Performance Index
Computex 2024
For details on Intel® Xeon™ 6 claims, please visit the Intel® Xeon® 6 page on the Performance Index.
For additional information about Intel at Computex 2024, including information about Intel® Gaudi® and Lunar Lake, please visit intel.com/newsroom.
Lunar Lake ITT (Intel Technology Tour) / Computex 2024 | Learn more at intel.com/newsroom. |
---|---|
Claim | Details |
Lunar Lake is 40% faster at generative AI compared to ARM systems. | As of May 2024, based on Intel analysis of currently available information; subject to change without notice. Based on Intel estimates of Lunar Lake on stable diffusion 1.5 with GIMP plugin against Snapdragon X Elite public claims. Generative AI workload with Intel's GIMP plugin and Stable Diffusion: This workload measures the total time in seconds to complete 20 inference steps using the GPU with Intel's GIMP 2.99.14 plugin, which includes Stable Diffusion support via OpenVINO 2023.1. Data in this presentation compares Intel's Lunar Lake reference platform estimates with Qualcomm's public statements on their GIMP Stable Diffusion plugin, noting that Qualcomm's optimizations are unknown. |
Up to 40% lower SOC power on Lunar Lake vs. prior generation. | Testing by Intel as of May 2024. Data based on Lunar Lake reference validation platform measurement vs Meteor Lake reference validation platform as measured by YouTube 4K 30 AV1. |
Lunar Lake delivers up to 50% better graphics performance vs. prior generation. | Testing by Intel as of May 2024. Data based on Lunar Lake reference validation platform measurement vs Meteor Lake reference validation platform as measured by 3DM Time Spy. 3DMark* |
Recent research shows that 40% of AI developers will choose to run their workloads on the GPU by 2025. | Based on internal Intel research as of May 2024. |
Lunar Lake delivers similar single-thread performance at half the SoC power vs. prior generation. | Testing by Intel as of May 2024. Data based on Lunar Lake reference validation platform as measured CBR24 ST vs. prior generation. |
Lunar Lake delivers up to a 35% power reduction when running Microsoft Teams with containment & power management optimizations enabled. | As of May 2024, based performance estimated with measurements on Lunar Lake reference platform with power optimizations enabled vs. power optimizations disabled. |
Lunar Lake P-cores (codenamed "Lion Cove") deliver 14% better IPC. | Iso frequency benefit estimate or the Lion Cove microarchitecture (ST) vs. the Redwood Cove microarchitecture (ST) across subcomponents of the following workloads: SPEC® CPU 2017 rate base components (estimated) running 1 copy (assuming Intel Compiler 23.2.3), Cinebench 1-CPU R23. R24, Geekbench 5.4.5 SC, Geekbench 6.2.1 SC, WebXPRT4 & Speedometer. |
Lunar Lake E-cores (codenamed "Skymont"). Skymont IPC on Lunar Lake Low Power Island delivers 1.38x integer and 1.68x floating point vs. previous generation low-power E-cores (codenamed "Crestmont"). | Results are based on Intel's internal projections/estimates as of 5.13.2024(+/- 10% Margin of Error) on SPEC® CPU 2017 rate base metrics (estimated), GCC12.1-O2 Linux at Fixed Frequency (ISO). |
Lunar Lake E-cores (codenamed "Skymont") deliver up to 2x peak single-thread performance or 1/3 the power at similar single-thread performance, and up to 4x peak multithread performance or 1/3 of the power at similar multithread performance. | Results are based on Intel's internal projections/estimates as of 5.13.2024(+/- 10% Margin of Error) on SPEC® CPU 2017 rate integer base components (estimated), GCC12.1-O2 Linux (ISO). Comparing a Skymont E-core cluster (4 Skymont cores) vs. Meteor Lake LP E-core cluster (2 Crestmont cores) to showcase workload coverage increase for the Lunar Lake Low Power Island. |
Lunar Lake E-cores (codenamed "Skymont") IPC is on-par with Raptor Cove IPC on LLC/Ring. | Results are based on Intel's internal projections/estimates as of 5.13.2024(+/- 10% Margin of Error) on SPEC® CPU 2017 rate base components (estimated), GCC12.1-O2 Linux (ISO) at Fixed Frequency (ISO) with similar memory + fabric for general (not vector heavy) workloads. |
Lunar Lake E-core (codenamed "Skymont") delivers more performance ISO power or ISO performance at lower power. | Results are based on Intel's internal projections/estimates as of 5.13.2024(+/- 10% Margin of Error) on SPEC® CPU 2017 rate integer base components (estimated), GCC12.1-O2 Linux (ISO). Comparing a Skymont E-core cluster (4 Skymont cores) vs. Meteor Lake LP E-core cluster (2 Crestmont cores) to showcase workload coverage increase for the Lunar Lake Low Power Island. |
Lunar Lake delivers up to 2x power efficiency vs. prior generation. | Testing by Intel as of January 2024. B ased on VPU-EM simulation. Power data is generated from the simulation tool based on power data that has been extracted from circuit simulation tools. This simulation, which is a ~100% utilization int8 network, is expected to correlate well with silicon. |
Xe2 IP performance per Xe-core is up to 1.2x to 12.5x higher than Xe1 IP across a set of various graphics functions. | Results are based on an internal suite of micro benchmarks and collected on a pre-release Xe2 engineering platform with pre-release GFX software. The comparison is a selected subset of micro benchmarks normalized for equal Xe-cores configuration and clock frequency. |
Lunar Lake delivers up to 4x NPU AI compute vs. prior generation | NPU TOPS Lunar Lake (up to 48) vs Meteor Lake (up to 11) |
Lunar Lake delivers up to 3.5x GPU AI compute vs prior generation | GPU TOPS Lunar Lake (up to 67) vs Meteor Lake (up to 19) |
Lunar Lake delivers up to 3.3x total AI compute vs. prior generation | Total TOPS Lunar Lake (up to 120) vs Meter Lake (up to 36) |
Lunar Lake's NPU4 delivers 4x vector compute vs. NPU3. | Based on 4x vector width increase vs. NPU3. NPU3 has 8 FP16 Vector ops/clock, NPU4 has 32 |
Lunar Lake delivers 12x overall vector performance. | Vector performance = 3x tiles and 4x the vector width (vs. NPU3) |
Lunar Lake delivers 2x IP bandwidth. | IP Bandwidth: Meteor Lake is 64GB/s; Lunar Lake is 136 GB/s. |
Experience up to 25% better read and write speeds with Thunderbolt™ 5 | Versus Thunderbolt™ 3. Based on estimated performance on Lunar Lake processors. Results may vary. Tested on Intel Reference Validation Platform (RVP) with Lunar Lake Ultra 5 processor running Windows 11 Enterprise (version 10.0.26100 Build 26100) with 16 GB RAM. Tested with Gen 4 NVMe 1 TB SSD used behind a Thunderbolt™ 3 dock (with Titan Ridge) and Thunderbolt™ 5 dock (with Barlow Ridge hub) connected to the Lunar Lake RVP. Storage Read and Write speeds were measured with CrystalDiskMark tool with settings of 5 test count and 32 GiB test size. |
Lunar Lake's integrated Bluetooth® new capability to use a PCIe interface enables up to ~55% reduced time to boot & from sleep compared to legacy USB implementations. | Results as of May 2024. Lunar Lake is the first Intel platform to integrate Wi-Fi 7 technology and enable use of PCIe vs. USB interface for integrated Bluetooth® technology routing. The use of PCIe frees up additional USB lanes for use by PC OEMs and enables improved Bluetooth® initialization and wake times. Intel engineering measurements using a LNL reference validation platform with PCIe-based Bluetooth® showed an average boot time of 0.648 seconds which was approximately 55% lower than the 1.439 seconds required for USB-based Bluetooth® designs. Similar levels of improvement were measured when comparing PCIe vs. USB Bluetooth® interface wake from sleep times for connected peripherals (Human Interface Devices) in both Classic Bluetooth ~55% (34.355 vs. 75.933 ms) and Bluetooth Low Energy ~56% (32.189 vs. 74.098 ms) modes. |
Up to ~50% lower throughput degradation caused by memory noise in channels used by Wi-Fi | Results as of May 2024. PC platform DDR memory can operate at many of the same frequencies as the 5 GHz and 6 GHz channels used by Wi-Fi technologies. When DDR frequencies interfere with Wi-Fi on these channels, the throughput degradation can be as high as 8 dB or about 50% lower, based on measurements by Intel platform engineering teams. Lunar Lake platform Wi-Fi integration enables Radio Frequency Interference Mitigation (RFIM), which coordinates simultaneous platform memory and Wi-Fi operation. DDR RFIM dynamically moves memory speed to clear the connected Wi-Fi channel from memory narrowband noise and helps avoid the potential Wi-Fi performance impact from DDR memory operating on the same frequencies. |
Up to 50% lower latency vs. typical single-radio solutions | Results as of February 2024. A typical Wi-Fi VR setup to a PC with a standard single-radio Wi-Fi card requires the HMD to make 2-hops and go through the router to access the PC, which adds latency to the user VR experience. High-end gaming PCs with premium dual-radio Intel® Killer™ Wi-Fi 6E AX1690 cards are able to dramatically reduce latency for wireless VR gaming using Intel® Double Connect Technology. This solution uses one Wi-Fi radio for the PC-HMD VR connection and a second Wi-Fi radio for the PC-router connection, allowing the Meta Quest headset to bypass the router for direct wireless access to PC games, resources, and Intel® Killer™ Networking software. To deliver the high-quality Wi-Fi connectivity that the Meta Quest HMDs need, working closely with Meta, Intel has enabled laptops configured with its connectivity software and Intel® Wi-Fi 7 (5 Gig) cards to deliver 1-hop single radio Wi-Fi PC VR performance that was similar to the high-end gaming PCs with its dual-radio Killer solution. Intel-Meta VR tested these three Wi-Fi PC VR configurations using 80 MHz channels in the 5 GHz band. In clean environments without channel contention and in a challenging environment with 70% channel contention, the 1-hop dual-radio solution enabled a -30% / -55% reduction in latency, and the new 1-hop single-radio solution enabled a similar -27% / -50% reduction compared to the standard 2-hop configuration. Test Platform: Dell Alienware x15 R2, Processor: 12th Gen Intel® Core™ i7, OS: Windows 11 (26052.1), Wi-Fi Cards: Intel® Wi-Fi 7 BE200, Intel® Killer™ Wi-Fi 6E AX1690, Wi-Fi Driver: 23.30.0.6; Wi-Fi Router: Netgear Nighthawk RAXE500; HMD: Meta Quest 3 (OS v57 496821007900511). |
Intel® Threat Detection Technology is the only AI-based silicon security in deployment across a billion PCs. | As of March 2023, Intel TDT provides the only silicon-enabled AI threat detection to help stop ransomware and cryptojacking attacks for Windows-based systems. Learn more at www.intel.com/tdt . No product or component can be absolutely secure. |
Intel delivers best-in-class security assurance | As of May 2024, based on ABI Report: Embracing Security as a Core Component of the Tech You Buy - https://www.intel.com/content/www/us/en/security/security-as-a-component-of-tech.html . This study was sponsored by Intel. Research independently conducted by ABI Research. |
Lunar Lake ITT presentation may contain references to product features that are currently under development. Information shown through the presentation is based on expectations as of May 2024 and subject to change without notice.
Results that are based on pre-production systems and components as well as results that have been estimated or simulated using an Intel Reference Platform (an internal example new system), internal Intel analysis or architecture simulation or modeling are provided to you for informational purposes only. Results may vary based on future changes to any systems, components, specifications or configurations.
Performance varies by use, configuration and other factors.
AI features may require software purchase, subscription or enablement by a software or platform provider, or may have specific configuration or compatibility requirements. Details at www.intel.com/AIPC.
All product plans and roadmaps are subject to change without notice.
Performance hybrid architecture combines two core microarchitectures, Performance-cores (P-cores) and Efficient-cores (E-cores), on a single processor die first introduced on 12th Gen Intel® Core™ processors. Select 12th Gen and newer Intel® Core™ processors do not have performance hybrid architecture, only P-cores or E-cores, and may have the same cache size. See ark.intel.com for SKU details, including cache size and core frequency.
Built-in Intel® Arc™ GPU only available on select Intel® Core™ Ultra processor-powered systems; OEM enablement required.
Built into the hardware, Intel® Thread Director is provided only in performance hybrid architecture configurations of 12th Gen or newer Intel® Core™ processors; OS enablement is required. Available features and functionality vary by OS.
Intel does not control or audit third-party data. You should consult other sources to evaluate accuracy.
While Wi-Fi 7 is backward compatible with previous generations, new Wi-Fi 7 features require PCs configured with Intel Wi-Fi 7 solutions, PC OEM enabling, operating system support, and use with appropriate Wi-Fi 7 routers/APs/gateways. 6 GHz Wi-Fi 7 may not be available in all regions. Performance varies by use, configuration, and other factors. For details on performance claims, learn more at www.Intel.com/performance-wireless.
SPEC®, SPECrate® and SPEC CPU® are registered trademarks of the Standard Performance Evaluation Corporation. See http://www.spec.org/spec/trademarks.html for more information.