WP_Term Object
    [term_id] => 109
    [name] => Imagination Technologies
    [slug] => imagination-technologies
    [term_group] => 0
    [term_taxonomy_id] => 109
    [taxonomy] => category
    [description] => 
    [parent] => 14433
    [count] => 37
    [filter] => raw
    [cat_ID] => 109
    [category_count] => 37
    [category_description] => 
    [cat_name] => Imagination Technologies
    [category_nicename] => imagination-technologies
    [category_parent] => 14433

Imagination’s New GPU Cores

Imagination’s New GPU Cores
by Paul McLellan on 01-06-2014 at 9:00 am

 This morning Imagination announced their latest GPU cores, including the world’s smallest fully-featured OpenGL ES3.0/OpenCL GPU core. More on that below. And it is the Internet…and graphics…so cats. Graphically rendered cats.

The first core is a a high end core new generation PowerVR Series6XT architecture delivers up to 50% higher performance and advanced power management. The Series6XT architecture features market-leading scalability, supporting implementations up to eight compute clusters that scale linearly in GFLOPS and texturing rates. With OpenGL ES 3.0 support across the range, Series6XT provides among the highest performance OpenGL ES 3.0 GPUs in the industry. Today they unveiled the first three cores in the Series6XT generation with two, four and six compute clusters respectively.

The PowerVR series is already the market leader, at least according to John Peddie research. There are now over 45,000 developers working on graphics applications using this portfolio of IP.

Here is how the whole family fits together for everyone who wants to see the whole (announced) roadmap:

 This achieves up to 50% performance increase on the latest industry benchmarks compared to their previous generation of cores, with both the best performance in terms of GFLOPS/mm[SUP]2[/SUP] and GFLOPS/mW. There are architectural enhancements such as streamlined instruction set for improved application performance, the next generation of the hierarchical scheduling technology (HST) for higher resource utilization, sustained polygon throughput and improved pixel fillrate along with other improvements for better parallel processing.

 The PowerGearing tile-based deferred rendering (TBDR) is the world’s most efficient. This enables fine-grained control of all GPU resources and dynamic demand-based scheduling, routing power only to needed resources to get the best in low-power performance.

The PVR3C triple compression provides for lossy texture compression (down to around 1 bit per pixel), but lossless image compression (about 2X) and lossless geometry compression. This reduces bandwidth requirements and thus leads to lower power.

The lower-end new Rogue GPUs are targeted at entry-level segments ideal for applications with limited area and bandwidth where the higher powered cores are inappropriate. There are four cores in the family:

  • PowerVR Series6XE G6050: takes advantage of the latest Rogue architectural developments and extends the scalability beyond one cluster to a half-cluster while maintaining full software compatibility, creating the smallest fully-featured OpenGL ES3.0 and OpenCL-capable GPU core available
  • PowerVR Series6XE G6060: This core augments the half-cluster design of the G6050 with the addition of second generation lossless image compression (PVRIC2), providing an optimal balance of small size and bandwidth reduction for products such as entry-level mobile products, tablets and full HDTVs and set-top boxes.
  • PowerVR Series6XE G6100: an updated and further optimized version of the original core and features a single Unified Shader Cluster (USC) combined with a high-performance texture mapping unit, enabling it to deliver the same raw fillrates as multi-processor GPUs from the previous generation.
  • PowerVR Series6XE G6110: single-cluster core extends the above G6100 design, adding PVRIC2 for improved throughput, reduced bandwidth, conserved power and improved system performance—key for products such as low-cost 4K UltraHD TVs and tablets where bandwidth is a limited, valuable resource.

Both cores come with physical design optimization kits (DOKs) to optimize power, performance and area (PPA). These include reference flows, tuned libraries from partners, characterization data and more.

More details on Imagination’s website here.

More articles by Paul McLellan…

Share this post via:


There are no comments yet.

You must register or log in to view/post comments.