Incompetence Inside.
Intel Corporation just can’t seem to catch a break. Under CEO Pat Gelsinger, the company has come under fire for encouraging motherboard vendors to run unverified BIOS settings that exceed default parameters. This issue has escalated into a global problem, with their Raptor Lake processors, encompassing 13th and 14th generation desktop CPUs are now considered defective.
These processors are running at high voltages even under stock configurations, leading to severe silicon degradation at best, or complete failure at worst.
With AMD’s Zen 5 processors recently launching to widespread disappointment, Intel’s last hope lies in its upcoming Lunar Lake and Arrow Lake processors. Despite early speculation suggesting only marginal performance gains over previous architectures, Intel desperately needs these new chips to combat AMD’s stagnate Ryzen 9000 series.
As Intel’s Core processors continue to either regress in performance or kill themselves, the company must deliver something substantial to restore its credibility and compete in a market where reliability has become a growing concern.
Back in June, Intel made a big deal about its upcoming Lunar Lake series, with Arrow Lake being nearly identical architecturally, though designed for desktop applications without the same efficiency constraints. Unlike Lunar Lake, which uses TSMC silicon, Arrow Lake was planned to be built on Intel’s advanced 20A processing node.
However, Intel’s recently spun-off foundry has been a major headache for the company. With the recent layoff of nearly 18,000 employees and the suspension of dividends, investor confidence has plummeted. Intel is scrambling to regain product leadership and stop the hemorrhaging of funds, as it faces mounting pressure to stabilize its operations.
Intel’s Foundry division, responsible for two-thirds of the company’s products, played a significant role in the company’s recent financial loss, bleeding billions. Intel is struggling to regain relevance as a semiconductor leader, having been overtaken by rivals like TSMC, which now produces chips for NVIDIA, AMD, Apple, and even Intel itself.
The foundry division has been problematic for years. Under former CEO Brian Krzanich, Intel overreached with an overly ambitious density target for its 10nm process. Despite heavy investments in research and development, as well as production at their new Israeli plant, the 10nm process faced substantial delays and issues with maturity, setting Intel back considerably.
This immaturity prevented Intel from mass-producing a significant number of processors. The issue was so severe that Intel had to essentially scrap its Cannon Lake microarchitecture, a 10nm version of Skylake, after barely meeting its commitment to investors by quote-on-quote “releasing” a limited batch of 10nm CPUs to the public.
Cannon Lake was quickly abandoned due to its impracticality for large-scale production.
Intel conveniently didn’t mention the fact that from 2015 until 2019, the only 10nm CPU they managed to release was the elusive Core i3-8121U, a 2-core, 4-thread monster that was so difficult to manufacturer its integrated graphics was fused off.
This underwhelming chip was seemingly shipped exclusive to the Chinese market, and after that, Cannon Lake quietly disappeared, never to be mentioned by Intel again.
In reality, Intel’s repeated failures with its advanced 10nm processing node left the company with no choice but to continue using its 14nm process. This resulted in multiple “refreshes” of the Skylake architecture, which began with Intel’s 6th-generation processors in August 2015 and extended through the 10th-generation “Comet Lake” chips released in August 2019.
However, the 14nm process wasn’t finished yet. For its 11th-generation “Rocket Lake” CPUs, Intel chose to backport the Sunny Cove microarchitecture designed for 10nm onto 14nm, creating “Cypress Cove.”
Rocket Lake turned out to be one of Intel’s most underwhelming CPU generations to date. While it delivered solid multi-threaded improvements over the aging Skylake core used in the previous four generations, it regressed in core count, offering a maximum of 8 cores and 16 threads compared to Comet Lake’s 10 cores and 20 threads.
More concerning for consumers, the 11th-generation Core processors actually saw a decline in gaming performance, a key area of interest for many buyers.
Eventually, Intel managed to regain its footing, finally achieving mass production of 10nm silicon and introducing a true architectural leap with the hybrid 12th-generation “Alder Lake” processors. However, the 13th and 14th generations were largely revised versions of Alder Lake, featuring a larger cache and an extra cluster of E-cores, significantly boosting multi-threaded performance.
In an effort to move past the stagnation on its 14nm process that lasted for nearly half a decade, Intel made the bold decision to rebrand its entire process roadmap. They introduced a new naming scheme, with the refined 10nm process being dubbed “Intel 7,” and their next-generation process, originally referred to as 7nm, now called “Intel 4.” This shift aimed to signal a fresh start and emphasize their progress.
Intel’s new naming scheme becomes increasingly bold and unconventional as it expands beyond Intel 4 to Intel 3, followed by Intel 20A and then 18A, processes that, theoretically, represent “2nm” and “1.8nm,” but the entire roadmap feels chaotic and hard to follow.
When Intel spun off its semiconductor division as Intel Foundry, opening itself up to external customers, it set ambitious targets. One such goal was to become the world’s second-largest foundry by 2030, an acknowledgment that they’ve been decisively outpaced by TSMC.
However, Intel seems content with aiming for second place to stay ahead of Samsung. CEO Pat Gelsinger made a bold declaration that Intel would produce four separate nodes, Intel 7, 4, 3, 20A, and 18A within a five-year timeframe, a lofty challenge that has proved impossible given the company’s recent struggles.
Intel previously scrapped its plans to release Meteor Lake processors for desktop, opting to limit the revised microarchitecture, built on Intel 4, to mobile platforms only. This decision became more understandable when considering the underwhelming performance gains, possibly even a regression compared to previous generation hardware and AMD’s Ryzen processors.
Meanwhile, Intel has yet to deliver on its Intel 3 process, and the company has now sent shockwaves through the industry by announcing it no longer plans to use its own 20A process node for the upcoming Arrow Lake processors in the consumer market. Instead, Intel will rely on external nodes, likely from TSMC or possibly even Samsung, if cost-cutting becomes a priority for all of Arrow Lake’s chip components.
Intel’s role in the production of Arrow Lake processors will now be limited to packaging the chiplets produced by external partners. This node change comes after Intel initially showcased a wafer of Arrow Lake processors manufactured on the 20A node at its Innovation 2023 event, suggesting that development was already advanced enough for manufacturing.
At that time, Intel had projected that Arrow Lake would hit the market in the latter half of 2024.
Intel claims that its crucial next-gen 18A node is still on track, and it has redirected engineering resources from the 20A node to the newer 18A node. However, given Intel’s history of missing its roadmap targets since the introduction of the 14nm process back in 2014, there is skepticism about whether the 18A process will stay on schedule. It seems Intel is skipping its advanced and costly 20A process entirely.
Intel’s 20A process was intended to showcase several significant advancements, including RibbonFet Gate-All-Around (GAA) technology, Intel’s first new transistor design since the introduction of FinFET in 2011. It also introduced PowerVia backside power delivery, which routes power through the backside of the processor die.
However, Intel has invested substantial resources in developing this technology, which now seems to be largely wasted. With Arrow Lake’s performance and expectations in question, the processor a desktop version of Lunar Lake shows minimal improvement. Leaked performance expectations for Arrow on 20A suggest only about a 10% increase over Raptor Lake.
This fiasco ultimately puts AMD in a more favorable position, even though they have little to gain from the situation. Despite the disappointing reception of Zen 5, which offers only marginal improvements in gaming performance and a slight boost in multi-threading with better efficiency, AMD Ryzen hasn’t seen significant advancements. Meanwhile, Intel’s reputation has taken a severe hit.
As much as I hate Intel and their deceptive and underhanded tactics such as the “Core Truths” fiasco or the use of VLC video footage to simulate real-time gaming, and their tactics of bribing OEMs to avoid superior AMD Athlon 64 processors in the early 2000s, this situation doesn’t benefit consumers.
It’s a lose-lose scenario where Intel’s struggles have not translated into meaningful gains for AMD or improvements for consumers.
Whether or not I have personal grievances with Intel is beside the point; it’s crucial for Arrow Lake to be competitive. Competition drives down prices and benefits end users who purchase these products. If Arrow Lake turns out to be underwhelming, particularly if it falls short of the Ryzen 7000 series X3D processors in gaming performance it would harm consumers.
This is especially relevant as we approach AMD’s Zen 6 architecture, which might require a new socket platform. Currently, AMD’s Zen 5 still uses the outdated IOD from Zen 4, leaving potential performance untapped.
Intel faces a difficult decision: should they go all in with TSMC for Arrow Lake on desktop, which might offer efficiency gains but come with higher manufacturing costs and potentially lower performance? Or should they save costs by using Samsung’s SF3 or SF4 processes, and leverage the fact that nobody wants Samsung’s silicon?
Given Intel’s financial strain, a partnership with Samsung would offer short-term benefits, such as better pricing arrangements for wafers compared to TSMC, who currently dominates the industry with an iron fist.
If Arrow Lake fails to meet expectations, AMD would dominate the DIY PC market, potentially leading to higher prices for consumers with the arrival of Zen 6 which is something that nobody wants.
