Final Words

With the insertion of the NVIDIA GeForce 7800 GT into the high end landscape, it becomes difficult for us to provide our readers with a one size fits all answer. Recommendations are complex because we must take into account current hardware, performance needs, price, and future upgrade prospects. Buying a graphics card right now poses quite a difficult situation.

From the numbers it is very clear that the 7800 GT is a better card than the 6800 Ultra. For the price, there is no contest between the two. The 6800 Ultra is no longer a viable solution except as an upgrade to SLI from a previously owned Ultra. That being said, paying about $400 for a second 6800 Ultra may be a good solution if everything else is already there and ready to go. For those who have a non-SLI PCI Express system and only a 450 W power supply, spending the money on a 7800 GTX is a much better upgrade than spending the money for a decent PSU, SLI motherboard, and 6800 Ultra. The best case scenario there will cost something like $650. If a system upgrade is in your future, it may be better to save your money until the new AMD processors and sockets come along.

We feel that the verdict for 6800 Ultra owners is to hold their cards and wait rather than go for the SLI upgrade. Those with less than last year's high end may be interested in the upgrade to the 7800 GT, but there are some cautions we would make in recommending this card. The current street prices show the 7800 GTX hitting as low as $500 (and a few OEM models are even lower). With many of the 7800 GT cards on the market hovering at $450, throwing the extra $50 at a GTX is a worthy upgrade for the performance increases gained if you need the extra push to get the quality and resolution desired. If high resolution and solid performance are needed, the GTX will not disappoint.

Here is where everything gets a little hairy. Will you be monitor limited by an upgrade? Many very good flat panels will only display up to 1280x1024 or 1600x1200. Owning one of these panels really negates the necessity for a 7800 GTX right now. From the numbers we have seen, the highest end card we would recommend for owners of a 1600x1200 flat panel (who don't get paid for gaming) is the 7800 GT. We would also recommend finding one for $400 rather than paying full price.

Owners of 1280x1024 panels don't really need to spend the money here as even the 7800 GT will run into walls at this resolution. 6800 GT prices are now dropping to $300 and below. At this price point, owners of older systems who haven't taken the high end plunge and don't want to spend the money on 1600x1200 and up panels would do very well to upgrade to a 6800 GT. Even the venerable 128MB 6800 is available at nearly $200.

We have entered an era where the graphics solution is often limited by the monitor more than anything else at the high end. It is important to pay attention to what is needed for a particular system rather than just grabbing the fastest thing out there. Even with all the bells and whistles, if nothing is going to be done at very high resolutions, current games do not stress the highest end cards enough.

Granted, when games based on technology like the Unreal Engine 3 come along, that 7800 GTX SLI system will come in handy at any resolution. However, we don't like recommending buying now for some promise of longevity or future performance. When games that push these cards come along, new cards will be out (or just around the corner). It is always best to buy for current needs.

That brings us back to the SLI situation and we have to stress that using SLI as an upgrade path isn't the best idea. SLI is best exploited as an ultra high end technology. For that reason, we won't recommend going out and buying a 7800 GT SLI system as spending less money on a single GTX (or a little more on two) are both more valuable options. This time around we can recommend 6800 Ultra SLI as a special case upgrade for current owners of SLI systems and a single 6800 Ultra who want added performance and need to choose between the 7800 GT and a second 6800U card. This recommendation is based on the fact that no more API features or technologies have been added to the GeForce 7 series that make it desirable over a 6 Series setup of the same speed.

Overall, our feeling is that the 7800 GT is a suitable replacement for the 6800 Ultra in the market. Keeping the relatively new, lower yield G70 silicon relegated to the very high end and eliminating (or severely reducing production of) the lowest yield NV4x part will help NVIDIA to maintain higher margins. The fact that plenty of 7800 cards are available now (including the GT) is a huge improvement over what we saw last year on the high end from NVIDIA with very low availability of high end parts. Keeping this trend rolling as long as they have is absolutely wonderful. We are very impressed with NVIDIA and we hope ATI will be able to follow this trend when they finally release Crossfire and their R520 based parts.

For now, NVIDIA's 7800 series own the high end graphics space. Now it's time for monitor makers to catch up and start releasing panels that can make use of the available dual-link DVI port at reasonable prices.

Power Consumption
Comments Locked

77 Comments

View All Comments

  • dwalton - Friday, August 12, 2005 - link

    "I would like G70 technology on 90nm ASAP, I have a feeling Nvidia didn't do a shift to 90nm for NV40 for a reason, as that core is still based on AGP technology, and Nvidia currently doesn't have a native PCI-E part for 6800 Line, they are all using HSI on the GPU substrate from the NV45 design."

    I believe Nvidia didn't want another 5800 fiasco. They probably determine a long time ago that 110 nm was a safer bet and used the 6600 as a guinea pig. Having a sucessful launch of the 6600 gave them confidence that manufacturing a 110nm g70 would be painless process.

    Futhermore, the 7600 will be a midrange card and will target a market segment that is more than likely dominated by AGP boards. So NV40 based 7600 would make perfect sense since the majority of the 7600 sold wouldn't require a HSI chip.

    "Let's faice it for the time being, were not going to be getting fully fucntional high end cores at the 199US price point with 256Bit Memory Interface, so far we have gotten things like Radeon X800, Geforce 6800, 6800 LE, X800 SE, X800 GT. Etc etc. It just doesn't seem profitable to do so."

    The X800 GT is a 256Bit Memory Interface Card and targets the 6600 GT segment.
  • coldpower27 - Friday, August 12, 2005 - link

    "The X800 GT is a 256Bit Memory Interface Card and targets the 6600 GT segment."

    I guess you missed reading the fully functional part, as the X800 GT does not comply with this statement.

    I guess I didn't get my meaning right, when I said G70 technology, I was talking about the mainstream cards going to 90nm not the 7800 GTX/GT.

    For a mid range part the risk would be reduced for going to 90nm as the core is not quite as complex, Nvidia did make safe bet to go to 110nm for their high end cards, I am asking for a G7x technology based performace (199US) card on 90nm technology. Not on the high end.

    Targeting PCI-E now would be a good idea as there are now boards on both sides that have PCI-E support for a decent amount of time, and it's the more forward thinking marchitecture, not to mention the possibility of power conusmption reduced enough on the 7600 GT to put it solely on the PCI-E bus if the Bridge Chip didn't exist. There isn't much point in designing a native AGP chip now, unless your talking about the value segment where margins are extremely thin per card.

    For the AGP users, I believe they can continue to use 110nm NV48, but I would like for PCI-E users to benefit from a 7600 GT 90nm PCI-E native card, with possible bridging to AGP if demand calls for it. There isn't much point of calling the mianstream card a 7600 GT if it's not based on G7x technology. We don't want Nvidia to follow ATI's lead on that kinda front. :)
  • neogodless - Thursday, August 11, 2005 - link

    I mainly agree with you, and who knows, such things could be in the work. But "simple process shrink"? I get the feeling that's a contradiction!

    Let us not forget the mistakes of the past... like the FX5800 and its "simple process shrink".
  • dwalton - Thursday, August 11, 2005 - link

    The FX5800 was Nvidia attempt to introduce a new high end architecture on a new process (130nm) it had never used before. Just like ATI is doing now. The 6xxx lines (130nm) is not new tech so producing a mature NV40 architecture on 90nm or 110nm should go alot smoother. Even at 130nm the NV40 is smaller than the G70 at 110nm (287mm² vs. 334mm²). Moving the NV40 to 90nm would reduce die size to ~200mm². Look at the 6600Gt: 110nm, 150mm², 8 pipes(?) vs. a 90nm NV40 ~200mm², 16 pipes.

  • JarredWalton - Friday, August 12, 2005 - link

    Unless they can make a 90nm "7600GT" part backwards compatible (via SLI) with the 6800GT, NVIDIA is in a position of "damned if you do, damned if you don't." As a 6800GT owner, I'd be rather sad to suddenly have the promise of upgrading to SLI yanked away.
  • dwalton - Thursday, August 11, 2005 - link

    Also, mature driver support at introduction.
  • Sunbird - Thursday, August 11, 2005 - link

    I don't know. At least they arent doing those ugly case reviews anymore, but they sure are still making me feel alienated.

    Thats first page smacks of elitism. Why can't we average people with a 5900XT (or even 5200) upgrade to say a 7600 that uses less power and thus is less noisy and easier to cool to than a 6600 or 6800?

    I wonder which of the 2 authors wrote that paragraph?

    I guess this could be a symptom of Anand and his Apple usage, cause Apple people are often very elitist. Or it could be that they want to be the upmarket tech website for people with lots of money and think Toms Hardware is better suited to us unwashed (FX 5200 weilding) masses.

    Actually, this whole new colourscheme smacks of cold sauve elitism! Not the warm yellowish homey feel of old....

    ;(
  • Shinei - Thursday, August 11, 2005 - link

    As has been pointed out in this very comments section, a 7600 release would be redundant because there already is a 16-pipe, 350MHz part with 6 vertex pipelines: The 6800GT. There is no elitism, it's the raw fact that a 7600GT would be identical to a 6800GT in specifications and (most likely) performance, rendering it pointless to spend time fabricating when the 6800GT serves just as well.

    As for the article, I noticed that the 7800GT was outperformed by the 6800U in some SLI applications (like UT2004). Is that related to memory bandwidth, or is that a driver issue because of the 77.77 beta drivers you tested with?
  • Sunbird - Friday, August 12, 2005 - link

    As has been pointed out in my very own comment, "upgrade to say a 7600 that uses less power and thus is less noisy and easier to cool to than a 6600 or 6800?"

    And anyway its about price point, will the 6800GT cost as much as the 6600GT then?

    I want 6800GT (aka 7600) performance at the $200 to $250 pricepoint and I want it now!

    I'd settle for a 7200 that performs like a 6600 too.
  • DerekWilson - Friday, August 12, 2005 - link

    Yes, the 6800 GT will come down to $250 and likely even more over the next few months. You can already buy a 6800 GT for $270 (from our realtime price engine).

    The 6800 GT is not a noisy part. The HSF solution for the 7800 GT is strikingly similar. A lower performance G70 part may run cooler and draw less power, but, again, the 6800 GT is not a power hog.

    There really is not a reason for us to want a lower performing G70 part -- prices on 6 Series cards are falling and this is all we need. Even if NVIDIA came out with something like a "7200 that performs like a 6600", the 6600 would probably be cheaper because people would the 7 means more performance -- meaning the 6600 would be a better buy.

Log in

Don't have an account? Sign up now