According to this article, Microsoft is pestering Intel to produce low-power Atom-based, x86 processors for server machines.
I am wondering why they would force this direction. Do they know server requirements better than Intel? Why do they think that low-power x86 server chips are so important?
I believe the answer comes from the following factors:
- Power efficiency is becoming more and more important in the server room. Intel processors (Microsoft’s home turf) have less than stellar watt/performance efficiency but they are the best in raw performance / cores.
- ARM provides the best watt/performance in general computing (far far better than Intel x86) and ARM is seemingly scalable to the server performance range (with multiple cores and coming to 28nm high-performance production processes)
- Microsoft doesn’t have a server operating system presence on the ARM architecture. Linux on the other hand runs on ARM, has optimized distributions for ARM SOCs.
- ARM licensees are actively pursuing server chips (Nvidia, Nufront…etc)
If 4-16 core ARM server processors appear in the near future, servers built with them would have superior watt/performance ratios so they may quickly gain acceptance.
These systems would be perfectly served by Linux distributions (Red Hat, Ubuntu, Suse) and Microsoft could not offer anything for them. Linux is already the strongest player in the datacenter and this would grow its market share considerably while reducing the market-share of Windows simultaniously.
Even if Microsoft manages to create a stable Win8 server OS solution with all the required additional Windows sw (database systems, application servers…etc) on ARM in 2-3 years, it will be pretty much too late. They will need to play catch-up with Linux. The Microsoft Win8 solution will have to sell for peanuts to be in the game which would make it very much unprofitable in the short-medium run. Moreover, as x86 server market share goes down, their x86 Windows Server OS profits also go down.
All in all: If ARM processors appear in the market in the near future, Microsoft may face a steep uphill battle in the datacenter. If x86 based Atom server can slow down the onslaught of ARM servers, Microsoft may gain enough time to come up with a Win8/ARM server solution and avoid serious loss of server market-share.
Reading about the likely launch of Tegra3 at Mobile World Congress 2011 and seeing this video, one cannot help wondering how big a mistake Intel made when denied Atom hardware interfaces from Nvidia some time ago. Doing that, it practically forced Nvidia to abandon mobile-x86 solutions and pour all of its resources into Tegra/ARM development.
Nvidia has recently announced its Project Denver effort which also shows how seriously the graphics company wants to transform into an all-out computer technology company shipping mobile, desktop and server processors as well not only graphics solutions.
As a result, Intel will have to face not only AMD in the desktop/server segment but a big-name ARM technologist as well. (And several smaller ones like Nufront)
Tegra3 is not well known yet, but some guesses can be made:
- Quad-core Cortex-A9 symmetric multi processing for generic application code execution
- Likely at least 1Ghz top, possible up to 1.5 Ghz, dynamic frequency scaling and individual core-power-off
- Geforce 8 or 9 level graphics core, likely with high-profile 1080p playback and encoding
- Support for Linux and Android
- Possibly produced on a <40nm process (GlobalFoundries 28nm anyone?)
If Nvidia can produce this on the GlobalFoundries 28nm process (or similar), we can be quite certain that the new SOC will still be viable for smartphones and will be an extremely appealing solution for tablets and Motorola Atrix-like phone/netbook/tablet modular solutions.
It will make Moorestown Atoms a very-very hard sell for Intel in the mobile phone and tablet space since the computing-power advantage of Moorestown is gone and Tegra3 will be much more efficient (being an all-out ARM solution). Android-centered OEMs will most likely go with ARM anyway and if there is a big-name producer like Nvidia with a powerful solution for their premium products, they will certainly pick that up instead of the Intel gear.
And this is only the mobile space. When Project Denver from Nvidia and Nufront start selling ARM based server SOCs, Intel will have to fight a battle in the datacenter which was absolutely home-turf so far.
All of this may not have happened at all (or would have happened years later, giving Moorestown a chance) if Intel had not chosen to deny Nvidia the hardware interfaces for building Ion2. They switched a huge threat and possible cut-throat competition in every computing segment for a very short-term gain in one segment.
Was it worth it Intel?
It is not an overstatement that the Motorola Atrix smartphone was one of the bright stars of CES 2011. An often-mentioned, breakthrough feature of the Atrix is its modularity, namely that it can be placed into a netbook dock which gives it work-time (and battery recharge) and a desktop-like work environment (Linux based).
It is worth mentioning that this concept is not brand new and that a smaller company called Always Innovating (AI) has a similar, even more modular product: the Smart Book.
The main difference between the two products is that the computing core of the Smart Book is only a MID, not a real mobile phone like the Atrix.
The advantages of the Atrix over the Smart Book (SB):
- The computing core of the Atrix is a real, usable mobile phone, not only a MID (IP phone as AI calls it) as with the SB. The Atrix phone is a high-end Android phone with beautiful, high-res screen (comparable to the iPhone4).
- The computing core of the Atrix has 1GB of RAM and a powerful Tegra2 (dual-core Cortex A9) instead of the last gen, slow Cortex-A8 SOC and only 512Mb RAM in the Smart Book.
- Computing core of the Atrix has a built-in 3G modem (with strong HSUPA and HSDPA) while the SB has only wifi radio and requires you to use an external 3G modem to connect to the internet when on-the-go.
The advantages of the Smart Book (SB) over the Atrix:
- Much more modular. The SB has tablet jacket AND keyboard/netbook dock for the tablet jacket, while the Atrix only has a netbook jacket for the phone. The SB’s tablet jacket has a capacitive touch interface
- The SB has real a real desktop operating system (Ubuntu) running when in desktop mode while the the Atrix has only Webtop (that only looks like a full blown desktop but it is only a Splashtop-like quick-linux OS, so it is limited to a selection of programs and is not easy to extend with apps).
- The SB has 2 inner USB ports for replaceable 3G modem or storage key which can always ship safely within the netbook dock (no protrusions)
- The SB’s netbook dock can be used as an independent bluetooth keyboard
- The SB’s tablet screen can be used as a secondary display of a desktop (DisplayLink)
- The SB has a dockable (into the talet) HDMI to USB adapter (DisplayLink)
In order to be the perfect companion, the Atrix needs to:
- Increase its modularity by separating the netbook dock into a tablet and a keyboard stand or at least release a tablet dock as well
- Upgrade the Webtop desktop environment to a real, powerful desktop Linux (aka Ubuntu 10.04) or at least ensure that Ubuntu can also be used in place of Webtop. It is important that the user be able to switch between Android and Ubuntu real time
The Smart Book could be a worthy contender to the Atrix by:
- Upgrading the computing core to a dual-core OMAP4 with 1GB of speedy RAM
- The computing core needs to be a real-word Android mobile phone with a strong HSPA data modem
I believe Motorola is in a better position to make the Atrix a one-stop computing solution but I also root for Always Innovating to make the Smart Book a successful product.
Both products clearly mark the future: modular, mobile computing for everyone.