More next gen ATI info at the Inquirer

McElvis

Regular
FOR A WHILE NOW, we’ve been suggesting that ATI’s R300 would be called the Radeon 10000. We were close, but can be accused of jumping the gun a little, I guess. We learned today that the card that should be introduced as soon as the middle of this month and will be named differently.
The R300 will be called Radeon 9700, where the 9 stands for DirectX 9 support and the 700 could be memory speed, we imagine, or maybe 9700 simply sounds better than 4600, which Nvidia uses for their Geforce 4.

Radeon 10000 will be the label for the next R300, which will be introduced as soon as November and which will be produced using 0.13 micron technology. And it’s likely be faster as well.

So, the NV30 will have a 0.13 micron mate to compete with and we will bring you more on this as soon as this summer ends and we get more details.

The Radeon 9700 will have everything that it needs to be the performance leader for some time, we’re told -- at least until the NV30 arrives. Still, the real battle comes with Radeon 9000 aka RV250 and NV 17 and later NV18.

The money’s in the mainstream…

Would they really introduce two high end 3D cards within 3 months of each other?
 
Could explain how they aim to get 350-400MHz out of 106mil transistor chip. Lower clocked .15 chip shortly followed by the .13 version - cheeper (more margin) and faster. Could have nVidia worried.

IHowever there are some inconsistencies as it states R300 will be the Radeon 9700 where 9 stands for DX9, but then goes on to state that RV250 will be Radeon 9000, but i though RV250 is a DX8 part?
 
don't get me wrong, I think its nice speculating about the names and stuff, but the cards themselves 'd get me a lot more interested... ;)
 
mboeller said:
Would they really introduce two high end 3D cards within 3 months of each other?


maybe :

Radeon9000 = mainstream

Radeon9700 = performance

Radeon10000 = highend / professional

My take from MuFu's posts is;

RV250 = R200 refresh and is to R200 what the 7500 is to the original R100, i.e. faster (questionable if its 2 pipes rather than 4, unless it is clocked much higher), cheaper, (maybe has some minor architecture changes ) plus has this daughter card option for VIVO/DVI etc that Mufu stated. I dont understand how it will get to 350mhz-400mhz though.

if 9700=R300 on 0.15micron @ 250mhz, then a New Year/Spring 0.13 micron R300 will be a cheap and 'fairly' quick counter to the NV30.

Maybe ATI are getting into 6 month cycles of chip & refresh to maintain their build up of momentum in competing with NVidia.

I dont see anyone claiming 350-400mhz for the R300 anymore now its fairly certain it is on 0.15micron.

Of course this all assumes MuFu's sources are sound :)
 
Hello. :)

Well I have been in the dark for ages now too (about a month), so I'm sure things may have changed since then. Back to speculation for me...

I never received any specific information about the actual GPU architecture of RV250 so I'm only going on performance estimates. They state that the R200 is "about equivalent to an R200 given a certain clock". I'd wager it being slightly higher due to further tweaking of the memory interface. Sounds to me therefore that it will be a 4x2 rendering array, the same as R200. I can't imagine a 2-pipeline part yielding at ~<325MHz being on a par with an 8500 in most situations and it seems dumb to be-fit a mainstream card with DX8.1 and then cut off a multitexturing testicle. Obviously something has happened to allow a clockspeed boost and the integration of a second RAMDAC while keeping production costs low, most likely a slight die re-map (getting easier now that ASICs are cell-based and have very little original logic) combined with a very mature 0.15u process.

I can envisage a 64MB RV250 clocked at 300/300MHz performing about on a par with the current 128MB 8500, so that's maybe where we get statements from Anand etc (who have been briefed under NDA) that the new card performs about the same as an 8500.

Anybody know the origin of the "2 pipeline" rumour?

And what happend with Radeon A22?! Is that a 0.15u part? If not, then we have a revised 0.18u Rage6C that manages to hit much higher clockspeeds than previous steppings of the same core. Something to think about when considering the relationship between RV250 & R200...

MuFu.
 
if 9700=R300 on 0.15micron @ 250mhz, then a New Year/Spring 0.13 micron R300 will be a cheap and 'fairly' quick counter to the NV30.

Actually that's one possibility I've been straving for a long time. Just a 15um/R300 wouldn't be sufficient against the competition. But a respin even if it's early next year (which makes sense since 3-4 months is a bit optimistic for that/see Inq rumours) will definitely give NV a hard time.
 
I can't see ATi calling any of the cards Radeon 10000, especially after they critisised nvidias naming policy.

Imo, rv250 will be 8xxx, probably 8600 or 8900 if its faster than r200.
The r300 out soon seems likely to be 9500 and if they do a die shrink later then that could be 9600-9900 depending on what the firegl chips get called.
 
I do hope that ATI will follow it's new naming policy since it's easy, simple and nice.

We don't need another GeForce 4 MX problem.
 
Bambers said:
I can't see ATi calling any of the cards Radeon 10000, especially after they critisised nvidias naming policy.

Imo, rv250 will be 8xxx, probably 8600 or 8900 if its faster than r200.
The r300 out soon seems likely to be 9500 and if they do a die shrink later then that could be 9600-9900 depending on what the firegl chips get called.

Well for sure there will be a "Radeon 9000" card. People are translating mauals for it, distributers are taking orders for it and marketing people are making little logos for it etc as we speak. I have that on good [Cartman]AUTHORITAH![/Cartman]. Whether it is the R300 or the RV250 I do not know, although believe it's more likely that it's the RV250. In which case they will be making a mistake if they call the R300 "Radeon 9700", IMO. Sounds weird anyway... Radeon 10000 would be much more appropriate. :-\

The big deal with the GF4MX was that the the nomenclature associated it with the Ti. As long as ATi differentiate sufficiently between the two products (e.g. 9xxx and 10xxx) I don't see what the problem is. DX association is not important, IMO. If it was, all graphics card manufacturers would have adopted such a scheme by now.

MuFu.
 
Back
Top