Intel stock tumbles as Dow status questioned

>An ice cream store in the desert should make money.

Because the Intel ice cream store is in a hot city and only sells the same vanilla ice cream they've been selling for decades. While it's on a busy street with Nvidia, arm and AMD selling a selection of chocolate chip cookie, creme caramel and rocky road ice cream.

Until now, the makers of Arm-based CPUs are the ones who have been unable to find an answer for x86, outside special markets where the competitivity of a device is determined primarily by the set of peripheral controllers included in the SoC and not by the included CPU, like smartphones, microcontrollers and various dedicated communication processors.

The Intel CPUs with Gracemont cores, i.e. Alder Lake N and Amston Lake, are the best for very cheap and small computers.

They have only an extremely feeble competition from Arm-based CPUs. Until very recently, the Arm CPUs with Cortex-A78 cores, which are approximately equivalent with the Gracemont cores, have been many times more expensive than the similar Intel CPUs. The older Arm cores, like the Cortex-A76 used by the latest Raspbery Pi are much slower, while the complete SoCs are not cheaper than the Intel CPUs.

Only during 2024 one cheap SBC with Mediatec Cortex-A78 cores has appeared and for early 2025 one cheap SBC with Qualcomm Cortex-A78 cores has been announced. These anemic Arm offerings are insignificant besides the deluge of small and cheap computers that use Intel Alder Lake N or Amston Lake CPUs.

Until now the manufacturers of Arm-based CPUs have never been able to compete with Intel in price and performance for general-purpose computers, unlike for microcontrollers and for smartphones. The makers of smartphone chipsets, i.e. Apple, Samsung, Qualcomm and Mediatec, together with NVIDIA, are the only companies that use up-to-date Arm CPU cores that can compete in performance with the x86 cores and their products do not make any attempt to compete in the market for cheap computing devices.

Lately Mediatec and Qualcomm have started to attempt to enter the market for embedded computers, but what they try to sell on this market is only their older chipsets, which have become obsolete for smartphones and which would have to be discontinued, unless a new market is found for them.

For several years, Qualcomm has attempted to introduce Arm-based laptops, but all of them have been simultaneously much more expensive and much slower than the x86 laptops. The new laptops with Qualcomm Oryon CPUs that have been launched this year are the first that are decent, but overall they are still inferior to the Intel or AMD alternatives.

Apple designs good Arm CPUs, but they could have as easily designed competitive x86 CPUs or POWER CPUs, had their licensing conditions been as convenient as those of Arm.

A modern ISA like Aarch64 simplifies a lot an instruction decoder that must decode many instructions in parallel, but even in an x86 CPU the area occupied by the instruction decoder together with the additional area occupied by a micro-operation cache in comparison with an instruction cache of the same size (measured in cached instructions) form a very small fraction of the total CPU core area.

Therefore, even if there exists an x86 penalty vs. Arm in core area and in power consumption, the penalty is only of a few percents, perhaps 10% at most, so when comparing an Arm core and an x86 core there are other microarchitectural details that are more important than what ISA must be decoded, because they can make one core several times better than the other, instead of providing only a few percents of advantage.

> Intel doesn’t have an answer for ARM

Is the Lunar Lake announcement not them starting to answer? https://arstechnica.com/gadgets/2024/09/intel-announces-firs...

Good non-Apple ARM laptops have not been out for very long, so it's not like they're that far behind schedulewise.

Intel is much more open about their future than most companies, by publishing a roadmap with years in advance.

Since Pat Gelsinger has become CEO, Intel has reached successfully all the milestones on their public roadmap, even if there have been a few predictable problems that have increased their losses, like low fabrication yields on their new Intel 4 CMOS process.

Unfortunately, even if their roadmap is ambitious, it is still not ambitious enough to ensure a quick return to profitability and some of the decisions on which it is based are very questionable.

The main problem is that it appears that they have a very big inertia in their design and validation processes, which prevents easy corrections of the projects that are already known to have no longer appropriate goals.

While after AMD finishes the design of a new CPU core, like Zen 5, during the next few months they introduce products using the new cores that update all their product lines, from low-power mobile to high-end servers, after Intel launches the first product using new CPU cores, like they are doing now with Lunar Lake, which includes Skymont and Lion Cove cores that are competitive with Zen 5, Intel may continue for a long time to introduce other new products that still use obsolete CPU or GPU cores, like with Granite Rapids for servers later this year or Arrow Lake H for bigger laptops at the beginning of 2025, and they may succeed to update all their product lines only after one year or even more, e.g. they may launch up-to-date server CPUs (Clearwater Forest and Diamond Rapids) only during the second half of 2025, at a time when they will introduce improved laptop CPU cores in Panther Lake, making again obsolete their 2025/2026 server CPUs, which might end competing against Zen 6, instead of against Zen 5, with which they would have been well matched.