Howdy, Stranger!
It looks like you're new here. If you want to get involved, click one of these buttons!
Quick Links
AMD announces Temash with better than expected specs
The specs on the "good" Kabini chips were known a while back when AMD launched the embedded version of it. They've now announced the names of the Kabini chips that unsurprisingly have the same specs:
A6-5200: 4 cores at 2 GHz, 128 shaders at 600 MHz, memory at 1600 MHz, 25 W TDP
A4-6000: 4 cores at 1.5 GHz, 128 shaders at 500 MHz, memory at 1600 MHz, 15 W TDP
To that, we can now add the two "good" Temash parts:
A6-1450: 4 cores at 1 GHz, 128 shaders at 300 MHz, memory at 1066 MHz, 8 W TDP
A4-1200: 2 cores at 1 GHz, 128 shaders at 225 MHz, memory at 1066 MHz, 3.9 W TDP
The CPU specs here aren't a surprise. But I am surprised to see a sub-4 W SoC with 128 shaders; I was expecting 64. The 300 MHz clock speed on the A6-1450 is also higher than I was expecting, though hardly shocking.
8 W is on the high side for a tablet, but manageable without needing outlandish cooling measures. AMD has detachables in mind with the A6-1450, as if you plug it in for extra cooling, you can clock the CPU up to 1.4 GHz and the GPU up to 400 MHz, albeit presumably at the expense of more power consumption.
AnandTech has some benchmarks up, and the new Jaguar cores at 1.5 GHz handily beat older Bobcat cores at 1.6 GHz. Thus, this is a substantial improvement in IPC.
I've previously said that if you want a Windows 8 tablet, then you want AMD Temash. Today's specs and benchmarks largely serve to confirm this. We'll see if Intel Silvermont is able to be competitive here, but I'm guessing that it won't be. Comparisons to previous generation Saltwell Atom cores are downright ugly, with Jaguar cores at 1.5 GHz roughly doubling the single-threaded performance of Atom at 1.8 GHz, and the Kabini GPU at 500 MHz offering about 10 (ten, and no, that's not a typo) times the GPU performance of an Atom Z2760.
Comments
AMD has been selling Kabini and Temash chips for a while, and it sounds like AMD has now given other companies the go-ahead to start selling laptops or tablets that contain those chips. Of course, just because AMD says it's okay to sell them doesn't mean that other companies actually have laptops or tablets ready for sale. Temash in particular is rumored to be used in the Microsoft Surface Pro 2. I'd also somewhat expect to see tablets based on it from HP, Asus, Acer, Vizio, and possibly others.
But the question is when. There will probably be several Temash tablets shown off at Computex in early June. But "vendor is ready to show off a demo" can be months away from "vendor is ready to sell large numbers of a tablet to the general public". If I had to guess, I'd say the first Temash tablet arrives around the start of summer, but wouldn't be surprised if it goes late into summer.
Kabini/Temash isn't just for tablets and detachables, of course. There are several other potential markets for Kabini.
The obvious one is cheap, low power laptops. It's basically the successor to Brazos, so the sort of cheap laptops that Brazos went into make a ton of sense for Kabini. Much higher performance, significantly lower power, and cheaper to build: what's not to like? Unless, of course, you had zero interest in Brazos because you wanted higher performance.
The 15 W version of Kabini could also be a nifty cheap for premium ultrathins. Tech Report found that it was competitive with a 17 W ULV Core i3 Ivy Bridge laptop:
http://techreport.com/review/24856/amd-a4-5000-kabini-apu-reviewed
It's a pretty safe bet that Kabini used less power in most of those tests, while beating the Core i3 about as often as it lost. Will laptop vendors be willing to use Kabini in a premium Ultrabook-like device? Say, a 1080p monitor, an SSD instead of a hard drive, good build quality, and so forth, all while costing hundreds less than an Ultrabook and being free of the dumb restrictions driven by Intel marketing and not reasons of engineering? Considering that laptop vendors weren't interested in using Intel hardware in Ultrabooks either until Intel paid them to--and that dismal Ultrabook sales figures have justified their reluctance--they may not be willing to make a premium AMD ultrathin. But if one does, it could make for an interesting product.
The 25 W version of Kabini should also have a place in extremely cheap, small form factor desktops. Think Mini ITX at the largest here, and going down in size from there. I don't know how much of a market there is for this, but if you want a $300 desktop with graphics that actually work, Kabini will probably be your best bet by the end of this year.
Kabini and Temash are going to rely on being low power as a major reason for their existence outside of nettops. You might think that a 15 W Kabini chip isn't going to be much lower power than a 17 W Intel ULV. It's only a 2 W difference, isn't it? But not so fast:
http://www.tomshardware.com/reviews/kabini-a4-5000-review,3518-13.html
A 2 W difference in TDP translates to a 14 W difference under a gaming load. Maybe the ULV Core i3 laptop has a bunch of other things using power? Well then, why is it only a 2 W difference under light web browsing or video playback loads?
What's going on here is pretty simple: a 17 W Intel chip will use turbo to use up that full 17 W. While a Core i3 doesn't let the CPU turbo, it does let the GPU turbo. Kabini doesn't have any turbo at all, so using its full 15 W TDP would mean that you have to simultaneously put everything under heavy load. That's hard to do.
That alone doesn't account for the full 14 W difference, as Kabini is surely using a lot more than 3 W. However, the Intel chip also has a separate chipset that can burn several watts. Kabini is an SoC, which means that the chipset is part of the same chip as the CPU, GPU, and so forth, and whatever power the chipset uses is included in the 15 W. Thus, a 15 W Kabini chip might well be lower on power consumption (at the platform level) than a "10 W" Ivy Bridge or Haswell chip. Think that will affect the form factors that Kabini can fit into?
That said, that the Intel chip has turbo and AMD doesn't is surely an advantage for Intel, not an advantage for AMD. Kabini would surely fare far better if that 15 W chip could turbo selected portions up to the speeds of the 25 W version while other portions were idle. Kabini can't do that, though.
If you're trying to do intense stuff on the battery, then you've probably got a laptop plugged in, so Kabini's lower power consumption isn't necessarily a major advantage. For a tablet, however, it can be.
The problem is that Intel doesn't have a clean competitor to Kabini. The Core i3-3217U that they used is the lowest power Intel CPU outside of Atom. They chose the Pentium B960 because it was the cheapest Intel-based laptop they could find. Temash can go in a tablet just fine, but Ivy Bridge really shouldn't--precisely because its power consumption is too high.
Clocking the CPU higher is a matter of cooling, not battery life. The quad core version of Temash will clock higher when you plug it into a dock that blows a fan across the back of the tablet. It would be good if they could dynamically adjust the CPU clock speed based on the load, though. It probably clocks down for idle or power gates off, but doesn't clock up from the stock speed.
Very cool!
Thanks