Microprocessor specialist Intel (NASDAQ:INTC) announced a product that it calls Skull Canyon a while back. It is a small form factor PC whose primary claim to fame is that it comes with a Skylake processor that packs the company's highest-end integrated graphics processor yet: Iris Pro 580.
Despite the fact that, on paper, Iris Pro 580 should be a substantial leap in performance from the three-year-old Iris Pro 5200, AnandTech's performance tests paint a less exciting picture.
How much better is Iris Pro 580 compared to Iris Pro 5200?
In the popular game BioShock Infinite, at 1680x1050 resolution, the Iris Pro 580 is just 26.3% faster than the Iris Pro 5200. In Tomb Raider, the Iris Pro 580 is able to pull ahead of the 5200 by about 55% at 1680x1050. In Sleeping Dogs, the delta is about 44.5% at the same resolution.
Obviously these figures represent a fairly large performance improvement, but when we consider all of the enhancements that Intel brought to the table, these results are quite underwhelming.
For one thing, Intel moved from a 40 execution unit design with Iris Pro 5200 to a 72 execution unit design with the Iris Pro 580 -- an 80% increase in execution units. Further, the Iris Pro 580 is based on a graphics architecture that's a full two generations ahead of the one powering the Iris Pro 5200 (Gen. 7.5 graphics versus Gen. 9).
And, on top of all of that, the Iris Pro 580 is built using Intel's 14-nanometer technology, which should provide fundamental performance and power efficiency improvements over the 22-nanometer technology that the Iris Pro 5200 is manufactured using.
Given the multiple generations of architectural advancement, the dramatic increase in hardware resources, and the move to the new 14-nanometer process, I expected a bigger leap in performance.
Intel needs to improve its execution in graphics
Intel has made it clear that it views graphics as an important investment area. Graphics performance is vital for things like 3D games, and it wouldn't be surprising to see more non-gaming applications take advantage of the computing capabilities of the graphics processor in time.
Whatever it takes, Intel needs to make continued, significant improvements in the performance, power, and area of the graphics architectures that it includes on its processors. I don't think anybody is asking for high-end discrete levels of graphics performance, but Intel clearly isn't producing the levels of performance from its integrated GPUs that its manufacturing technology should entitle it to.
Sadly, it doesn't look as though we will see a significant leap in graphics performance in the company's upcoming Kaby Lake family of processors. As I understand it, the main graphics improvements in this coming processor architecture will actually be in the media engine. This should improve video playback and encoding, but for 3D games as well as GPU computing applications, I don't think Kaby Lake will make major strides over the current generation Skylake-based processors.
The next big jump in graphics performance for Intel should come with the Cannonlake processor family that's scheduled to debut in late 2017. It will be interesting to see if Intel delivers the significant leap that it should be able to.