e Innovator's Dilemma
Mar. 17, 2021 11:34 AM ETIntel Corporation (INTC)AMDSSNLFTSM73 Comments12 Likes
EnerTuition profile picture.
EnerTuition
Marketplace
Follow
Summary
Intel's latest desktop CPU, Rocket Lake, has been widely panned by reviewers. The article discusses the reasons.
The article also discusses the mindset at Intel that brought this suboptimal CPU into existence.
Intel needs to snap out of its current IDM mindset or the Company's greatness will slip away over time.
In his famous 1997 book "The innovator's dilemma", author Clayton Christensen showcased how successful, outstanding companies can lose their market leadership - or even fail - as new, unexpected competitors rise and take over the market. Intel's (INTC) Rocket Lake CPU launch on Tuesday ties neatly into the concept and shows how successful companies can go wrong as competitors redefine the market.
Rocket Lake, for readers who may not be aware, is the latest Intel 14nm desktop CPU that launched to unflattering reviews from the enthusiast community. Harsher critics have piled on with scathing narrative about the new Intel entry. The core of the criticism is that Rocket Lake falls short of top-of-the-line chips from Advanced Micro Devices (AMD) even as it consumes considerably more power than AMD alternatives.
The reason for the lower performance and higher power has to do with something the technical world calls "backporting". Backporting, in this case, means that the next generation desktop CPU was originally supposed to be manufactured using Intel's 10nm process but, due to Intel's process challenges, the chip was redesigned in, or ported back to, Intel's older 14nm process. As a matter of chip design, a newer CPU made using 10nm process depends on the goodness of that process - goodness that is not available in the older 14nm process.
Manufacturing a design intended for 10nm process on 14nm process means the design will result in a much bigger die, take more power to run, and may not reach other design goals. That is effectively what we see with Rocket Lake. To make the 10nm design work on 14nm, Intel had to cut back the design to 8-cores and reduce L3 cache to 16MB - compared to 10 cores and 20MB in the previous generation Comet lake. Rocket Lake, as most reviewers note, also consumes considerable amount of power and runs hot. In addition to these functional issues, Rocket Lake, because of backporting, sports a large die size (image below from Tom's Hardware).
As can be seen in the image above, at 276.4 mm2, Rocket Lake chip is larger than any of its recent predecessors. Larger die means Intel can produce fewer of these chips for a single wafer thus reducing the effective unit throughput.
Due to such major disadvantages, backporting, or redesigning a chip into a more primitive process technology, is rarely done in the industry.
So, why is Intel taking such regressive path?
Innovator's Dilemma: Intel Is Stuck On Its Dated IDM Model
When Intel was struggling with its process roadmap, especially the 10nm process, Intel had a choice to make. As much of the semiconductor industry, including rival AMD, is wont to do, Intel could have outsourced its new CPU manufacturing to a foundry like TSMC (TSM) or Samsung (OTC:SSNLF). Or, Intel could stay with its Integrated Device manufacturer, or IDM, model and delay its new products until the 10nm process worked as expected.
Intel chose the latter. This is a classic case of Innovator's dilemma. As discussed previously in Intel And Taiwan Semiconductor: A Tale Of Two Cities and Intel And Taiwan Semiconductor: A Tale Of Two Cities Revisited, the semiconductor industry has fundamentally changed and, Intel's IDM model is now a story of yesterday. Regardless of how hard Intel tries, it is extremely unlikely that Intel can regain its process technology leadership that made the Company an industry giant during the last several decades. The future belongs to the foundry model and TSMC. By sticking to the IDM model and stubbornly trying to regain leadership, Intel is throwing the proverbial good money after bad.
Not surprisingly, the outcome, as predicted in the linked articles, has been dismal. Intel 10nm process suffered from delay after delay and has yet to become a mature node. As Intel stays stuck on yesterday's paradigm, the Company has now incurred long delays in product introductions. Not just on the desktop CPU front with Rocket Lake but also on the server front with Ice Lake and Sapphire Rapids. The delays meant that rival AMD has been gaining product advantage and thus market share in the x86 CPU market.
Over time, Intel had some limited success with the 10nm process and was able to manufacture its laptop CPUs on the new process - including the Tiger Lake CPU released during 2020. However, it appears that Intel is unable to satisfactorily manufacture the high power dissipation desktop chips with the new process. As AMD market share gains became significant and as 10nm delays became untenable, Intel could no longer wait for the 10nm process to work. Given Intel was not outsourcing the CPU fabrication to TSMC or Samsung, it only had one option: to backport its new desktop design to 14nm process. Which is precisely the path that Intel has taken.
Implications
Unfortunately, backporting leads to a sub-optimal CPU like Rocket Lake. While Rocket Lake has been near universally panned, it is something that Intel had to do given the decision to not outsource. Without Rocket Lake, Intel's desktop CPU product line would be weaker, and AMD would go nearly uncontested in some market segments. Competitively, Rocket Lake is a necessity.
The large die of Rocket lake means Intel will yield fewer chips and these CPUs will cost much more to produce than previous generation Comet Lake CPUs. However, this is mitigated somewhat due to the mature nature of 14nm process. This process, which is now over 5 years old, is nearly fully depreciated and the production costs, even for large die CPUs are not exorbitant. Intel can further reduce the capacity and cost impact by limiting Rocket Lake only to the high-end of the market where ASPs are high and the volume needs are low. In other words, despite its limitations, Rocket Lake partly plugs a significant product hole.
However, the problem for Intel here is not Rocket Lake but the mindset at Intel that produced Rocket Lake. Unless Intel escapes the paradigm that brought it its current success, the Company's future is in jeopardy.
We do not view Intel stock as being worthy of a long term investment until the mindset at Intel changes.
