There Will Be Two 32nm Atom SoCs in 2012: Medfield and One Other
by Anand Lal Shimpi on September 17, 2011 12:49 AM ESTIntel did a good job of showing off what it promises to deliver with Medfield at IDF: a reasonably looking smartphone and tablet built around the platform. As a quick recap, Intel introduced Atom to the ultra mobile market with Menlow in 2008. Menlow was a three chip solution (CPU, NB, SB) built at 45nm. Moorestown came in 2010, integrating the North Bridge but still requiring a separate South Bridge. Most solutions in this space offer a single chip for the majority of the logic and interfaces. Medfield delivers exactly that. Built on Intel's 32nm process, Medfield finally integrates the CPU, North and South Bridges - putting it on par, from an integration standpoint, with offerings from TI and NVIDIA (Qualcomm still offers an integrated modem with its SoCs).
Intel has actually been relatively quiet about the specifics of a Medfield SoC. It's unclear whether there are one or two cores, with or without Hyper Threading. Even the GPU is somewhat unclear, although it's unlikely that it's something SGX 543MP2-class given Medfield's original launch target was sometime this year.
The goal of Medfield seems to have shifted to simply finding itself in a smartphone. We were supposed to see such a thing by the end of this year, but instead we've been told to expect it early in 2012. If Medfield is now aimed at simply showing up, what can Intel do to remain competitive in 2012?
Silvermont will be available in 2013, based on Intel's 22nm process. Between Medfield and it however will be another 32nm Atom design that has, until now, remained out of the public eye. We were first made aware of the design at IDF this week, however we weren't given any more details. If I had to guess, I'd say this 2nd 32nm Atom SoC would definitely be a dual-core solution (perhaps with HT enabled?) with an upgraded GPU core. Intel is playing its Atom strategy close to its chest until the first design win is announced.
The Google announcement at IDF was critical. This actually reminds me a lot of what happened to NVIDIA's Tegra 2 in 2010. Announced at CES 2010, Tegra 2 was plagued by design wins that never surfaced and work on OSes that never went anywhere. It was only after focusing exclusively on Android that NVIDIA was able to finally execute. Intel is at that stage now with Atom, now it's all up to execution and optimization.
Facebook
Twitter
RSS
7 Comments
View All Comments
jjj - Saturday, September 17, 2011 - link
Given how slow Google is when releasing a new Android version on just 1 SoC and how long it takes others to make it work on their own SoCs, i wouldn't put too much weight on this announcement. Intel might get to be the first at some point ,when they'll have a decent SoC, but until then they should be at a disadvantage.As for Nvidia they only got lots of wins in tablets because everybody (or almost) wanted to avoid more delays in getting tablets to market. If it wasn't for that they would have very few wins.Looking forward they might have a crappy year if dual core Krait clocked north of 1.5GHz arrives in time. The key for Nvidia wasn't their focus on Android,Tegra 2 was the fastet SoC on the market and a good choice for Google.
Going back to Intel,they don't even really care much about phones since there is little cash in it for them and the competition is tougher than in x86.They got to enter this space at some point,pretty soon,but there is no rush.
Exophase - Monday, October 10, 2011 - link
If they don't care about phones what are they doing Medfield for? Look at Moorestown - the only press we ever got for it was how it'd be in phones, and in the end it got almost no design wins (I can think of exactly one, an oddball hybrid that runs Windows 7 while also having an ARM SoC for running Symbian w/o killing your battery life). I'm sure it's not economical to make an entire platform that doesn't end up getting used in any products. But if Intel isn't aggressive with getting Medfield in phones the same thing is going to happen.Back with Menlow there was at least a highly niche UMPC market but that's been mostly killed off, and MIDs as Intel envisioned them were subsumed at both ends by high end ARM phones and small ARM tablets.
Aries1470 - Saturday, September 17, 2011 - link
Having been looking at this segment, Intel will miss it this year, but they have already started in Japan with the Fujitsu F-07C http://www.fujitsu.com/global/news/pr/archives/mon...It has a Intel Atom in it and comes with a full version of Win 7 32bit Home Premium SP1 and Office.
The battery life isn't that good in Win7 mode ~2h, but still...
So, when they start shipping them with 2Gb or more of ram, a mini 1.8" either ssd or hdd with 128GB or (up to 320Gb for the hdd), that would be more interesting :-)
You will then notice more people going to these devices, since all they would need for the general joe is a keyboard/ mouse and a hdmi out. they k/b & mouse could even be BT that uses less energy if they forgot to pack their charger with them.
This is the exciting time for ARM vs Intel, like about a decade ago was AMD vs Intel for the pc / desktop arena. Now it is the mobile arena, i.e. phones, tablets, ultralights etc. basically SFF's.
Just my thoughts at least.
Gondalf - Saturday, September 17, 2011 - link
Interesting thing ! it will be head or faster (likely faster) than 2012 quad core Arm devices...So the customer will be able choose between a fully x86 capable Tablet running Win 8 and a so so x86 capable (recompilation) Arm tablet running the same OS :). Will see the winner.
From a perf. point of view DC Medfield will be surely faster running native x86 SW :). Arm has a competitor....finally !!!
RU482 - Monday, September 19, 2011 - link
menlow was a 2 chip solution (CPU + PCH), not a 3 chip solution (CPU + NB +SB)Exophase - Monday, October 10, 2011 - link
Menlow didn't have an IMC/IGP, you're thinking of Moorestown.zhangqq - Monday, October 31, 2011 - link
http://ygn.me/bTf7p