CHIPMAKER Intel has used the added transistor budget afforded by its 22nm process node to significantly expand the GPU in its Ivy Bridge processor line and make it comparable to AMD's Radeon HD 6550D found in the firm's Llano processors.
Intel told journalists that although Ivy Bridge is a tick in its tick-tock model, parts of the processor represent more than a tick, specifically the GPU. Chipzilla has spent a significant amount of the increased physical die space afforded by its 22nm process node shrink to improve the GPU and it shows.
Intel has named its Ivy Bridge GPU as Intel HD Graphics 4000 and the most visible update is Microsoft DirectX 11 support, a major feature that is missing in its Sandy Bridge integrated graphics. Intel has also added hardware tessellation, improved sampler throughput for better anisotropic filtering quality and triple display support.
Although Intel claims its HD Graphics 4000 core is a major upgrade from the previous generation, it is hard to believe it by looking at the specification sheet. Although Intel has added four execution units to bring the total up to 16, the maximum clock speeds in Intel HD Graphics 3000 and Intel HD Graphics 4000 are identical at 1,350MHz, suggesting that Intel has made significant architectural changes in order to increase performance.
The Intel HD Graphics 4000 GPU supports stereoscopic 3D and an improved Quick Sync transcoder. Intel was particularly keen to highlight the Quick Sync performance increase in applications that support it, with applications such as Arcsoft's Media Convertor being one of the first to make use of Quick Sync.
Intel has relatively quietly introduced OpenCL support in its Intel HD Graphics 4000. Although it is a member of the Kronos consortium that develops the OpenCL standard, Intel has been slow to support GPU acceleration of OpenCL, however with the Ivy Bridge launch, it finally released a software development kit for OpenCL that takes advantage of the Intel HD Graphics 4000 GPU in Ivy Bridge.
Ironically, Intel's support for OpenCL could benefit AMD the most, as it has desperately been trying to promote OpenCL development in order to push its Fusion accelerated processor units (APUs). Intel, perhaps with more to lose, didn't play up Intel HD Graphics 4000 GPGPU support to journalists other than to refer to supporting Microsoft's Direct Compute shared local memory and vector input/output support.
Ultimately Intel HD Graphics 4000 should be compared against AMD's Fusion APUs. AMD has bet the farm on its Fusion architecture that brought DirectX 11 and more importantly OpenCL support to help what is generally regarded as a mediocre CPU mated to an impressive integrated GPU.
While Intel's Sandy Bridge graphics languished behind AMD's Llano APUs, on paper at least, Intel HD Graphics 4000 should bring graphics performance up to par with AMD's top-end Llano chip, the A8-3870K, which features the AMD Radeon HD 6550D GPU running at 600MHz. Intel has clocked the GPU on its range topping Core i7 3770K GPU between 650MHz and 1,150MHz, meaning that it can overcome architectural inefficiencies with raw clock speed.
Intel has said its GPU architectural improvements introduced with Ivy Bridge set the foundation for further developments. By that the firm is referring to its Haswell processor set to debut in 2013, however it also suggests that perhaps there won't be such a large performance delta, as the firm admitted most of the changes came from the physical die space afforded by the move to its 22nm tri-gate process node.