Instead of just reposting my opinion again, I'll just link to the many times we (as a forum) have discussed this before and my opinions have been stated:
Impressions of the GF 4 MX...
Feb 2002
Feb 2002
Mar 2002
First impression of the 9000...
Jul 2002
My slightly different impression after a chart showing that ATI had described the number as not being based on DX 9...
Nov 2002
Note that my "defense" of the 9000 kicks in when people say it is the "same as" or "worse than" the GF 4 MX...I pretty much tend to agree with criticisms otherwise, especially as based on performance (see my comments on the "9100" below).
Now, to put it in the context of the 4800 naming problems from nvidia...
Nov 2002
Note that the "8500 OEM" I'm criticizing is not the 8500LE, but the reduced speed cards released before that naming effort took effect. ATI's only "saving grace" (as far as them being negligent rather than intentionally malicious in the lack of identifying the down-clocked cards) in that regard that they have a history of going "both ways" (i.e., they released the original Radeon with an unannounced speed
increase at one point...dubbed by the community as "SE" cards).
And to repeat my outlook on the "9100" name...
Dec 2002
My view on the 9200 name, besides its being a continuation of the above problems, will depend on release clock speeds, as I've said recently. Makes the parallel to the 4800 SE even stronger, because it goes against their own naming consistency.
Oh, and IMO, the 5200 is similar to the 9000 (and likely 9200) problem (theoretically, depending on the vertex shader issue), but better since the feature set is consistent with the "GF FX" monicker. The performance relative to the prior generation high end (4600) may be lacking like the 9000 was (though we don't know to what degree as of yet), and therefore
some of the features it offers may only be theoretical, but this is nowhere in the same league as the GF 4 MX! See the above posts for how I view such arbitrary equivalency.
Of course, a name like GF 5 MX might have been best to illustrate something was lacking, since the vertex shader
assumptions we are making at this point could make the card a significant problem for some people. Also, GF FX 5000 would have been less misleading than the "4200" parallel implied, but I think we should put the GF 4 MX parallels behind us (Hmm...though I wonder if we can put the GF 4 MX itself behind us as of yet?)
For some odd reason my point on the 9000 (RV2xx) was twisted to mean the 9100 and in my hurried graph throw-together I didn't proofread the reply about the 9100 when if ya see my previous post, was entirely referencing the 9000.