The Inquirer-Home

NV40 will give Nvidia performance lead by year end

Pricing and other machinations of the graphics folk
Sat Apr 19 2003, 11:41
I know that Fudo can be a little eccentric, but throwing him off the Nvidia stand at CeBIT was really going far too far - unnamed AMD executive

THIS IS WHERE we see pricing and technology from ATI and Nvidia for their family of products close to the end of this year and the beginning of next year.

Remember that by this time next year, PCI Express is bound to upset the applecart and change all the rules of the game.

Up to $100 - R9200 family based on RV280 (still DX8 chip)
Up to $200 - R9600 family on RV350
Up to $300 - R9800 family on R350
Up to $400 - R9900 family on R350 with DDRII memory

Up to $100 - NV34 in a lot of different core/memory speeds (DX9!)
Up to $200 - NV36 the true half of NV35
Up to $300 - NV35 256 bit DDR + twice the floating power of NV30
Up to $400 - NV40 - probably we see it, as a Christmas present.

But we don't see any NV30/31 at the end of this year. Seems like this was a kind of error from Nvidia's point of view, and Nvidia has attempted to correct this "error" just as fast as it possibly can.

And it also seems as if NVDA is doing a sort of half step forward, not only will it have DX9 in all segments but it will also try to introduce the new generation chip NV40 sometime around the end of the year - a couple of months earlier than ATI's R400 schedule.

Let's explore this a little further.

The NV35 - DX9 chip with PS/VS 2.0+ functionality and base OpenGL 2.0 fragment shading capabilities. 4?2 architecture with 8?1 turbo mode for single texturing and with 2 universal floating/integer ALU and two TMU on each pipe. Also this will have twice the width internal register file, and this fact removes the known NV30 bottleneck and busts the complicated 2.0 shaders performance a lot.

NV36 is a true half slice of the the NV35 - with 2?2 or 4?1 modes. Pixel power is half as low but the vertex power is pretty much the same as the NV35.

The NV40-DX9 chip will have PS/VS 3.0 functionality and sophisticated OpenGL 2.0 fragment shading capabilities. 8?1 (only) plain architecture with two universal floating/integer ALU and one TMU on each pipe. So no more sophisticated schemes like 4?2 or 8?1 depending on the application, only raw power.

Yes, it's hard to believe right now, but probably by the end of the year history will repeat itself, and Nvidia stands to be a bit faster and a bit more revolutionary, but ATI will be less expensive and will give practically the same benefits.

So developers will be able to rest a little. Only NV40 will create problems for them, everything else will remain pretty similar right up to the end of the year.

So, guys, now's the time for to sit down and write a couple of 2.0 shaders, eh? µ


Share this:

blog comments powered by Disqus
Subscribe to INQ newsletters

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

INQ Poll

Heartbleed bug discovered in OpenSSL

Have you reacted to Heartbleed?