more bad info from the inq?

[plat]

http://www.theinquirer.net/?article=17112

We didn't know about the R430 chip but then again it's just the R420 built at a 110 nanometres. It's entirely natural that ATI would switch to a smaller process and save some space on its wafers as an ultimate goal and to have as many dies as possible out of a silicon sliver. It's not safe, though, saying that 110 nanometres means cooler chips and we learned this lesson from Mr Prescott.

i thought that low-k wasn't offered on 110nm so wouldn't that be a step backwards?

 

[Nebuchadnezzar]

Maybe they just want to produce cheaper.

 

[MuFu]

Dunno about the rest, but this bit is "bad info":

It's not safe, though, saying that 110 nanometres means cooler chips and we learned this lesson from Mr Prescott.

There is a performance-optimised 0.11u process. Hard to say how it compares to 0.13u w/a low-k dielectric. Perhaps a combination of the dice-per-wafer benefits and thermal characterstics have brought it into favour.

Meh.

 

[Pete]

"Mr Prescott" is 90nm and packs a ton of cache, no? It seems a bit of a stretch to generalize from Northwood to Prescott, given all the previous ones (e.g., R360 to R420 = ~ same power draw, double performance).

 

[jvd]

I'm sure it will be cheaper . Perhaps with refinement they will be able to get the chips running as fast and as cool as the 130nm low k ones.

 

[JCLW]

"Mr Prescott" is 90nm and packs a ton of cache, no? It seems a bit of a stretch to generalize from Northwood to Prescott, given all the previous ones (e.g., R360 to R420 = ~ same power draw, double performance).

Banias (130nm)-> Dothan (90nm) also did extremly well.

 

[Tim Murray]

Bad info from L'Inq? Perish the thought!

Seriously, folks, how can you buy any of this stuff? Fudo also said that the GeForce 6800, the NV40GL, would be capable of outputting 32 color-only no-Z pixels per clock.

 

[Pete]

The same Fudo who didn't know Parhelia had a 256-bit bus? GET OUT!

 

[Dave Baumann]

Fudo had a bit of a rash of dodgy reports that day. For instance his report on RV410's positioning had this:

Fudo[/url]]Marchitecturally, the RV410 will be a RV380 build on 110 nanometre process. It's still Shader Model 2.0, if you are interested in such things, rather than seeing pictures on a nice screen.

This is somewhat silly since we already know that RV370 is functionaly idendical to RV380, the only difference being the process - this being the case, why produce Rv410 as another equivelent? Of course, the code name should suggest some differences as well.