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."
I might need to dust off my textbook from "Parallel and Distributed Computing", but I'm pretty sure that getting double the performance from two cards is about as likely as getting double the performance from two processors. It's just not likely unless the graphics routines can split up jobs perfectly and not suffer from any overhead for communication. I imagine there will be a noticeable performance increase from 2 cards working in parallel since graphics algorithms do have a tendency to be very parallelizable, but claiming double performance in naive at best and dishonest at worst.
Ignorance warning: I don't know much about the technology, but scanline interweaving seems like it's difficult to pull off with present day technology because of anti-aliasing algorithms, temporal AA, etc. These things have to be calculated, and available on both cards if they are generating the image line by line (alternating turns). It's "obvious" in that it makes sense intuitively, but technologically, it seems like a more impressive feat.
We are one consciousness experiencing itself subjectively. Back to you with the weather, Bob!
Since I don't keep up with things.. Is PCI Express way better than AGP, for bandwidth on graphics cards? If so - is there anything new from the AGP camp planned?
.
I fear something like AGP EXTREME
I'm all for advancing technology, but when it comes to video cards, it's all a matter of who can keep up with Microsoft's DirectX demands the best.
Meanwhile, OpenGL, the industry standard graphics library, is getting left behind because every video chip maker wants to show off how well it supports GlibFlobber() DirectX 27 API.
Won't someone please think of the industry standard instead of the proprietary (and very small market) "standards" of Windows?
Why would it?
So long as game companies turn out new games that make existing systems cry for mercy, (and we choose to buy them) we will always need to buy newer video cards in order to stave off choppy video for another generation of games.
Same goes for CPU... although much of the difference is that most of those people buying a Geforce 6 are gamers and will use most/all of the power at their disposal... I'd wager only a fraction of those using the latest and most powerful CPU's from AMD or Intel use them to their full potential.
Help Brendan pay off his student loans
All I want is DirectX 9 support in hardware, not the kludges which the current NV's have. The GPU makers churn these things out so quickly, yet they can't keep up with an industry standard a year old...
"We know what happens to people who stay in the middle of the road. They get run over." - Aneurin Bevan
Based on recent cinema experiences, you would have to say were still a hell of a long way from this. I just saw Spiderman 2, and a lot of the CG work still looks totally artificial. Likewise, the trailer for I, Robot made me cringe with its computer-generated aura. Even LOTR looked fake in places.
Considering these movies are using the absolute cutting edge of pre-rendered graphics technology, I would suggest we're still a decade or so from anything like 'real' looking PC graphics.
Read Pynchon.
There still are a number of things that are way out in the future for graphics processors, especially polygon based - for instance, ray tracing has the ability to reflect off multiple surfaces (you could create a house of mirrors, for instance, with true curvature reflections), while polygon models have just started to make decent reflections on a single flat(ish) surface. Radiosity and similar effects are usually mapped beforehand because they are so processor intensive to calculate in real time, but could be used to cast "foggy" shadows and create other creepy effects. Another possibility is to offload the entire graphics model to hardware and do everything (e.g. frustum polygon culling and quadtree/oct-tree culling) inside the hardware instead of in software.
It seems to me graphics hardware has a long ways to go still. There are also probably newer, more photorealistic models that have appeared since I studied computer graphics, as well. Virtual reality in a sense depends on audio and AI in a true form, but a virtual visual (and perhaps audio) reality is probably on the horizon. AI is probably 15-20 years down the line (at least for something that stands a chance at passing a Turing test, IMO).
They might concentrate on getting their CURRENT high end card (6800 Ultra) on the retail shelves instead of "pre-announcing" crap in the pipeline.
NVDA has just reported a HORRIBLE quarter. Many are wonder what the F is going on with that company. This is a PR release. They need to say these things. They need to say they have native PCIe despite not a SINGLE OEM design win. They need to say 6800 volume will ramp up and product will be driven down to the low end. Will this actually happen? I have no idea, but this is the least I would expect NVDA to say on this horrible week for NVDA longs. ATI has really put the hurt on. This next 12 months should be pivotal for NVDA's future.
As I understand it, you're real close here.
From what I understood (when I read an article about it around what, a month back?) is that yes, each card renders a seperate portion of the screen, but the spiffy thing about this new implimentation is that the ratio is dynamic; if there's a lot going on in one half of the screen, and not much in the other portion, the under-utilized card starts rendering more of the screen to allow more focus on the "action-intensive" area by the other card.
Then again, I could just be talking out of my ass here - like I said, its been a while since I read the article, and I may have gotten some of the details mixed up with some other dual-card rendering schemes I've read about.
There is no mod option "-1: Disagree" for a reason. "Overrated" is not an acceptable substitute. Post something instead.
From the article I took that nVidia will be filling out the rest of the 6xxx line (6500, 6200, etc). Your 6800 will still be the cream of the crop for the next 180 days or so.
"SLI was such an obvious way to make graphics rendering parallelized! I'm glad they're bringing it back... I've been missing it."
:)
From an economics point of view, it sounds pretty cool. Spend a few extra $$$ to get a top of the line card. Then, in a year or two, pick up a second card when the prices are considerably lower, then you get 2x the performance without tossing hardware. Bitchin.
Unfortunately, I wonder if that puts NVidia in an ugly place. It does set the bar for what the Geforce 7's have a minimum to do. But... that aint bad for us, now is it?
"Derp de derp."
Does it have low power consumption or does it include a nuclear powerplant?
Jesus Christ, I just want to play the game, not let it rule my life.
Then you'll be fine. The game runs great on my P4-2.0 GHz w/ 512MB RAM and a similar video card to what you have (I've long since forgotten the exact number, I bought it to replace a GF4Ti that stopped working). I can't crank the graphics and resolution, but it still looks better than most other games out there.
The box says a 1.5GHz machine is the requirement with a GeForce 3 so I'm sure I'm fine, but why make a game that you need a top of the line machine to play at high settings? That's stupid and alienates 90% of your fan base for years until they upgrade to machines eventually that will run it.
Though I doubt it will run well on a 1.5GHz machine with a GF3 card, you have to realize that new machines sell with significantly better specs than that. My machine is over 2 years old (except for the graphics card, which still isn't top of the line) and runs the game fine.
The game is clearly targeted at the same audience any previous id game was released to, meaning that the requirements are not going to work for the person that doesn't really play many games, but the game will work fine for almost anyone that plays any 3D games at all. For those that must have the absolute top performance, there is a significant benefit for owners of top-of-the-line machines, which is clearly a selling point to some individuals.
The opposing view-point is to make a game that runs at its best on a mid-range machine. This satisfies many gamers that can't justify the purchase of expensive PC hardware just for gaming, but irritates many gamers that want to see what their hardware can really do before its out-dated by new cards and CPUs. It also leads to some level of decline when 2-5 year old consoles can compete graphically with up-to-date computers, as technology tuned for TV resolutions doesn't need to push as many polygons to look good (comparatively).
At which time I'm sure Doom 4 will be out and need a quad processor system and 4 video cards.
Probably. The one thing I've always liked about id's games is that they tend to push for the game to run on whatever top-of-the-line system is available at the time they start development, rather than developing for the most common system specs available at the start. The most obvious reason for this is that, by the time the game is released, those systems will be much closer to the most common system. Add in the significant work done in optimizations for the newest GPUs and processors as they were released (during the game's development), and there're benefits to be had for those that bought the newest stuff, while the people that aren't buying new hardware at release can still play. Considering this game's had one of id's longest development cycles, at about 4 years, most of the people that were aware of the game before its release have done some upgrading in the development period.
-PainKilleR-[CE]
You're wrong. SLI in the Voodoo 2 sense died with the Voodoo 2.
Please read up about the current nVidia solution before commenting again. Kthx.
What part of virtual reality requires AI and what does it have to do with the Turing test?
I think this is the reason that SLI is only available on the PCIe cards. PCIe provides an independent bus for each component. This means that, not only can components communicate with the processor very quickly, but also each other. My understanding is that the processor is also connected to the rest of the components in your system using a PCIe bus. So, due to PCIe, both the cards can communicate with each other as quickly as they are communicating with the processor. This should make it possible to have the data requisite for AA present on both cards.