We took our first look at AMD's Fusion series earlier this year when we tested the company's dual-core Zacate 18W processor, formerly known as the E-350 APU. Fusion is the marketing name for a number of APUs (Accelerated Processing Units) that AMD has developed since 2006 when it acquired ATI.
As you should already know, Fusion is all about combining general processor execution with 3D geometry processing and other functions of modern GPUs in a single chip.
The E-350 examined in February was part of AMD's "Brazos" platform and combined two Bobcat cores with a Radeon HD 6250 graphics unit to achieve truly impressive performance. Not only was the E-350 more powerful than Intel's Atom when it came to raw computing power, the GPU was also much faster than anything we've seen with the Atom before, including Nvidia's Ion.
AMD follows in Brazos' footsteps and is now ready to unleash its bigger brother. The new addition, code-named "Llano", is in the "Lynx" platform from AMD and contains four Husky cores (very similar to those in Athlon II processors – more on this later) as well as a robust GPU based on the redwood architecture the Evergreen family.
The Evergreen GPUs supplied the latest generation of AMD's Radeon HD 5000 graphics cards with a refresh. Llano's graphics core has 160 to 400 stream processors, which should make them significantly more powerful than any other integrated graphics solution we've seen.
Since the Llano APUs are part of a completely new mainstream desktop platform (Lynx), they need a new socket. The new format, called "Socket FM1", supports A-series processors and uses PGA packaging (pin grid array) with 905 contacts.
AMD plans to ship many different A-Series APUs, which are divided into three key families. High-end models will be part of the A8 series and will be quad-core processors using the Radeon HD 6550D graphics core at 600 MHz. The A8-3850 runs at 2.9 GHz and the A8-3800 does its job at 2.4 GHz. However, this speed is increased by AMD's Turbo Core technology and can therefore be operated at up to 2.7 GHz. The A8-3850 has a TDP power of 100 watts, while the A8-3800 only has 65 watts.
The APUs A6-3650 and A6-3600 – also quad-core processors – are in the middle. These were downgraded to a Radeon HD 6530D, which is about 40% slower on paper. The A6-3650 operates at 2.6 GHz without the support of Turbo Core technology and still receives a TDP power of 100 watts, while the A6-3600 operates at 2.1 GHz with a Turbo Core frequency of 2.4 GHz and gets a more modest TDP of 65 watts. All quad-core chips have a 4 MB L2 cache.
The A6 series will also include a triple-core A6-3500 operating at 2.1 GHz with dynamic boosts up to 2.4 GHz, 3 MB cache and 65 W TDP. Below, AMD's affordable dual-core APUs A4-3400 and A4-3300 operate at 2.5 GHz and 2.7 GHz (neither of them support Turbo Core). Both have 1 MB cache and a Radeon HD 6410D, with the first clocked at 600 MHz and the second at 443 MHz.
Now that you're up to date with Llano's desktop product range, let's analyze the new architecture and welcome AMD's new flagship APU with our usual test battery.