NVIDIA Tegra 4 Details Leak; Promises An Impressive Punch
When dual core phones first hit the smartphone market several years ago, Tegra 2 was perfectly positioned across the smartphone and tablet market. Tegra 3 has done extremely well in tablets this past year; Nvidia has won high-profile designs with everyone from Google to Microsoft. In the past few months, however, it's become clear that Tegra 3's quad-core, Cortex-A9 CPUs and older GPU technology wouldn't be able to compete with the latest designs from Apple or Qualcomm.
A leaked slide from a Chinese site has shed light on what Team Green plans to answer with and the next-generation SoC is pretty darn sexy. Here's what Wayne (aka Tegra 4) purportedly looks like:
The new Tegra 4 packs 72 GPU cores compared to 12 cores in Tegra 3. Conveniently, NV claims T4's performance is 6x that of Tegra 3. In this case, comparing cores against cores is a really bad idea. Tegra 2 and Tegra 3 are based on a GeForce 7-derived architecture. They have specialized hardware for pixel shader and vertex shader programs. The 12 cores inside Tegra 3 <em>are</em> cores -- they are capable of executing programs -- but they are far more limited than the CUDA architecture Nvidia debuted in 2006, to say nothing of the company's latest CUDA-capable hardware.
Nvidia's G80, and all of its desktop/mobile parts since 2006, are far more programmable and flexible. So why didn't the company bake that tech into Tegra from the beginning?
Power and die size. Nvidia's second-generation GeForce 7, built on 90nm, had a die size of 196mm sq. The first generation GeForce 8800 GTX, also built on 90nm, weighed in at 484mm sq. Since we're talking about total area, that makes G80 6x the size of G70. Adding programmable architectures with flexible resource allocation isn't the cheapest way to increase performance, but in the long run, it's the smartest.
But figuring out how to build a programmable GPU that's faster than a fixed-function model while fitting in your pocket? That took awhile. The timing makes more sense when you factor in the fact that NV started working on Tegra before Fermi had even taped out. The GeForce 7 was a thoroughly known quantity in 2008, the low-power future of CUDA was rather murkier.
What all this means is that the 72 cores inside Tegra 4 aren't going to be exactly comparable to T3s, and I don't expect a 6x linear increase -- but they should pack quite a punch. The CPU side is similarly aggressive; a quad-core Cortex-A15 should be able to match anything at the high end of the spectrum.
If the leak is true, T4 should hit the ground running in 2013. Apple, Qualcomm, Samsung, and Intel all have their own plans for the segment, but Nvidia should be well equipped to fight them.
A leaked slide from a Chinese site has shed light on what Team Green plans to answer with and the next-generation SoC is pretty darn sexy. Here's what Wayne (aka Tegra 4) purportedly looks like:
The new Tegra 4 packs 72 GPU cores compared to 12 cores in Tegra 3. Conveniently, NV claims T4's performance is 6x that of Tegra 3. In this case, comparing cores against cores is a really bad idea. Tegra 2 and Tegra 3 are based on a GeForce 7-derived architecture. They have specialized hardware for pixel shader and vertex shader programs. The 12 cores inside Tegra 3 <em>are</em> cores -- they are capable of executing programs -- but they are far more limited than the CUDA architecture Nvidia debuted in 2006, to say nothing of the company's latest CUDA-capable hardware.
Nvidia's G80, and all of its desktop/mobile parts since 2006, are far more programmable and flexible. So why didn't the company bake that tech into Tegra from the beginning?
Power and die size. Nvidia's second-generation GeForce 7, built on 90nm, had a die size of 196mm sq. The first generation GeForce 8800 GTX, also built on 90nm, weighed in at 484mm sq. Since we're talking about total area, that makes G80 6x the size of G70. Adding programmable architectures with flexible resource allocation isn't the cheapest way to increase performance, but in the long run, it's the smartest.
But figuring out how to build a programmable GPU that's faster than a fixed-function model while fitting in your pocket? That took awhile. The timing makes more sense when you factor in the fact that NV started working on Tegra before Fermi had even taped out. The GeForce 7 was a thoroughly known quantity in 2008, the low-power future of CUDA was rather murkier.
What all this means is that the 72 cores inside Tegra 4 aren't going to be exactly comparable to T3s, and I don't expect a 6x linear increase -- but they should pack quite a punch. The CPU side is similarly aggressive; a quad-core Cortex-A15 should be able to match anything at the high end of the spectrum.
If the leak is true, T4 should hit the ground running in 2013. Apple, Qualcomm, Samsung, and Intel all have their own plans for the segment, but Nvidia should be well equipped to fight them.
