
I should go get one
I'll keep using my HP-41 CX, thank you very much.