Slashdot Mirror
NVIDIA Gives Details On New GeForce 6
An anonymous reader writes "According to Firingsquad, NVIDIA will be announcing a new GeForce 6 card for the mainstream market at Quakecon this week. Like GeForce 6800, this new card will support shader model 3.0 and SLI (on PCI Express cards), so you can connect two $199 cards together for double the performance. NVIDIA will also be producing AGP versions of this card as well."
SLI was such an obvious way to make graphics rendering parallelized! I'm glad they're bringing it back... I've been missing it.
Does anyone have any idea how many PCI Express ports this uses? It's my understanding that you have a total of 20 and most motherboards are allocating 16x to the video... will this card require 8x? Or do you need a special motherboard for this?
Anyone know?
I am disrespectful to dirt! Can you see that I am serious?!
Is there a point where graphics cards get so advanced that humans can't even tell the difference anymore? Or is that virtual reality?
Only $200?
This should be interesting to see and good for competition to say the least.
From the article:
Price points and product names weren't discussed
So where did $199 come from?
Uttering logically derived and empirically supported truths to the disciples of the orthodox establishment.
...But does this mean you have to load the same texture data into both cards in order to obtain this parallel processing? Isn't that rather inefficient?
This does raise an interesting point, I think. NVIDIA seems to have let the cat out of the bag. A display card that can coordinate with another display card, perhaps doubling performance. Why buy next year's card that doubles your performance when you can buy last year's card and add it to your existing duplicate card for way less than paying the premium for the bleeding edge?
BTM
That was the turning point of my life--I went from negative zero to positive zero.
Hold the horses. SLI almost perfectly doubles the raster speed (Z-buffer checks, texture lookups, frame buffer writes), but pretty much everything else is performed at normal speed because it is executed redundantly on both cards. Especially with the ever increasing complexity of (vertex) shader programs, parallelization isn't as effective as you may think it is.
If you're going to use highly complex vertex shaders and lots of geometry information, you're better off with one of the other ways of splitting the workload.
Then go petition MS to create and distribute cards that supports their gd standard in hardware. I don't use Windows and have no interest in paying a fee to MS for having DX9 embedded into a card when I'll never be able to use it. If MS wants to pay for it and it's a zero cost addition for nVIDIA and it doesn't adversely affect OpenGL performance, then it would be inconsequential to me if it were included or not. Btw, what companies are in the consortium that controls the DirectX industry standard?
Good people do not need laws to tell them to act responsibly, while bad people will find a way around the laws-Plato
What's weird is that nVidia already _does_ have a $200 variant of Geforce6 - the Geforce 6800LE. It's essentially a lower-clocked (GPU and RAM) 6800 with only 8 pipes (so, half of what the 6800GT/U has). One of the hardware sites did a review of it (t-break?), and it performed pretty nicely - almost always beat the 5950. It's supposedly only for OEMs, but that's never stopped the online vendors from selling a card.
If they are indeed talking about a 6600, it's going to need to go under $170 to have any sales value whatsoever. SLI is nice and everything, but most people simply don't have PCIe mobos to take advantage of it, so it's going to be a non-issue for the next year and a half.
Still, it'll be nice to see nVidia actually try to deliver a better price/performance ratio than ATI for once.
-Erwos
Plausible conjecture should not be misrepresented as proof positive.
Unfortunately, this is actually something GOOD for the video card companies (from a sales standpoint). Because the consumers need blazing DX speed, but the workstation market needs OGL, they can still charge a hefty premium for the better OGL support in a workstation version of the card, even though the hardware is 99% the same.
Recursive: Adj. See Recursive.
In my Phys III class ages ago, we did the calculation for the resolvable limit for the human eye given a certain distance from 2 points. I can't recall the formula, but it seems that at some point in the near future a 8000 x 6000 screen will look exactly like a 80,000 x 60,000 screen unless your 2 cm away from it.
"I must not fear. Fear is the mind killer." -Bene Gesserit Litany Against Fear
Direct X is the whole API that includes Direct3D, Direct Input, Direct Music or whatever they are in the current release. Basically it covers more than just 3D. However the component of DirectX that deals with 3D is Direct3D.
OpenGL is a different API that allows you to do a similar set of things to Direct3D. However many people on Slashdot seem to think that OpenGL is the standard 3D API. This is (unfortunatley) not the case in practice, where the 3D API used for games is almost always D3D. There are a number of reasons for this but to sum them up from the top of my head:
*) Open GL is extended by IHV extensions - so that NVidia and ATI for example have different extensions that do the same thing but need exposing and programming differently for the different hardware.
*) Open GL doesn't support many of the newer features of GPU's. By newer I mean anything in pretty much the past 2-3 years isn't really supported by OpenGL. As an example there isn't yet a HLSL or even a platform independant shader language at all.
This is unfortunaltey a huge problem (IMHO) for OpenGL and as such for an open 3D standard for gaming - the only up to date hardware independant layer that we have is D3D. This is one of the main barriers for producing native Linux games too.
I wish DX would die. Quickly. Then maybe I could get some games native to linux...
:)
I know, I know. There are a few, but if everyone used OpenGL, it would be so much easier for them to port.. right? That "Sorry, we used DirectX" excuse most game makers throw about drives me crazy.
Why, yes, I *am* waiting for the release of the Linux Doom 3 binaries.
"An infinite number of monkeys typing into GNU emacs would never make a good program."
Not only that, but there's no such thing as Acura outside the U.S. - they just call them all Hondas, because that's what they really are.
Does this mean we'll see a price drop for the GeForce line? I've been putting off buying a new card, I don't want to end up buying it a couple of weeks before a price drop.
WURD!!
Which is exactly why it's going to take a while for video game graphics to look like current movie technology. You have 1/60th of a second to render a frame for a PC or console game, but a given frame in a movie like LoTR or Spider-Man may have taken hours or days to render, often using more powerful computers (and/or clusters).
-PainKilleR-[CE]
Oh, but it is good value.
You spend $200 now, so you can play the latest games, and then a year and half down the road, you spend $50 more for the second card, and still play the latest games.
Sign me up!
How depressing. I just droped $500 on my 256mb Geforce FX 5900 Ultra back in September and, less than a year later, it's going to be two or three revisions behind? Usually a top model card takes a year to be replaced, let alone be replaced twice or thrice over.
Graphics cards have been on a 6-month cycle since nVidia took over the market from 3dfx, with only occasional (maybe 2 or 3 in the last 6 years) breaks in that cycle. Therefore, a top-model card is about $100 cheaper 6 months after its release, because there's a newer top-model card.
Of course, in many cases there isn't a new generation of cards on a 6-month period (usually every 12 months), but there are faster versions of the same generation of cards.
On the other hand, the only reason I spent so much on a card was I wanted to build a new power house machine in preperation for Half Life 2 which at the time was slated with a SPECIFIC DAY of September 29th for release (I had already bought my copy in preperation on Amazon).
Well, in the future, make sure you set aside that money to buy a new video card on the day that copy of HL23 or Doom51 comes out, instead of buying a new card in anticipation of a (even an announced, verified by the developer) release date. New video cards come out all the time, and each release tends to drive down the prices on perfectly good cards. Whether the delay is on the scale of Doom 3 (what was it, 6-9 months?) or HL2 (who knows any longer), you will either be able to buy the same card for less money, or a better card for the same money. Then again, you could have been enjoying your card on other games that actually shipped in the interim.
I don't really care to connect two cards together anyway. Just build me a single super kick ass card.
On the other hand, what if you could buy 1 really super kick ass card, and then buy another 1 in 6 months when HL2 is supposed to come out (no, not really, I have no idea when it's coming out) to get better performance without having to spend nearly twice as much on the latest, greatest card? I know it helped me out significantly when I bought my first computer that I could buy 1 Voodoo 2 card for $300 and then buy another one a few months down the line (still at $300) to get better performance than anything available at the time (for consumer-level graphics anyway), rather than wait another year for the $300 cards to match the performance of the SLI setup. It matches quite well with the reason I build my computers myself anyway: distribution of cost. Anyone can afford to buy a better computer if they can distribute the cost over time (one reason credit is so popular) than if they have to cough up all of the money at once (of course, when you're buying $300-800 parts, even the distributed costs can be a significant hit to the wallet).
-PainKilleR-[CE]
flying-rhenquest died a couple weeks back (The fan base may have noticed that the web page is down,) so I upgraded to a system with a ATI X600 PCIE card. You can force the system to recognize it as a radeon for 2D, but apparently PCIE is not yet supported by the ATI proprietary driver nor the Xfree86 radeon driver. Rumor has it the Nvidia proprietary drivers have PCIE support, but I haven't had any solid confirmation of that yet. So does anyone know for sure that if you drop this card into a Linux system, you'd be able to get 3D acceleration?
I'm trying to teach myself to set people on fire with my mind... Is it hot in here?
Are you sure you know what "T&L and other stuff" means? Transform & Lighting... nowadays, you need to think of it this way: Transform ~= Vertex Shader (vertex-level lighting is done here), Lighting ~= Pixel Shader. Given the advances in "T&L" with GPUs, do you really think that "that's a negligable part of the work nowadays"? So basically, because of the non-pixel-specific nature of Vertex Shading, each card needs to run the appropriate vertex program on each vertex that it might need to have data for in rasterization. They could do some neat stuff with sending particular fragments to another card for pixel shading, which may be what they are doing. That'd make this new SLI system likely to perform about as well as a single card with twice the memory bandwidth and twice the pixel shader pipelines. It'll certainly be faster, but IMHO, until it matures it won't get near 2x speed.
All Your Memory Are Belong To Java
You bring up a very good point, and it was quite an ingenious move on Microsoft's part. DirectX, for various reasons, has become the standard. This is, pretty much in and of itself, the problem.
Remember FrontPage extentions? And those silly non-standard tags IE can use, but nothing else can? We have good, decent (although not perfect) web standards (CSS, XML, etc.) and many web developers strive to comply with them. There's a big push from the W3 to advertise and get the word out.
We need the same type of thing with game development, IMO. Be it with OpenGL, or perhaps something else. When companies set standards, they do so with their own self-interests in mind. Those interests may be harmful for us users! I'm not saying that any organization should dictate what developers choose to use, but an organization that would push for the advancement and perhaps standardization between/among of different ways of doing things would be idea.
Ohh well. Just a passing thought. Off to lunch.
"An infinite number of monkeys typing into GNU emacs would never make a good program."