Intel Core Ultra 7 255H lands 32% faster than the 155H in PassMark's single-core benchmark

Intel Arrow Lake
(Image credit: Intel)

Intel's Arrow Lake-based Core Ultra 7 255H appears to have been tested in PassMark, where it outperformed its Meteor Lake equivalent by 32% in single-threaded tests (via x86deadandback at X). Across a wide variety of metrics, the 255H is roughly 15% faster in CPU Mark, which is PassMark's proprietary metric for gauging a CPU's performance.

Intel revealed its Arrow Lake-H family of processors at CES, shortly followed by a user review at Bilibili, which left much to be desired from these CPUs. Dubbed Core Ultra 200H, these processors employ Intel's Arrow Lake architecture featuring Lion Cove P-cores and Skymont E-cores. What separates them from desktop chips, apart from the power envelope, is that these CPUs feature an LPE (Low Power Efficient) core cluster on the SoC Tile, a feature reused from Meteor Lake.

Unlike Lunar Lake which uses Battlemage (Xe2) graphics, Arrow Lake-H is armed with up to eight Xe-LPG+ (Alchemist+) cores with support for XMX. As the SoC Tile remains unchanged, Arrow Lake-H's NPU is capable of dishing out just 13 TOPS of INT8 performance, versus 45 TOPS on Lunar Lake. What is similar to Lunar Lake is the process node: TSMC's N3B, a step-up from Intel 4 used with Meteor Lake.

The Core Ultra 7 255H in question packs 16 cores, divided into six P-cores, eight E-cores, and two LPE-cores with 16 threads in total, as Arrow Lake lacks hyperthreading support. The Core Ultra 7 155H on the contrary is equipped with a similar layout but 22 threads. In PassMark's single-core benchmark, the 255H blazes past its predecessor, scoring 4,631 points compared to the 155H's 3,500 points for a 32% lead. This is a direct result of the updated Lion Cove P-cores and N3B process, allowing a 300 MHz bump in boost clocks. When aggregated, the CPU Mark rating puts Arrow Lake ahead by around 15%.

Core Ultra 7 255H vs 155H PassMark

(Image credit: PassMark)

Relatively speaking, efficiency remains the Achilles' heel of these chips, as the 16-core Core Ultra 9 285H failed to beat the 10-core Ryzen AI 9 365 when limited to 50W of power. While Arrow Lake-H offers an updated Compute Tile and a slightly modified Graphics Tile, the SoC, and IOE Tiles are largely carried over from Meteor Lake.

It all comes down to how these laptops are priced since Strix Point devices still have an entry price of around $1,000. On that note, it is important to mention that the 15W variant of these Intel chips, Core Ultra 200U, is reported to be based on Meteor Lake with Redwood Cove+ P-cores and Crestmont+ E-cores fabbed on Intel 3, a node once reserved for Intel's server counterparts. This will allow Intel to extract higher margins with possibly lower prices for us, though we haven't exactly found affordable Arrow Lake laptops to be abundant, at least not yet.

TOPICS
Hassam Nasir
Contributing Writer

Hassam Nasir is a die-hard hardware enthusiast with years of experience as a tech editor and writer, focusing on detailed CPU comparisons and general hardware news. When he’s not working, you’ll find him bending tubes for his ever-evolving custom water-loop gaming rig or benchmarking the latest CPUs and GPUs just for fun.

  • Gururu
    In light of AMD and nVidia giving us more of the same, with small forward increments in hardware design, its nice to see Intel experimenting with design. Though some designs didn't pan out so well, some appear to. This year looks like a bridge to better architectures in the future.
    Reply
  • usertests
    Gururu said:
    In light of AMD and nVidia giving us more of the same, with small forward increments in hardware design, its nice to see Intel experimenting with design.
    What exactly is AMD not giving you in laptops? They moved mainstream APUs to 12 cores, have 16 desktop cores with 3D V-Cache, and soon mega APUs.
    Reply
  • PixelAkami
    usertests said:
    What exactly is AMD not giving you in laptops? They moved mainstream APUs to 12 cores, have 16 desktop cores with 3D V-Cache, and soon mega APUs.
    Technically its a 16 core processor with 8 cores only with V-Cache so you'd still need to core park anyways. Also mega APUs are nice but that means they'll also run toasty.
    Reply
  • Alvar "Miles" Udell
    As the SoC Tile remains unchanged, Arrow Lake-H's NPU is capable of dishing out just 13 TOPS of INT8 performance

    Decidedly a pro, no "Copilot+" "AI" features like Recall to worry about.
    Reply
  • bit_user
    The author failed to note the obvious problem of looking at random scores in a benchmark DB, which is that you don't know what chassis they were tested in, or other parameters of the test. This is especially true for laptops, since Intel gives the OEM more control over power & boost, basically to match the system's thermal solution.

    To be a true apples-to-apples test, you'd want to see these two CPUs tested in the same chassis, from the same vendor, with the same settings.
    Reply
  • bit_user
    PixelAkami said:
    mega APUs are nice but that means they'll also run toasty.
    Strix Halo (Ryzen AI Max) is specified for laptops with a TDP of anywhere from 45W to 120W. So, they're not small, but they also don't clock as high as the Fire Range ones (i.e. the ones to come in a X3D option), which will be paired with a dGPU.

    Ryzen AI Max is clearly aimed at the upper-middle of the market, not the high-end.
    Reply
  • usertests
    PixelAkami said:
    Also mega APUs are nice but that means they'll also run toasty.
    In some form factors they're going to be replacing both a high-TDP CPU and dGPU, which could easily exceed the 120W cTDP. I think 60-120 Watts is typical for one of these laptop dGPUs (low end ones like 3050/4050 might be configured to use 35W).
    Reply
  • PixelAkami
    usertests said:
    In some form factors they're going to be replacing both a high-TDP CPU and dGPU, which could easily exceed the 120W cTDP. I think 60-120 Watts is typical for one of these laptop dGPUs (low end ones like 3050/4050 might be configured to use 35W).
    Fair enough. Let's just hope the encoders are up to snuff for those that use their encoders
    Reply
  • watzupken
    PixelAkami said:
    Technically its a 16 core processor with 8 cores only with V-Cache so you'd still need to core park anyways. Also mega APUs are nice but that means they'll also run toasty.
    It depends on what you are using the laptop for anyway. I think its been proven that games don't require more than 8 cores, and non game software don't really benefit from the 3D V-cache. So if you game, the extra cache on the 8 cores is sufficient, while you still get to enjoy the benefits of the 16 cores when you need it to run multi-threaded software/ workflows.

    As to running toasty, I agree. But this is generally a problem with any mobile CPUs due to the lack of cooling. To try and mitigate this issue, mobile CPUs operate with lower power limit and aggressive throttling. Hence, don't expect it to run as fast as the desktop counterpart, though you gain mobility in a laptop form factor. So it is down to individual requirements.
    Reply
  • bit_user
    watzupken said:
    I think its been proven that ... non game software don't really benefit from the 3D V-cache.
    Not quite, but true to a large extent. Most PC productivity stuff has either too small or too large of a working set to gain more from a larger L3 than the X3D models traditionally have sacrificed in clock frequency. Phoronix has tested 3 generations of Ryzen and EPYC with 3D VCache and found some wins in certain compute tasks. Here are the benchmarks for the 9800X3D:
    https://www.phoronix.com/review/amd-ryzen-7-9800x3d-linux
    The TL;DR is that the following sorts of workloads sometimes gain a substantial benefit:
    AI (CPU layers)
    Some compression/decompression
    AV1 encoding
    Some server apps & database workloads
    Some HPC & simulation
    Sadly, it doesn't really seem to benefit rendering or compilation. However, in the Ryzen 9000 series, there's very little downside to the 3D cache (other than cost). So, you can either buy it if you're the type who always wants the best at any price or if you know one of your common workloads stands to gain from it.

    watzupken said:
    As to running toasty, I agree. But this is generally a problem with any mobile CPUs due to the lack of cooling. To try and mitigate this issue, mobile CPUs operate with lower power limit and aggressive throttling.
    Some laptops have bigger, heavier thermal solutions and fans that spin way up. I don't want that sort of thing, in my laptop. As you say, I'd rather have a laptop that throttles more aggressively, in multicore workloads.
    Reply