Nvidia: Tegra 3+ ''Wayne'' Arrives Later This Year

[exfileme]

An improved version of Nvidia's Tegra 3 SoS will be available in 3Q12 and capable of working alongside current LTE modems.

Nvidia: Tegra 3+ ''Wayne'' Arrives Later This Year : Read more

 

[Guest]

Right on time for Windows 8!

 

[dragonsqrrl]

"Also thrown into the mix will be 24 GPU cores with support for DirectX 11.x, OpenGL 4.x and PhysX."

Wow, that's pretty big. The GPU in Wayne is supposed to be based on a new architecture. Some have even speculated it's a Kepler derivative. This should bring the GPU performance bump many have been waiting for in Tegra SOC's.

 

[victorious 3930k]

a 1.8 GHz "flagship" version, a 2.0 GHz "flagship" version

WTF?

 

[Tab54o]

They're calling it wayne.... Tegra 4 Jimmy, tegra 5 Bob and tegra 6 jose.

 

[Parsian]

[citation][nom]frozonic[/nom]system on ship??[/citation]


I am not kidding you, i did read it as "SoC" and then I read your "system on ship" as "system on chip". Brain's interpolation is bad sometimes

In regard to the post, I wish AMD had also entered this arena.

 

[ashesofempires04]

[citation][nom]frozonic[/nom]system on ship??[/citation]

It's a message in a bottle!

 

[alfaalex101]

Buy the 1.2ghz version - get a custom kernal + rom - OC to 2.0 ghz no problemo. Thou mad?

 

[gilgamex]

[citation][nom]alfaalex101[/nom]Buy the 1.2ghz version - get a custom kernal + rom - OC to 2.0 ghz no problemo. Thou mad?[/citation]

It may be lazer cut to reduce performance, like all your 680 GTX compute dreams.

 

[joytech22]

[citation][nom]gilgamex[/nom]It may be lazer cut to reduce performance, like all your 680 GTX compute dreams.[/citation]

My dreams live on.. With my GTX580

 

[tipoo]

The current Tegra 3 has an 8 "core" (if people insist on calling them that) GPU, 24 is a massive step up. They would need enough memory bandwidth to feed it though, the current one is already starved on that single channel controller. But assuming they get memory bandwidth to match core speed, 8 to 24 on an improved architecture would put it well past even the A5X graphics.

 

[tipoo]

I'd love to see Tomshardware at least try to compare the performance of chips like these to x86, I know its hard given different operating systems but there must be at least some ways to do it. 1.8GHz quad sounds impressive but the IPC is no doubt lower, I want to know how much lower than say a single core desktop Atom at the same speed.

 

[dragonsqrrl]

[citation][nom]tipoo[/nom]The current Tegra 3 has an 8 "core" (if people insist on calling them that) GPU, 24 is a massive step up.[/citation]
The GPU in Tegra 3 has 12 SIMDs, Tegra 2 had 8.

 

[dragonsqrrl]

[citation][nom]tipoo[/nom]I'd love to see Tomshardware at least try to compare the performance of chips like these to x86, I know its hard given different operating systems but there must be at least some ways to do it. 1.8GHz quad sounds impressive but the IPC is no doubt lower, I want to know how much lower than say a single core desktop Atom at the same speed.[/citation]
We should get a better idea of how the Cortex A9 performs in comparison to Atom once the first Medfield based phone launches. Further down the road, Windows 8 should give us an opportunity to do more direct comparisons between ARM and Intel mobile SOC performance.

 

[zanny]

[citation][nom]tipoo[/nom]I'd love to see Tomshardware at least try to compare the performance of chips like these to x86, I know its hard given different operating systems but there must be at least some ways to do it. 1.8GHz quad sounds impressive but the IPC is no doubt lower, I want to know how much lower than say a single core desktop Atom at the same speed.[/citation]

It is actually painfully easy to do it. GCC works on ARM, so you can cross compile anything open source that doesn't depend on OS or Instruction Set quirks of x86 and benchmark thusly. Problem is none of the benchmark tools are open source to do that with.

 

[saturnus]

Quad core ARM Cortex A15? A15? Are you sure about that?

 

[olaf]

I wish they'd make better batterys and stop with these insane powerhog of a chips..... I still can't get more the 12 of use out of my phone if I actualy use it and not just keep it in my pocket with data turned off....

 

[aznshinobi]

Seriously wish AMD would enter the ring. Really would be interested in their offerings. Always loved supporting the underdog. Plus Llano showed they had some spark still left in them!

 

[vision33r]

You people think this is like designing for a desktop PC, nobody has thought about power efficiency.

Device OEMs do not want a powerful SOC if it drains power too quickly. That's why the Tegra 3 really runs on a single core until it needs the other 4.

 

[vilenjan]

If the new gpu is indeed based on Kepler, and we assume that the 8 core version in the initial terga 3 is based on fermi, than it will not be anywhere near 3 times as powerful, maybe a 30-40% boost. Why? Because the GTX 580 has around 500 cores (fermi), while the GTX 680 has around 1500 cores (kepler).

24/3=8 Sure Kepler is a little better core for core than fermi, but thats about it. It will be more power efficient though, which is good.

 

[threehosts]

[citation][nom]frozonic[/nom]system on ship??[/citation]
No, it's "Save our Soles".

 

[fseonart]

@vilenjan,
The current Tegra 3 GPU is not based on Fermi but on G90. Tegra 3 graphics is about 5 generations behind. Kepler was built from ground up with power efficiency in mind so it was decided that it would be integrated into Tegra.

 

[jonnyboyC]

Wow pretty huge updgrade higher clock and A15 core. Also the gpu was meh at 24 cores until i read that it'll do dx11 and all the other goodies. So this means it's essentially like a scaled down kepler that was speculated eariler. That will have some serious power. Probably some serious power drain too lol