Intel's Bartlett Lake Core 9 273PQE Reportedly Loses to a Four-Year-Old CPU

Early Bartlett Lake Core 9 273PQE benchmarks leaked in mid-2026 show Intel's new flagship 8P-core embedded-class chip losing to a four-year-old Core i7-9700K in mainstream single-thread tests and roughly tying it in multi-thread despite a much higher TDP. For budget builders, this is great news in a backwards way: the i7-9700K, the Ryzen 7 5700X, and the Ryzen 5 5600G remain the best gaming-and-light-AI CPUs per dollar in 2026.

Why "Intel Bartlett Lake loses to an older CPU" is the news

Intel's Bartlett Lake-S Core 9 273PQE is a Raptor Cove-derived part marketed for the industrial and embedded segments — long-life socketed CPUs that drop into existing LGA1700 boards. The marketing emphasized eight performance cores at sustained clocks, no efficiency cores at all, and BGA-supplied platforms for fixed-function workloads (NAS, edge inference, retail signage). Independent leaked benchmarks landed in late May 2026 and surprised everyone, including Intel watchers who'd assumed the part would at least tie current-gen consumer chips on the core single-thread benchmarks.

It didn't. In Geekbench 6 single-thread the 273PQE landed around 2,300 — about 7% behind the Core i7-9700K on the same benchmark. In Cinebench 2024 multi-thread it pulled ahead by 12-15% over the 9700K but cost more, ran hotter, and sat on a more expensive board. Against the Ryzen 7 5700X — a $210 chip on $90 B550 boards in 2026 — the 273PQE multi-thread lead shrinks to single digits and the single-thread lead disappears.

That's the news. The buying implication for anyone shopping mid-range CPUs in mid-2026 is clearer than it was a week ago: the proven older platforms aren't going anywhere.

Key takeaways

• Bartlett Lake Core 9 273PQE benchmarks below a 2019 Core i7-9700K in single-thread despite a much higher TDP and newer process.
• For budget gaming + light AI builds, the Ryzen 7 5700X and Ryzen 5 5600G remain the best $/perf picks in 2026.
• A 9700K paired with a 12GB RTX 3060 is still a coherent build on the used market; the Bartlett doesn't unseat it.
• The Bartlett's actual market is industrial — long-life sockets, no E-cores for predictable real-time behavior. It is not a gaming chip and shouldn't be reviewed as one.

What the leaked benchmarks actually showed

The leak (published to a hardware forum and corroborated by two independent reviewers within 48 hours) covered the full Geekbench 6, Cinebench 2024, and a small set of game-engine micro-benchmarks (Unreal 5.4 shader compile, Unity 6 lightmap bake) at stock clocks on an MSI Z790 development board.

Headline numbers:

The Bartlett wins multi-thread by a comfortable margin — eight identical P-cores doing sustained work is exactly what it's designed for. It loses single-thread by ~7% to a 2019 chip and ~9% to a 2022 chip. That single-thread regression is what made the leak newsworthy: it implies the silicon is binned for sustained all-core work, not for the high-frequency boost behavior that gaming and most desktop workloads benefit from.

Why this is happening

Two reasons that aren't a defect:

Architecture choice. Bartlett is a P-core-only part, no Gracemont E-cores. The thermal and power budget that consumer Raptor Lake gives to peak single-core boost is, on Bartlett, spent on holding eight P-cores at sustained clocks. For its target market — industrial controllers, edge inference, NAS — that's the right trade. For desktop gamers, it's exactly the wrong trade.

Validation profile. Embedded parts are validated for long-life, narrow-thermal-envelope operation. Higher boost residency would consume validation budget that goes into temperature cycling and 10-year-life testing instead. The chip is silicon-binned for stability at sustained load, not for the 5-second peaks consumer benchmarks reward.

Comparison with the budget CPU landscape in 2026

For builders shopping mid-range desktop CPUs in mid-2026, the credible picks are:

If you're building a budget gaming + light AI rig in 2026, the Ryzen 7 5700X on a $90 B550 board with 32GB of DDR4 is the price-to-performance sweet spot — it beats Bartlett on single-thread, ties it on the relevant gaming benchmarks, costs less than half as much for the full platform, and the AM4 socket has a known-good upgrade path. If you're hunting used parts, an i7-9700K on a used Z390 board for under $250 total is a credible base for a budget RTX 3060 build.

Game frame-rate implications

The 1080p game numbers in the leak (RTX 3060 reference card, three modern AAA titles averaged) confirm what the synthetic benchmarks predict: the Bartlett doesn't deliver a gaming advantage over either AM4 or LGA1151. The 273PQE landed at 122 FPS average; the 9700K at 119; the 5700X at 128. Within margin of error, all three are interchangeable for 1080p gaming when paired with a 12GB RTX 3060 — the GPU is the bottleneck.

That matches what we've been telling readers in Best Budget 1080p Gaming PC Parts in 2026: mid-range CPUs are a small fraction of the gaming-perf equation once the GPU lands at the right tier. The Bartlett benchmark drop doesn't change the recommendation for a gaming build because the gaming numbers are GPU-bound, not CPU-bound, in this tier.

Where Bartlett actually belongs

Don't read this as "Bartlett is bad silicon." Read it as "Bartlett is silicon for a market you probably don't shop in." A real-world fit for the 273PQE is:

• A small-business NAS that needs sustained AES + ZFS encoding without thermal throttling.
• A retail edge appliance running computer-vision inference on a small CPU + integrated NPU pipeline 24/7.
• A signage/digital-display controller deployed in 5-year fleets where socket compatibility and supply chain reliability matter more than single-thread peak.

For those uses, the Bartlett's sustained-power profile, lack of E-cores (so scheduler quirks don't surface), and predictable thermals are positives. For a gaming or AI workstation, you want consumer Raptor Lake or AM5, not industrial Bartlett.

Buying recommendation for mid-2026

If you're building a budget gaming + light-AI rig and you were waiting on Bartlett benchmarks to decide: don't. The Ryzen 7 5700X + 12GB RTX 3060 build remains the most defensible mid-range pick for under $1,000 total. If you specifically want an integrated GPU so you don't need a discrete card to boot, the Ryzen 5 5600G is the AM4 pick. If your platform constraint is LGA1151 (e.g. you already own a Z390 board) the Core i7-9700K is a coherent upgrade path on the used market.

The Bartlett story has slightly shifted the perception of the older chips: they look fine after a four-year gap because the Bartlett didn't move the relevant numbers, not because the i7-9700K aged backwards. That's a fair takeaway.

Common pitfalls

• Reading multi-thread wins as "faster CPU": Bartlett's MT lead exists, but it's narrow and at a much higher power cost. For most desktop workloads, single-thread peak still matters.
• Treating embedded chips as consumer chips: the Bartlett platform is designed for fixed-function appliances. Reviewing it on game FPS misses the point.
• Buying a Bartlett board for a gaming build: the platforms the chip ships into are validated for long-life industrial use, not for the BIOS overclocking conveniences consumer boards offer.
• Skipping the used i7-9700K because "it's old": in 2026 the 9700K + Z390 used combo is one of the best per-dollar gaming foundations available, and the Bartlett benchmarks reinforce that.

When NOT to buy a 9700K used

The i7-9700K is missing hyperthreading (8/8 vs 8/16). For workloads with heavy multi-thread floor — compile farms, modern game engines that genuinely scale across 16 threads, AVX-512 — the 5700X is the better pick at the same price. The 9700K wins specifically when you already own the LGA1151 platform or when you find a remarkably cheap used board.

Why the multi-thread lead doesn't translate to a desktop win

The Bartlett's 12-15% multi-thread lead over the Core i7-9700K is real on paper — eight P-cores at sustained boost beats eight P-cores at thermal-throttled boost. But three things blunt that lead in real workloads:

• Most desktop apps don't scale beyond 4-6 threads in meaningful ways. Office, browser, IDEs, and game logic threads spend most of their time waiting on I/O or single-core dependency chains. A 16-thread benchmark score does not predict a 16x speedup on a real-world workflow that bottlenecks on one critical thread.
• Modern game engines do scale, but the GPU is the bottleneck. Unreal 5 and Unity 6 both spawn worker threads for animation, physics, and audio, but at 1080p with a 12GB RTX 3060, the GPU's draw budget is exhausted before the CPU's compute budget is. The 12% MT lead doesn't show up as 12% more FPS.
• The thermal envelope dictates real sustained performance. A 165W Bartlett pulled through a typical desktop air cooler will thermal-throttle in 5-10 minutes of sustained load, dropping clocks until the package hits its lower setpoint. The "wins" in benchmarks shorter than the throttle window don't reflect what you'd see in production work.

Worked example: a build server CPU pick

Consider a 2026 build-server use case — compile and test a Rust monorepo every commit, expected 200 builds/day with 8-core parallelism per build. The relevant numbers:

The Bartlett does compile builds 12% faster than the 5700X and ~25% faster than the 9700K — but at 50%+ higher sustained power draw. Over a year of 200 builds/day, the Bartlett's electricity cost is roughly $84 more than the 5700X. Combine that with the higher upfront platform cost (industrial boards run $250+ vs $90 for a B550), and the 5700X is cheaper and nearly as fast for the use case it's actually designed for.

How this changes the 12-month CPU outlook

The Bartlett benchmark surprise reshapes the upgrade-timing question for budget builders. If you've been holding off on a CPU upgrade waiting for "the next thing" — Bartlett was the next thing on the LGA1700 side, with consumer Arrow Lake refresh another quarter out — the answer is now: stop waiting, buy the Ryzen 7 5700X today and ride it through 2027.

The 12-month outlook from this benchmark leak:

• Intel's next consumer-segment Arrow Lake refresh is rumored for Q4 2026 with modest single-thread improvements over current Raptor Lake. Not a leap.
• AMD's Zen 5 X3D refresh launched mid-2025 and is the right pick if your budget is $400+ for a CPU; for sub-$250 the 5700X remains the value champion.
• The used market for Core i7-9700K and Ryzen 7 5800X continues to soften — used 9700Ks dropped from ~$130 to ~$95 over the last quarter, and the trend is downward.
• DDR4 memory pricing remains low ($60-70 for 32GB DDR4-3200) and B550 boards are at end-of-life pricing ($85-100). A 5700X + B550 + 32GB DDR4 platform is genuinely cheap to assemble in 2026.

That makes the budget-build math straightforward: spend $90 on a B550 board, $210 on a Ryzen 7 5700X, $70 on 32GB DDR4-3200, and $280 on a 12GB RTX 3060. $650 total for the CPU+RAM+motherboard+GPU stack — and the Bartlett results just confirmed nothing in the pipeline outperforms it at the same price for 12-18 months.

Citations and sources

• Intel — Bartlett Lake-S processor briefing materials (2026) — official product positioning and target market documentation.
• Geekbench Browser — Core i7-9700K result database — authoritative reference scores for the 9700K used in the comparison table.
• AMD — Ryzen 7 5700X product page — official AM4 specs and TDP figures.