The NVIDIA GTX 1080 Ti – An Introduction
Today at GDC, while we were all working on our reviews of AMD’s Ryzen processors, NVIDIA surprised no one by officially announcing their GTX 1080 Ti graphics card. This is the card that, if rumors were to be believed, was supposed to launch before Christmas last year, then at CES and now finally at the Game Developers Conference.
However, before the festivities fully got underway, there was one small announcement: the GTX 1080, which was originally launched in May of last year (yes, it’s going on a year old now) is getting a $100 price cut to $499 USD. Naturally, that discount was making some room within NVIDIA’s lineup in preparation for the highlight announcement….the GTX 1080 Ti.
Billed as being about 35% faster than the Titan X, the new 1080 Ti will be hard launched next week for a stately sum of $699. According to our own testing of the Titan X, that 35% number puts it in direct competition for the title of the fastest GPU on the planet, but not by much.
Much like the mighty but now-supplanted Titan X, the GTX 1080 Ti will be using NVIDIA’s GP102 die with 3584 cores, 11GB of 11Gbps GDDR5X memory and it should have a good amount of overclocking headroom. NVIDIA showed it reaching about 2GHz in their demo but we will have to see what that means for fan speeds.
In order to offer some amount of differentiation, NVIDIA has disabled a single ROP partition and memory controller on this version of the GP102 core. This results in a somewhat odd 352-bit wide memory bus, 88 ROPs and 11GB of memory. It seems like the L2 caching hierarchy has remained intact though.
With that being said, in order to actually make this thing faster than its more-expensive cousin NVIDIA has upped the core frequencies by a reasonable amount to compensate. The same can be said about the memory which may have a narrower interface but its 1GHz speed bump over the Titan X should pay dividends in some situations.
In order to keep things cool under the hood, NVIDIA has supposedly revised their internal heatsink design so it has double the amount of surface area while also updating the PWM to a more advanced 7-phase all-digital layout. This is a beastly card and even though the Pascal architecture is known to be extremely efficient, it still boasts a TDP of 250W.
With launch coming next week and the card already under full production, one has to wonder where AMD stands with their Vega architecture. Hopefully we know soon.