This week was CES, where companies make a big splash and try to show what’s going to happen in the consumer electronics space in the coming year.
In the coverage of CES, I saw three things that seem interesting, but only one of those three was a surprise.
Windows on ARM, and Nividia’s CPU ambitions
The big surprise wasn’t Windows on ARM, per se, as there’ve been rumblings of that for a few weeks now. The stunner was that Nvidia came out and said that the CPU it’s been not-so-secretly working on would be ARM based.
Nvidia’s interest in CPUs is no bigger secret than Iran’s interest in nuclear weapons, given that all of its remaining competitors also make CPUs. But everyone expected them to build some kind of mainstream, x86 CPU completely compatible with its competitors. And this announcement doesn’t exactly eliminate that possibility. They could still have a skunkworks x86 project, or plans to buy struggling VIA to gain its x86 products, design teams, and critical CPU patents.
Legally, this is less risky. But will anyone buy the thing?
Granted, there have been people–specifically, the initiatives who build machines for the developing world–who have been asking for an ARM port of full-blown Windows for years. And now Microsoft is obliging, and Nvidia is going to offer a compelling hardware platform for it to run on.
But Windows availability doesn’t guarantee success. Windows NT 3.1, 3.51, and 4.0 ran on MIPS, Alpha, and PowerPC. But with spotty hardware support and a severe lack of native applications, it wasn’t much different from running Linux at the time. Windows 8 running on ARM is going to run into the same problems. Not everything that runs on Windows 8 x86 is going to run on ARM, and there are going to be people who are going to be unhappy about not being able to run mainstream software on it, whether it’s $5 card games from Wal-Mart or Adobe Creative Suite.
The same people who use Adobe Creative Suite as an excuse not to run Linux–even though they don’t run it on Windows either–are going to use it as an excuse not to buy ARM.
Now, if certain critical applications become available and ARM-based systems are cheap enough, they could catch on as secondary PCs for jumping on to check Facebook and stuff like that. But AMD may have a monkey wrench to throw into that, which leads us to…
Fusion is AMD’s answer to Intel’s Atom. There’s not a lot of concrete information out there yet about what performance Fusion will really give or what it will cost. But it appears that AMD is trying to at least match the performance of Intel’s Atom when paired with Nvidia’s Ion chipset. And it’s scalable. At the top end, it’s a dual-core, 1.6 GHz chipset that uses 18 watts of power, and it suitable for use in netbooks, notebooks, and low-end desktop computers. But there are lower-power versions that target all of the niches Intel sees Atom potentially playing in too. Whether they can compete with ARM and MIPS in the embedded space remains to be seen of course.
Depending on how AMD decides to price Fusion, it’s going to be trouble for Intel and Nvidia. At the very least, Intel is now going to have to share the netbook and low-end laptop market. And if a device running Windows on Fusion only costs $20 or $30 more than a device running Windows for ARM on Nvidia hardware, the overwhelming majority of people are going to gladly pay the 20 or 30 bucks, because it’s cheap insurance.
Personally, I’m interested in Fusion because two companies displayed mini-ITX boards based on Fusion that offer PCIe, a maximum of 8 GB of RAM, and SATA 3 and USB 3. None of those things are available on current Atom-based mini-ITX boards. That’s a system with the potential to be useful for many years, and it’ll fit in a shoebox.
And of course, there were new SSDs. Crucial showed its next-generation RealSSD product, which offers a modest 17% speed increase, due mostly to a process shrink on the memory chips.
And OCZ showed a shocker: a next-generation Sandforce-based SSD that gives sustained sequential I/O in excess of 500 megabytes per second. SATA 3 isn’t even commonplace yet, and here’s a drive that can, under ideal circumstances, come close to saturating the bus.
Random I/O makes a bigger difference in overall system performance, and this drive is a stunner in that regard too, at around 180 megabytes per second. That’s a significant drop, but all drives, platter and SSD alike, suffer considerably when doing random I/O. The random performance of this drive is comparable to the sustained sequential performance of the first-generation OCZ Vertex.
This drive won’t hit the market until the second quarter at least. But OCZ won’t be the only ones offering drives based on this controller, and makers of competing controllers aren’t holding still either, so 2011 promises to be an interesting year for SSDs as their performance advantage over conventional drives widens even further.