Intel's Slipping Tick-Tock Schedule

So Sandy Bridge's NDA lifted today and I'm reading the review (even though I should be sleeping so I can wake up for my flight tomorrow), and I saw that there would be no enthusiast parts until Q4, which hasn't happened for a "tock" since Intel adopted the Tick-Tock schedule. So I'm gonna take a look at history and see if I can spot a trend.

The story began in 2006. Intel's Pentium 4 was losing to AMD's Athlon 939 series on the desktop side, but was enjoying success with the Pentium M/Core Duo line on the mobile side. As a reaction to how power hungry the Pentium 4's were, Intel decided to abandon the Pentium 4 line and instead evolve the Core Duo line to serve both desktop and mobile markets. In July 2006, Core 2 Duo(Conroe) was unleashed to the world. It was epic pwn, and marked the start of the Intel's new Tick-Tock schedule.

The schedule was put to test a year later with the 2007 tick; a 45nm shrink. Penryn/Wolfdale was not release until October, a 3-month delay. Negligible, I thought, blame it on the difficulty with the shrink or whatever. Next year's release should be closer to the July time frame.

July 2008 came, and the much anticipated Nehalem tock was nowhere to be seen. October came, and there was still no tock. It wasn't until November did Nehalem come out. At this point it became obvious that Intel wasn't keeping a strict one-year release cycle. And why would they? There was no market pressure. Since being beat by Conroe in 2006, AMD had not been able to keep up with Intel. Athlon's successor, Phenom, was delayed way too much due to the technical difficulty of it being the first monolithic quad core. When Phenom did come out, midway through the Penryn lifecycle, it merely equaled the performance of Penryn's quad-core variant, Yorkfield. So at this time Intel had already dominated much of the market for 1.5 years, except for one segment, the HPC/Server market. In the high performance market, AMD's processors, with its high memory peformance thanks to the integrated memory controller, has kept itself relevent, trading blows with Intel's processors depending on the type of workload. This was actually the focus of the Nehalem release. Intel added an integrated memory controller to Nehalem to help it capture the HPC/server market, and the first models of Nehalem's released were the enthusiasts/HPC/1P/2P models.

Wait, what? A different die for the high performance segment? Didn't Intel just say they were keeping one architecture for the entire line-up back in 2006? Well, architecturally the cores are the same, but the rest of the dies varied between the server and consumer versions. The consumer version of Nehalem, with Intel still not feeling any market pressure in that space, did not see the light of day until 10 monthes later. But all is well as far as the enthusiasts are concerned; we got our processors during the initial release.

When's the tick for 2009? Did not happen until December. With this release, Intel focused on the mobile segment. There were no enthusiast part, no mainstream part, just the dual core mobile part. Intel's dominant market position allowed it to not only not keep the yearly tick-tock, but also initially release models in only the market segment of interest. 32nm shrink lowered power consumption, and the mobile segment could use the lower power consumption. The 6 core server part would be released some time later, but no enthusiast or mainstream part was ever planned. It was okay though, I thought. Ticks aren't exciting anyway; as long as we got our enthusiasts parts during Tocks then we're happy.

That brings us to now, the 2010 Tock. But wait, Intel actually waited till January 3rd 2011 to lift the NDA. So any illusion of an yearly release schedule is gone now. It was obvious from the previous 3 years, but I held out hope that Intel would not miss the Christmas shopping season. The thing I did not see coming, which is less surprising now that I typed out this post, is the lack of enthusiasts models until Q4. This Tock was for the mainstream and mobile market. The mainstream was skipped over last time so a new model was long overdue. And the mobile market can really use the improved integrated graphics performance. The enthusiasts? Well, we were never important to begin with. We got our parts in 2006 and 2007 because Intel had just one die for the entire family. We got our parts in 2008 because Intel was attacking the HPC/server market. We didn't get our parts in 2009, and we're not getting it now in 2010/2011.

In conclusion, we have 3 trends from Intel: 1) Longer than 12-month Release cycles, 2) diversification of feature set in different dies, and 3) targeting certain market segments during each initial release. Honestly though, only the hardcore enthusiasts should be affected this time around. Most enthusiasts (me included) should be fine with the mainstream Sandy Bridge parts, which has plenty of performance. The only thing I don't like is the fact that it contains an integrated graphics cores that we'll never use because despite performance improvements it's still way slower than a dedicated graphics card, so it's kind of a waste of space/transistor/money. Looking forward, the 2012 Ivy Bridge Tick will probably happen around February or March. There will still be different dies containing different feature sets. But which market Intel will target initially isn't as easy to predict, though I suspect it will be the mobile segment again, as it had been the past 2 years.