The Inquirer-Home

More NV40 details emerge

From Germany with love
Wed Jan 28 2004, 11:22
OUR GERMAN COLLEAGUES at 3dcenter.de have a lot of details about NV40 and most make perfect sense to and match the details we've published ourselves.

According to the site, Nvidia's NV40 will use 175 million transistors developed at 130 nm with an 8x2 architecture capable of doing 16 passes without rendering Z values.

The chip is still DX 9.0 with PS 3.0 functionality but has more efficient shaders then NV38. I t will support DDR 1 GDDR2 and GDDR3 memory and will work as an AGP 8X part, with the chip clocked from 500 to 600MHz and memory at 600 to 800 MHz. Nvidia will announce it at either GDC or CeBIT - with market availability in April/May. The site reckons the name will be the GeForce FX 6XXX

Let me comment on these numbers if I may. The number of transistors is about the right sum, and while it will be a 130 nanometre chip, I am not sure about 8x2 or even 16x1 without Z passes. I don't say no to that possibility, but I would rather believe it will be 8x1 then 8x2. The claim that the chip is only DX 9.0 and not 9.1 capable is another matter, as Nvidia knows how how important this is for its marketing. As for memory support, I believe it will have support for all memories including DDR 1, GDDR2 and GDDR3 just like ATI's R420, while the cards will be AGP8X with future PCI Express versions under development.

As for frequencies this is what I would expect from the chip, but this depends on many things Nvidia and TSMC/IBM cannot control. As you can buy DDR 2 that will be used by Nvidia for NV40 running at 600 to 800 MHz this information is likely to be highly reliable. We know that Nvidia received thousands of samples of chips from Samsung in late 2003 and we heard that Nvidia is struggling to make it stable at 750MHz or, should we say, 1500MHz DDR 2.

Nvidia is likely to launch the first day of SnoBIT since it wants its partners to show off their cards. That is, if all is ready by then. µ

L'INQ
3dcenter.de

 

Share this:

blog comments powered by Disqus
Advertisement
Subscribe to INQ newsletters

Sign up for INQbot – a weekly roundup of the best from the INQ

Advertisement
INQ Poll

Heartbleed bug discovered in OpenSSL

Have you reacted to Heartbleed?