Archive for April, 2010
Intel isn’t the only chipmaker to have a record first quarter—AMD’s Q1 revenue of $1.57 billion set a first-quarter record for the company. Like Intel, AMD saw a huge boost in year-over-year processor shipments, and an average sales price jump as well. But the CPU unit’s 23 percent year-over-year jump was nothing compared to the GPU division’s 88 percent year-over-year increase. Average sales prices jumped as well, as did shipments of mobile GPUs. Clearly, it’s good to have the top GPU on the block.
In addition to the general PC market rebound and the strength of its smaller, more focused product mix, AMD also benefited from the fact that this was the company’s first quarter to fully reflect the deconsolidation of the Globalfoundries spinoff, and it’s the first full quarter after the chipmaker’s settlement with Intel. Free from its fab burden, its legal battles, and the effects of some of the practices that Intel agreed not to engage in, the company looks to have turned itself around, at least initially.
It remains to be seen how far AMD’s Fusion strategy will take it—this is partly because certain elements of the strategy are unclear, and partly because some of the elements that are clear are new and risky enough that they may not work. The company’s upcoming high-end Bulldozer core is a departure from anything else that’s out there, and its Llano mobile part is a bit of a gamble that this level of GPU/CPU integration at this early of a stage will really pay off. Still, if the IT rebound holds together, things look better now for AMD than they have in quite a few years.
One of the ways to boost single-threaded performance on multicore machines is to shut down the cores that aren’t in use and divert power to whichever cores are running the single-threaded workload. That’s essentially what Intel’s Turbo Boost does—it can dynamically “overclock” one or more individual cores, based on the needs of the system and the amount of power and thermal headroom available.
This “brute force” approach is actually the opposite of making use of multicore—it’s about deliberately cutting back on the processor’s core count, in some cases all the way down to one active core, in exchange for a single-threaded boost. In short, it amounts to a temporary, strategic retreat from the multicore era.
Now AMD has revealed new details of how it plans to endow its multicore processors with the same dynamic power optimization capabilities, but (at least initially) via a much cruder implementation.
The new feature in AMD’s Phenom line of desktop processors is called Turbo CORE, and it will make its debut with the upcoming six-core Phenom II X6 (“Thuban”) line. Today, AMD gave the first details of Turbo CORE, and they are pretty straightforward.
When the processor’s power consumption drops below a certain level because three or more cores went idle, the chip downclocks the idle cores by a fixed amount and cranks up the power and clockspeed on the remaining cores by a fixed amount. So the chip is either in a turbo state or it’s running normally—there are only two options. And the idle cores don’t actually go into a sleep state—they’re just downclocked by a few hundred MHz.
Turbo CORE’s functionality is so basic that it appears that AMD really hasn’t put much effort into doing the necessary plumbing for more fine-grained power optimization—multiple power planes and extensive gating are needed to actually make this kind of strategy work to the degree that it does on the new Nehalems. Even so, Turbo CORE is better than nothing in this regard.
It’s a bit surprising that this new feature wasn’t included in the recent Magny-Cours processor. Servers need power optimization even more than desktops, although in the server space it’s more critical to underclock than it is to overclock (underutilization is a huge problem in datacenters). AMD has to be working on this for the next iteration of its Opteron line, though; I can’t imagine it going beyond 12 cores and not including such technology.
AMD hasn’t announced when it will actually launch the Phenom II X6; today’s reveal was just about Turbo CORE. Anandtech has an unofficial look at the likely launch lineup, if you’re dying to know right away. When the new chips do launch, though, they’ll be drop-in compatible with Socket AM2+ and AM3 boards—only a BIOS update is needed.
Apple’s long anticipated iPad has finally arrived and while it may be too early to receive comprehensive comments from public, some users has already started to complain about inability to charge iPad through commonly available USB ports on certain laptops. When this occurs, a message with “Not Charging” will be displayed in the status bar next to battery icon.
This root cause is pointed to under-powered USB ports that is insufficient to provide enough juice to charge the relatively power hungry iPad as compared to other mobile devices such as iPhone. And based on actual feedback from certain online users, apparently many Windows based laptops as well as older Macs machines like MacBook Pro are facing such phenomenon while it seems that latest MacBooks and iMacs users have no complaint on this so far. If you notice, Apple’s official support site does provide some clues and based on support article from its website, typically all older type of Macs computers can only able to provide 5V with 500mA on each USB port as similar to other Windows based PC. Whereas some newer Intel-based Macs like MacBook is capable to ramp up to 12V at 1200mA whenever needed. For instance: When a more power hungry device such as the 9.7-inch iPad is connected and that may explain why such phenomenon doesn’t happen on those newer MacBook.
And as a quick workaround, consumers are recommended to utilize the 10-watt USB power adapter that will able to fully charge the iPad within hours. And even when they connect to those low-powered USB port, iPad may still able to get some extra power (even through at a slower rate) to work without any issue.