Final Thoughts

When we were first informed about the GeForce GTX 560 Ti With 448 Cores, I approached the matter with a great deal of skepticism. 3rd tier products have not been impressive in quite some time, and NVIDIA’s previous effort with the GTX 465 is a very good example of this. So imagine my surprise once we had a card in hand and benchmark results to work with. NVIDIA has both impressed me and disappointed me at the same time.

The hardware is impressive enough. GTX 570 is a good base to work off of both with respect to performance and operational characteristics – it’s well balanced and the GTX 560-448 directly inherits this. Perhaps most importantly NVIDIA didn’t make their 3rd tier product significantly worse than their 2nd tier in terms of its performance targets, and that makes a world of difference. As a result the GTX 560-448 is what we’d happily call a GTX 570 LE or GTX 565 in any other universe, because it’s certainly not as slow as a GTX 560 Ti.

On a larger scale, once we factor in AMD’s products things get a bit more murky. The GTX 560-448 is definitely faster on average, but as with every other GF100 card, this is heavily dependent on the game being tested. Throwing out CivV – a game where NVIDIA has a distinct advantage due to driver features – leaves things much closer between the GTX 560-448 and the Radeon HD 6950. The 6950 is on average $40 cheaper, and this cannot be ignored. As fast as the GTX 560-448 is, unless you’re specifically using it for games NVIDIA has an advantage in or need their ecosystem for, it’s just not $40 faster. AMD has made the 6950 a good value, and this can’t be ignored.

So if we’re generally impressed with the performance, what are we disappointed about? As you can probably expect however, the disappointing aspect is the name. Even if performance really was close to a GTX 560 Ti it still wouldn’t excuse the poor name. GF110 isn’t GF114, the SM layout and superscalar execution features make these distinctly different GPUs whose differences cannot be reconciled. This is particularly evident when it comes to things such as FP64 performance where the GTX 560-448 is going to be much, much faster; or in cases where the architecture differences mean that the GTX 560-448 isn’t going to pull well ahead of the GTX 560 Ti.

NVIDIA is purposely introducing namespace collisions, and while they have their reasons I don’t believe them to be good enough. The GeForce GTX 560 Ti With 448 Cores is not a GeForce GTX 560 Ti. Most of the time it’s much faster, and this is a good thing. But it also requires more power and generates more heat, and this is a bad thing. My greatest concern is that someone is going to build a system around the operational attributes of a GTX 560 Ti, an then pick up one of these cards, ending up with a system that can’t handle the extra load. This is one of the many benefits of a clear, concise, non-conflicting namespace. And it only gets worse once you see the GTX 560 Ti OEM, a much lower-performing GF100 part that nevertheless shares the GTX 560 Ti name. NVIDIA can and should do better by their customers.

Ultimately NVIDIA has thrown us an interesting curveball for the holidays. We have a GTX 560 Ti that isn’t really a GTX 560 Ti but rather is a card trying hard to be a GTX 570.  At the same time it’s a 3rd tier product but unlike other 3rd tier products it’s actually quite good. Finally as good as it is it will only be available for a limited time. It’s a lot to take into consideration, and a name alone doesn’t do the situation justice. The GeForce GTX 560 Ti With 448 Cores isn’t going to significantly shake-up NVIDIA’s product lines – it’s not meant to – but for the budget-minded among us it’s a chance to get performance near a GTX 570 for just a bit less for Christmas, and that’s as good a reason as any to exist.

Finally, to wrap things up we have the matter of Zotac’s GeForce GTX 560 Ti 448 Cores Limited Edition. If the regular GTX 560-448 is nearly a GTX 570, then Zotac’s card is a GTX 570’s fraternal twin. It’s close enough in performance that the differences in performance cease to matter, and the power consumption doesn’t suffer for the factory overclock. At $299 there’s a greater risk of running into the actual GTX 570, which is what makes the Zotac card a GTX 570 substitute rather than something immediately more or less desirable than the GTX 570. On the plus side if you're in North America and don’t yet have Battlefield 3, the choice becomes much clearer.

Power, Temperature, & Noise
Comments Locked

80 Comments

View All Comments

  • Jamahl - Tuesday, November 29, 2011 - link

    In your 560 Ti review you said that it was "a bit faster" than the 6950. What's changed? Maybe AMD's drivers are helping to pull the card away because it's clearly ahead here with the same games being tested.

    http://www.anandtech.com/show/4135/nvidias-geforce...

    "The GTX 560 Ti ultimately has the edge: it’s a bit faster and it’s quieter than the 6950"

    Perhaps you should do an article on that one? You know you were one of the very few sites on the web who actually found the 560 Ti to be faster than the 6950 in the first place?

    I wonder why that was.
  • Ryan Smith - Tuesday, November 29, 2011 - link

    If you haven't already, I'd invite you to take a look at Bench, our online benchmark database. The video card numbers are periodically revised for newer articles, which is what you're seeing here.

    The latest data we have for the 6950 vs. the GTX 560 Ti: http://www.anandtech.com/bench/Product/293?vs=330
  • Jamahl - Wednesday, November 30, 2011 - link

    Glad to see you are keeping those updated and thanks for the reply.

    My point was, what happened to the 560 Ti's lead from your initial review? Looking at that bench now the 6950 is a good bit ahead.

    Drivers?
  • Ryan Smith - Wednesday, November 30, 2011 - link

    Yes, I'd say that's a fair assessment. Looking at 1920 between January and November

    Crysis: 48.6->51.4
    BF: 58.3->68.9
    HAWX: 108->119
    CivV: 34.8->40.1
    BC2: 61.8->69.2
    Etc.

    Note that the 560 Ti was launched only a month after the 6900 series, so AMD only had a short amount of time to optimize their 6900 drivers between the 6900 launch and then. Whereas they've had another 10 months since then to work on their drivers further. Given the similarities between VLIW4 and VLIW5, if you had asked me for my expectations 10 months ago it's actually more than I thought AMD would get out of optimizations.

    Meanwhile the 560 Ti has shifted very little in comparison, which is not surprising since the Fermi-lite architecture had been around for over half a year by that point.

    The 560 Ti and 6950 still trade blows depending on the game in a very obvious way, but the 6950 is now winning more games and on a pure numerical average is clearly doing better.
  • Jamahl - Friday, December 2, 2011 - link

    Yep that looks like a pretty fair assessment. I was suprised to see the gap open up so clearly.
  • pixelstuff - Tuesday, November 29, 2011 - link

    Are these all of the crappy GF110 processors that had manufacturing defects?
  • Ryan Smith - Tuesday, November 29, 2011 - link

    Correct. Technically speaking NVIDIA could take perfectly good GF110 GPUs can still make products like this, but it wouldn't make any sense for them to do so. All of these cards would be using GF110 GPUs with 2 defective SMs.
  • Duwelon - Wednesday, November 30, 2011 - link

    Your image shows 3 defective SMs. At least i'm assuming it's supposed to be the "new" chip.
  • Ryan Smith - Wednesday, November 30, 2011 - link

    Indeed, that was a diagram error on my part. It's been fixed.
  • piroroadkill - Tuesday, November 29, 2011 - link

    I'd still regard the 6950 2GB as the best value proposition card, and it has been ever since the launch of the card almost a year ago.

    Even though I only bought one recently, and heard the extra shaders had been lasered off, this thankfully proved wrong, and one BIOS update later and I have a 6970.

    You can't ignore value like that.

Log in

Don't have an account? Sign up now