DescentBB

You're pretty wrong then:

http://www.theinquirer.net/?article=17889

D'oh. Any sponsors around?

Stop reading shady "news" sites and go back coding

The only trouble with this kind of card, apart from the cost, is that we are not sure there are there 256 Platinum chips available.

Heh!

wasn't SLI 3dfx's not nvidia's?

nVidia bought 3dfx, if you recall.

SLI 6800's getme all horny....

3DFX's SLI was Scan Line Interleave, basically each card drawing every other line on the screen. NVIDIA's new SLI is Scalable Line Interleve, it will load balance across two cards to try and and get the best results. (ie: one card draws skybox and far away objects while the other performs closer, more detailed objects.)

no, not exactly. Each card draws half the screen.

One draws the top half, the other the botton half.

That makes a lot more sense than every-other-line... in 3dfx's method, doesn't each card have to caluclate every polygon that's visible in the scanline it's currently rendering? And when was the last time you saw any 1-pixel-tall polygons? Essentially both processors were doing the same work twice, correct?

STRESSTEST wrote:no, not exactly. Each card draws half the screen.

One draws the top half, the other the botton half.

Then it load balances according to how long it took each one to draw, letting one draw more or less the screen.

http://www.neoseeker.com/Articles/Hardw ... sli/2.html

I assume this will be across an AGP card and a PCI card in the slot next to it....will actual display data then end up going over the PCI bus, or will the AGP card relay data to the PCI card, thus bypassing the PCI bus completely?

This is a PCI-X dual slot opperation vader

Mmm http://img.neoseeker.com/v_thumb.php?ar ... 64&image=9

FORBIDDEN!

cut/paste

woah....hmm, how many motherboards could you actually use that with?

being that it is a new formfactor, it's getting there. So, more and more

excerpt from CPU mag article

======paste=====
One card is set up as the master, rendering the top half of the screen, and the second one is set up to render the bottom half. The amount of rendering done by each card is not limited to 50% each. Using NVIDIAâ??s dynamic load balancing, the sharing can be optimized so that more complex scenes rendered by one card at the bottom of the screen where there are multiple objects, can be done by one card and less complex â??skyâ?

How does that work, though? I am thinking maybe the cards determine which part of the screen is a "complex" picture when it has a certain number of vertices/pixels.

So, basically, with two GF6800 Ultras in SLI mode, one should be able to sm0ke a z00b. ^_~

0wnage! ^_^

Seriously, though, would two GF6800 Ultras in SLI mode in two 16x PCI-X slots be a lot better than just one? Is it 2wice as nice? ^_~

Oh, by the way, I have a question: I've heard that setup called GeForce 6800 Ultra Extreme. Is that actually the name that nVIDIA gave it, is it just something that people call(ed) it, or is it just some name that some n00b (I'm not saying any names ^_~) called it?

Neo wrote:Seriously, though, would two GF6800 Ultras in SLI mode in two 16x PCI-X slots be a lot better than just one?

The marketing people say that it is nearly double the speed since each card does roughly half of the work in a relatively smart manner. Of course that's talking about the video card itself; if you're CPU bound you'll still be CPU bound with a nicer video card, although you could get away with much higher AA and AAF with little performance hit.

And correct me if I'm wrong, but I don't think you even need two 16x pci-e slots.

2wice as nice? ^_~

...

Tetrad wrote:And correct me if I'm wrong, but I don't think you even need two 16x pci-e slots.

Okay, then I will. =P

Then thereâ??s the fact that none of Intelâ??s PCI-Express chipsets support more than 24 PCI-Express lanes so even prototype dual-PCI-Express 16 interface motherboards act as 1x16 and 1x8 lanes.

I'd rather have two 16x PCI-X slots than a 16x and an 8x slot. =P