Nvidia News: Shield Gets New Software, and the $3,000 Titan Z GPU

Nvidia News: Shield Gets New Software, and the $3,000 Titan Z GPU

Nvidia Titan Z 310x

Nvidia pushes out a feature-rich OTA update for the Shield in April, and unveils a new $3,000 super GPU.

Nvidia is dropping plenty of news today on both the mobile and desktop GPU fronts, from $3,000 graphics cards to Shield software updates.

Starting in the wonderful realm of the desktop GPU, Nvidia is bringing a new version of it's super-duper-ludicrous-speed-high-end GPU to market. The Kepler-based, dual-GPU Titan Z packs 12 GB of memory and 5,760 cores. In other words: The Titan Z is two Titan Black GPUs on one PCB. Nvidia says the Z is for those looking for 5K and multi-monitor gaming bliss.

The Titan Z is coming shortly ahead of an expected dual-GPU card reveal from AMD -- the R9 295X (my working title, not AMD's). My guess is AMD's offering won't be $3,000, but it also won't be as powerful, assuming it's based on two R9 290X GPUs.

Shifting gears to the Nvidia Shield, the Tegra-powered gaming portable will be getting a feature-rich OTA update on April 2nd. The update will upgrade Android on the Shield to 4.4.2 (KitKat), first and foremost, but plenty of Nvidia-specific updates are in tow.

GameStream is getting a major update, and the service will now allow for PC-to-Shield gaming outside of the PC's network. Want to play Borderlands 2 on you Shield while your surviving another trip to the in-laws house? That's totally possible now. The Shield will be able to wake up a sleeping PC remotely, and connecting to password-protected machines will be possible as well.

You'll need a "robust" WiFi connection (5 Mbps or better) to use the Remote GameStream feature, but you're no longer chained to your home network.

Streaming will now be possible from Nvidia-equipped notebooks, too. If your gaming laptop has a GTX 600, GTX 700, or GTX 800 GPU, you're likely in the clear ("likely" because some lower GTX 600 GPUs might not cut the mustard).

Other enhancements include Bluetooth keyboard and mouse support, a revamped TegraZone app, and upgrades to the Gamepad Mapper tool.

Last but not least: From right now through the end of April, the Shield will see a price drop from $249 to $199. That's a standalone price -- no GeForce GPU purchase required.

Permalink

Dual titan on one PCB ? Its a bit overkill IMO but hey way to kick AMD when its down :P. Personally i am waiting to see what the do with Maxwell it is supposed to be even more power efficient will be nice to see what they do with the 850ti or whatever they call it this time around, that will make a good candidate for a steam-box card, the 750ti is already efficient enough that it does not need additional power from the auxiliary PCIE power, i am looking forward to a super small card that can be passively cooled :).

The Titan Z is two Titan Black GPUs on one PCB.

So wait...
It's 3000$ yet a single Titan is around the 1000$ mark.
So shouldn't it actually be around the 2000$? Where does the other 1000$ magically appear from?

That is... overpriced IMO. 2200 maybe, not 3000...
Though even Titan Black is already overpriced as sh*t. That is 850, max.

Besides, there is... questionable point in such a GPU even now. Especially with Direct X 12, new OPenGL and Mantle on the horizon.
No game requires that much power. Even a NORMAL 780 is pushing it. A 760 is more then enough a 750 TI is sweet...

Haha, I love the way he's holding it up.
Someone needs to make a caption that says something like...
LOOK UPON THE FACE OF YOUR NEW GOD, AND DESPAIR!

I'm really excited about the new Shield updates since I'm the only one who seemingly owns one. That might make the shield a lot more functional if you can get on a guest wifi network with a reasonable speed.

VincentX3:

The Titan Z is two Titan Black GPUs on one PCB.

So wait...
It's 3000$ yet a single Titan is around the 1000$ mark.
So shouldn't it actually be around the 2000$? Where does the other 1000$ magically appear from?

All the bits that allow the 2 of the largest GPU's to not burst into flames while on a single card plus the higher density memory to allow 12GB to fit on a normal sized PCB and all the additional engineering.

Plus the workstation Tesla and Quadro cards are $5000 for the equivalent of the regular Titan Black.

This card will sell to those who have a Quadro or Tesla and it's specialist drivers but need more grunt and the Titan Z is perfect as it's nearly twice the performance of a Tesla for the same price.

Kumagawa Misogi:

VincentX3:

The Titan Z is two Titan Black GPUs on one PCB.

So wait...
It's 3000$ yet a single Titan is around the 1000$ mark.
So shouldn't it actually be around the 2000$? Where does the other 1000$ magically appear from?

All the bits that allow the 2 of the largest GPU's to not burst into flames while on a single card plus the higher density memory to allow 12GB to fit on a normal sized PCB and all the additional engineering.

Plus the workstation Tesla and Quadro cards are $5000 for the equivalent of the regular Titan Black.

This card will sell to those who have a Quadro or Tesla and it's specialist drivers but need more grunt and the Titan Z is perfect as it's nearly twice the performance of a Tesla for the same price.

Yet wouldn't it be better just to buy 2 Titans and save 1000$?
I Imagine the people that buy a Titan have enough space for 2 of them anyway.

Good thing I did not wait and purchased myself two Titan Blacks for hardware accelerated CGI rendering. That card would be even more awesome if the memory was unified instead of two separate pools of 6GB, assuming it is not already. None the less it is no longer on my radar anymore given that it would be a grand more than what I payed for my two babies.

CAPTCHA: Very doubtful - Touche.

VincentX3:

Yet wouldn't it be better just to buy 2 Titans and save 1000$?
I Imagine the people that buy a Titan have enough space for 2 of them anyway.

The only advantages would be if they figured out how to unify the two 6GB pools of memory into one large 12GB that non-SLI programs can make use of and more than twice the CUDA count which scales well in CGI rendering.

Judging by the price of that GPu I'd imagine it's not going to sell for many millions if not maybe a few hundred or thousand, I seriously have to question what that GPu is specifically meant for because I sure as hell don't see that being anywhere near suitable for "PC gaming", if anything it's meant for CGI rendering and other much bigger tasts that require more power than a mere game.

Though I'm already expecting some super rich with money to throw away enthusiasts with their mouths watering at a piece of hardware that will eventually become dated within at least a years time.

The Shield whoop I suppose, I don't really see a need for it for my own gaming habits and it feels more expensive than say my 3DS or a Vita.

VincentX3:

The Titan Z is two Titan Black GPUs on one PCB.

So wait...
It's 3000$ yet a single Titan is around the 1000$ mark.
So shouldn't it actually be around the 2000$? Where does the other 1000$ magically appear from?

Saulkar:

Having the single fastest setup on a single PCB is probably more relevant to workstation use. I guess they decided to make a gaming version of this because... well because why the hell not? Personally I'd like to see how price vs. performance goes for this. The 7990 isn't an ideal card but its price is dropping considerably. There is also the caveat that you can pretty reliably convert it into a Firepro s10000 with a bit of driver and firmware tweaking.

I use CAD packages, do high quality rendering and even run complex simulations like those done by Ansys and i think that's where having all this sheer power is best suited.

A lot of work goes into putting two high-end, workstation-class GPUs onto one PCB. $3,000 is a lot of cheddar, but for those who can afford it, the performance is (probably) going to be insane.

I have the biggest techie/nerd boner right now. I could never justify laying out that kind of money just for a gpu, but it is nice to think that in my life time we have gone from something like the TRS-80 to this tiny monster.

Imagine in 10 years or so when something like this is the standard.

Also, the Shield is sounding more intriguing every update I hear.

I have the biggest techie/nerd boner right now. I could never justify laying out that kind of money just for a gpu, but it is nice to think that in my life time we have gone from something like the TRS-80 to this tiny monster.

Imagine in 10 years or so when something like this is the standard.

Also, the Shield is sounding more intriguing every update I hear.

VincentX3:

The Titan Z is two Titan Black GPUs on one PCB.

So wait...
It's 3000$ yet a single Titan is around the 1000$ mark.
So shouldn't it actually be around the 2000$? Where does the other 1000$ magically appear from?

Well its not like you can cram two Titans into a space slightly larger than one Titan and call it a finished product.

That aside, chances are if you have the other hardware capable of spitting out 4K resolutions for multiple screens, then you must be so loaded that $3000 would be pocket change.

And of course, theres certainly no need to buy the Titan Z if youre happy at 1080p.

Scrumpmonkey:

I use CAD packages, do high quality rendering and even run complex simulations like those done by Ansys and i think that's where having all this sheer power is best suited.

Sweet! I am primarily a 3DS Max and Blender User, supplemented by Zbrush, Photoshop, and Painter.

Have you ever tried doing the BMW test scene for Blender Cycles with your current build?

http://blenderartists.org/forum/showthread.php?239480-2-6x-Cycles-render-benchmark
image

With 2 Titan Blacks with the render buckets set to 480x270 (quarter HD resolution) I can get render times as low as 12 seconds where as the number crunching behemoths that were the 500 series get almost three times that.

Saulkar:

Scrumpmonkey:

I use CAD packages, do high quality rendering and even run complex simulations like those done by Ansys and i think that's where having all this sheer power is best suited.

Sweet! I am primarily a 3DS Max and Blender User, supplemented by Zbrush, Photoshop, and Painter.

Have you ever tried doing the BMW test scene for Blender Cycles with your current build?

http://blenderartists.org/forum/showthread.php?239480-2-6x-Cycles-render-benchmark

With 2 Titan Blacks with the render buckets set to 480x270 (quarter HD resolution) I can get render times as low as 12 seconds where as the number crunching behemoths that were the 500 series get almost three times that.

Well I'm more on the functional side of the number crunching as a Mechanical Engineer.

Renders are only really used for presentation purposes and since those can be done at a central location i just have a middling (by workstation standards) home setup and tend to render at better workstations i have access to or use non local services (God I hate the term 'cloud'). As long as i have enough power to show complex assembles in Autodesk Inventor (which is largely a RAM issue) I'm usually okay.

Computational fluid dynamics and high detail mesh stress analysis are what I'd love to turn these kinds of cards towards though. I've already seen how an Firepro S10000 can chew through numbers and generate readouts in 10 or so minutes that would usually take hours. The GPU support for these programs is in its infancy though so we'll have to see where it goes. Before you literally needed racks upon racks of CPUs in a server to do these kind of operations. Now they are moving workstation side which is not only exciting but also helps democratize access to high precision 'in cilico' testing.

How long would it take to render if the ray-tracing was better? 12 seconds is impressive but i can still see quite a few artifacts (and what i like to refer to as "The swarm of angry bees" :P) in that image.

Scrumpmonkey:

How long would it take to render if the ray-tracing was better? 12 seconds is impressive but i can still see quite a few artifacts (and what i like to refer to as "The swarm of angry bees" :P) in that image.

Yeah, the render samples are set by default to 200 for this test scene. To completely eliminate them you need to set said samples to 1500 which increases the render time by 3 to 4 minutes. I am on my laptop right now, I will get back to you on how long it takes to render in about half an hour.

EDIT: It takes a minute and ten seconds to render a single half 1080p frame with 1500 samples but 4 minutes and 18.5 seconds to render full HD resolution.

I'm pretty sure this would be great for dogecoin mining.

Look, i know titan is a cool name and all, but let it rest. everyone knows its just a 780ti with more ram and not worth the money by now. you may be better off cramming two 780s together. but then, how can you mark up the price to twice the worth, right?

Evil Smurf:
I'm pretty sure this would be great for dogecoin mining.

Unlikely. Cuda mining is not as good as AMD miners and the only thing titan got for it is unhealthy amouts of RAM which is only useful when your rendering video on multiple screens.

 

Reply to Thread

Log in or Register to Comment
Have an account? Login below:
With Facebook:Login With Facebook
or
Username:  
Password:  
  
Not registered? To sign up for an account with The Escapist:
Register With Facebook
Register With Facebook
or
Register for a free account here