No doubt if you’re reading this you already know that Nvidia released the GTX 680 late last week. With little more than a single card and the rest of its family tree just rumors at this point Nvidia has managed to change the graphics market.
Regardless of whose review you read it’s hard to deny the fact that the GTX 680 is the card we’ve been waiting for ever since the first space heaters known as Fermi rolled off the assembly line. Toss aside the marketing hype about specialized features, custom cooling solutions and questionable bandwidth claims because Nvidia finally has a real horse in the race.
Don’t get me wrong, I’m nobody’s fanboy as I consider Nvidia’s Fermi predecessors burnt offerings. AMD/ATI has been on a tear for the past few years with the price/performance crown squarely in their camp with little competition from team green.
Nvidia’s claimed performance advantages over ATI offerings in the past few years were easily erased with little more than a driver update. With Fermi, Nvidia saddled itself with products whose signature was high power consumption, high operating temperatures and a high price to match. Compared to their competition, those factors ceded the market to AMD/ATI when performance never rose above the level of equivalence.
Think of it this way, If you had a choice between two cars with the same performance attributes but one was less fuel efficient and subject to overheating it’s likely you wouldn’t choose it.
Enough about what Nvidia did wrong because it seems they’ve finally done something right.
The 680 GTX is a video card based on the new Kepler GPU architecture which is claimed to be “Fermi evolved” but where Fermi from its inception was meant to be more GPUGPU this new card has actually scaled back that functionality. A functionality that most gamers could care less about by the way. It is reasonable to consider Fermi as a square peg in a round hole as its core design was more about compute performance than pixel pushing. When you repurpose an architecture like Fermi to run contrary to its primary purpose it’s no wonder that heat and power consumption result. Performance will similarly suffer.
Addressing the deficiencies of Fermi is exactly why the 680 GTX takes the win from the 7970. A $50 to $100 lower MSRP only drives the point home. Further, a TDP of 195 Watts versus the 250 Watt TDP of the 7970 puts the ATI offering in the unenviable position of being the new king of hot, power hungry enthusiast video cards formerly occupied by Nvidia.
I’ve only found one review that claimed the 7970 was the better card and then it was based only on memory bandwidth which has less value on graphics processing than CPU workloads. GPU’s have multiple avenues to tweak performance. Just as an Nvidia GTX 260 with its 448 bit memory bus is easily bested by an ATI 128Bit bus on a 5770 due to factors other than memory bandwidth.
Much like Intel took back the CPU crown from AMD with Bloomfield (Socket 1366) in 2008 based at least in part on a technology heritage flowing from mobile platforms it appears that Nvidia’s more efficient design may be borrowing to some degree from their forays into the mobile market where power efficiency and thermal control while maintaining performance is a priority.
AMD/ATI may be facing a similar situation in the graphics realm to its experiences with Intel in the CPU realm. Recent reports suggest a price drop is likely but not certain for the 7970 due to the full line of 7xxx series cards and prices already announced. A price drop at the high end would have to extend down the entire line. With thin margins, AMD isn’t in a good position to pull the trigger.
Who’d have ever thought Nvidia would be the new price/performance leader?