Pascal Titan X announced.

[Tenknics]

It began with a bet.

Brian Kelleher, our top hardware engineer, bet our CEO, Jen-Hsun Huang, we could get more than 10 teraflops of computing performance from a single chip. Jen-Hsun thought that was crazy.

Well, we did it. The result is crazy. And, as of today, Jen-Hsun now owes Brian a dollar.

The new NVIDIA TITAN X, introduced today, based on our new Pascal GPU architecture, is the biggest GPU ever built. It has a record-breaking 3,584 CUDA cores.

We said our GTX 1080 delivers an ?irresponsible amount of performance.? It was a bit reckless. But this is even more reckless.

So forget words. Here are its numbers:

11 TFLOPS FP32
44 TOPS INT8 (new deep learning inferencing instruction)
12B transistors
3,584 CUDA cores at 1.53GHz (versus 3,072 cores at 1.08GHz in previous TITAN X)
Up to 60% faster performance than previous TITAN X
High performance engineering for maximum overclocking
12 GB of GDDR5X memory (480 GB/s)
Did we go too far? Your call. Just don?t call us crazy, or you might owe Brian a dollar, too.

TITAN X will be available Aug. 2 for $1,200 direct from nvidia.com in North America and Europe, and select system builders. It is coming soon to Asia.

https://blogs.nvidia.com/blog/2016/07/21/titan-x/

https://www.youtube.com/watch?v=Y02kptTItj8

[StyM]

that's early if true..
also no hbm2

[zanzabar]

that's early if true..
also no hbm2

hmb2 has no real supply until a tleast september.

[StyM]

so where's the Ti ...

[StyM]

NVIDIA Announces “NVIDIA Titan X” Video Card: $1200, Available August 2nd

Let’s dive right into the numbers, shall we? The NVIDIA Titan X will be shipping with 3584 CUDA cores. Assuming that NVIDIA retains their GP104-style consumer architecture here – and there’s every reason to expect they will – then we’re looking at 28 SMs, or 40% more than GP104 and the GTX 1080.

It’s interesting to note here that 3584 CUDA cores happens to be the exact same number of CUDA cores also found in the Tesla P100 accelerator. These products are based on very different GPUs, but I bring this up because Tesla P100 did not use a fully enabled GP100 GPU; its GPU features 3840 CUDA cores in total. NVIDIA is not confirming the total number of CUDA cores in GP102 at this time, but if it’s meant to be a lightweight version of GP100, then this may not be a fully enabled card. This would also maintain the 3:2:1 ratio between GP102/104/106, as we saw with GM200/204/206.

On the clockspeed front, Titan X will be clocked at 1417MHz base and 1531MHz boost. This puts the total FP32 throughput at 11 TFLOPs (well, 10.97…), 24% higher than GTX 1080. In terms of expected performance, NVIDIA isn’t offering any comparisons to GTX 1080 at this time, but relative to the Maxwell 2 based GTX Titan X, they are talking about an up to 60% performance boost.

Feeding the beast that is GP102 is a 384-bit GDDR5X memory bus. NVIDIA will be running Titan X’s GDDR5X at the same 10Gbps as on GTX 1080, so we’re looking at a straight-up 50% increase in memory bus size and resulting memory bandwidth, bringing Titan X to 480GB/sec.

[Tenknics]

so where's the Ti ...

if you read the anandtech article, it sheds some great insight. Took Nvidia 6 months to get the last titan out on Maxwell. It took Nvidia less than 60 days to do it with Pascal. I'm sure the trend will continue and we will get a 1080ti by the end of August, if not Sept, which rumors are already strongly pointing toward. Rumors were right about an early Titan X, I don't see it being any different for the ti...


I also love how you questioned if it was true when I pasted official Nvidia links. And then you post it later with Anandtech links and now its true. Lol

[StyM]

lol, thought it was a rumor, missed the blog link...
then i saw jen hsun's twitter...

[zanzabar]

if you read the anandtech article, it sheds some great insight. Took Nvidia 6 months to get the last titan out on Maxwell. It took Nvidia less than 60 days to do it with Pascal. I'm sure the trend will continue and we will get a 1080ti by the end of August, if not Sept, which rumors are already strongly pointing toward. Rumors were right about an early Titan X, I don't see it being any different for the ti...


I also love how you questioned if it was true when I pasted official Nvidia links. And then you post it later with Anandtech links and now its true. Lol

the TI has only ever come out when amd beats them. so unless they have yield problems and no quatro cards that need gddr5 i dont see it coming for a while. but late august was supposed to be interesting so we will see if they have a high end card out that is not a titan. buy then i dont see this card lasting long as it looks to be a 100 card reworked with a traditional memory controller. so, when hmb2 finally ships in large quantities so the machine learning parts can get stock and then have some left over i bet the 100 comes to the desktop.

[StyM]

so if pascal comes out this year, volta will come out next year..
so no point upgrading this year...

[SADS]

i'd prefer to have a wait for the Ti and then see if with HMB i think!

i've just ordered a second GTX980 since they're going off quite cheap on ebay and will try SLI for a while before i think about changing

[SADS]

I realise this is very off topic, but we also seem to be running on data from 2014 after the update?

need to update my sig!

[StyM]

I realise this is very off topic, but we also seem to be running on data from 2014 after the update?

need to update my sig!

http://www.xtremesystems.org/forums/...24#post5243024

[SADS]

http://www.xtremesystems.org/forums/...24#post5243024

ah thank you!

can delete the posts if required

[hennyo]

So Aug 2, does anyone have any idea what time that day I should be refresh camping the Nviidia site to purchase a Titan?

[SADS]

i've never understood why people go for the titans.

sure they're really really fast, but the price of them compared to what you get from a Ti just doesn't seem right!

Am I not right in thinking that the Ti offers around 80% of the performance for around 50% of the price? (roughly speaking!)

[StyM]

looks like no Ti for pascal, no more space for Ti.
titan @ 11 TFlops and 1080 @ 9 TFlops
doubt a 10 TFlops will be released

[Ket]

It was obvious this was coming, and once enough chips build up that don't make the new Titans grade, then the 1080Ti will be released or the chips used on a "refresh" 1080 grade part, either with 10 or 8GB GDDR5X.

[StyM]

earlier rumor was that two pascal titan will be released,
a hbm2 & gddr5x versions...

[Ket]

How anyone bought in to that rumor I'll never know. Consumer Pascal was always going to be GDDR5X as to make it compatible with HBM2 would of required significant architectural redesign which simply was never going to happen given how blindingly obvious it is that Pascal is just a slightly tweaked and castrated variant of Maxwell II. Which is to say, it would of cost nvidia too much money to tap out cores like that across the board, plus HBM2 is hardly available in large volume right now which would of pushed said GPU prices up even further, all of which equal lower sales for NV on top of lower profit margins, not a smart business move.

[kemo]

looks like no Ti for pascal, no more space for Ti.
titan @ 11 TFlops and 1080 @ 9 TFlops
doubt a 10 TFlops will be released

Maxwell Titan X had 6.6 T Flops and 980 GTX Ti had 6.0 T Flops

The question is , where is the competition to release 1080 TI.

[StyM]

pascal tesla already uses hbm2, gp100

[EternityZX9]

looks like no Ti for pascal, no more space for Ti.
titan @ 11 TFlops and 1080 @ 9 TFlops
doubt a 10 TFlops will be released

Yeah...I'm beginning to think that maybe a HBM2 based Ti won't come out this year and instead will be saved for whatever refresh nvidia comes out with for next spring.

[Chrono Detector]

Meh, was hoping for HBM2 and more than 12GB, guess that HBM2 isn't ready yet. Hope AIB cards will come out shortly as well, but since NVIDIA won't allow it to happen so soon like their Titan predecessors I guess this one won't be any different.

[Ket]

pascal tesla already uses hbm2, gp100

I said consumer Pascal, not Tesla. NV can charge basically whatever they want for Tesla cards and still rack up the sales from big corporations so it makes sense to have HBM2 on those cards, so the extensive architectural rework in that scenario is viable.

[Tenknics]

so if pascal comes out this year, volta will come out next year..
so no point upgrading this year...

there is if you need to upgrade now..