found this..ripped from xbit before hand
"As we (and not only we) have already told you, in the beginning of February, to be more exact on February 5 in the USA and some time on February 6 in Europe, NVIDIA will launch its new solution for hardcore gamers – GeForce4 chip aka NV25. Today we got the opportunity to tell you a bit more about this new product.
Just like in case of Titanium chips family, the graphics cards makers are most likely to launch several products on NV25 at a time, which will differ by the chip and memory working frequencies as well as by the size of the graphics memory used. We will now touch upon the fastest model of this family, as we believe it deserves most attention and describes all the peculiarities of NV25 architecture.
So, the new NVIDIA graphics chip, just like GeForce3, will be manufactured with 0.15micron technology. However, the number of transistors used in it will reach 63 million, and the working frequency will be increased up to 300MHz. Here the natural question arises: how on earth NVIDIA managed to increase the chip working frequency so greatly without shifting to a new finer manufacturing technology, especially since NV25 should be more architecturally complex than GeForce3? The answer to this question is pretty original: we will see a new packaging in NV25 – PBGA with integrated heatsink. In other words, NVIDIA decided to enhance the heat dissipation process in order to be able to speed up its chips. For the same purpose, NVIDIA will provide its NV25 chips with a special NVIDIA thermal control system.
NVIDIA will increase the memory (DDR SDRAM) working frequency as well, which will work at 650MHz on the fastest solutions. This way the memory bus bandwidth will make 10.4GB/sec. Memory chips used on the graphics cards (at least on those following the reference design) will feature BGA-packaging. NV25 based graphics cards will get up to 128MB DDR SDRAM.
NV25 will feature 4 rendering pipelines with 2 TMUs each. This will make the new chip similar to GeForce3. However, due to higher working frequencies and improved technologies, the fillrate is claimed to equal 4.9 billion AA samples/sec.
Besides significantly higher working frequencies, NV25 will boast some fresh new features and technologies, though it will remain very much like its predecessor, GeForce3. First of all, we have to mention Lightspeed Memory Architecture II, the technology aimed at unloading the memory bus. It includes loss-free Z-buffer compression with 4:1 ratio, 2nd generation Occlusion Culling and Quadcache Architecture, which saves even more memory bandwidth.
Thanks to another significant performance and memory bus efficiency increase, NVIDIA managed to implement a new more progressive anti-aliasing method in its NV25. It is called Accuview AA. This method will use new subpixel locations mask and feature NVIDIA’s patented and optimized for the best performance and quality algorithm.
As for the T&L unit, it has been changed significantly compared with that of GeForce3. Now it is called nFinite FX II and features 2 vertex shader pipelines. Higher clock frequency will triple the performance in those applications, which use the second computing pipeline, making it three times as fast as that of GeForce3. Moreover, increasing the chip frequency will speed up the operations with pixel shaders by 50%.
Besides that, NVIDIA also claims that its new solution will support Z-Correct bump mapping feature, which is still a mystery to all of us. We will not speculate here now, as you will learn everything very soon: on February 5
The performance solutions from NVIDIA have now also got dual-monitor support. NV25 will boast nView technology, the next stage in the TwinView evolution. It will have richer features and will be much easier to use.
Attention! This news story will be soon removed upon NVIDIA’s request"