Introduction

Today marks the launch of NVIDIA's newest graphics cards: the 7900 GTX, 7900 GT and the 7600 GT. These cards are all based on an updated version of the original G70 design and offer higher performance for the dollar. Today we will see just how much faster the new NVIDIA flagship part is. But first let's take a look at what makes it different.

At the heart of this graphics launch is a die shrink. The functionality of the new parts NVIDIA is introducing is identical to that of the original G70 based lineup. Of course, to say that this is "just a die shrink" would be selling NVIDIA a little short here. In the future, if either NVIDIA or ATI decide to move to TSMC's newly introduced 80nm half-node process, all that would be involved is a simple lithographic shrink. Sure, things might get tweaked a little here and there, but the move from 90nm to 80nm doesn't involve any major change in the design rules. Moving from 110nm to 90nm requires NVIDIA to change quite a bit about their register transfer logic (RTL), update the layout of the IC, and verify that the new hardware works as intended.

The basic design rules used to build ICs must be updated between major process shrinks because the characteristics of silicon circuits change at smaller and smaller sizes. As transistors and wires get smaller, things like power density and leakage increase. Design tools often employ standard components tailored to a fab process, and sometimes it isn't possible to drop in a simple replacement that fits new design rules. These and other issues make it so that parts of the design and layout need to change in order to make sure signals get from one part of the chip to another intact and without interfering with anything else. Things like clock routing, power management, avoiding hot spots, and many other details must be painstakingly reworked.

In the process of reworking the hardware for a new process, a company must balance what they want from the chip with what they can afford. Yield of smaller and smaller hardware is increasingly affecting the RTL of a circuit, and even its high level design can play a part. Making decisions that affect speed and performance can negatively affect yield, die size, and power consumption. Conversely, maximizing yield, minimizing die size, and keeping power consumption low can negatively affect performance. It isn't enough to come up with a circuit that just works: an IC design must work efficiently. Not only has NVIDIA had the opportunity to further balance these characteristics in any way they see fit, but the rules for how this must be done have changed from the way it was done on 110nm.

After the design of the IC is updated, it still takes quite a bit of time to get from the engineers' desks to a desktop computer. After the first spin of the hardware comes back from the fab, it must be thoroughly tested. If any performance, power, or yield issues are noted from this first run, NVIDIA must tweak the design further until they get what they need. Throughout this entire process, NVIDIA must work very closely with TSMC in order to ensure that everything they are doing will work well with the new fab process. As microelectronic manufacturing technology progresses, fabless design houses will have to continue to work more and more closely with the manufacturers that produce their hardware in order to get the best balance of performance and yield.

We have made quite a case for the difficulty involved in making the switch to 90nm. So why go through all of this trouble? Let's take a look at the benefits NVIDIA is able to enjoy.

NVIDIA's Die Shrink: The 7900 and 7600
POST A COMMENT

97 Comments

View All Comments

  • DigitalFreak - Thursday, March 09, 2006 - link

    I saw that as well. Any comments, Derek? Reply
  • DerekWilson - Thursday, March 09, 2006 - link

    I did not see any texture shimmering during testing, but I will make sure to look very closely druing our follow up testing.

    Thanks,
    Derek Wilson
    Reply
  • jmke - Thursday, March 09, 2006 - link

    just dropped by to say that you did a great job here, plenty of info, good benchmarks, nice "load/idle" tests. not many people here know how stresfull benchmarking against the clock can be. keep up the good work. Looking forward to the follow-up! Reply
  • Spinne - Thursday, March 09, 2006 - link

    I think it's safe to say that atleast for now, there is no clear winner with a slight advantage to ATI. From the bechmarks, it seems that the 7900GTX performs on par with the X1900XT with the X1900XTX a few fps higher (not a huge diff IMO). The future will therefore be decided by the drivers and the games. The drivers are still pretty young and I bet we'll see performance improvements in the future as both sets of drivers mature. The article says, ATI has the more comprehensive graphics solution (sorta like the R420 vs/s the NV40 situation in reverse?), so if code developers decide to take advantage of the greater functionality offered by ATI (most coders will probably aim for the lowest common denominator to increase sales, while a few may have 'ATI only' type features) then that may tilt the balance in ATI's favor. What's more important is the longevity of the R580 & G71 families. With VISTA set to appear towards the end of the year, how long can ATI and NVIDIA push these DX9 parts? I'm sure both companies will have a new family ready for VISTA, though the new family may just be a more refined version of the R580 and G71 architectures (much as the R420 was based off the R300 family). In terms of raw power, I think we're already VISTA games ready.
    The real question is, what does DX10 bring to the table from the perspective of the end user? There were certain features unique to DX9 that a DX8 part just could not render. Are there things that a DX10 card will be able to do that a DX9 card just can't do? As I understand it, the main difference between DX9 and 10 is that DX10 will unify Pixel Shaders and Vertex shaders, but I don't see how this will let a DX10 card render something that a DX9 card can't. Can anyone clarify?
    Lastly, one great benefit of Crossfire and SLI will be that I can buy a high end X1900XT for gaming right now and then add a low end card or a HD accelerator card (like the MPEG accelerator cards a few yyears ago) when it is clear if HDCP support will be necessary to play HD content and when I can afford to add a HDCP compliant monitor.
    Reply
  • yacoub - Friday, March 10, 2006 - link

    quote:

    The future will therefore be decided by the drivers and the games.


    And price, bro, and price. Two cards at the same performance level from two different companies = great time for a price war. Especially when one has the die shrink advantage as an incentive to drop the price to squeeze out the other's profits.
    Reply
  • bob661 - Thursday, March 09, 2006 - link

    I only buy based on games I play anyways but it's good to see them close in performance. Reply
  • DerekWilson - Thursday, March 09, 2006 - link

    The major difference that DX9 parts will "just not be able to do" is vertex generation and "geometry shading" in hardware. Currently a vertex shader program can only manipulate existing data, while in the future it will be possible to adaptively create or destroy vertecies.

    Programmatically, the transition from DX9 to DX10 will be one of the largest we have seen in a while. Or so we have been told. Some form of the DX10 SDK (not sure if it was a beta or not) was recently released, so I may look into that for more details if people are interested.
    Reply
  • feraltoad - Friday, March 10, 2006 - link

    I too would be very interested to learn more about DX10. I have looked online but I haven't really seen anything beyond the unification u mentioned.

    Also Unreal 2007 does look ungodly, and I didn't even think to wonder if it was DX9 or 10 like the other poster. Will it be comparable to games that will run on 8.1 hardware sans DX9 effects? That engine will make them big bux when they license it out. Sidenote-I read they were running demos of it with a quad SLI setup to showcase the game. I wonder what it will need to run it at full tilt?

    BTW Derek I think you do a very good job at AT, I always find your articles full of good common sense advice. When U did a review on the 3000+ budget gaming platform I jumped on the A64 bandwagon (I had to get an AsrockDual tho, instead of an NF4 cuz I wanted to keep my AGP 6600gt, and that's sad now considering the 7900 gt performance/price in sli compared to a 7900gtx.) and I've been really happy with my 3000+ at 2250 it runs noticeably better than my XP2400M oc'd 2.2) I'm just one example of someone you & AT have made more satisfied with their PC experience. So don't let disparaging comments get you down. You thorough committment to accuracy of your work shows how you accept criticism with grace and correct mistakes swiftly. I think the only thing "slipping" around here are peoples' manners.
    Reply
  • Spinne - Thursday, March 09, 2006 - link

    Yes, please do! So if you can actually generate vertices, the impact would be that you'd be able to do stuff like the character's hair flying apart in a light breeze without having to create the hair as a high poly model, right? What about the Unreal3 engine? Is it a DX9 or DX10 engine? Reply
  • Rock Hydra - Thursday, March 09, 2006 - link

    I didn't read all of that, but I'm glad it's close becasue the consumer becomes the real winner. Reply

Log in

Don't have an account? Sign up now