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
Alx - Wednesday, June 22, 2005 - link
Face it, this launch isn't gonna hurt anyone except people with minds too small to accept that there is simply one more option than there was before. If you liked pc gaming yesterday, then there is no reason why this launch should make you stop liking it today. Unless you're a retarded buttbaby who can't handle choices. In that case please get a console and stop coming to this site.mlittl3 - Wednesday, June 22, 2005 - link
#82, WesleyWell that sucks that ya'll have lost your web editor for awhile. Especially when there is so much cool hardware coming out around now. In our research lab, we pass around our publications and conference posters to others in the group so that a fresh pair of eyes see them before they go live or to the journal editor. But of course, everyone else at AT is also busy so oh well.
Good work guys and I look forward to the "new CPU speed bump" article (or FX-57 for those not under NDAs).
Mark
PS. If ya'll have an opening for another web editor, you should hire #84 (ironchefmorimoto). I hear he can cook really well.
AtaStrumf - Wednesday, June 22, 2005 - link
Nicely corrected Derek, I think there are just a few typos left, like this one (**):Page 20
Power Consumption
We measured power consumption between the power supply and the wall. This multiplies essentially amplifies any differences in power draw because the powersupply is not 100% efficient. Ideally we would measure power draw of the card, but it is very difficult **determine** to determine the power draw from both the PCIe bus and the 12V molex connector.
AND a few double "Performances" in the title (Performance Performance) starting with page 10.
Nice card nVidia!!! I hope ATi isn't too far behind though. Crossfire --> cheap SLi ;-) I need a nice midrange product out by September when it'll be time to upgrade to a nice E6 stepping S939 A64 and something to take the place of my sweet old GF2 MX (I'm not kidding, I sold my 6600GT AGP, and now I'm waiting for the right time to move to PCIe).
IronChefMoto - Wednesday, June 22, 2005 - link
Amen -- you guys work hard on your articles. Keep up the great work. And don't f*cking bother the web editor. We...er...they don't get enough vacation as it is.IronChefMorimoto
(another web editor who needs a break)
Wesley Fink - Wednesday, June 22, 2005 - link
Derek was too modest to mention this in his comments, but I think you should know all the facts. Our Web Editor is on vacation and we are all doing our own HTML and editing for the next 10 days. In our usual process, the article goes from an Editor to the Web Editor who codes the article, checks the grammar, and checks for obvious content errors. Those steps are not in the loop right now.The next thing is NDA's and launches. We are always under the gun for launches, and lead times seem to get shorter and shorter. Derek was floating questions and graphs last night at 3 to 4 AM with an NDA of 9AM. Doing 21 pages of meaningful commentary in a short time frame, then having to code it in HTML (when someone else normally handles that task), is not as easy as it might appear.
I do know Derek as a very conscientious Editor and I would ask that you please give him, and the rest of us, a little slack this next week and a half. If you see errors please email the Editor of the article instead of making it the end of the world in these comments. I assure you we will fix what is wrong. That approach, given the short staff, would be a help to all of us. We all want to bring you the information and quality reviews you want and expect from AnandTech.
IronChefMoto - Wednesday, June 22, 2005 - link
#79 -- But why wouldn't it be a high quality article, mlittl3? I thought you told me that AT was infallible? Hmmm? ;-)Houdani - Wednesday, June 22, 2005 - link
Thanks for the refresh, Derek. I went back and took a peek at the revised graphs. I have a couple of comments on this article before you move on to the next project.>> When the Splinter Cell page was refreshed, the graph for 20x15x4 apparently disappeared.
>> When you removed the SLI's from the Guild War page, it looks like the 7800GTX changed from 50.5 to 55.1 (which is the score previously given to the 6800 Ultra SLI).
>> Several of the pages have scores for no AA benches listed first, while other pages have scores for the 4xAA listed first. While the titles for the graphs are correct, it's a little easier to read when you stay consistent in the ordering. This is a pretty minor nit-pick, though.
>> Thanks for updating the transparency images to include mouseover switches ... quite handy.
fishbits - Wednesday, June 22, 2005 - link
"They priced themselves into an extremely small market, and effectively made their 6800 series the second tier performance cards without really dropping the price on them. I'm not going to get one, but I do wonder how this will affect the company's bottom line."The 6800s were "priced into an extremely small market." How'd that line turn out? I don't imagine they've released this product with the intention of losing money overall. Why do you think retailers bought them? Because they know the cards won't sell and they're happy to take the loss? It's already been proven that people will pay for you to develop and sell a $300, wait $400, wait $500 video card. It's already been proven that people will pay a $100+ premium for cards that are incrementally better, not just a generation better. Sounds like this target is a natural, especially knowing it'll eventually fall into everyone else's purchasing ability.
Being able to say you have the bar-none best card out there by leaps and bounds is certainly worth something. Look at all the fanboys that are out there. Every week or month you're able to stay on top of the benches means you get more people who'll swear by your products no matter what for years to come. Everyone you can entice into buying your card who sees it as a good product will buy your brand in the future as a preference, all other options being equal. I could be wrong, but suspect Nvidia's going to make money off this just fine.
-----------------------
"I am proud that our readership demands a quality above and beyond the norm, and I hope that that never changes. Everything in our power will be done to assure that events like this will not happen again."
See... that's why I'm a big fan of the site.
mlittl3 - Wednesday, June 22, 2005 - link
#78, I bet you didn't even read the article. How do you know it demonstrated editoral integrity?IronChefMoto - Wednesday, June 22, 2005 - link
#23 (mlittl3) still can't pronounce "Penske" and "terran" right, regardless of the great editorial integrity demonstrated by the AT team today. Thanks!