From Portland, Oregon, the lab bench of a Linux enthusiast looks like an archaeology dig. Arranged in a rough timeline sit a dozen battered laptops, their hinges worn, their keyboards yellowed by years of use. The oldest, a Lenovo ThinkPad T61 from 2008, still boots. That fact alone is remarkable. What it reveals when put head-to-head against a cutting-edge 2026 machine is more remarkable still.
Phoronix, one of the most respected independent Linux benchmarking publications, has spent months running more than 150 individual tests across 15 Intel mobile processors launched between 2008 and early 2026. As reported by Tom's Hardware, the results place a concrete number on something most people experience only as a vague sense that their old computer was slower: the Core Ultra X7 358H, codenamed Panther Lake, was up to 95 times faster than the Core 2 Duo T9300, codenamed Penryn, in OpenSSL cryptography tasks, and 93.9 times faster in OpenVINO AI workloads.
That headline figure deserves some unpacking. The geometric mean across all benchmark results showed that the Core Ultra X7 358H outperformed the Core 2 Duo T9300 by 21.5 times, which is a more representative measure of everyday gains than the peaks achieved in specialised tasks. In tasks such as web browsing or photo manipulation, the improvements were up to 10 times. The 95-times figure reflects workloads that specifically exploit instruction sets and dedicated silicon that simply did not exist in 2008.
A Chip Built for a New Era
The Core 2 Duo T9300, launched in 2008, is the oldest chip in Phoronix's tests and resides inside a Lenovo ThinkPad T61. It features two physical cores without any Hyper-Threading, clocked at up to 2.5GHz, with a 35-watt thermal design power. Its successor at the other end of the timeline is a fundamentally different kind of device. Panther Lake is Intel's first product built on its Intel 18A process node, and the first client system-on-chip built on that node. Intel 18A is the most advanced semiconductor node developed and manufactured in the United States.
New hardware instruction sets, such as AVX-512, and AI-driven workloads favour Panther Lake and the neural processing unit it contains. Those AI workloads are where the performance gap stretches to its widest. A 2008 processor was designed with no conception of on-device machine learning; the Panther Lake chip has dedicated silicon that exists solely to accelerate it. The NPU has been upgraded to the NPU 5 architecture, delivering 50 standalone TOPS for local AI tasks. Comparing the two on AI benchmarks is, in one sense, like comparing a horse-drawn cart to a freight train on the same stretch of road.
The Power Question
For those inclined to see only a triumphal march of progress, the efficiency numbers introduce a more complex picture. Panther Lake was not only 9.7 times faster than Sandy Bridge but also consumed, on average, 7.8% less power, which is impressive given that the Panther Lake chip comes equipped with eight times more cores than the Sandy Bridge part. That is a genuine engineering achievement. But the comparison with the leaner chips of the early 2010s tells a different story.
The Core i7-3517U, codenamed Ivy Bridge, had the lowest average power consumption among the 15 processors tested. Compared to Ivy Bridge, Panther Lake consumed 1.92 times more power but delivered 9.1 times more performance. For anyone who values battery life above all else in a laptop, the Ivy Bridge era still has a certain appeal, at least in raw efficiency terms. The industry has largely chosen performance over miserliness, and the market has rewarded that choice.
The Phoronix study also found that both Penryn and Clarksfield lacked support for processor power sensors, preventing the collection of power efficiency data for those processors. That gap in the record is a reminder that the data we have tells a partial story, not a complete one.
A Competitive Market and a Company Under Pressure
Context matters for Intel in 2026. The company has faced sustained criticism over manufacturing delays and its loss of process leadership to Taiwan's TSMC and South Korea's Samsung. Panther Lake represents a symbolic recovery. All of the compute tiles are fabricated by Intel on the 18A process, marking the return of manufacturing of the most critical portion of Intel's chips to its own fabs, after Lunar Lake infamously relied on TSMC for its compute tile.
Global retail availability of laptops carrying the new processors is scheduled for January 27th, 2026, with more than 200 designs from partners including MSI, Lenovo, and ASUS expected to reach the market in the first half of the year. Intel faces competition on multiple fronts. The launch of Panther Lake creates immediate pressure on rivals including Qualcomm, whose Snapdragon X Elite series had cornered the market on Windows-on-ARM efficiency for the past two years. AMD's Ryzen AI lineup also competes aggressively for thin-and-light laptop market share. The benchmark study, whatever its merits as a historical document, arrives at a commercially convenient moment for Intel's marketing department.
Critics of the benchmark methodology would note that testing over 18 years using a single operating system, Ubuntu 26.04, and a fixed set of workloads inevitably favours the newest hardware. Software frameworks for AI acceleration, database management, and cryptography have been rebuilt from the ground up to exploit modern instruction sets. Running them on 2008 silicon and measuring the result is not quite the same as benchmarking equivalent workloads under conditions that were contemporaneous with each chip. The numbers are real; what they mean requires some care to interpret.
Compound Progress and Its Limits
The deeper significance of the Phoronix study is that it documents, in granular detail, the compounding nature of technological improvement. On average, the Core Ultra X7 358H was up to 9.7 times faster than the Core i5-2520M, codenamed Sandy Bridge, a chip that was already considered powerful when it launched. Each generation built on the last, not just in transistor counts but in architecture, memory bandwidth, instruction sets, and the addition of entirely new compute units. The ThinkPad T61 that still boots in someone's spare room is not just slower than a 2026 machine; it is a different kind of machine, designed for a different vision of what computing was for.
For Australian consumers and businesses, the practical implication is less about headline multipliers and more about the genuinely expanding capability of devices that fit in a bag. A laptop that can run AI inference locally, handle video encoding without cloud assistance, and last through a working day represents a meaningful shift in what portable computing can do. Whether it represents value for money depends entirely on what you need it for, and how recently you last upgraded.
The numbers are extraordinary. The lesson they carry is more modest: progress is real, it is uneven, and it rarely arrives without trade-offs. The old ThinkPad still works. The new one is 95 times faster at some things. Reasonable people can argue about which fact matters more. For the Australian Bureau of Statistics, which tracks technology adoption across Australian industry, the broader trend of compute capability outpacing software demands is one worth watching. For the rest of us, the benchmark is a useful reminder that the laptop collecting dust in the cupboard is not broken. It is simply from a different world.