Intel Core Ultra 7 265F vs Intel Core Ultra 9 275HX Comparison

The Intel Core Ultra 9 275HX and the Intel Core Ultra 7 265F look like siblings on paper: both are Arrow Lake parts, both built on TSMC's 3 nm process, both carry 17,800 million transistors on a 243 mm² die, and both sit in the 93rd to 94th percentile of all CPUs in the database. But they target different machines entirely. The 275HX is a mobile flagship with a soldered BGA 2114 package, while the 265F is a desktop chip on Socket 1851. The benchmark data shows a split personality: the mobile part dominates sustained multi-core work, while the desktop part wins several single-threaded and modern-workload tests. The overall record is 12 wins for the 275HX against 5 for the 265F, yet the margins tell a more nuanced story than that tally suggests.

The Verdict

Pick the Core Ultra 9 275HX if throughput is the priority. It wins 12 of 17 head-to-head tests, including every Passmark compute category: data compression by 20%, encryption by 19.4%, extended instructions by 19.8%, integer math by 12.4%, floating point math by 10%, and the overall Passmark multithread score by 12.8%. With an average benchmark score of 67469 versus 64438 for the 265F, and a percentile rank of 94 versus 93, the data places the 275HX slightly ahead overall. It also lands in rarefied company: its nearest rivals by average score are workstation and server parts, including the Intel Xeon w5-3525 (67673, just 0.3% ahead of it), the AMD EPYC 4484PX (67822, 0.5% ahead), the AMD Ryzen Threadripper PRO 5955WX (67868, 0.6% ahead), and the Intel Xeon 6515P (67006, 0.7% behind).

Pick the Core Ultra 7 265F if the workload leans on single-threaded or lightly threaded performance. It wins Cinebench R23 single-core outright, 5926 to 2204, a 62.8% gap in its favor, and Cinebench R15 single-core 597 to 334, a 44.1% win. It also takes Cinebench R23 multi-core, 41980 to 35589, a 15.2% advantage, and edges the Passmark single-thread test 4750 to 4713, a 0.8% win. Its nearest rivals include the Intel Core Ultra 7 265 (64640, 0.3% ahead), the AMD EPYC 7343 (64202, 0.4% behind), the AMD EPYC 4464P (64823, 0.6% ahead), and the Intel Core i9-13900KS (64051, 0.6% behind). Both chips ship with a launch MSRP stated once here: the 265F carries a launch MSRP of $379, while no launch MSRP is recorded for the 275HX.

Note that these two parts are not direct shopping alternatives. One is a mobile BGA processor destined for high-end laptops, the other a socketed desktop processor. The comparison is useful for understanding what mobile silicon now achieves against desktop parts in the same architecture family.

Where Each One Wins

The 275HX owns anything that saturates its 24 cores. In Cinebench R15 multi-core it posts 5619.5 against 4231, a 32.8% win. In Cinebench R20 multi-core it leads 19899 to 17631, 12.9%. Across the entire Passmark suite it never loses a single compute test: data compression 608381 versus 507018, random string sorting 74320 versus 62439 (19%), find prime numbers 448 versus 416 (7.7%), physics 3338 versus 3172 (5.2%). Rendering, encoding, compression, simulation: all go to the mobile flagship.

The 265F owns bursty and single-threaded work. Its Cinebench R23 single-core score of 5926 is nearly triple the 275HX's 2204. Its Cinebench R15 single-core score of 597 beats 334 by 44.1%. It also wins Cinebench R23 multi-core, which suggests the 275HX's advantage does not hold uniformly across all rendering engines; the R23 multi-core result flips 15.2% toward the desktop chip. Games, office applications, and anything latency-sensitive with one busy thread favor the 265F. The Passmark single-thread test is essentially a tie, 4750 versus 4713, so the single-thread picture depends heavily on the test used, with Cinebench strongly favoring the desktop part.

Architecture Differences

Both chips belong to Intel's Core Ultra Series 2 and use the Arrow Lake architecture, but they are different implementations. The 275HX uses the Arrow Lake-HX codename and the mobile BGA 2114 socket, with 24 cores and 24 threads, a 2.70 GHz base clock, and a 5.40 GHz boost clock, at a 55 W TDP. The 265F uses the Arrow Lake-S codename and the desktop Socket 1851, with 20 cores and 20 threads, a 2.40 GHz base clock, and a 5.30 GHz boost clock, at a 65 W TDP. Neither chip uses hyperthreading-style doubling: threads equal cores on both.

Cache differs in one tier. Both share identical per-core L1 (192 KB) and L2 (3 MB) sizing, but the 275HX carries 36 MB of shared L3 versus 30 MB on the 265F. The 275HX also includes integrated graphics, Arc Xe-LPG with 64 EU, while the 265F has none, meaning the desktop build requires a discrete GPU. The 275HX has an unlocked multiplier; the 265F does not.

Everything else lines up. Both are on TSMC's 3 nm node with the same transistor count and die size, both support dual-channel DDR5 at 102.4 GB/s of bandwidth, both offer PCIe Gen 5 with 20 CPU lanes, neither supports ECC memory, and both were released in early January 2025, the 265F on the 6th and the 275HX on the 12th.

FAQ

Q: Which chip is faster overall? A: The 275HX, on aggregate. It wins 12 of 17 head-to-head tests and holds an average benchmark score of 67469 versus 64438, roughly 4.7% higher, with a 94th percentile rank versus 93rd.

Q: Which one is better for single-threaded tasks? A: The 265F. It wins Cinebench R23 single-core 5926 to 2204 (62.8% ahead), Cinebench R15 single-core 597 to 334 (44.1% ahead), and Passmark single-thread 4750 to 4713. The 275HX does win Cinebench R20 single-core, 2809 to 2488, so results vary by test.

Q: Do these fit the same motherboard? A: No. The 275HX uses Intel BGA 2114, a soldered mobile package. The 265F uses Intel Socket 1851, a standard desktop socket.

Q: Does either chip have integrated graphics? A: The 275HX has Arc Xe-LPG graphics with 64 EU. The 265F has no integrated graphics at all, so a discrete graphics card is mandatory.

Q: Can either be overclocked? A: Only the 275HX has an unlocked multiplier. The 265F's multiplier is locked.

Q: How much cache does each have? A: Per-core cache is identical (192 KB L1, 3 MB L2). The difference is shared L3: 36 MB on the 275HX versus 30 MB on the 265F.

Head-to-Head Benchmarks

Start with the 275HX's biggest statement: Cinebench R15 multi-core, 5619.5 versus 4231, a 32.8% win driven by its 24 cores. Cinebench R20 multi-core goes the same way, 19899 versus 17631, 12.9%, and it adds a single-core win there too, 2809 versus 2488, also 12.9%.

The Passmark suite is a clean sweep for the 275HX. Data compression: 608381 versus 507018, up 20%. Random string sorting: 74320 versus 62439, up 19%. Extended instructions: 47016 versus 39235, up 19.8%. Encryption: 47112 versus 39468, up 19.4%. Integer math: 155218 versus 138078, up 12.4%. Multithread: 55759 versus 49410, up 12.8%. Floating point math: 191186 versus 173855, up 10%. Prime finding: 448 versus 416, up 7.7%. Physics: 3338 versus 3172, up 5.2%.

Then the 265F strikes back where it counts for responsive computing. Cinebench R23 single-core: 5926 versus 2204, a 62.8% margin, the largest gap in the entire dataset either way. Cinebench R15 single-core: 597 versus 334, 44.1%. Most surprisingly, Cinebench R23 multi-core: 41980 versus 35589, a 15.2% win for the desktop chip despite four fewer cores. Passmark single-thread is a near-tie at 4750 versus 4713, 0.8% to the 265F. The pattern is consistent: sustained compute and memory-bound integer work favor the 275HX, bursty per-thread work and the newest Cinebench engine favor the 265F.

Specification Differences

• Cores: 24 (275HX) versus 20 (265F)
• Threads: 24 versus 20
• Base clock: 2.70 GHz versus 2.40 GHz
• Boost clock: 5.40 GHz versus 5.30 GHz
• TDP: 55 W versus 65 W
• Socket: Intel BGA 2114 (mobile, soldered) versus Intel Socket 1851 (desktop)
• Codename: Arrow Lake-HX versus Arrow Lake-S
• L3 cache: 36 MB shared versus 30 MB shared (L1 and L2 per-core sizes are identical)
• Integrated graphics: Arc Xe-LPG 64 EU versus none
• Market segment: Mobile versus Desktop
• Release date: January 12, 2025 versus January 6, 2025
• Multiplier unlocked: Yes versus no
• Part number: SRVFK versus SRQCV

Identical between the two: manufacturer, series, architecture, process node (3 nm, TSMC), transistor count (17,800 million), die size (243 mm²), memory support (dual-channel DDR5 at 102.4 GB/s), PCIe (Gen 5, 20 CPU lanes), ECC support (absent on both), and production status (both active).