nv31 future

[vb]

just my thoughts.

I don't quite see a reasoning for the three card line-up in nv products (nv34, nv31, nv35), and each with two core clocks. Most surprising are nv31 and nv34 ultra with very comparable performance. yet for a low end card nv 34 ultra is very expensive (bga memory, high cost pcb, high power consumption) and with nv36 very soon it is getting very crowded.

here's what I think: having so much problems with nv30 on 0.13µ, they ran two low-end projects, pin-compatible, one on 0.15µ, one on 0.13µ, nv34 only purpose to have something to sell on nv31 pcb until yelds ramp up. considering nv31 is all but totally missing, as they get good yelds on it, and (sorry if i repeat myself) they run on the same pcb it will soon be cheaper for nvidia to sell nv31 chips then nv34.

All that makes nv34 a very temporary product with sub-par, "crappy", performace even by nv standards, soon to be replaced by nv31 on low-end as it was originally designed for. that makes understandable the performance gap to the gf4 ti.

of course all that holds if they can get good yelds on nv31. if it is the same "quality" design as nv30 they may not, and it may get cancelled like the other .13µ contemporary chip. in that case do you think they'll keep the nv34 or get replaced with a cut-down/lower-speed nv36?

 

[Arun]

Won't happen I fear...
1. The NV31 die size is bigger than the NV34's, even when using 0.13u
2. nVidia is getting not-so-good yields on the NV31, but they ARE getting good yields on the NV34

What will happen, I suppose, however, is that the NV34U will be replaced by a NV31. Still, the role of the NV36 is still very messy in all this, since we don't even know exactly what market segments it is in - we know it's mid-end, and that it'll *partly* replace the NV31 - but the flipchip one ( the NV31U ) will still be on the market.

Eh, sooo messy...

 

[kemosabe]

Won't happen I fear...
1. The NV31 die size is bigger than the NV34's, even when using 0.13u
2. nVidia is getting not-so-good yields on the NV31, but they ARE getting good yields on the NV34

What will happen, I suppose, however, is that the NV34U will be replaced by a NV31. Still, the role of the NV36 is still very messy in all this, since we don't even know exactly what market segments it is in - we know it's mid-end, and that it'll *partly* replace the NV31 - but the flipchip one ( the NV31U ) will still be on the market.

Eh, sooo messy...

There's just been too much back-and-forth respinning and repositioning of these mainstream/performance cores lately. First the NV31 revision leaves the average consumer not knowing whether they're getting the slower card or the newer 400/400 version. Now ATI is reportedly going to deliver a boosted RV350 of its own, probably leaving the current 9600 Pro somewhere in the twilight zone as well.

It's also getting quite tiresome to see major OEMs (who complain about poor PC sales) still pushing the ancient GF4MX on their new desktop configurations. NV34 has started winning some of those slots, but it's been a slow transition. If RV350 yields are so good, I still wonder what ATI is waiting for to position the vanilla 9600 as a low-end/mainstream DX9 solution and completely relegate the 9200 core to the integrated market. :?