Slashdot Mirror
NVIDIA Makes First 4GB Graphics Card
Frogger writes to tell us NVIDIA has released what they are calling the most powerful graphics card in history. With 4GB of graphics memory and 240 CUDA-programmable parallel cores, this monster sure packs a punch, although, with a $3,500 price tag, it certainly should. Big-spenders can rejoice at a new shiny, and the rest of us can be happy with the inevitable price shift in the more reasonable models.
Yes, AMD's Stream technology. I don't think it is used as much as CUDA in practice.
Or maybe there are companies that need high end cards with 4GB of RAM. This isn't some trick to get consumers to pay more for a low end card. This is now Nvidia's highest end workstation card.
excuse me but this is total bullshit. oldest trick in the book. if you are behind in technology, pop out a card with huge ram and try to get some sales.
lets face it. nvidia has fallen behind ati in the chip race. you can place any number of 4870s in a setup as much as you like to equate the power of any monolithic nvidia card and they always kick the living daylights out of that nvidia card in terms of cost/performance per unit of processing power.
In case the $3,500 price tag didn't tip you off, this isn't a gaming/enthusiast card. This is a Quadro - a professional card for high-end 3D rendering. Stuff like generating film-grade 3D or insane CAD stuff. Actually, due to the design of the card, it'd be pretty horrible at playing games.
This thing is aimed at high-end scientific calculation and professional-grade rendering.
ATI may, or may not, have something comparable. ATI may even have something better. I don't know, I don't follow the GPU industry very closely. But claiming that they're just slapping a bunch of RAM on a card to drum up sales is just plain wrong. Hell, the blurb here on Slashdot even mentions the fact that it has 240 cores.
"Work is the curse of the drinking classes." -Oscar Wilde
These were being sold in the first half of August for 10500$ - containing 2 of those cards. Only 3 months late.
"Violence is the last refuge of the competent, and, generally, the first refuge of the incompetent" - Thing_1
Two reasons:
One is simply that the cards use memory that isn't available normally. They don't use normal DDR RAM, they use special RAM for graphics cards, called GDDR. It is similar but not the same as memory in systems. Thus you can't just go out and buy sticks of RAM for it. So they'd have to be made special for the cards (and each gen of card uses different RAM), and thus would be expensive.
The bigger one is that the RAM is really pushed to the limit. You start to run in to all sorts of shit you never thought about. The electrical properties of the connection are highly important and there is a difference between what you get soldered on to traces and in a socket.
It's a nice thought, but not practical these days. Graphics cards are heavily dependent on high RAM bandwidth and you get that by really pushing the envelope. That means new RAM technologies all the time and the chips being pushed to the max.