Older article here pointing out nVidia claiming that DirectX 11 Won't define GPU sales.

Just wanted to get some of your thoughts on this and the upcoming DX11 release and next generation of video cards.

I would put forth the notion that nVidia doesn't really have a lot going on for DX11 GT300 cards. There have been several rumors that the GT300 single GPU flagship card should be faster than the leaked 5870 but going off of rumors is a bad idea until some benchmarks are actually done. nVidia seems to think that GPGPU will be the future for graphics card sales and production. While ATi is pushing OpenCL, they are still mainly focusing on games as seen with their quick adoption of DX10.1 while nVidia refused to make a compatible product. Their focus on games is again highlighted by the Eyefinity thing they just previewed where 6-30 inch monitors are powered by a single graphics card and still pushing over 60 frames per second in several games.

I think nVidia has the wrong business plan of pushing proprietary stuff like CUDA and PhysX urging game makers to stay with older technology and limiting their markets to one graphics manufacturer.

Don't get me wrong, I would love the idea of more GPGPU applications for everyday use (more than video encoding, SETI, and folding@home) and stronger video game applications like physics rendering but at the rate that CPU's and other hardware (*cough*SSD's*cough) are increasing in speed and bandwidth, I almost do not see it being totally feasible. I suppose this comes to another chicken and egg debate where program writers are faced with either writing and optimizing code to work on more cores/threads or work better on the GPGPU or both (as suggested by Intel's larrabee).

Well i would love to see advances in a single GPU doing everything necessary in a system. Obviously not for a tower, but for low-end laptops (netbooks, for example) and specifically handheld devices. When a major performance-means-everything company like nvidia create hardware for next-gen phones and the such we'll actually be able to call them next-gen. That's what i'd really love to see.

Keep in mind i do not urge this for anything except small and/or weak devices. A tower is a tower and deserves each component on it's own bus. If they're trying to make it for general computing i doubt they'll get too far in the future with it.

Why would a relatively low end item like a netbook, PDA or phone need more graphics power? None of those things were designed for gaming which is what graphics cards' primary purpose is, nor were they designed for GPGPU utilization. For what little processing power they need, they certainly do not need another integrated component increasing power. From what I can understand as well, GPU's and CPU's execute code very differently making each excel in different tasks, but rarely the same ones such as physics (general physics and nVidia's PhysX), floating point calculation, and some other 3 dimensional calculations.

I think what you may be referring to is Intel's larrabee where its (from my understanding) basically a CPU and GPU on the same die, but optimized to work really well together. However, it will be a long time before anything like that is put into netbooks and other small devices due to power consumption. Intel's Atom processor has a TDP of about 2-4 watts while some celerons consume about 8 when underclocked and undervolted. Cell phones probably use less than half that of an Atom because while it is a different architecture, it is simply designed to be low consumption and thus low computing power.

Yes, CPU and GPU coding is very different. Intel's Larrabee would be more into what I'm thinking in fact, I've heard of it but never looked at what its about. Looks like nvidia has no real verdict here compared to that, unless it can somehow succeed and simplify the instruction set of GPU processing. I was looking more into portable gaming consoles (PSP, DS) and specifically a smoother operating OS for newer phones. I've seen many phones have dismal performance, whether its just lagging or slow to process/find data.

I've been using Win7 for a while now, moved onto it being my main OS. It's overall smoothness, ease and quickness of finding practically anything and the fact that it's a pleasure rather than a pain to work with is something that i believe phones should adopt. No matter how you look at it, the majority of the population living within 50km of a city use a phone more than most other things, more than a computer for many people. The issues i see is of course that apps and OS will need to start from scratch, power consumption and heat, but something that is feasible as a phone *cough iphone cough* deserves these features, surely.

Most of the actual processing of a phone doesn't necessarily need too much CPU power, so integrating both into a GPU might not be so detrimental to performance since CPU's are more accustomed to that work. Time goes on and expectations become higher, this may be the best route to take (a single GPU rather than just a CPU or CPU + GPU) for later technologies.

edited to add some more

nVidia is speaking about their "new technology" and how it will "revolutionize the industry" when really, they have been behind ATi in terms of new technology since before the xbox 360 was produce (as i'm sure you know ATi did the GPU for that, and it was indeed dx10 compatible)

ATi was the first to produce dx10, dx10.1 and now dx11 and nvidia is finally feeling the burn from it, directx adoption drives GPU sales and that's the bottom line, they can try and push 3d vision all they want, and physx and their CUDA line, but they will lose the real market, gamers with a low budget that need performance, which ATi has always provided.

now i've seen the ability of nVidia products, they can do some amazing processing, and their SLI is probably superior (i have no experience, so can't say for sure) but we all know that software is the one that needs to catch up with hardware, that's always been the issue, it always will be from what i can see, so the hardware that supports the latest development tools (read: directx, opengl, shader models) is the one that will win

nVidia has nothing to offer the directX 11 market, so it's natural that they would say something like, DX11 is nothing special, simply because that makes them look like they are still doing ok, but when DX11 is released and adopted (as directx versions ALWAYS are, regardless of failed OS releases) nVidia will be in the dust

as for GPGPU availability, AMD has had that idea in the books since over a year ago, and with their acquisition of ATi, it will only go smoother

I think you are correct in that the company supporting the latest innovations will win, but I think you need to modify it to include support of legacy innovations along with the new ones. This is what ATi has done actually; since DX11 is evolutionary of DX10 and 10.1, their DX11 cards support DX9, DX10, DX10.1, and DX11.

I think nVidia hasn't had any innovations since their 8800 days. Supposedly the GT300 is supposed to be radically new, but it sounds like their gt200 chip with a programmable tesla core, just like ATi did.

Since the new 5800 series cards just released today at a fair price, I cant wait to see nVidias marketing team sweat and try to figure out how much to charge for their cards now that they do not have the performance crown...

well when it comes to CPU i can understand new legacy item support, SSE4 v. SSE5 and so on, but when it comes to GPU support, it really matters that you support the main ideas first, like i said

ATi has done it in the past with their tesselation support, which they have had since x800 series (and maybe the 9000 series) and just now it's being supported, so they have that advantage, but honestly the world of 3D is still too young for 3d vision to be adopted IMO, even the new idea of 3d movies is a joke

get the fundamentals supported and top notch before you try adding on extra things that may or may not ever be supported i believe

nVidia is trying everything to draw people's attention away from ATi's new HD5xxx series -- that's understandable, of course.

This is what their new GT300 (Fermi) cards will look like: NVIDIA 'Fermi', Tesla Board Pictured in Greater Detail, Non-Functional Dummy Unveiled - though keep in mind that this is just a dummy, not a working card yet as they'll likely be very late to the Windows 7/DirectX11 party.