NVIDIA's GeForce GTX TITAN X Becomes First 12GB Consumer Graphics Card 110
Deathspawner writes: When NVIDIA announced its GeForce GTX TITAN X at GTC, no one was surprised that it'd be faster than the company's previous top-end card, the GTX 980. But what did impress many is that the company said the card would sport a staggering 12GB of VRAM. As Techgage found, pushing that 12GB is an exercise in patience — you really have to go out of your way to come even close.
Additional reviews available at PC Perspective and AnandTech. The latter notes, "...from a technical perspective, the GTX Titan X and GM200 GPU represent an interesting shift in high-end GPU design goals for NVIDIA, one whose ramifications I’m not sure we fully understand yet. By building what’s essentially a bigger version of GM204, heavy on graphics and light on FP64 compute, NVIDIA has been able to drive up performance without a GM204-like increase in die size. At 601mm2 GM200 is still NVIDIA’s largest GPU to date, but by producing their purist graphics GPU in quite some time, it has allowed NVIDIA to pack more graphics horsepower than ever before into a 28nm GPU. What remains to be seen then is whether this graphics/FP32-centric design is a one-off occurrence for 28nm, or if this is the start of a permanent shift in NVIDIA GPU design."
12Gig (Score:5, Funny)
Re: (Score:1)
Congratz! You win one internet!
Re: (Score:2)
To Quote Darth Vader... (Score:2)
Impressive! Most impressive. [youtube.com]
What portion of the memory is usable this time? (Score:1)
So, is this a 12GB card or "12"GB card? Will it have 2GB of fast RAM and 10GB of crap behind a I2C bus which only exists for marketing reasons?
Re:What portion of the memory is usable this time? (Score:5, Insightful)
On the other hand, as I remarked to a friend when discussing graphics cards the other day, it seems to me that none of the numbers on the cards really matter - RAM, Clock Speed, etc. There are so many variables that you might as well say that it has 12 GigaDrawing Cores and 256 SuperShaders with a 1.5 TeraTransit Hop Interval. What really matters is the final performance, using the universal standard measure of FoC(3), aka "Frames (per second) of Crysis3".
Honestly, when I look at graphics cards (and I'm currently doing so for my next system), that's what matters to me most in comparing them - looking at benchmark results among all the major games. Certainly Nvidia should be telling us up front how the architecture works, and that you have 3.5GB of main/fast memory or whatever, but benchmark statistics from 3rd parties seem to me to not only be more important, but not exactly (or at least not easily) fakeable.
That said, it's interesting to me that the benchmarks I've seen for this card mark it as inferior to the 295X2, which is not only almost a year older, but significantly cheaper. I'm personally hesitant to go with AMD, as my current experience with AMD drivers on my desktop has been downright horrible, compared to a relatively painless experience with my current laptop running an Nvidia card, but I have to wonder what's going on when Nvidia comes out with something like this, a year later, and at seemingly worse performance.
Re: (Score:2)
What really matters is the final performance, using the universal standard measure of FoC(3), aka "Frames (per second) of Crysis3".
Except it's not. Not even close. Some of us have never even played a Crysis game.
But it's still a good benchmark even if you never play any of the Crysis games.
If the card gives a good performance in "Frames (per second) of Crysis3" then that's a reasonably good indicator that you'll get good performance in other applications -- it's at least better than AMD or Nvidia's meaningless marketing doublespeak.
Re: (Score:2)
But it's still a good benchmark even if you never play any of the Crysis games.
It may be a good benchmark for graphics performance. But many of us use GPUs as compute servers. If you are using it for cryptanalysis, than raw performance is important, and memory performance and I/O is not important. If you are using it for neural net training, then memory and raw FLOPS are important. Nvidia's CEO specifically mentioned neural nets as a potential big market for these devices.
Re: (Score:2)
Re: (Score:2)
So unless you're a GPGPU developer, you should buy the card that the application you use was designed for, since the variance in performance is huge, with some GPGPU video encoders performing worse than CPU implementations.
Comment removed (Score:4, Interesting)
Re: (Score:2)
Yes, there is a point in having 12GiB RAM on the card: There are people who can make use of it, even with existing applications. It's also a good way of driving development. This also ties in with your cluelessness about Compute performance: while DP is low, not all Compute tasks need DP. Signal processing for example is a field where DP is a waste. Take music for example. Softsynths that do physical modelling require a lot of resources, both in terms of processing and RAM, but you only need single precisio
Re: (Score:2)
Re: (Score:2)
Re: (Score:2)
Re: (Score:1)
Off the top of my head right now..
Video screen capture / recording with 0 delay
Raycasting
High resolution procedural textures
Re: (Score:2)
Trivial to fake. They put things into the drivers that if it recognizes a benchmark game it turns down or disables certain features to boost frame rates, even where the driver and/or game claims different settings.
Both AMD and Nvidia have been shown to do this.
Re:What portion of the memory is usable this time? (Score:4, Informative)
RAM is very important if you use high resolutions. If you game in 1080p, then yeah this won't tell you much. If you have a 4k monitor though, 3GB isn't enough so you can at least look at the RAM to narrow your selection, then look at benchmarks.
Re: (Score:2)
You are talking about game benchmarks.
For people editing, doing F/X and CG work, having all those CUDA cores helps (Photoshop, Adobe Premier, Avid, and AfterEffects all can use the CUDA cores for crazy fast rendering).
And, all that VRAM helps when working with HD and 4K video, especially multiple layers of video.
AMD doesn't yet offer CUDA processing.
Re: (Score:2)
I'm personally hesitant to go with AMD, as my current experience with AMD drivers on my desktop has been downright horrible, compared to a relatively painless experience with my current laptop running an Nvidia card, but I have to wonder what's going on when Nvidia comes out with something like this, a year later, and at seemingly worse performance.
I've actually found the reverse case lately, for my usage. I have a R9 290 in my HTPC and a Geforce 770x. Every time I update my 770x driver, it moves my taskbar to the other screen and resets window sizes, incredibly frustrating and something that shouldn't happen in 2014. Its been a problem for at least 4 years through multiple OSes and different installs. I had this problem on a GF 250, and I've reported the bug multiple times. I also experienced CTDs in Battlefield 3 and about 3 other games, usuall
Re: (Score:2)
12GB card apparently, because this seems to be upsized 980 rather than 970.
But if they were to ever make Tital Z LE or similar cut down version to utilize their not fully operational dies, one would have to look very hard into just how much of real memory and how much "marketing memory" such a card would have.
Sigh (Score:2)
So basically the card is overspecced for no sensible reason and you can't fill that amount of VRAM even at 8K with 8xFSAA (and when you do that, you get 9fps). Even SLI'ing 4 of those together won't get you 40fps at those res.
So the extra VRAM is entirely, completely pointless and they could have just supplied it with 8Gb (or less!), reduce the price slightly and had done with it and nobody would have noticed any difference.
Selling point of our product: "We've put in useless shit that you'll never be able
Re: (Score:2)
You're not doing anything useful on a 12 GB Titan that couldn't do on 8 GB or 6 GB.
If you've bought some integrated solution that requires/prefers CUDA / nVidia cards you buy nVidia and piss the money away on Quadros.
Otherwise you go with AMD for compute.
Nvidia's double precision performance on their workstation cards is a joke (and almost completely absent on the Titans). There's a reason Bitcoin mining was done on AMD GPUs.
Re:Sigh (Score:4, Informative)
The reason Bitcoin was done on AMD GPU's was because AMD had a bitshift instruction, and that due to how simple Bitcoin is, AMD's rather retarded memory handling didn't cause any choking. Double precision had absolutely zero to do with Bitcoin running well on AMD.
AMD's memory handling means that the actual Double Precision performance you get out of a given card is extremely different from the theoretical maximum. In fact, for many real-world tasks, Nvidia+CUDA delivers better DP performance than the AMD paper tigers, especially per watt, and so the HPC world heavily leans towards Nvidia, even with DP heavy tasks. For tasks that doesn't need SP, like many forms of signal analysis etc? Forget AMD, not worth the bother of shitty drivers, piss-poor memory handling, worse performance per watt.
As for RAM, there are things it's useful for, especially since the Titan also has its Universal Physics Solver, meaning that together with the RAM amount, you can put together very detailed sims at decent resolutions and framerates, all on a single card.
Re: (Score:2)
And that was supposed to be "For tasks that don't need DP, like many forms of signal analysis etc?"
Re:Sigh (Score:4, Interesting)
The rest of your complaints, "shitty drivers", "piss-poor memory handling" and "worse performance per watt" are also bogus. I own or manage machines using a large number of Nvidia and AMD video cards, and have seen as many driver issues between the two that neither has come out worse. This is a typical fanboy stereotype that keeps being repeated with no real fact behind it.
Your second complaint is seen a lot in programming forums, but I have never seen anyone do a proper write up of any memory issues with any of AMD's generations and most of the conversations lead me to believe it was an issue of the programmer's personal preference not wanting to learn a second platform with less market share than an actual technical issue. Most of these issues would be alleviate if the programmer would just use a common optimized library and stop trying to redo the work themselves.
Lastly, AMD's offerings have historically produce more performance per watt and their latest offerings continue that trend. This, besides the bit shift ability you mention, is also one reason why AMD was used for Bitcoin mining and supercomputers.
http://www.tomshardware.com/re... [tomshardware.com]
http://www.green500.org/news/g... [green500.org]
Now, my latest personal computer has an Nvidia GeForce GTX 980 inside because I more often need to fix CUDA code and noticed some of the games I wanted to play ran better on it (again, from the game designer's preference and not a technical merit). I personally own eight other video cards across AMD, Nvidia and Matrox (who use AMD GPUs these days) and three generations for testing.
And I am only sticking up for AMD because I admire their push to get people to code for multi-core better. Nvidia has been too conciliatory in the last six years in that respect, which is fine for their revenue stream and market share but not a good thing in the big picture for the broader computer industry. Since Moore's law has begun to slow, we are going to need a massive shift to multi-core optimized applications and we need programmers ready for that day.
AMD seems to be ready with the tough love to get everyone there while Nvidia keeps enabling bad behaviors.
Re: (Score:2)
640k is enough for anyone, right? lol
there is a LOT you can do with 32-bit operations, and oftentimes it's better to push more data through than to get precision you don't need anyway. don't mistake your pedantic ignorance for insight.
Re: (Score:2)
Blender would find it useful.
It doesn't need double floats for rendering, and it has to load the data for the whole pass into the video RAM to use CUDA. That puts a limit on your scene complexity.
There are various issues using monolithic kernels on AMD cards. Until Blender's GPU renderer (Cycles) is redesigned to use microkernels (work is in progress, but it'll take a while), NVidia is where it's at.
I just do the stuff as a hobby, but if I were a professional, I'd have a couple of Titans in my machine rig
Re: (Score:1)
Re: (Score:3)
Deep Neural Nets will use it, at least the very big, wide, deep ones will.
Re: (Score:2)
Re:Sigh (Score:5, Informative)
In case people are confused it is important to point out that the Titan cards aren't aimed at gamers. They are partly a PR stunt for Nvidia (look, we make the biggest, baddest GPU out there), and partly of interest to developers working in graphics research (either developing tech for next gen games, GPGPU research, fluid simulations, and other projects). When you are raycasting massive voxel scenes for example, the 12GB can look rather attractive.
At the end of the day it is very much a niche product, and calling it a "consumer card" is perhaps a bit of a misnomer. If all you are looking to do is to consume content (i.e. play games) this isn't the card for you, just SLI as many GTX980's as you can afford together and be done with it.
Re:Sigh (Score:4, Interesting)
I know some people who are drooling all over this card for various real-time simulations. The Unified Physics Solver and 12GiB RAM on the card will allow them to push a combination of good graphics and good enough physics, at decent frame rates in real-time, all on a single card. It also makes it easier to develop sim solutions.
Re: (Score:2)
and then probably have sex with your girlfriend.
No one wants to have sex with his hand.
Re: (Score:2)
it's not a bad strategy for the gaming market either (not great, but not bad). it's a developer gateway to including physics simulations in games. rather than cramming in even more pixels (at the point where most people won't even notice them) or cosmetic effects, even adding fluid simulation can create a more significant market differentiation than more anti-aliasing. pushing these cards on the "early-adopter" (read: sucker) crowd opens that door.
Re: (Score:2, Funny)
Titan cards aren't aimed at gamers
Correct. They are aimed at suckers.
Re: (Score:2)
Just because you don't understand the difference between the Titan's FP64 (float64) = 1/3 FP32 (float32) performance versus the FP64 = 1/24 FP32 performance of regular cards doesn't imply that everyone who buys Titan cards is clueless.
Re: (Score:1)
Re:Sigh (Score:4, Insightful)
Yeah I noticed that nVidia is playing games with the FP64 with the Titan X. :-(
Conflating high end gamers using Titan X with the scientific performance of the previous Titan Black and Titan is rather dodgy.
Re:Sigh (Score:5, Insightful)
The Titan cards ARE aimed at gamers.
Have you seen the reviews? Did you see the presentation today? Did you see the double precision compute performance? This is not a compute card.
Quadro = Workstation
GTX = Gamer
Titan = Gamer with more money than sense
Re: (Score:2)
Re: (Score:1)
No matter what Nvidia themselves say this isn't a gamer targeted card. It simply doesn't make sense for a gamer to buy, they are far better off with GTX980's in SLI (which outperforms the Titan X in gaming benchmarks significantly).
Lacking in DP performance does make it less attractive to most GPGPU researchers, but some cases do just fine using SP and most graphics related research certainly does. So let me reiterate: if you are working on research for next generation games and you are working with massive
Re: (Score:3)
In case people are confused it is important to point out that the Titan cards aren't aimed at gamers.
Yes, the Titan series is an odd bridge between consumer price/ performance, and professional reliability (ECC RAM) and unhampered double precision performance at painfully professional level prices of the CUDA / GPGPU oriented Quadro and Telsa cards. (Telsa K20 to K80 cost $3500 - $5000 USD approx AFAIK)
So they are great alternatives for CUDA aware 3D graphics application users who traditionally can't afford a Quadro or Telsa card (are not professional movie / video-game studio artists or CAD designers), a
Re: (Score:2)
Actually that's exactly what this card is not, they've dropped FP64 performance and basically made a bigger, more badass gaming card. If there's any others who can use it, that's because they don't really need the compute features. This is like Intel's "Extreme Edition" CPUs, the performance/dollar is abysmal but it's not an Xeon. Neither is this a Tesla. This is for the "I can drop $3k on a gaming rig" market. I know people with more expensive hobbies than computers.
Re: (Score:3)
Just to correct a misunderstanding many people have:
Not everyone who does compute needs double precision. Many tasks within signal processing only need single precision, for example.
Re: (Score:2)
Intel has been doing the exact same thing for the past 20+ years ...
nVidia finally started taking lessons. :-/
Re: (Score:2)
Compute might use it. The Titan was traditionally both the top-tier graphics card, as well as the entry-level compute card. They've slashed FP64 performance with this iteration, but FP32 performance is still sky-high, and they're pushing situations like that for using this as a compute card.
Re: (Score:2)
For compute, multiple GTX cards gets you more performance/$ than the Titan, with the same "LOL DOUBLE PRECISION NOT GONNA WORK" caveat.
If you're not stuck with CUDA, AMD's offerings are a better choice.
This isn't a compute card - it's an idiot card. For both games and compute, both Nvidia's and AMD's standard "gamer" cards are a better value.
The only thing novel about the Titan X is 12 GB accessible by 1 GPU.
Re: (Score:2)
I do wonder if ridiculous amounts of VRAM may end up being useful in game engines that are currently only on the horizon - for example, Outerra's 1:1 scale planet engine [outerra.com] renders mostly in GPU [moddb.com]. One wonders too about RAM demand of something like Euclideon's "unlimited detail" engine [rockpapershotgun.com] (assuming it isn't vapourware). If we're moving to games that do more in the GPU, then maybe stupid amounts of VRAM might actually get used? Then again, I expect if that were to happen (using complex GPU-based world generators),
Re: (Score:2)
So basically the card is overspecced for no sensible reason and you can't fill that amount of VRAM even at 8K with 8xFSAA (and when you do that, you get 9fps). Even SLI'ing 4 of those together won't get you 40fps at those res.
You do realize that 8K @ 12GB is the same as 4K (25% the pixels) @ 3GB? Are you claiming the GTX970/980 is equally overspecced at 4GB(-ish)? I don't think we have 8K monitors yet but make a triple UHD monitor setup and I think you'll make quad-Titan X sweat.
Re: (Score:2)
"We've put in useless shit that you'll never be able to use anyway! And charged you a premium for it!"
That's called Marketing.
Re: (Score:2)
They should have kept the price at $1000 and only put on 8GB, keep the money for profit, it's not like the people who buy them care or make sensible price/performance decisions.
Re: (Score:2)
Yeah but will the drivers handle exceptions gracefully
When available as 16nm, yes (Score:1)
Re: (Score:2)
Yeah, I'm sticking with my GTX780 until the 20 or 26nm chips are out. Hopefully something from that range will support top-end graphic settings with decent frames per second on a 4k screen, in which case I'll consider buying one.
Oh great (Score:1)
Damn it (Score:1)
It doesn't really help me. I've been contracted to build a high level gaming system with no stutter/jitter etc. I've been delaying the build looking for Nvidia's answer to the AMD 295x2. The Titan X is good, but the vram is ridiculous. I'll need 2x 980's or 970's to beat the performance of the 295x2 and I don't want to go SLI until the game developers write better code (never) and Nvidia sort out their drivers (improbable). Even the 295x2 has minor stutter issues. I just can't wait for the AMD Fiji chip whi
Useless for budget scientific computing (Score:2)
Which is really a shame, numerical simulations would easily make full use of the memory.
Re:Useless for budget scientific computing (Score:5, Insightful)
I was at the keynote at GTC this morning and it really depends on what you are doing. If you want to do numerical simulation, it is not very useful because double precision performance is terrible. But if you do data mining, you mostly care about bandwidth and single precision performance. And then 12GB isn't too much. Actually I find it still a bit on the low side. Intel Xeon Phi are featuring 16GB this days. And in the realm of data analysis fitting the data on the accelerator is what make the difference ebetween the accelerator is great and the accelerator is useless.
Re: (Score:2)
True, when I wrote 'scientific computing' I meant fp64 numerical simulations. Data mining is filed under 'business intelligence' in my mind :-)
Re: (Score:2)
I do lots of scientific computing (radio astronomy signal processing) on consumer-level NVIDIA cards and it is *all* FP32. These kinds of cards (and the 12GB of mem) could be very useful for my work, and others who do DSP-type of *scientific* computing.
Re: (Score:2)
Glad to hear we have a taker who can make some use of them.
Where do you put the card? (Score:2)