Intel’s Tick-Tock strategy has been a model the company adopted since 2007, and had has been as reliable as the tides. The Tick represents the shrinking of the manufacturing process (say 32nm to 22nm), while Tock signified the introduction of a new architecture.
The keywords in the previous paragraph was “had been.” Since 2014 Intel’s Tick-Tock’s reliability had started to folder, and started making use of refresh cycles after a Tock. Some folks in the industry had even started referring them to Tick-Tock-Tock. The reasoning behind this decision was the delay of the shrinking of process technology. 10nm and beyond have been notoriously challenging, and this has forced Intel to re-evaluate their strategy.
Intel confirmed the end of their Tick-Tock strategy in a Form 10-K report to the SEC and will now adopt a 3 year cadence. This will see Process, Architecture and Optimization.
This new strategy will fully leverage a single process node, and is almost the formalization of the Tick-Tock-Tock strategy mentioned above. Taking the current 14nm process as an example, this would mean Broadwell (2014), followed by Skylake (2015) and then finally the yet to be released Kaby Lake (2016).
Goodbye Tick-Tock, hello Process, Architecture and Optimization
Process is the arrival of Broadwell on 14nm, architecture was Skylake – which was a new Tock (architecture) on the previously established 14nm process. Finally we see optimize, which in the case of Kaby Lake takes Skylake and refreshes and optimizes its performance.
Intel have already confirmed they are delaying the launch of Cannonlake until the second half of 2017, about a year of the introduction of Kaby Lake. A year after cannonlake, we’ll see Ice Lake (in 2018) and then 10nm Tiger Lake (2019).
Intel abandoning Tick-Tock and moving towards Process, Architecture and Optimization is a symptom of the ending of the silicone era – and in just a few years (about 2020) we will likely have reached about the end of the road as we start on 5nm processes. Intel are exploring technologies such as Spintronics, but just what the future holds is a bit of a mystery.
When Intel first adopted Tick-Tock the company were producing NetBurst architecture processors such as the Pentium 4 on a Tick. It then released its first Tock in the form of the well received Core line of processors on the same 65nm process.
Just what differences we’ll see in Kaby Lake over Skylake remains a bit of a mystery, but it’s likely that we’ll see a modest IPC gain and a few additional features such as the native support of USB3.1 (Skylake uses an add-on chip to do this) and also additional hardware decoding, such as VP9 10-Bit.
Unfortunately, it would appear given the early rumors that we’ll see a similar core / thread configuration as found in Skylake and earlier processors. So the 6700K equivalent will still feature 4 cores and Hyper Threading, meaning it will offer support of up to 8 threads.
How Kaby Lake and Cannonlake will compare to AMD’s Zen range of processors will be interesting over the next couple of years. Especially if the rumors concerning the IPC gains of Zen are accurate, and the greater number of processor cores.
Thanks Fool
