Some numbers comparing the 8800 GTX & the X2800 XTX.

Awesome! But are these simple or complex shader op's you counted?

X2800XTX just KILLS. 'nuff said.

Um...I took the numbers from the thread I linked to for X2800 XTX and then I used our overclocker tool thingy to get the numbers for the overclocked 8800 GTXs. That is all I did, and that is all I know.

Ok sorry man did'nt mean to be rude.

1. The Shader Operations on the G80 are NOT 87,200. They are much higher because they use the Shader Clock.
2. Does the G80 really have more TMU's? If so, it should be better than he R600 already.

I dunno why, but i think its bs again. like all the other tables, what I have found...

Track. The chart does not say the G80 can perfrom 87,200 Shader Ops/sec. I don't know what a TMU is. Texture Mapping Unit maybe?

Nowhere are the G80 shader ops listed as 87,200, Track.

Uh, I mean 83,200.

Well i think this will be interesting. From what i've read it seems to me that ati is stalling for some reason.

They are Float 1 only to get the float 4 shader op's on the G80 divide by 4 the stream procs and multiply by the stream proc clock(I dont think it can really do float 4 unless you're talking hypothetically) And to get the float 1 op's of any 7-series card or x1k and before series multiply by 4 so the x1950xtx has 124800 float 1(or simple shader op's) and the 7900gtx has 62400 float 1. The R600 should have 192000 float or 48000 float 4 op's.(Man i gotta stop being so precise with everything all the time)

So ur saying that FOUR scalar pipes are the same as a SINGLE Vec4, and thus it would be possible to know how many shaders a Scalar has by dividing it by 4 and multiplying it by the core clock?

Well, i dont know if thats true.. cause things arent coming in 4's anymore. So its probably not accurate.

But it wouldnt have been accurate anyway, since u must use the shader clock and not the core clock.

I'm talking hypothetically. The g80 cant do real vec4 shader op's but it can do simple shader which are 1/4 of a complex shader.

They are not exacly 1/4 of a Vec4. They are better than that.

Sorta right but scalars can clock higher then vec4.