Titan X Announced

[n0x1ous]

What does a backplate do...?

It can offer some heat dissipation but primarily they are for protection and aesthetics.

The negatives are they encroach on the space above the pci-e slot which hurts the air that a card in the above slot would get in a multi-gpu setup. On the 980 and 970, nvidia made a small removable piece in the spot that another cards fan would be to allow for more airflow to alleviate this issue.

Perhaps having VRAM on the back of the Titan is a factor?......Titan Z had a backplate too. Was there VRAM on the back of Titan Z?

 

[destrekor]

What does a backplate do...?

Mainly provide support to a longer and heavier card, and as a secondary benefit, help protect the components on top from getting dinged. It also serves as a heatspreader for any memory and VRMs on top, and that may be the chief reason for its presence when compared to preventing "dings", as that is rarely something any manufacturer actually has in mind. You either know how to handle components or you don't, amirite?

And as noted in the reply before mine, they look cool and serve as a bonus marketing technique for those who are easily impressed by how cool things look. :awe: Much like glowing lights that show the card name.

 

[imaheadcase]

Feel bad for the people who buy that when you'll have the consumer version of GM200 in the autumn for half the price if not less with probably only a very, very minor performance delta.

Why would you feel sorry for someone who can afford it now and have it for months before you, and be faster than what you get?

 

[destrekor]

Why would you feel sorry for someone who can afford it now and have it for months before you, and be faster than what you get?

I'm not him but I can respond. It's their money so I won't cry a river for them, but I would pity them for spending excessively and approving of this kind of release schedule that is meant to do exactly that, take advantage of people with more money than sense.

Now, if I was rolling in money and buying that card is the same financially as buying a 290X priced as they are today, well... shoot, why not? Buy 2 or 3 of them! I'm impatient so if I had that kind of money, I would. And I would expect people to have a certain level of pity regarding my expenditures. That said, I will almost admit I will be partially jealous of anyone who could afford such a beast of a card.
But I really can't wait to see how the top 300 series card performs compared to GM200, because if it comes in as its equal at even half of the price of the Titan X, I won't be able to help myself and will have a fit of hysterical laughter and truly pity the actual gamers who bought it.
For anyone who needs CUDA, well, it might be the right decision. But we don't truly know the FP64 performance. We know GM204 is pitiful, but that could have been a purposeful design decision, much like crippling FP64 performance in the 780/780Ti compared to the Kepler Titan series. They might yet demonstrate good compute performance on the Titan X. And if it performs worse than the Titan, then I hope people who need CUDA ignore the Titan X.

But it's not my money, so I truly don't care. I can feel pity and/or jealousy, but ultimately, I don't care.

 

[njdevilsfan87]

What mad scientist setup would one need that can actually use 12gb of vram?:awe:

I remember being so amazed when i jumped from 512mb on my 8800gts to the 1gb gtx280,i was blown away but 12gb of vram wow.What the hell can one do with this much vram?Launch nuclear warheads?Guessing the best usage of this much memory is with configurations consisting of multiple cards?

I do admire the beast though.

These are like the first steps to building our own Matrix!

(and I only say this half jokingly)

 

[Subyman]

http://www.legitreviews.com/hands-nvidia-geforce-gtx-titan-x-12gb-video-card_159519

It was a fake and mockup-2

I thought they handed them out to reviewers at the event?

 

[Creig]

12GB is a ridiculous amount of memory for a consumer level video card at this time. There is no way that all 12GB will ever be used within its expected useful lifetime. I would imagine that 12GB will be the exception, not the norm. 6GB cards will most likely comprise the majority. Well, the majority of a $1,000+ card line, however many they manage to sell at that price.

 

[destrekor]

12GB is a ridiculous amount of memory for a consumer level video card at this time. There is no way that all 12GB will ever be used within its expected useful lifetime. I would imagine that 12GB will be the exception, not the norm. 6GB cards will most likely comprise the majority. Well, the majority of a $1,000+ card line, however many they manage to sell at that price.

I would expect Nvidia to release a 6GB GM200 as a GTX card before the end of the year at no more than $700. Perhaps in the summer a $700 cut-down GM200, and then in the fall that drops to $500 and a full GM200 GTX again uses the $700 price level.

Right now they aren't focusing on gaming with the Titan, the real purpose of Titan, when it came to gaming, was to only capture the absolute extreme enthusiasts willing to pay for it, otherwise it targets developers and others who work with Cuda and OpenCL.

A consumer and less astronomically-priced 6GB version will surely be released. I'm curious if they'd actually release a 12GB variant in the consumer line. It would be possible considering 1GB RAM chips are basically the standard now, but as you said, completely unnecessary in a consumer gaming card.

 

[raghu78]

12GB is a ridiculous amount of memory for a consumer level video card at this time. There is no way that all 12GB will ever be used within its expected useful lifetime. I would imagine that 12GB will be the exception, not the norm. 6GB cards will most likely comprise the majority. Well, the majority of a $1,000+ card line, however many they manage to sell at that price.

The Titan is a money grab. It makes sense only for CUDA professionals who need the double precision perf. 6GB is more than enough for 4k and multi monitor. We will see a GTX 980 Ti or GTX 990 with 6GB VRAM. btw within 12 months (Q2 2016) we will see 14/16nm GPUs so these GPUs will have a short reign at the top.

I am sure the R9 390X will make the Titan-X price look silly. AMD will price their products aggressively to gain back the lost GPU market share. So we will see some very good prices for the consumer. :thumbsup:

 

[skipsneeky2]

12GB is a ridiculous amount of memory for a consumer level video card at this time. There is no way that all 12GB will ever be used within its expected useful lifetime. I would imagine that 12GB will be the exception, not the norm. 6GB cards will most likely comprise the majority. Well, the majority of a $1,000+ card line, however many they manage to sell at that price.

The nature of the beast,you get 4 of these things for 4k and there it is.I honestly doubt any game could push this much memory.Would be fun to see though.

Maybe not exactly for gaming but you know the diehards will buy them 4 of them anyways.

 

[n0x1ous]

I honestly doubt any game could push this much memory.Would be fun to see though

4k surround perhaps?

 

[n0x1ous]

AdamK do you have a pre-order in for 4 with Jensen?

 

[Headfoot]

It is my understanding a sizeable portion of CUDA workloads don't need high DP throughput, but require literally as much RAM as you can possibly give it.

It could be that nVidia is breaking down the line like this:
Quadro = Professional Drivers, Full DP, maximum RAM size, Colossal price tag
Tesla = Compute focused (no video outputs), no professional drivers, just CUDA, Full DP, slightly less colossal but big price tag, maximum RAM size
Titan = Gamer drivers; Gamer DP but still runs CUDA fine; expensive for gaming, cheap for compute; max RAM size
GeForce = =Gamer drivers, Gamer DP, half-density RAM size (normal RAM size), cheapest but still runs CUDA

 

[skipsneeky2]

4k surround perhaps?

I think Adamk47 or someone here with a youtube channel had i think a few 5k monitors tank the 6gb Titan cards lol.If i could find the video i will edit the post with a link.

I am sure it was Adam.

Edit:It was a single 5k monitor and it was tanking the 4gb 980s

https://www.youtube.com/watch?v=AYIFDm9iyKk skip to about 13 minutes,he goes into detail about the memory issue with games.

 

[destrekor]

It is my understanding a sizeable portion of CUDA workloads don't need high DP throughput, but require literally as much RAM as you can possibly give it.

It could be that nVidia is breaking down the line like this:
Quadro = Professional Drivers, Full DP, maximum RAM size, Colossal price tag
Tesla = Compute focused (no video outputs), no professional drivers, just CUDA, Full DP, slightly less colossal but big price tag, maximum RAM size
Titan = Gamer drivers; Gamer DP but still runs CUDA fine; expensive for gaming, cheap for compute; max RAM size
GeForce = =Gamer drivers, Gamer DP, half-density RAM size (normal RAM size), cheapest but still runs CUDA

The GeForce cards have less DP than Titan cards, by a wide margin. And it looks like the Titan cards had basically the same DP performance as the Tesla cards based on the same chips, which is the same as the Quadro.

They market them different and have different business warranty support and whatnot for the different lines. The Titan is basically a consumer Tesla or Quadro that can also play games perfectly fine, but it is likely supported the exact same as the GeForce brand. Drivers and toolkits for the professional lines are developed more slowly with a major focus on stability and uptime, I reckon, and they don't bother with optimization for gaming of course.

 

[Alatar]

It was a fake and mockup-2

Hopefully just joking.

There's nothing to indicate that it's a woodscrews card. Just looks like the cooler has been hand installed (or reinstalled) and someone did a sloppy job. Wouldn't be the first time I saw someone forget the screw that supports the pci-e bracket.

 

[destrekor]

Hopefully just joking.

There's nothing to indicate that it's a woodscrews card. Just looks like the cooler has been hand installed (or reinstalled) and someone did a sloppy job. Wouldn't be the first time I saw someone forget the screw that supports the pci-e bracket.

I wouldn't find fault in showing a non-functioning card, be it an engineering sample or some other assembly that is either dead or never functioned. The ones in working systems were the real demonstration and showcase, what he holds up in the air is simply a flashy model. I think it may be common to see such public demo "hold in the air" cards as basically dead or fake.

Again, they had functioning ones in demonstration systems, and that is where you truly show off your new card. This isn't like a car show where you roll out a working model and can show off the full product and it is self-propelled by means of the parts inside the product. This is a sort of like showing off a transmission, you can't really demonstrate it in motion when you show it on stage. In fact, you usually get cut-away models to show what it looks like inside, designed to not function like it would in the final assembled car.

 

[xthetenth]

I think it may be common to see such public demo "hold in the air" cards as basically dead or fake.

Especially if they're going to let Linus wave it around.

Joking but the whole point of the wood screws thing was that they didn't have a working card. What does it matter if he waves a dead card or even a dead chicken around by its neck if they have a working card that they can show the performance on?

 

[Creig]

Especially if they're going to let Linus wave it around.

Joking but the whole point of the wood screws thing was that they didn't have a working card. What does it matter if he waves a dead card or even a dead chicken around by its neck if they have a working card that they can show the performance on?

Well, the difference back then was that while Jen-Hsun was happily waving that dead chicken... I mean, dead card around, he proclaimed "This puppy here... is Fermi!" when it actually was just some random PCB cut down to fit a Fermi cooler. It's the PCB and its components that make a video card, not the cooler.

 

[AdamK47]

AdamK do you have a pre-order in for 4 with Jensen?

No pre-order. I'll just clean out the local Micro Center of Titan X cards shortly after the store opens on release day. No waiting on them being shipped. Did that with my 980s.

 

[Baasha]

No pre-order. I'll just clean out the local Micro Center of Titan X cards shortly after the store opens on release day. No waiting on them being shipped. Did that with my 980s.

LOL... Wish the Micro Center near me was still open - they shut down due to slow business(?) - At the heart of Silicon Valley!

Anyway, can't wait for this beast of a card to come out!

Witcher 3 @ 5K w/ 4x GTX-Titan X !!

 

[Headfoot]

The Titan is basically a consumer Tesla or Quadro that can also play games perfectly fine, but it is likely supported the exact same as the GeForce brand. Drivers and toolkits for the professional lines are developed more slowly with a major focus on stability and uptime, I reckon, and they don't bother with optimization for gaming of course.

I am commenting on how this might be changing. Rumor has it the Titan X will not have full DP capability, further segmenting their line where you can get large RAM size without DP or professional drivers in the form of Titan

 

[skipsneeky2]

LOL... Wish the Micro Center near me was still open - they shut down due to slow business(?) - At the heart of Silicon Valley!

Anyway, can't wait for this beast of a card to come out!

Witcher 3 @ 5K w/ 4x GTX-Titan X !!

No more issues with 4gb at 5k either,maybe bump those settings in BF4 to high or even ultra now and not be limited to medium?

 

[escrow4]

The Titan is a money grab. It makes sense only for CUDA professionals who need the double precision perf. 6GB is more than enough for 4k and multi monitor. We will see a GTX 980 Ti or GTX 990 with 6GB VRAM. btw within 12 months (Q2 2016) we will see 14/16nm GPUs so these GPUs will have a short reign at the top.

I am sure the R9 390X will make the Titan-X price look silly. AMD will price their products aggressively to gain back the lost GPU market share. So we will see some very good prices for the consumer. :thumbsup:

Anything AMD isn't Titan though. That is the whole point. Nvidia has luxury GPUs covered. Doesn't matter if its a rip off its a Titan. And it will sell. A lot.

 

[caswow]

Anything AMD isn't Titan though. That is the whole point. Nvidia has luxury GPUs covered. Doesn't matter if its a rip off its a Titan. And it will sell. A lot.

i dont know. if someone comes to me and says he got a titan gpu i wouldnt be jealous as it would be more feeling pitty for their poor choice to impress someone with electrical waste that gets replaced 5 months after its purchase.