Earlier this week I told you about Nvidia's GeForce GTX Titan, the $1,000 graphics card with supercomputer power, tons of tweakability and an astounding acoustic footprint which keeps it whisper-quiet under the heaviest of loads. What I couldn't tell you about were the benchmarks—those all-important numbers.
Nvidia claimed this was the card to kick Crysis 3's ass. Today I can tell you Crysis 3 isn't going to be sitting comfortably anytime soon.
It's a little odd, announcing a brand new advanced piece of graphics hardware, putting it in the hands of the critics and then making them wait two days, but it gave me plenty of time to reflect on the technology I had at my disposal. My GTX Titan came inside the latest version of Digital Storm's Bolt, a mere slip of a computer at only 3.6 inches wide. This is the sort of card Digital Storm's engineers designed their case to hold, powerful enough to handle the most taxing games, while efficient and acoustically sound enough to do so without making the tiny case roar like a lion the minute 3D rendering starts happening.
The GTX Titan never got loader than a low hiss during my testing period, and that's incredibly important in this sort of system. Keep that in mind while looking at the numbers to come.
The Titan comes with third-party tools to tweak the hell out of performance, should the user so choose.
I mention this because I did not choose to do so. I am a bit of a coward when it comes to fiddling with settings. I've been burned before, quite literally, so I leave the tweaking to the professionals. These benchmarks reflect the default settings.
I did, however, fool with the card's innovative display overclocking, which allows the GPU to override the monitor's refresh rate, allowing for higher frame rates with VSync enabled. Results differ for every monitor, but my 23-inch AOC i2367fh jumped up to 75Hz without a problem—things got janky higher than that. It was a small difference, but an incredibly noticeable now.
I began my benchmark odyssey with Unigine, my new favorite method for judging graphic performance. First I went into Heaven and turned everything up to DirectX 11.
Solid scores, as expected, but the Heaven benchmark is child's play compared to Unigine's latest, the Valley.
Valley integrates all the latest bells and whistles—dynamic sky, volumetric clouds, sun shafts, depth-of-field, ambient occlusion—giving today's advanced graphics cards a serious workout. Performance dipped a little during the dynamic weather portion of testing, but that's what happens. That the Titan managed to stay above 20 frames per second during that nightmare is impressive.
After spending an hour wandering through the Valley benchmark (it's really addictive, and much easier than real walking), I went to my library of benchmarking games, making sure to mainly pick games with the number three in them. Batman: Arkham City slipped past me while I wasn't looking. It's sneaky that way.
Remember back when Battlefield 3 was a system-taxing monster? Mind you I performed the benchmark during the game's opening story sequence on the train, but I've seen plenty of video cards stumble through that bit. Most cards I've tested the sequence with hovering between 30 to 50 FPS. Titan rendered it while doing the crossword and catching up on A Game of Thrones.
The Arkham City numbers might not seem as impressive, but for this particular test I cranked up the PhysX as high as it would go. I'd say 78 FPS with ice particles bouncing all over the place is nothing to scoff at.
Far Cry 3 ended up on the low end of the frame rate scale, which I attribute to my mad dash through the jungle on a four-wheeler, jumping it off a cliff into the ocean. I cannot sit still in Far Cry 3, sorry.
Thanks to an issue with my cloud save, my Assassin's Creed III test was spent playing Connor's asshole dad in the early bits of his trip to Boston. I gave the jerk quite a workout, running him through crowds, over rooftops and ultimately into a group of redcoats. The benchmark ended in the middle of the fight, so I let him die. That was pretty satisfying.
And then we have Crysis 3. The supposed system-strangler was no match for the Titan's prowess. My benchmark was performed in the middle of a vast open area, in which a dozen enemies were attempting to end my life with extreme prejudice while I sprinted in and out of them like a madman. After the benchmark completed I spent a good ten minutes bragging to Kirk Hamilton about how much better the game was running for me than it did for him. I guess this counts as bragging as well. Oops.
I would have spent more time benchmarking threequels, but I figure the five I've run plus the Unigine tests presents a pretty solid picture of the sort of performance a cowardly non-tweaker can expect from Nvidia's GeForce GTX Titan. I'd imagine that someone with the courage to adjust settings could get those numbers even higher. I salute the brave men and women that will be spending the next few weeks potentially jeopardizing $1,000 video cards. They are true heroes.