Twin engine GPU?

flounder · April 16, 2002 10:23PM

Juicy little article over at Architosh claiming apple is coming out (possibly in conjunction with Nvidia) with a twin engine graphics card with 128 Megs of DDR for each GPU.

<a href="http://www.architosh.com/news/2002-04/2002c1-0412-applegraph1.phtml"; target="_blank">http://www.architosh.com/news/2002-04/2002c1-0412-applegraph1.phtml</a>;

If I'm not mistaken, it also insinuates Dorsal M is a source of theirs.

Well reasoned out little article

Probably a crock, but certainly plenty entertaining

Discuss......

eugene · April 16, 2002 10:29PM

3Dfx did this with their VSA-100 chips. It was too bad they had lost so much with Voodoo3 and Banshee by that time. Voodoo4 and Voodoo5 looked pretty nice.

cdhostage · April 16, 2002 11:10PM

Prolly a crock.

airsluf · April 16, 2002 11:28PM

tsukurite · April 17, 2002 1:06AM

could this supposed new card explain the alleged performance increase for aqua we've been told about in posts about the beta boxes, a la Dorsal M?

Heh, all those posts should be categorized into something like:

"Son of G4, it's back and it's pissed!"

or

"Night of the living G5's"

dennis the phantom menace · April 17, 2002 1:16AM

You know, I'd be just as happy with a geforce 4 ti 4600. It too has 128 MB DDR RAM per GPU.

ipad · April 17, 2002 2:03AM

Well I doubt Apple themselves will come out with this card, but the upcoming NV30 (which will probably be around a year from now if nVidia keeps to their moore's law cubed plan) will probably incorporate SLI, better quality AA, and other technologies purchased from 3dfx.

From recent announcements, nVidia seems to be committed to mac now so I'm sure if a card like this does come out, it'll be available for the Mac within weeks of announcement of the PC version (or in the case of the GF3, announced for mac first, but in actuality the GF3 shipped first on PC afterwards).

alpha mac · April 17, 2002 6:01AM

[quote] Jobs apparently said Apple was prepared to do whatever was necessary to satisfy the needs of this elite user-base.

<hr></blockquote>

Now thats good news.

[quote] Chief 'wants' among the studios is for more powerful hardware. Specifically, they want 'best of breed' graphics performance and options there. They also want a 1U or 2U industry standard, rack-mountable Power Mac for render farm building and servers. And they want only duals and quads -- single processor Macs are not desired except at the portable level.

<hr></blockquote>

So thay might get the massage at last

[ 04-17-2002: Message edited by: Alpha Mac ]

xype · April 17, 2002 6:18AM

[quote]Originally posted by Alpha Mac:



So thay might get the massage at last

<hr></blockquote>

We'll see. Also I am not sure whether that "going after the pro market" means any more reasonable pricing. <img src="graemlins/hmmm.gif" border="0" alt="[Hmmm]" />

mattyj · April 17, 2002 8:25AM

Damn, if this is true then perhaps the contract with motorola will end, and apple will sign up with IBM to ptoduce the next generation of PowerMacs. If this is true then long live apple (as long as starting prices don't start at $3000)!

matsu · April 17, 2002 8:43AM

duals at least shouldn't creep much (or at all) over the cost of current DP machines. A top line PPC has traditionally been cheaper than a top line Intel. Last time PPC prices were published on the net, Ghz G4's were 290 (when they first came out). Northwood P4's start at 450. So the price-performance of PPC's is pretty good. Smaller process should make them cheaper to fab, but any quad machine would be a VERY VERY VERY expensive.

murk · April 17, 2002 8:49AM

[quote]Originally posted by Flounder:

If I'm not mistaken, it also insinuates Dorsal M is a source of theirs.

[/QB]<hr></blockquote>

Dorsal, you little whore! <img src="graemlins/lol.gif" border="0" alt="[Laughing]" />

rickag · April 17, 2002 9:15AM

[quote]AirSluf

They might have had me if they hadn't said "In fact, the G4 architecture has always been cited for its strong floating point performance". <hr></blockquote>

I've read this elsewhere also. I could be wrong but I think these references to a strong floating point performance refer to it's efficiency. The woeful fp performance compared to Intel and AMD might have more to do with clock speed and don't Intel chips have 2 floating point units??

Any one have the correct information?

programmer · April 17, 2002 9:22AM

While much of that article sounds like wishful thinking, the parts about the "twin engine GPU" can be interpreted any number of ways -- and all of which are likely to be delivered by nVidia in the not-too-distance future. It would nice to Apple working closely with nVidia to deliver top-notch graphics, but I'll believe it when I see it. Their track record in this department is not stellar.

PowerPCs have traditionally been stronger at floating point, but the G4 simply has not kept pace over the course of its lifetime and the Athlon & P4 have overtaken it in terms of scalar floating point. The AltiVec unit's single precision floating point is still superiour, but is now so far behind in clock rate as to have lost much of its advantage. The next PowerPC design could change that and put it back on top, at least temporarily.

mattyj · April 17, 2002 9:42AM

Well, here at Appleinsider, someone posted that a dual 1.8Ghz G5 (the chips by themselves) produce 80GigaFLOPS of power. If this were true, where would it put apple against the likes of AMD?

preston · April 17, 2002 9:53AM

IBM and Motorola make the G3 and G4, but they are Apple Products.

Nvidia makes the architecture for XBox and Gamecube(?)...

could this become an Apple videocard manufactured by Nvidia?

(is my research completely accurate? hell no, but I raise an interesting question no?)

airsluf · April 17, 2002 10:50AM

amorph · April 17, 2002 1:11PM

[quote]Originally posted by AirSluf:

I have also read some of the same concerns relative to the OS X math libraries. Folks seem happy with the accuracy of Apples but rumors of the uber-no-debug-code-fast-mathlib-that-fixes-performance-but-gives-wrong-result still pop up every now and then. I think it will take a design difference to break out of that rut.<hr></blockquote>

The uber-no-debug-code-fast-mathlib-that-fixes-performance was rolled into 10.1.2. The difference (besides No Debug Code(TM), for Snappier(TM) performance!

) is highly optimized PPC assembler code (as OS 9 had) instead of off-the-shelf, cross-platform C code borrowed from BSD.

I haven't heard any complaints about accuracy.

[ 04-17-2002: Message edited by: Amorph ]

paul · April 17, 2002 2:14PM

[quote]Originally posted by preston:

IBM and Motorola make the G3 and G4, but they are Apple Products.

Nvidia makes the architecture for XBox and Gamecube(?)...

could this become an Apple videocard manufactured by Nvidia?

(is my research completely accurate? hell no, but I raise an interesting question no?)<hr></blockquote>

ATI makes the gamecube gfx card

Nvidia doesnt manufacture anything... they just design boards and outsource the heavy lifting to other companies--like apple so i think what you meant to say was backwards.... although i didnt do any research either... so who knows...

Also, if apple made a propritary video card, they would probably only sell it with new machines to pad their bottom line, which would make a lot of people angry.... not that it would matter to apple anywyas... they rape the "faithful" all the time.... <img src="graemlins/oyvey.gif" border="0" alt="[No]" />

programmer · April 17, 2002 4:18PM

[quote]Originally posted by mattyj:

Well, here at Appleinsider, someone posted that a dual 1.8Ghz G5 (the chips by themselves) produce 80GigaFLOPS of power. If this were true, where would it put apple against the likes of AMD?<hr></blockquote>

Yeah well, lots of people post lots of things here. Many of them make good jokes.

Current P4/Athlon performance levels are about 8-20 gigaflops per processor, and this is peak vector single precision.

crusader · April 17, 2002 8:00PM

How about a quad engine GPU? 512 megs of pure VRAM

Twin engine GPU?

Comments