The recent Nvidia roadmap leak shows us what Nvidia has in store for late 2011 to late 2012 period. We can see Kal-El in there, a never before seen Kal-El+ (most likely an upgraded version of Kal-El), Wayne (Kal-El’s next-gen successor), and Grey (seemingly a low-end chip).
Kal-El+ was to be expected, although many people seem surprised that it exists. We learned last year when we first heard about the dual core 1.2 Ghz Tegra 2 3D that Nvidia was planning on releasing a “new” chip every 6 months, instead of just every year. That’s not very shocking considering all the other chip manufacturers are currently doing that, too: Samsung with the dual core 1.2 and 1.4 Ghz Exynos chips, Qualcomm with the dual core 1.2 Ghz an 1.5 Ghz Snapdragon chips, and TI with their dual core 1 Ghz and 1.5 Ghz chips (all for 2011).
But unfortunately for Nvidia, Tegra 2 was already too late in the market, and already in planned products. It wouldn’t have made sense to push Tegra 2 3D to appear in tablets for fall 2011, when they already had Kal-El ready for that. So they ditched that mid-life kicker to Tegra 2. But if everything goes according to plan this time (so far Kal-El was already pushed 2 months, though), we should see Kal-El+ before Wayne arrives, in both tablets and phones.
My guess is Kal-El+ will be just an upgraded Kal-El from quad core 1.5 Ghz Cortex A9 to quad core 2.0 Ghz Cortex A9, while Wayne will probably be a quad core 2.5 Ghz Cortex A15 chip, manufactured at 28 nm.
It may look like Wayne will be just a slight improvement over Kal-El and Kal-El+, but it won’t be. It will be an improvement just as big as from Tegra 2 to Kal-el – that is to say, huge. Cortex A15 is supposed to be 40% to 100% more powerful than Cortex A9, so a quad core Cortex A15 at 2.5 Ghz should be 2-3x as powerful as a quad core 1.5 Ghz Kal-El, while Kal-El+ will only be about 33% more powerful than Kal-El. Kal-El+ is just a mid-life kicker, while Wayne will be Kal-El’s true successor.
Now, about Grey – Nvidia is finally looking at the low-end market. If the graphic at the top is accurate, Grey is supposed to be less powerful than a Tegra 2, and obviously much cheaper, too. It’s meant for low-end to mid-end smartphones, and also tablets (you know, the under $200 ones I’ve been asking for?).
Why is Nvidia doing this and why is it important to them? I believe it’s very important they should get in the lower end markets. Why? Because if they really want to (finally) become the “Intel of mobile chips”, it’s not enough to just deliver the “best” chips on the market, as Intel also does. It’s also important to “dominate” that market, from low-end to high-end. If they want to eventually become as powerful as Intel, but in the mobile world, they need to follow their full strategy.
Nvidia is completely right when they believe ARM is the future. Everyone knows it. Microsoft knows it, Apple knows it. Only Intel doesn’t know it (or just refuses to accept it and embrace it). As smartphones start replacing all phones, there’s a lot to gain from owning the low-end of the chip market, too.
At the moment, Qualcomm sort of has that covered, making chips from a measly 600 Mhz ARM11 chip to a dual core 1.5 Ghz Cortex A9 one, coming this year. Qualcomm has a lot more design wins than Nvidia right now- something like 150-200. That’s probably more than 50% of the market.
So how powerful will Grey be? My guess is it will be either a single core, or possibly a dual-core chip, and it will be either Cortex A5 or a custom made chip by Nvidia. Cortex A5 is a new chip (not to be confused with Apple’s A5, or Cortex A15). It’s meant to replace the now ancient ARM11 we can find in low-end smartphones. It’s supposed to offer up to 80% more performance than ARM11, for the same footprint and energy consumption.
It will be almost as powerful as a Cortex A8 (Hummingbird CPU, for example), but probably for half the price or lower. It’s also built on the ARMv7 architecture, which Cortex A8, A9 and A15 also share, as opposed to ARM11 which was built on ARMv6 architecture. This should also mean that it will support Flash, because Adobe made Flash work on the ARMv7 architecture, but not ARMv6.
Nvidia is in a very good position right now with their brand (although it would’ve been even better if they actually delivered Tegra 2 on time, so it doesn’t have to look like it’s weak when it appears 6 months later than it should have on the market). Kal-El is getting people excited again, Kal-El+ should continue that, and Wayne will be the first ARM chip to actually hurt Intel in the clamshell notebook market (if you look in the graph, it’s intended for both tablets and clamshells, equally).
Where’s Nvidia heading after this? In 2013 they should have their most likely custom-made Project Denver chip, which should have anywhere from 8 to 32 cores, and be meant for servers mainly, but also for high-end ARM PC’s and notebooks. The downside to that is that it shouldn’t have a tiny power consumption like Tegra 2, Kal-El, and Wayne, but one that is several times bigger. However, it should still be a low TDP compared to x86 chips, perhaps under 10 W, which would actually make it compete with Atom chips at that level on power consumption, but be much, much more powerful.
Project Denver is the reason I’m uncertain whether Grey is a Cortex A5 chip, or a custom-made one. I think Nvidia might be moving to a custom-made ARM design, sort of like Qualcomm, rather than use and improve on ARM’s own stock Cortex designs.
With AMD and Intel squeezing Nvidia out of the notebook market, AMD by owning the old ATI graphics cards, and Intel by using their integrated graphics, which will gradually become more and more “good enough” for normal users, Nvidia is committed to “win” the ARM mobile chip market, no matter what. With the whole Post-PC shift, I think Nvidia will have the last laugh, as neither Intel nor AMD seem to be ready to move to ARM, so they will be the ones to ultimately lose, when most “personal computing devices” will be based on ARM.