NVIDIA's GeForce 7800 GTX Hits The Ground Running
by Derek Wilson on June 22, 2005 9:00 AM EST- Posted in
- GPUs
Inside The Pipes
The pixel pipe is made up of two vector units and a texture unit that all operate together to facilitate effective shader program execution. There are a couple mini-ALUs in each shader pipeline that allow operations such as a free fp16 normalize and other specialized features that relate to and assist the two main ALUs.
Even though this block diagram looks slightly different from ones shown during the 6800 launch, NVIDIA has informed us that these mini-ALUs were also present in NV4x hardware. There was much talk when the 6800 launched about the distinct functionality each of the main shader ALUs had. In NV4x, only one ALU had the ability to perform a single clock MADD (multiply-add). Similarly, only one ALU assisted in texture address operations for the texture unit. Simply having these two distinct ALUs (regardless of their functionality difference) is what was able to push the NV4x so much faster than the NV3x architecture.
In their ongoing research into commonly used shaders (and likely much of their work with shader replacement), NVIDIA discovered that a very high percentage of shader instructions were MADDs. Multiply-add is extremely common in 3D mathematics as linear algebra, matrix manipulation, and vector calculus are a huge part of graphics. G70 implements MADD on both main Shader ALUs. Taking into account the 50% increase in shader pipelines and each pipe's ability to compute twice as many MADD operations per clock, the G70 has the theoretical ability to triple MADD performance over the NV4x architecture (on a clock for clock basis).
Of course, we pressed the development team to tell us if both Shader ALUs featured identical functionality. The answer is that they do not. Other than knowing that only one ALU is responsible for assisting the texture hardware, we were unable to extract a detailed answer about how similar the ALUs are. Suffice it to say that they still don't share all features, but that NVIDIA certainly feels that the current setup will allow G70 to extract twice the shader performance for a single fragment over NV4x (depending on the shader of course). We have also learned that the penalty for branching in the pixel shaders is much less than in previous hardware. This may or may not mean that the pipelines are less dependent on following the exact same instruction path, but we really don't have the ability to determine what is going on at that level.
Facebook
Twitter
RSS
127 Comments
View All Comments
Diasper - Wednesday, June 22, 2005 - link
oops posted before i wrote anything. Some of the results are impressive, others aren't at all. In fact results seem to be all over the board - I suspect drivers are something of the culprit and are to be blamed. Hopefully, as new drivers come out we'll see some performance increases or at least more a uniform distribution of good resultsDiasper - Wednesday, June 22, 2005 - link
Live - Wednesday, June 22, 2005 - link
Derek get cracking, the graphs are all messed up! And the Transparency AA Performance section could use some info on what game it is tested on and some more comments. I also think that each benchmark warrants some comments for all of us that have a hard time remembering two numbers at the same time. Keep it simple folks….Johnmcl7 - Wednesday, June 22, 2005 - link
I agree something is wrong with these results, I thought they were odd but when I got to the Enemy Territory ones they seem completely wrong - at 2048x1536 and 4xAA the X850XT is apparently getting 104 fps, while the 6800 Ultra gets 48.3 and the SLI 6800 Ultras are only getting 34.6 fps! Especially bearing in mind it's an OpenGL game.John
rimshot - Wednesday, June 22, 2005 - link
This has got to be an error by Anandtech, all other reviews show the 7800GTX in SLI at those same settings hammering the 6800Ultra in SLI.Lonyo - Wednesday, June 22, 2005 - link
The benchmarks are all a load of crap it seems.Check the Wolfenstein benchmarks.
The X850XT goes from 74fps @ 1600x1200 w/4xAA to 103fps @ 2048x1536 w/4xAA
A 33% increase when the res gets turned up. Good one.
There also seem to be many other similar things which look like errors, but they could just be crappy nVidia drivers, or something wrong with SLI profiles.
Who knows, but there's definately a lot of things which look VERY odd/suspicious here.
Dukemaster - Wednesday, June 22, 2005 - link
My Iiyama VMP 454 does 2048 no prob so i'm game :pvanish - Wednesday, June 22, 2005 - link
oh and in several of the benchmarks, the 6800U SLI more than doubles the performance over the single 6800U. Is that normal? I thought SLI gains were generally about 45% or so.rimshot - Wednesday, June 22, 2005 - link
Is it just me or is it a little strange that the 6800Ultra SLI outperforms the 7800GTX SLI at 1600x1200 with 4xAA in every benchmark ???PrinceXizor - Wednesday, June 22, 2005 - link
No comment on the fact that in virtually every game it LOSES to the 6800 SLI at 1600x1200 at 4XAA.All other scores look very impressive. But, in this particualar group of settings, the 6800 SLI eats it for lunch.
P-X