13 
Today Nvidia is pulling the wraps off the GK110-based GeForce GTX Titan, a single-GPU card that is expected to easily capture the title of Baddest Ass GPU in the world when benchmarks are released this Thursday, February 21st. The Titan is Nvidia’s “Big Kepler” GPU, and has double the transistors and almost double the CUDA cores of the mid-range GK104 chip found in its flagship GeForce GTX 680 GPU. Though it runs at a lower clock speed in stock trim, it should still offer a sizable performance improvement over the already capable GTX 680.
The GK110 has been in use for over a year in the super-computing world, most notably in the Tesla K20X GPUs used by the Oak Ridge National Laboratory’s Titan supercomputer, which is currently the fastest super computer in the world. Though benchmarks for the Titan haven’t been posted just yet, Nvidia tells us it should be neck-and-neck with the dual-GPU GeForce GTX 690, which is currently the fastest single card GPU in existence. In other words, the Titan should be almost fast as dual GTX 680s in SLI, but from a single GPU. Oh, and guess what? It’s actually quieter than both of them, even under load (yes, we’ve heard it).
So let’s talk specs. The GTX Titan features a whopping 2,688 CUDA cores in 14 SMX units, 6GB of GDDR5 memory running at 6GHz, a core clock speed of 836MHz with a boost clock of 876MHz, 7.1 billion transistors, a 250w TDP, 384-bit memory interface, and a single-fan cooling unit. Power is provided by an eight-pin and a six-pin connector, and the price is a surprising 1000 euros, so hopefully around £900 (around the same as the GTX 690). The card is 10.5″ long and comes wrapped in an aluminium shell, as opposed to the magnesium alloy shell used by the GTX 690, and like that card, Nvidia is not allowing any changes by add-in board manufacturers, so all Titans will look just like the one shown here.
Perhaps even more impressive than the card’s raw specs is how quiet it is thanks to technology Nvidia is calling GPU Boost 2.0. When we first sat down to check out the card in an undisclosed location in San Francisco, we were shocked to find out that not only was a Titan running a mere 3 feet from where we sat and made no noise whatsoever, but another machine just a few feet from us to the right was running Battlefield 3 using three cards in SLI, and was also as quiet as a whisper. The secret isn’t just how Nvidia designed the single-fan cooling apparatus, though that obviously plays a huge role, but rather in how it designed the card’s software to work with the card. Put simply, you can now control the maximum temperature you want the card to reach, and it will throttle itself to remain at that temperature though a combination of fan speed, voltage and clock speeds. This lets you tweak the card to run any way you want – overclocked with a loud fan, stock speeds and relatively quiet, or underclocked and totally silent. You can adjust all these variables in real-time too through the software provided by the add-in card manufacturers, most likely Asus and EVGA if the same precedent set by the GTX 690 is followed.
Aside from the new temperature, voltage and clock speed controls (yes, you can over-volt the card), a new feature that is exclusive to the Titan will let you overclock your display’s refresh rate when running Vsync, so you could theoretically achieve 80fps even when using a 60Hz panel, and the same goes for 120Hz panels as well. Nvidia told us the GPU simply “lies” to the panel and tells it to run at a rate higher than it’s supposed to, with the caveat that not all panels will allow this, but the only way to find out is to try.
Another interesting feature is the colour and status of the GeForce GTX logo that is lit up on the side of the card can now be controlled through software. We didn’t get to see it in action, but Nvidia tells us you’ll be able to tweak it to perform a range of stunts including changing color based on temperature, you can have it “breathe,” or just be a different colour to match your rig’s mood lighting.
We never thought Nvidia would do this, simply because tradition dictates mild, incremental advancements in its GPU lineup year-to-year, especially when it’s not introducing a new architecture. Plus, since Nvidia already held a narrow performance advantage over the AMD HD 7970 card (at least in terms of power consumption), it would stand to reason that we’d see a modest 20 percent performance boost in the GTX 780 card, or something along those lines. Nobody expected a card to be released that promises almost double the performance. In fact, we stated as much in our 2013 Tech Preview article in our Holiday 2012 issue of the magazine when we wrote about the prospect of the GK110 showing up in gaming trim, stating, “It would be totally uncharacteristic of Nvidia to double performance in the next generation, particularly when the competition doesn’t warrant it.” It looks like we were wrong, and we couldn’t be happier about it.
So why would Nvidia release this card, especially when it’s already largely considered to be the leader in the single-GPU space? We have two opinions on that issue. The first is that it wanted to have a definitive lead, not just the lead in heat output, noise and energy consumption. We all know the GTX 680 consumes less power and produces less heat and noise than the HD 7970, but the two cards are so close in benchmark performance now that AMD has the GHz edition of the card circulating that in most tests it’s either a tie or AMD has a slight advantage. That obviously didn’t sit well with Nvidia, especially when it had all this horsepower lying around in the form of the GK110. Our guess is that Nvidia decided to end this pussy-footing around with the mid-range GK104 and just release the hounds once and for all.
]
The second reason for the Titan’s deployment is the sudden interest in the new generation of Small Form Factor rigs such as the Falcon Northwest Tiki, iBuyPower Revolt, Digital Storm Bolt and others. The GTX 690 was a bit too long to fit inside these cases, but the Titan fits perfectly being about an inch shorter despite having roughly the same width. This will allow system builders to construct tiny boxes with no-compromise graphics performance, and will also allow home builders to stuff the Titan into most ATX cases without any issues as well, providing they survive the sticker shock and can find one in stock.
Specs
That’s our quick glimpse of the new GeForce Titan. Check back in two days for full benchmarks in both single-and-SLI configs. We can hardly wait to show them to you.
Maximum PC brings you the latest in PC news, reviews, and how-tos.
Unless they have inproved the power consumption over the 680 then it will use about 713W at load.
TDP of 250W, so 2 of them might
TDP is how much heat it produces, so that card produces 250W of heat.
OK. So where’s the energy that creates that heat coming from?
If it produces 250W of heat then it has to be drawing more then 250W otherwise what would be powering the chip.
Right, but heat accounts for the vast majority of what it’s drawing. It is possible for power draw to be greater than TDP, but not by much.
I’m sure this will suck down more juice than any single GPU card, but it should be significantly better than the 690, perhaps around 400W at load.
Guru3D reckon 372W at load.
http://www.guru3d.com/articles_pages/geforce_gtx_titan_preview_reference,9.html
A 680 can pull about 230W if you push it to its limits, a lot of these tests seam to just max one part of the card and leave the rest alone.
Sure, but this card will draw significantly less than 2x what a 680 draws.
Keep in mind that the numbers in that graph reflect overall usage, not just max usage, so they will be lower. However, the article does quote the max draw of the Titan as 372W.
In fact, that’s not even right because that number is full system draw. PrinterElf’s TDP number looks more likely.
£900? Right, so this doesn’t change anything in practical terms. The fastest cards are still crushingly expensive and are only bought by a handful of money-no-object types.
Cards like this don’t change the landscape in any way, they’re just a marketing exercise to see who can claim to have the fastest card in their arsenal, creating a halo effect over the more mainstream cards that people actually buy.
Yes and no – the fact that they’re at the point of being able to ‘mass’ produce these chips means that they’re also at a point whereby the tech therein can trickle down over the next couple of generations to the mainstream cards.
Sure, but I don’t see much in the way of new tech – it’s just a bigger chip, and one that has already existed in low volumes for a while now.
If it were a new architecture I’d agree with you.
The price is a literal joke. You could build two mid-range *systems* for that money, and I say this having formerly owned a GF295 a couple of years ago – which cost, BTW, slightly more than 1/3rd what this “Titan” costs.
I wouldn’t exactly call a £450 system mid range, especially for gaming for which this card would imply the system is designed for, but I see your point. However I don’t agree with your point, for the £300+ you paid or even the £220 I paid for my card you could still get a full system but it wouldn’t be anywhere near as capable, the same can be said for this card. If someone can fully utilise the card’s power it makes complete sense to buy it, and if they’ve paid the money for monitors with a high enough resolution to push it the cost of the card likely won’t be as large to them.
Yeah, but will it run Crysis?
Will it run Theme Hospital?
Probably not enough complex computing power to run “Theme Hospital: NHS Edition”
That crashes regardless of what hardware you run it on
And always force closes Lewisham Hospital…
The only question that truly matters of course.
“…mid-range GK104 chip found in its flagship GeForce GTX 680 GPU.”
Since when is the GTX680 a mid-range card? Granted, the Titan and 690 are both faster, but the 680 is still a fast-as-hell card. I always thought the 660s and 670s of the world made up the mid-range while 650 and less made up the low-end of the spectrum.
Yeah, I thought that was a bit odd.
He’s not actually saying that the GTX 680 is a mid-range card as he describes it as the ‘flagship’ card in the same sentence. He’s referring to the GK104 chip, which was never intended to be the top dog in the Kepler series. That honour goes to the GK110 in the Titan, hence the ‘mid-range’ comment.
I feel like it was a joke, as in he’s saying that though it is ‘high-range’ really, the Titan has raised the stakes so significantly that the ridiculously powerful GTX 690 now looks ‘mid-range’ by its standards.
I’m due a GPU update this month.. think i’ll wait on this and just get a HD 7970.
You could get two 7970′s for the price of this, and they’d probably be faster too.
I just got a free grand today, might actually get one and sell my 670SLI.
Would you hit me if I offhandedly mentioned that 670SLI is already excessive for game performance and that you could, I dunno, give the grand to charity thereby providing a huge improvement in quality of life to a ton of people. That just seems logical given you’d be going from impractically powerful to imfuckingpractically powerful.
Lol. May have a point
I wouldn’t call it excessive it can be fully utilised. To play at just 1920×1080 or even lower it would indeed be excessive, but some people, me included, expect far more from their gaming experience.
Even at a higher resolution 670SLI is ample, I don’t know how you could say that tbh. What game aside from the literally just released (and startlingly average) Crysis 3 couldn’t be run on max at 1440p or higher with them?
Metro 2033 at 2560*1600 with no AA won’t break 60fps. However, with the next generation of consoles coming out games will be getting a lot better just like they did when the current gen came out. Then there are people running larger resolution than you seem to think, such as 4960*1600 or even triple 2560*1600 monitors, 2560*1400 displays overclocked to 100Hz+, and of course 4k.
‘Metro 2033 at 2560*1600 with no AA won’t break 60fps’. There’s the rub I suppose, I don’t demand a rock solid 60fps, and usually push my games down to about 30 for them to look as nice as possible. I would perhaps say that your expectations are a tad high if you expect a triple monitor setup at the best graphics possible with 60fps. That’s guaranteed to break the bank every time, and I can’t imagine a worse money sink than having to constantly maintain that performance by buying into the newest hardware every time something is released.
Also I am not aware of any even slightly practical gaming size screen that is 4k. Plus can we easily perceive more than about 60fps? I’ve never been able to tell when my games are above about 45-50 methinks (as evidenced by me pushing games down to roughly 30 for graphics).
That’s were future proofing comes in. There is currently a 32″ 4k display although ridiculously expensive, but that will (hopefully)come down. There also no requirement for a small screen for gaming, especially at higher resolutions. I’d be happy with the same pixel density as my 27″ 2560*1440 display but with 4k (so around 40″), that way I could bump up the field of view whilst keeping the same amount of detail and viewing the display at the same distance.
Like I said, some of us expect more. No problem if you’re fine with 30+ but that doesn’t mean we should all be stuck with that, especially competitive players. Also there’s the issue of V sync requiring a frame rate that is consistently above the monitor’s refresh rate, not as much of an issue with Nvidia’s adaptive V sync (although that does have its own drawbacks) but for us AMD users it’s a must.