NV30 transitor count still at 120 million or has it changed?
- Thread starter Luminescent
- Start date
Well the HOS support was dropped on nvidias DX8 cards and they were still considered a DX8 cards. Nvidia would be making a massive mistake IMO if they don't support displacement mapping. HOS are cutting edge in terms of graphic chip design AFAIK. This would show some sort of incompetence if nvidia should overlook supporting displacment mapping which is of course the primary reason it will support it.... err I hope.
Humus,
IIRC, VS2.0 has a register that allows you to same the current displacement map's value. Since manually doing a render to VAR/glReadPixels (which would not be slow at all according to the link you provided) would not give you the ability to sample and displace in the vertex shader using the special VS register, that means NV30 wouldn't be PS2.0 compliant.
Surely, the register itself isn't optional!
IIRC, VS2.0 has a register that allows you to same the current displacement map's value. Since manually doing a render to VAR/glReadPixels (which would not be slow at all according to the link you provided) would not give you the ability to sample and displace in the vertex shader using the special VS register, that means NV30 wouldn't be PS2.0 compliant.
Surely, the register itself isn't optional!
Luminescent
Veteran
Maybe this is why the transistor count maybe lower on the NV30 than on the 9700. Like the p10, maybe the NV30 relies on programmable pipelines to achieve the effect of a HOS-specific processor without one whatsoever, as opposed to the 9700. If Nvidia chose to not implement the programmable primitive processor they might have thought it would cause more harm than good to introduce a hardwired HOS implementation, which would defeat their programmable intent.
Any thoughts on this?
Any thoughts on this?
Maybe 120 million was just an estimate and it ended up being smaller than expected after everything was said and done?
Nah, there's no drama in that.
Nah, there's no drama in that.
Luminescent
Veteran
I don't know, maybe the NV30's tranny count is not lower than the R300's, but 100+ million - unless Nvidia was still getting design straight - sounds like just above 100 million to me. Oh well, in the end, it is all speculation, but it is fun for the time being, till the supposed Comdex launch.
DemoCoder said:
Hmm, I've been searching the DX9 beta2 documentation, but I haven't been able to find anything such. I know I have heard about it though, but I suppose if it might have been dropped, unless of course the documentation isn't complete, or if I'm just not finding it.
The documentation does however mention support for displacement mapping through presampling the displacement map.
Luminescent said:
That's the problem though. How can you be so sure that this technology is going to be so much more advanced solely on the basis of its longer implementation time. Remember 3dfx's VSA-100? Nvidia is running into the same kind of problems 3dfx faced. By going with a complex design and a new manufacturing process they took a huge gamble, and for the looks of it, it is going to cost em a little more than they calculated.
Don't get me wrong, I want competition for the R300 so that it will drive its price down and consumers will benefit by being able to afford it. What I don't like is assumptions based on rumours and unconfirmed reports.
Luminescent
Veteran
I'm not saying a longer or more current development time will assure a superior product for Nvidia, it very well may not, but I sure wish the NV30 would serve as a catalyst for the industry in gpu design, rather than a plateau.
Luminescent
Veteran
It would be good for development if both the R300 and NV30 were close in capabilities, less wasted silicon. This would, most likely, ensure greater use of each respective chips' feature set, in contrast to a situation where 1 processor completely outdid the other. In a situation where there was unequal ground, end users would have to conform to the lowest common denominator, regardless. Realistically, as stated before on this forum, there seems to be less disparity between the R300 and NV30 processors than in previous architectures, thus no one will be overwhelmed one way or the other. Hopefully, as we saw with the introduction of the R300, the consumer will begin, not only to witness increased features, but also increased practical performance.
Bigus Dickus
Regular
GetStuff said:
I must have missed that press announcement from NVIDIA.
KnightBreed
Newcomer
Confirmed, when? Link or source?
Luminescent said:
Again, why can't R300 be considered that catalyst or plateau?
This time around Nvidia is not bringing anything really new compared to what R300 has already brought. They are merely catching up and according to rumors adding no new features, but only slightly expanded of what's already available on R300
Similar threads
