NVIDIA Finally Gets Serious with Tegra

NVIDIA’s Logan will introduce the Kepler architecture to mobile products.

Tegra has had an interesting run of things.  The original Tegra 1 was utilized only by Microsoft with Zune.  Tegra 2 had a better adoption, but did not produce the design wins to propel NVIDIA to a leadership position in cell phones and tablets.  Tegra 3 found a spot in Microsoft’s Surface, but that has turned out to be a far more bitter experience than expected.  Tegra 4 so far has been integrated into a handful of products and is being featured in NVIDIA’s upcoming Shield product.  It also hit some production snags that made it later to market than expected.

I think the primary issue with the first three generations of products is pretty simple.  There was a distinct lack of differentiation from the other ARM based products around.  Yes, NVIDIA brought their graphics prowess to the market, but never in a form that distanced itself adequately from the competition.  Tegra 2 boasted GeForce based graphics, but we did not find out until later that it was comprised of basically four pixel shaders and four vertex shaders that had more in common with the GeForce 7800/7900 series than it did with any of the modern unified architectures of the time.  Tegra 3 boasted a big graphical boost, but it was in the form of doubling the pixel shader units and leaving the vertex units alone.

While NVIDIA had very strong developer relations and a leg up on the competition in terms of software support, it was never enough to propel Tegra beyond a handful of devices.  NVIDIA is trying to rectify that with Tegra 4 and the 72 shader units that it contains (still divided between pixel and vertex units).  Tegra 4 is not perfect in that it is late to market and the GPU is not OpenGL ES 3.0 compliant.  ARM, Imagination Technologies, and Qualcomm are offering new graphics processing units that are not only OpenGL ES 3.0 compliant, but also offer OpenCL 1.1 support.  Tegra 4 does not support OpenCL.  In fact, it does not support NVIDIA’s in-house CUDA.  Ouch.

Jumping into a new market is not an easy thing, and invariably mistakes will be made.  NVIDIA worked hard to make a solid foundation with their products, and certainly they had to learn to walk before they could run.  Unfortunately, running effectively entails having design wins due to outstanding features, performance, and power consumption.  NVIDIA was really only average in all of those areas.  NVIDIA is hoping to change that.  Their first salvo into offering a product that offers features and support that is a step above the competition is what we are talking about today.


Logan Gets Sampled

The next iteration of the Tegra line is code named Logan.  Little is known about this part in general.  We are assuming that it will be NVIDIA’s second Tegra on 28 nm.  Tegra 3 was still fabbed on 40 nm and Tegra 4 is the first 28 nm part.  We do not know if it is a 4+1 setup like previous versions or if they will adopt the big.Little setup that is popular now with other manufacturers.  These things are certainly unknowable because NVIDIA has not disclosed them.  They have disclosed one thing though.  NVIDIA is finally getting with the program when it comes to graphics on mobile.

Kepler was introduced last year with the GeForce 600 series of products and it was a great success for the company.  There were certainly some tradeoffs to the architecture, but overall it was extremely power efficient at the high end and exceptionally focused on the primary workload that it was being used for.  It was a graphics processing machine.  NVIDIA sacrificed GPGPU performance in the name of rendering efficiency and speed.  Kepler could still do the work, but it was hindered as compared to previous architectures in the name of graphics processing power.  This is not necessarily a bad thing as it scaled nicely throughout the entire product line; though perhaps not as efficiently as hoped with the GTX 660 and GTX 650 Ti BOOST lines.  Still, Kepler is certainly a success for NVIDIA as it has translated also to the extreme high end with Titan, Quadro, and Tesla.  These products are not hindered when addressing GPGPU, OpenCL, and CUDA workloads.

If the reader has not guessed by now, Logan will be utilizing Kepler for the integrated graphics on this particular SOC (System On a Chip).  NVIDIA will be providing a licensed ARM processing core and will be pairing it with a single 192 CUDA core SMX.  These are unified shader cores, which are a big step up from the separate pixel/vertex shaders of the previous Tegra parts.  NVIDIA is also jumping from 72 pixel/vertex units in Tegra 4 to a full 192 unified units in Logan.  These units are also much more advanced than the first generation of unified shaders contained in the now legendary GeForce 8800 GTX.  That particular product (a powerhouse for the time) featured 128 unified DX10 shaders.  On paper, this GPU is more powerful than the 8800 GTX.

We were not given the clock speeds that this product will be running at, but we do know that it is aimed at the 2 watt area.  In handheld devices 2 watts will be total SOC power, while tablets and other larger form factors will scale up in wattage as needed.  Consider that the original 8800 GTX was in the 225 watt range, this is a pretty impressive shrink.  Obviously the mobile products will be at a disadvantage when it comes to memory bandwidth and ROP partitions, but in pure floating point output the single Kepler SMX in Logan should outrun the old 8800 GTX.

What Kepler really brings to the table is a part that is fully compliant with all of the latest industry standards.  It goes not just to OpenGL ES 3.0, but jumps to OpenGL 4.4.  It is fully CUDA 5.0 compliant, which is a massive jump from the non-compliancy of previous Tegra parts.  It is also DX11 compliant, which is more than not a boon for Microsoft and their Windows RT platform.  As such it fully implements a strong tessellation engine.  From my understanding, the latest generation competing units do not support this feature.

NVIDIA is finally showing that it can compete in the feature department in the mobile space.  This is a big jump up from what NVIDIA was doing, much less what the rest of the industry is shipping.  NVIDIA is now sampling Logan to its partners for testing purposes.  We do not have a time frame upon which these products will be shipping to consumers, but we would expect it to be around 9 months from now with first available products.

There are still a lot of questions about Logan that cannot be answered at this time.  We do not know how well it will perform against the competition, we do not know what clock speed the GPU portion will be set at, and we certainly do not know what other “secret sauce” that NVIDIA has implemented into this new product.  All we know is that NVIDIA is a very aggressive company and they now have several years of experience in the ARM market.  Samsung and Qualcomm could certainly use some better competition, especially with the specter of Intel really pushing into the mobile market with their Silvermont based Atoms later this year.  There is some thought that Kepler might in fact show up in licensed form in the x86 market, but that rumor is neither confirmed nor denied by NVIDIA at this time.

What can be said for certain is that Logan will utilize the Kepler architecture for graphics and GPGPU functionality.  This is a huge step up for NVIDIA in the Tegra line, and one which could prove to be a big selling point for manufacturers.  Finally NVIDIA has not only matched, but leapfrogged over the competition in terms of graphics and capabilities.  Perhaps this will be the product that we have all been waiting for from NVIDIA?  Time will tell, but so far with NVIDIA sending samples to partners at this early date is a very good sign.  Considering how much distance NVIDIA needs to cover to catch up to Qualcomm and Samsung, this is a good thing.  The world needs more competition in this field.