Next gen graphics and Vista.
- Thread starter kyetech
- Start date
Jawed
Legend
The kinds of algorithms that benefit from XB360's unified architecture are going to make an awful lot of developers biased towards the first PC GPU that also implements a unified architecture.
In that race, I just don't see any advantage for NVidia.
The only saving grace is that these games are prolly 2 years away from the PC platform.
Jawed
In that race, I just don't see any advantage for NVidia.
The only saving grace is that these games are prolly 2 years away from the PC platform.
Jawed
I think its almost a certainty that ATI would feel their approach is better for games as well, otherwise they woudln't do it - this is, afterall, where they make their money. On the otherhand, you have to wonder why NVIDIA are now saying the opposite and suggesting that Unified artchitecures probably will become a necessity.Chalnoth said:
xbdestroya said:
You say it yourself, it's a big question mark. I'm pretty sure nVidia is doing EVERYTHING they can to kill any competition. And I also see nV as the more aggressive party. They also openly stated some time ago that they're about to kill ATI. I can't remember where it was, though.
But whatever, I'm pretty sure that the unified parts will take some time to establish and to get enough software to show off it's capabilities. It may be playing "older" (non-SM4) games much slower than the traditional architectures and actually hurt in the beginning.
Just a few stupid guesses, all I want to say is that such clames are totally off base.
I don't see that as being the point of unified at all, I'd see it as benefitting any shader title as the point of its design should be to be able to better balance whatever workload is given to it._xxx_ said:
It'll be of benefit to anything that uses the programmable pipeline - given that every architecture will be dedicating most of its silicon to the programmable pipeline, they are all likely to be in similar boats where the fixed function element are leant on most of the time.
I think your argument can be flipped around entirely for DX10, in fact. Lets take the case of a unified architecture that unifies the VS, PS and GS, and a non-unified architecture that has discrete VS, PS and GS; load balancing between VS and PS operations aside, most apps are not going to be using that GS unit for a long time, so that an element of the die that goes unused, which is not necessarily the case for a unified architecture.
The biggest gamble on the unified front is not "some application using it", because thats transparent, its the cost of the control mechanism in relation to the scalability and effectiveness - we've seen that ATI have implemented more ALU's with the unified control structure in a non-enormous chip, we've got to get more of an indication as to its payoff.
Dave Baumann said:
I suspect this has more to do with behind the scenes politics and long term projections than any short term performance considerations. NVIDIA already knew ballpark performance figures for a non-unified DX10 architecture when they made their initial comments, so it seems pretty unlikely that they've now done a u-turn because they think they'll be significantly behind any time soon. Perhaps it's been made clear to them by MS that they're not going to be able to continue to follow a non-unified approach beyond DX10, and so they're changing their stance? It won't have been a change to DX10 itself that's forced their hand, because we're way too far down the road for Microsoft to be making those kinds of changes to the spec.
Oh, absolutely, of course it is to do with their own internal roadmap - if they have started down a unified approach already then its nothing to do with competetive performance as performance figures for a unified platform would only just be getting back to them now. It got to do with their own internal projections, analysis and testing, but if they really were set on their initial statements then that should probably still be the case even as the unified API matured.PaulS said:
Well, since nVidia's first DX10 architecture will at least be unified in software, the worst case scenario would probably one in which the pixel pipelines are used exclusively for the "unified" shading, so I don't expect that it should be that bad at such algorithms.Jawed said:
As a side comment, though, while I still don't yet see any reason for games to really benefit dramatically from a unified architecture, it would have tremendous benefits for GPGPU type stuff. For example, I have some code running now whose most intensive algorithm is one that uses log/exp instructions and does 1-d integration via a recursive algorithm (simply dividing the integration region in to N parts takes 100x more time for similar accuracy).
So it'd probably be really fun to get my hands on some DX10 hardware to see if I can make the code that much faster. It'd be especially nice for GPU implementation because the number of inputs is very small, and there is just one output per integration.
_xxx_ said:
Well NVidia is certainly the more brash company, that's for sure. They may say they want to end ATI's existence, but I think it's going to take more than words and one gen beyond the current one to acheive it. To tell you the truth I'm a big NVidia fan (all my cards have been NVidia), but I've always found ATI's solutions to be the more elegent ones (architecturally, since R300) - and I think that the unified shaders follow brilliantly in that tradition.
And as for NVidia now growing 'warmer' to the unified shader idea, I truthfully just think they were dead set against it - publicly mind you - while they knew ATI was ahead on the DX10 front, and now that they're feeling a little more comfortable with their current situation, I think they feel they can afford to be a little more honest about the whole thing. It behooves a company to talk down on the architectural landscape their competitor is talking up, regardless of what one truly feels about it.
Last edited by a moderator:
? There is no such thing as "unified shading" - in DX10, would have VS, PS and now GS operations and at the very least the VS and PS will have the same programmable capabilities and the same instruction set; in a non-unified hardware platform the VS will be performed on the VS units and the PS on the PS units, but on a unified hardware platform the same hardware is shared for the operations, thats it.Chalnoth said:
So efficent utilisation of the available execution units is not of use to games? Again, take my prior example of a GS unit in a hypotetical non-unified architecture - how often will that just be die unused, but would be used (for non GS operations) in a unified platform?
Jawed
Legend
I'm thinking of algorithms that are heavy on vertex shading. I assume it's going to lead to a sophisticated, phased rendering of each frame, with some phases being solely vertex shader work and it's also going to obviate techniques in which pixel shaders are used to generate data for the vertex shaders.Chalnoth said:
Sure, I hardly know about this stuff since I'm not a developer. That's what I've picked up, though.
Jawed
Given PowerVR have a unified platform for handheld devices, I would guess the costs aren't that big.
It occurred to me that NVIDIA may have another solution to the unified shader scenario, and part of that can be seen in the multiple clocks of the G70. For example: one VS running at 1GHz and 16 PS running at 700MHz. With this solution, they can mantain dedicated units to the vertex and pixel shaders and yet save some space in the units they use...
If this is possible or not, I don't know...
If this is possible or not, I don't know...
Jawed
Legend
Generally, I don't think it makes sense. The complexity of the hardware to execute a MADD in a vertex shader is roughly the same as the hardware to execute a MADD in a pixel shader - certainly looking forwards to DX10.Sigma said:It occurred to me that NVIDIA may have another solution to the unified shader scenario, and part of that can be seen in the multiple clocks of the G70. For example: one VS running at 1GHz and 16 PS running at 700MHz. With this solution, they can mantain dedicated units to the vertex and pixel shaders and yet save some space in the units they use...
If this is possible or not, I don't know...
Although to be fair there are significant differences in numbers of objects processed by each half (way more pixels than vertices, per frame).
VS is MIMD, whereas PS can get away with being SIMD, too.
Jawed
I wonder if future unified architectures are going to continue to be SIMD or if MIMDs are on their way instead in the foreseeable future.
By the way I think it's a good time to remind that this demo here also contains geometry textures:
http://www.pvrdev.com/pub/PC/eg/h/Cloth.htm
I can see at least twice the performance on G70 compared to NV40 in that one.
By the way I think it's a good time to remind that this demo here also contains geometry textures:
http://www.pvrdev.com/pub/PC/eg/h/Cloth.htm
I can see at least twice the performance on G70 compared to NV40 in that one.
however, unlike NVIDIA who basically only have one current high end architecture (G70~RSX).. ATI also have their Xenos architecture in addition to their R520 architecture, and I imagine they would like to get the technology developed for Xenos out into the PC space as soon as possible, and take further advantage of the R&D they spent on developing that part.. I think this may especially be the case if Xbox 360 does well & 'unified shaders' receives positive hypeChalnoth said:
Last edited by a moderator:
Similar threads
