Conclusion
Conclusion
Remember, if you are interested in our GTX 470 review, click here.
Writing a conclusion such as this one is never an easy endeavour since while the benchmarks themselves give a clearly defined numerical value to the performance of a certain product, things like final retail pricing and availability can only be educated guesses at this point. Before we wade face first into either of those sticky situations, let’s look at NVIDIA’s GF100 architecture from an outsider’s perspective and as someone who is familiar with the technology behind it. Most people won’t give two hoots about any of the real technological aspects behind the GTX 480 but will rather look closely at the performance numbers and make a hopefully informed decision based on those. So, as we know some of you will likely skip to this point in the review straight away, let’s start the next part of this conclusion with something which sums things up quite nicely.
There is no denying the fact that the Fermi architecture is extremely ambitious to the point where some would call it over-ambitious. Not only is NVIDIA pushing the limits when it comes to power consumption and heat production but the complexity of their new chips has led to delay after delay. Nonetheless, as you can see in the chart above the GTX 480 redefines performance for single GPU cards and at times does so in a wholly convincing manner. In our opinion the most crucial litmus test for any card in this price category is performance when IQ settings such as anti-aliasing are enabled and the GTX 480 passes here with flying colours. Basically, what is a mere (average) 8-11% lead increases to an average of 15 – 19% when AA is enabled which is impressive to say the least. As we saw, this performance with AA enabled also allows the GTX 480 to draw even with the HD 5870 when it comes to the critical performance per dollar category. Minimum framerates are all-important as well and tell an extremely important story since it is more than obvious NVIDIA’s current flagship provides a much less drastic average / minimum performance delta than the HD 5870.
As with any new architecture, there are still obviously areas for improvement and in the case of the GTX 480 there was a straw that almost broke the camel’s back. Our comparative testing charts were extremely eye-opening since they showed exactly what they were meant to: issues with resolution scaling. In Left 4 Dead 2, Aliens versus Predator, BattleField: Bad Company 2 and Unigine Heaven, we saw what looked to be an insurmountable lead at lower resolutions all but vanish at 2560 x 1600. Thankfully for NVIDIA, the GTX 480 was able to pull its butt out of the fire with strong AA performance. Nonetheless, this is particularly worrying since high resolution gaming is what the GTX 480 was supposedly built for and if it can’t maintain a sizable lead over the HD 5870 in exactly this area, many enthusiasts may question its price premium. We can’t state for a fact whether some of the performance drop-offs we saw at 2560x1600 were due to an architectural issue or immature drivers but it seems everyone we talked to (from fellow editors to NVIDIA themselves) had their own explanations. For now, we’ll keep an open mind and side with NVIDIA’s explanation which stated there are certain driver optimizations which have yet to be implemented.
We have a feeling that heat and power consumption will be issues that will continue to plague this architecture well into the future, at least until a die shrink. There will of course be smaller, much more efficient versions released in the coming months, but they likely won't compete with ATI’s impressive HD 5000-series when it comes to overall power efficiency. Having said that, when it comes to the GTX 480, power consumption isn’t nearly as bad as we thought it would be, but that’s partially because NVIDIA isn’t using a fully decked-out core. As such, performance per watt isn’t in this card’s favour at all. Even with an SM disabled, this is a seriously hot-running piece of silicon that pushes temperatures into the scary-hot 95 degree range, which is shocking considering how fully endowed the heatsink is. Luckily, noise wasn’t too much of an issue on our card when in an enclosed case. We just all have to remember that when push comes to shove, us enthusiasts are usually the first to bitch about power consumption and heat but then look the other way when it comes to the best possible gaming performance.
While pricing may seem good when looking at NVIDIA’s stated SRP of $499 USD, heaven only knows what kind of mark-up retailers will put on these puppies. Talking to our retailer contacts resulted in more waffling about than we care to admit but we are confident that some cards will hit that magical $499 mark or even a bit below as stock becomes readily available. When talking about availability, things become far less tangible since there MAY be some VERY limited stock in the retail channels today. General availability will likely begin somewhere around April 12th but stock of the GTX 480 will continue to be extremely tight for the foreseeable future. To us, this smells like a paper launch but we’ll just have to play the wait and see game.
The GTX 480 isn’t necessarily a resounding success but we consider it to be a good stepping stone towards some much needed competition in the DX11 marketplace. It’s a great performer in exactly the areas where it will be most utilized while being backed up by an extremely respectable price. But, it just doesn’t deliver the much-needed knock-out blow to ATI’s offerings. The main problem we see is understandably high expectations which may be tempered by NVIDIA’s failure in two areas: to deliver a flagship product that fully utilizes the architecture’s full, 512-core potential and actually launch the card with real retail volume in Q1 2010. Instead, it looks like we will still be waiting to see what this architecture can really offer with 512 cores and potential customers may not get their cards for another two weeks. Heat and power consumption are also two areas which raise red flags. To us, this isn’t exactly the greatest way to kick off the next generation.
