RV250 And R300 Pics Speculations ...

[David G.]

A very interesting hypotesis surfaced Rage3D a while ago from some very smart dude that the COMPUTEX card is a RV250 and not a R300.

He said that if the latest specs rumors about Rv250 said it was just 300/600 he speculated that the COMPUTEX card could be , in fact a faster setup for RV250 like 350Mhz CORE + 700 Mhz Memory ...

The latest pictures on the web present 2 cards that look very similar to the Radeon 8500 ... at least the R300 pic looks very similar to the Radeon 8500 as the new setup for RV250 is quite different after looking at the trace density ...

Could this be so ? After all ATi said that the first card to be anounced is RV250 in different types of setup like 200/200 , 250/250, 300/600 ...

I'm very dissapointed by the benchmark results and I don't think a company like ATi would release an inferiour product .

The only think that puzzeles me is the fact that the R300 benchmark pics were presenting the card as a R300_E44 ....

 

[David G.]

Isn't there two version of the Radeon 9000 ? 9000 and 9000Pro ?

Maybe that other card is a Radeon 9000 Pro on 350/700 ....

 

[MuFu]

I will eat my own faeces if the Computex board on VIA's stand was an RV250. Actually, wait, no I will buy an R300 when they are available and send it to you for free (and eat my faeces... and yours too).

MuFu.

 

[MuFu]

BTW, the "R300_E44" driver is initialized through an ASIC ID association; R300's registered ASIC ID is 1002 (ATi's Vendor ID) - 4E44, for the non-DC board anyway... I think the DC configs have different IDs; 4xxx.

Look at the pics; PN 109-94200-00A - that's the Computex board. The pics from PCPOP show PN 109-94200-00B - this is the second PCB revision, which I believe is receptive to both the A11 and A12 steppings of R300, hence the floppy-style power connector. Anyway... none of this matters, the crucial thing is that 942 IS the R300. RV250 boards are 958 and 964, G5000 (TYAN) and RV25A (PowerColor), of those I know about. You can also tell it's an R300 by looking at the serial and active termination resistor packets (4 SMT resistors in a pack) inline with the memory bus traces - there are twice as many as on a card with a 128-bit memory bus. The trace density is also pretty astonishing, bearing in mind some traces are on internal layers. Another thing is the sticker; RV250 is not AGP 8x compliant, and the timing; by the time Computex started there were already production RV250 boards ready to go. The R300 boards weren't even using respun silicon (A12) yet.

I'm very dissapointed by the benchmark results and I don't think a company like ATi would release an inferiour product .

Inferior to what? It's superior to the GF4MX in every way. 25% cheaper to make, about 40% faster, has the daughtercard options, is DX8.1 compliant, etc etc...

MuFu.

P.S. Here is a shot of the R300 PCB (00A), close-up:

http://muthafunker.homestead.com/files/miscmessageboard/r300bga.jpg