Slashdot Mirror
NVIDIA Drops Surprise Unveiling of Pascal-Based GeForce GTX Titan X (hothardware.com)
MojoKid writes from a report via HotHardware: Details just emerged from NVIDIA regarding its upcoming powerful, Pascal-based Titan X graphics card, featuring a 12 billion transistor GPU, codenamed GP102. NVIDIA is obviously having a little fun with this one and at an artificial intelligence (AI) meet-up at Stanford University this evening, NVIDIA CEO Jen-Hsun Huang first announced, and then actually gave away a few brand-new, Pascal-based NVIDIA TITAN X GPUs. Apparently, Brian Kelleher, one of NVIDIA's top hardware engineers, made a bet with NVIDIA CEO Jen-Hsun Huang, that the company could squeeze 10 teraflops of computing performance out of a single chip. Jen-Hsun thought that was not doable in this generation of product, but apparently, Brian and his team pulled it off. The new Titan X is powered by NVIDIA's largest GPU -- the company says it's actually the biggest GPU ever built. The Pascal-based GP102 features 3,584 CUDA cores, clocked at 1.53GHz (the previous-gen Titan X has 3,072 CUDA cores clocked at 1.08GHz). The specifications NVIDIA has released thus far include: 12-billion transistors, 11 TFLOPs FP32 (32-bit floating point), 44 TOPS INT8 (new deep learning inferencing instructions), 3,584 CUDA cores at 1.53GHz, and 12GB of GDDR5X memory (480GB/s). The new Titan X will be available August 2nd for $1,200 direct from NVIDIA.com.
I thought it had been surpassed by C++, but this is great for everyone.
So bloody fast it actually made the Kessel run in 12 parsecs.
WARNING: Smartphones have side effects--most of them undocumented.
... can it run Crysis?
The card looks like it will fit in a standard case, but the cooling tower will be the size of a small house.
"The ferrets, they're every where I tell you!"
"By the time Skynet became self-aware it had spread into millions of computer servers across the planet. Ordinary computers in office buildings, dorm rooms; everywhere. It was software; in cyberspace. There was no system core; it could not be shutdown. The attack began at 6:18 PM, just as he said it would" ...at an artificial intelligence conference in California. Judgment day has arrived. Now we just need to perfect time travel.
For the ones wondering: http://www.nvidia.com/object/gpu-architecture.html
So where are those sub-$100 Pascal-based videocards?
Your affordable is another man's can't afford one. Same thing goes for the Titan X.
The card is designed for data mining and neural network research; it's not for games or even remotely intended to be used for them.
basically a supercomputer on a card. I'd be *really* interested in finding out if those cores are individually addressable, etc and the memory setup. I remember the computer my Dad did his PhD calculations on -- an IBM 704 with memory expansion to a whopping 48K
C|N>K
A lot of researchers are using GPUs for things very different than graphics. A professor was telling me just last week that the boundary between a machine learning training algorithm being interesting was to train to deal with a problem in a week or less [note one trained it does its job much faster, that's just the get-it-ready-to-go time], and that GPUs were often used for that training. The bit width requirements are modest, but the amount of data to process is huge.
Of course, he went on to show how the approaches his students had come up with were faster and more power efficient by orders of magnitude for many common algorithms, but still they were trying to improve a normal way of doing things, which is to get up and running fast using GPUs are a source of number crunch.
In summary, people who don't actually need so more horsepower buying it helps keep it being developed for the smaller number of people who get it who are actually doing something useful with it.
Great Scott! 1.21 jiga flops! 1.21 jiga flops?! What was I thinking!!
Oh, it was on a dare...
"File to fit, pound to insert, paint to match" - Aircraft Maintenance 101
So its loudest supporters include a notorious spammer known for making poor choices (like spamming Slashdot). I don't know if you thought your input would help Pascal's image, but it certainly hasn't :)
There are *other* uses for GPU's. They make great compute processors for specific kinds of problems some of which are NOT related directly to pushing out pixels on a screen.
"File to fit, pound to insert, paint to match" - Aircraft Maintenance 101
The editing here becomes dummerer every day...
love is just extroverted narcissism
An affordable video card is totally capable of outputting 4k or even 8k video with no problem.
Encoding that or playing video games or using it for totally unrelated things is a different story though.
As you're posting as AC maybe you don't even WANT an answer. And now when you will likely get multiple will that help educate you and reconsider? Not likely. Because it's not a card FOR YOU and that's all that matter for you. That doesn't mean it's useless for everyone else.
Of course people will buy this and can afford it.
A friend just bought a GTX 1080 and a 34" 21:9 100 Hz G-sync screen (and the rest of the computer including the much more expensive Samsung Pro 950 drive and he will likely end up having the HTC Vive or another VR product too), he can afford this. Do he need this? NEED? Guess no. But of course not being able to pull off 4K gaming is an issue for him and as such the plan have seemed to included two GTX 1080 though I think it's better with one card which mean that he actually "need" something BETTER than this. Maybe this get close to 60 FPS gaming on 4K though.
You first adopters that have fat wallets, please start buying.
The Nvidia 1060 6GB 192-bit video cards are supposed to start at $250. That's $100 more than the Nvidia 950 2GB 128-bit video cards. I'm set up for auto-notify at Newegg.
nVidia set up us the surprise.
For great justice.
(I think they meant that nVidia dropped the surprise at the unveiling (on the audience), not dropped it from the unveiling.)
I have discovered a truly marvelous proof of killer sig, which this margin is too narrow to contain.
The card is designed for data mining and neural network research; it's not for games or even remotely intended to be used for them.
Bullshit.
http://www.geforce.com/hardwar...
"With the DNA of the worldâ(TM)s fastest supercomputer and the soul of NVIDIA® Keplerâ architecture, GeForce® GTX TITAN GPU is a revolution in PC gaming performance."
I admit I don't completely know who they focus on with the Titan cards.
10-11 Tflops single precision performance with this one.
317-343 Gflops double precision.
159-171 Gflops half precision (shouldn't that one be higher?)
The idea with the more professional card is to hit 5+ Tflops of couble precision performance?
http://wccftech.com/nvidia-pas...
I don't really know where the Titan cards fall between the consumer cards and the professional cards.
Once re-released as the GTX 1080Ti it will definitely be a gamers card.
I guess without further evidence saying "yes it is!" is just as good as saying "no it isn't!"
It's an expensive gaming card but those who want this performance now only have this option.
An affordable video card is totally capable of outputting 4k
So where are those sub-$100 Pascal-based videocards?
https://www.youtube.com/watch?v=gfvM5JX1Mk4
This is Intel Atom:
http://liliputing.com/2015/04/...
"Intel says its new chips can play 4K videos at 60 frames per second when theyâ(TM)re encoded at up to 250 Mbps bitrates. 1080p videos can play at up to 240 frames per second."
No need to buy a graphics card at all for 4K playback.
There's little reason to release sub $100 cards.
For those purposes you've got integrated graphics.
Nvidia uses scientists names for their products: Tesla, Fermi, Kepler, Maxwell, Pascal and Volta (next version after Pascal).
If the leading edge of "consumer" graphic cards is of any interest to someone, they'd know what Pascal was since it's been announced now for over 2 years.
I didn't see the "video" at the end of his comment. Since we're talking about GPUs I assumed we were talking about 4K gaming.
Well, sub-$200 cards then.
Why GDDR5X? HBM2 triples the memory throughput. If they want a monster card that is overkill for today, it should at least incorporate the king of memory buses.
The card you should be comparing to in the hiearchy is the GTX 960 4GiB
PROTIP: it's means it is.
It's been nice proving you wrong.
> Nobody has the money to afford one of these things
Speak for yourself. I'll be getting one for sure.
> You don't need one, either. It serves no purpose.
Completely not true. You need something as powerful as this or even more to play AAA games like Elite: Dangerous with maximum image quality (i;e. including say 2x supersampling) in high definition VR at 90 frames sec X2 (eyes) without dropping frames (i.e. making you feel nauseous)
The card you should be comparing to in the hiearchy is the GTX 960 4GiB
I stand corrected. Thank you.
You do know that Windows 10 ignores its host file when it comes to telemetry, right?
The world's burning. Moped Jesus spotted on I50. Details at 11.
Please now back up your assertion that there is some magical affordable GPU out there that can render modern 3D software at 4K or 8K at a constant 60 fps with a link to some kind of... what do we call it? proof.
This isn't for simple video playback, numb nuts. This is for 3D render, and massively parallel floating point math (read: CUDA apps).
Slashdot still doesnâ(TM)t support Unicode after it was added to the HTML standard in 1997.
Yeah, now let's have that Intel GPU do some actual 3D work and see how it performs at 4K. Hint: it will be terrible, which is why Nvidia is still in business.
Slashdot still doesnâ(TM)t support Unicode after it was added to the HTML standard in 1997.
Virtual Reality will consume all the GPU you can possibly throw at it, and will continue to do so for the next 5-10 years. Hell, Trials on Tattoine was built for a Quad Titan setup (3 GPU, one PhysX). I will be ordering one of these day one.
Good-bye
The 960 is power inefficient as hell
Since when is Elite: Dangerous an AAA game?
...gis sdrawkcab (usually not responding to ACs; don't bother posting as AC)
Actually the Tesla line of cards are for data mining and neural network research.
Nvidia has 3 PC lines of cards
Geforce Gaming
Quadro CAD/CAM and professional graphics.
Tesla for GPU compute. AI, data mining and other GPU compute functions.
The Titan is part of the GForce line and is a bit of an odd duck. It is a gaming card but like every other gaming card it can be used for AI, data mining, CAD, and even professional graphics but it is a gaming card. A very high end expensive gaming card.
See my blog http://ilovecookes.blogspot.com/ for light hearted technical information.
FIY, intel's "skylake" has nothing to do with the sky or lakes.
https://devblogs.nvidia.com/parallelforall/how-build-gpu-accelerated-research-cluster/
Aww damn really? I guess I'll have to cancel my order.
dammit really? There goes that idea....
Huh? No it's not. It's pretty decent in power/performance ratio, comparing favourably to slower cards from AMD from the same era
So read it again:
I didn't see the "video" at the end of his comment. Since we're talking about GPUs I assumed we were talking about 4K gaming.
"An affordable video card is totally capable of outputting 4k or even 8k video"
Yeah, but this person thought 4 and 8K video was all that mattered.
Why the fuck he think 8K video is important but not games is beyond me. How many have an 8K display? How many have 8K content to watch? How many would suffer with settling for 4K content right now?
I just have to say... Scientists have awesome names.
Nvidia have cheap cards in that category.
I assume you'll see a GTX 1050 at some time too.
Only the GP100 board has a 1:2 ratio of FP64 cores, and 2:1 of FP16 cores (albeit in a vector of 2, since it's processed by an FP32 core). The GP102, GP104, and GP106 have a 1:32 ratio of FP64 cores, and 1:128 of FP16x2 cores; which honestly is just to be compatible, not powerful. If you really want FP16, wait for GP100 in a Tesla card, or promote to FP32. The GP102 however does support INT8 at 4 times the speed of FP32; so 44 TOPS. Probably also vectorized like FP16, but done by the cuda cores.
If he already has 2560 cores with a GTX 1080, he'd be better off buying another GTX 1080 and link them for 5120 cores. The Titan X has 3584 cores at a slower clock, for the price of two.
But that's not $100 or even $150 territory for those who want it.
But there's the RX 480, 470 and 460 for them so far. But I can only assume Nvidia will release cards for that market too.
Then again DX12 dual Nvidia cards doesn't seem to work in Hitman and Rise of the Tomb Raider (but does in Ashes of Singularity.)
I assume eventually more games will have support but dual cards isn't the best and I'd suggest he just wait and upgrade to a single Volta card instead.
The first Titan had high speed double precision (FP64), so did the regular GTX 480 / 580 before it. The Maxwell Titan didn't, but it offered a very large 12GB RAM instead, a size which was available previously but only on highest end Quadro. And today we get a new one with fast support for 8bit integers as a differentiating feature, which comes as a surprise. 1080 Ti ought to have 12GB RAM (because of a 384bit bus, and because 1080 has 8GB already)
That's to say the idea of what a Titan is for is evolving a bit. One main niche is for 3D artists and off-line rendering, because a Quadro is more for actual CAD and otherwise a geforce is like 5x faster at a given price.
Well I'm sure Nvidia could have named it after scientists named Smith, Jones, Doe, Smith (a different one), and Johnson, but those names aren't very exciting.