Post your breaking NV30 news links here!

Toms shows it as being an 8 Pipe/1 Unit architecture.

show much for the 8/2

 

Another one --> http://www.nvnews.net/previews/geforce4_fx/index.shtml

and another --> http://www.anandtech.com/mb/showdoc.html?i=1749

So far, not impressed....

 

So far...pretty unimpressed.

I see no evidence of any better bandwidth savings techniques relative to R300. (That doesn't mean it can't be more effective than R300 GB for GB of course, but I see no evidence to suggeset some radical step change in bandwidth efficiency...)

And did I read Tom's correctly, that the "OATEs like" cooling solution will basically be the reference design?!

And poeple thought the R-300 was expensive to make?

 

http://www.extremetech.com/print_article/0,3998,a=33705,00.asp

"Number of Texturing Units per Pipe - Undisclosed"

There's hope......


WOW.......What's that!!!!! http://common.ziffdavisinternet.com/util_get_image/1/0,3363,sz=1&i=17841,00.jpg

 

Nvidia is keeping things to itself...

 

Those tech demos look good!

 

http://www.hothardware.com/hh_files/S&V/nv30_preview.shtml

 

The GeForce FX's occlusion culling algorithm has the capability to cull objects with a depth complexity of 1. - nV news

Explain. I'm serious, Is that cull objects to a depth complexity of one? Because if an object has a DC = 1, How can you cull an obect if it's the only thing there? This wording is confusing me.

 

NVIDIAs website has/is been updated

 

Not only what I expected, but maybe even a bit more disappointing.....

 

The cooling solution is huge!!!!
What are they thinking!!!!
I'm not going to buy that monster......
If that cooler stays, i'll buy a 9700

http://www.hardocp.com/images/articles/1037482125yomDI2gHPV_5_9_l.jpg

 

Agreed.
According to the most-biased NV-site, Tom's HW, GFFX sports 1 TMU per pipe - same as R9700...

 

Cool, I don't think I've ever been quoted from IRC before

 

Neither can I put it into my packed system. That's a really BIG fan.

 

aaah but buy a nForce2board with onboard sound/LAN etc and free up those PCI slots

 

From ExtremeTech's article:

"The GeForceFX weighs in at an imposing 4Gpixels/sec in the key pixel fill-rate area. It can't write that many pixels to memory though, because trying to move that many pixels around would outstrip memory bandwidth. But to give you an idea as to what it can move around, consider this example:

Let's start with an extreme case where 4X FSAA and trilinear filtering are enabled. Assume we're running with DXTC texture compression and getting a compression ratio of 4:1. We're assuming single-cycle processing for trilinear filtering and 4X FSAA for simplicity's sake. For the GeForceFX to just write 4Gpixels/sec into the frame buffer, here's what we calculate:

4 (number of Gpixels) * 4 (number of bytes per pixel using DXTC compression) * 8 (trilinear samples) * 4 (4X FSAA) = 488.3GB/sec.
Even dropping down to 32-bit color and using DXTC compression so we'd only have one byte per pixel, you'd still need 122.1GB/sec of bandwidth to write that many pixels into the frame buffer. Clearly impossible.

Now having shown you what GeForceFX can't do, let's show you what it can:

Let's assume a resolution of 1600x1200x128 at 60 fps with 4X FSAA and trilinear filtering enabled. Assume we're running with DXTC texture compression and getting a compression ratio of 4:1. for the GeForceFX to supply the needed pixels to meet this demand, here's what we calculate:

1600 * 1200 (resolution) * 4 (number of bytes per pixel using DXTC compression) * 8 (trilinear samples) * 4 (4X FSAA) * 60 (frame rate) = 5.1GB/sec.
Note: to get GB/sec we took the product of all these numbers and divide by 1024 three times. Without doing this you wind up with bytes/sec, and it's a huge number.
Going to 32-bit color with DXTC where you'd now have one byte per pixel, the memory bandwidth requirement drops to 1.3GB/sec. "

Heh...

 

Not that I am horribly upset or anything ... it just isn't the Radeon 9700 killer that I had originally imagined. I think given what I have seen here ATi may very well be able to dampen the retail launch of the NV30 using something based on the R300 core.

 

... and of course!

http://www.beyond3d.com/previews/nvidia/nv30gfx/

Interesting as they mainly compare the NV30 to the GF4...

One quick remark: 500 mHz core speed looks damn fast on paper, but we need much info on LMA III and FSAA + AF implementations before we can judge whether we have a GF2 GTS (massive fillrate vs limited memory bandwidth) issue at ours hands...

 

well we know nothing about performance so how can we jump to conlcusions?

 

I will not upgrade motherboard/CPU in the near future as I have just done it (P4 2.66GHz + Gigabyte 8IHXP).

If I have to find a place for it, for sure I can. But I have to remove a card or 4 USB outputs to do it.

Anyway, wait until the product hit the shelves first.