Intel Lunar Lake Previewed: Over 100 AI TOPS, 50% Faster Battlemage “Xe2” GPU, Faster CPU Performance Versus AMD Zen 4 & Snapdragon X Elite

Intel Lunar Lake Previewed: Over 100 AI TOPS, 50% Faster Battlemage “Xe2” GPU, Faster CPU Performance Versus AMD Zen 4 & Snapdragon X Elite

Intel has previewed its next-generation Lunar Lake CPUs ahead of Computex, which will launch in Q3 with the latest AI NPU and Battlemage “Xe2” GPU cores.

Intel Lunar Lake Will Set the Bar for Next-Gen Thin & Light AI PC Platforms: New CPU Cores, 40% Faster AI NPU, Battlemage “Xe2” GPU and More

Ahead of its announcements at Computex 2024, Intel has given the press a preview of its next-generation Lunar Lake CPUs. The company unveiled that Lunar Lake CPUs will underpin the next generation of its AI PC platform and is already rolling out production wafers with plans for the third quarter of 2024.

Before going into the specifics of Lunar Lake, Intel says its first-generation “Core Ultra 100” CPU lineup, codenamed Meteor Lake, has shipped more than 7 million units. The blue ship is confident of its kill. Target to ship 40 million units by end of 2024 through its Meteor Lake, Arrow Lake and Lunar Lake CPUs and will eventually ship 100 million units by 2025. The Intel Meteor Lake CPUs themselves are gaining a lot of recognition on the developer side of things with more experimentation and optimization coming to Intel's “AI PC” platforms. Including AI-enhanced security, client-optimized LLMs, local personal assistants, and more. Their long-term goal will be to deliver a robust AI PC experience on next-generation CPU platforms such as Arrow Lake, Lunar Lake, Panther Lake, and beyond.

So starting with the details, Intel will ship Lunar Lake CPUs with a completely new architecture on all fronts. You're looking at new CPU cores, new GPU cores, and new NPU cores, all made possible by the latest innovations in 3D packaging technology and power innovations.

The main changes include the use of the latest Lion Cove P-Cores and Skymont E-Cores which should bring some decent IPC improvements. Intel says it's too early to talk about performance and IPC figures for its new P/E CPU cores, but they're being evaluated and more information will be shared soon. The company also touts “significant” IPC and perf/watt improvements with the new CPU cores, along with an improved thread director to better manage those cores.

On the GPU side, Intel is adding its brand new Xe2 GPU architecture codenamed Battlemage. It's been a long time coming and Lunar Lake is the first CPU to take advantage of the Xe2-LPG design, which is designed for integrated graphics solutions. The biggest advantage of this brand new graphics chip is the addition of Intel XMX engines for AI. Currently, Intel's Xe-LPG iGPUs featured on Meteor Lake chips use DP4a acceleration and don't use the first-gen XMX engine, but with Lunar Lake, you're not only getting XMX acceleration, but 2nd Gen XMX cores are also getting, enabling much faster performance.

On the NPU side of things, Intel's Lunar Lake CPUs are boosting AI performance by more than 3x with an NPU that can deliver 45 tops alone compared to 11 tops on the current NPU on Core Ultra “Meteor Lake” chips. Below is a breakdown of Intel's Meteor Lake and Lunar Laketops:

  • Lunar Lake (GPU/NPU/Total) – 60+45= 105 AI tops
  • Meteor Lake (GPU/NPU/Total) – 19+11= 30 AI tops

If you do the math, the Lunar Lake GPU offers a 3.15x boost over the Meteor Lake GPU while the Lunar Lake NPU offers a 4.09x boost over the Meteor Lake NPU. That's quite an increase and with the CPU added to the mix, you're looking at more than 105 AI peaks from the NPU and GPU alone but keep in mind that the GPU and NPU are doing most of the work. will Heavy lifting when it comes to specific tasks related to AI. Lunar Lake CPUs will also offer support for mixed precision and several other data types, enabling optimization in more than 500 models.

The chip used for performance testing was an early prototype with a 4+4 configuration meaning 4 P-Cores and 4 E-Cores. The chip had two SKUs, one rated at 17W PL1 and the other rated at 30W PL1. Lunar Lake takes advantage of on-package memory and this sample uses single-rank LPDDR5 memory rated at an unknown clock speed but this sample is said to have a capacity of 16 GB.

We know that the Lunar Lake-MX “Core Ultra 200V” chips will offer up to 32 GB of on-package memory capacity.. The reference performance used Intel Filmore Peak connectivity and had a 14″ 2880×1800 panel with eDP 1.4, PSR2, and LRR2.5 capabilities.

Intel Core Ultra 200V “Lunar Lake” CPU SKUs (Initial):

CPU name architecture Core / threads GPU core On-package memory NPU TDP
Core Ultra 7 2XX Lion Cove + Skymont 4+4 (8/8) 8 Xe2 (Battlefield) 32 GB 3x Meteor Lake 7-15W
Core Ultra 7 2XX Lion Cove + Skymont 4+4 (8/8) 8 Xe2 (Battlefield) 16 GB 3x Meteor Lake 7-15W
Core Ultra 5 238V Lion Cove + Skymont 4+4 (8/8) 7 Xe2 (Battlefield) 32 GB 3x Meteor Lake 7-15W
Core Ultra 5 234V Lion Cove + Skymont 4+4 (8/8) 7 Xe2 (Battlefield) 16 GB 3x Meteor Lake 7-15W

Intel also shares preliminary performance data for its Lunar Lake CPUs against Meteor Lake (Core Ultra 7 165U), AMD's Ryzen 7 7840U/8840U, and the upcoming Snapdragon X Elite chips. In AI TOPS, the Lunar Lake CPU is about 40% faster than the Snapdragon X Elite in Stable Diffusion 15. In graphics performance, the Lunar Lake Battlemage “Xe2” GPU (17W) offers a 50% increase over the Meteor Lake Alchemist “Xe1”. GPU (15W) and CPU core performance in 3DMark Time Spy are said to be “faster” than Ryzen 7 8840U and Snapdragon X Elite.

In all of these tests, Intel referred to the power plan as “Balanced Mode”, the Power Mode as “Best Power Efficiency” and the Performance Mode as “Best Performance” in its and competitor AI PC platforms.

Well, moving on to the power figures, Intel has added a brand new “Advanced Low Power Island” to its Lunar Lake CPUs that deliver 30% less power than the Ryzen 7 7840U and Ryzen 7 8840U chips. 20% lower than Snapdragon 8cx. Gen3 chips.

2024-05-20_2-58-21
2024-05-20_2-58-25

This is just a brief overview of what Intel has in store for the second half of this year and there's no doubt that Lunar Lake is looking like a fantastic solution for thin and light PC buyers. In these official tests, Lunar Lake CPUs look pretty good compared to the competition and while it's always best to wait for third-party tests and independent reviews, especially for the CPU and GPU aspects, we'd recommend Lunar Lake “Core Looking forward to Ultra 200V”. CPU family.

Intel Mobility CPU Lineup:

CPU family Panther Lake Lunar Lake Arrow Lake Meteor Lake Raptor Lake Alder Lake
Process Node (CPU tile) Intel 18 A Intel 20 A Intel 20 A Intel 4 Intel 7 Intel 7
Process Node (GPU Tile) TSMC 3/2nm? TSMC 3nm? TSMC 3nm TSMC 5nm Intel 7 Intel 7
CPU architecture Hybrid Hybrid Hybrid (four cores) Hybrid (Triple Core) Hybrid (dual core) Hybrid (dual core)
P core architecture Cougar Coo Lion Cove Lion Cove Redwood Cove Raptor Cove Golden Cove
E Core Architecture Skymont? N/A Skymont Crestmont Gracemont Gracemont
LPE Core Architecture (SOC) Skymont? Skymont Crestmont Crestmont N/A N/A
Top configuration TBD 4+4 (MX Series) TBD 6+8 (H-Series) 6+8 (H-Series)
8+16 (HX-Series)
6+8 (H-Series)
8+8 (HX-Series)
Maximum cores/threads TBD 8/8? TBD 14/20 14/20 14/20
Planned lineup H/P/U series V series H/P/U series H/P/U series H/P/U series H/P/U series
GPU architecture Xe3-LPG (sky) Xe2-LPG (Battlemage) Xe-LPG (Alchemist) Xe-LPG (Alchemist) Iris Xe (Gen 12) Iris Xe (Gen 12)
GPU execution units TBD 64 European Union 192 European Union 128 EUs (1024 cores) 96 EUs (768 cores) 96 EUs (768 cores)
Memory support TBD LPDDR5X-8533 TBD DDR5-5600
LPDDR5-7400
LPDDR5X – 7400+
DDR5-5200
LPDDR5-5200
LPDDR5-6400
DDR5-4800
LPDDR5-5200
LPDDR5X-4267
Memory Capacity (Maximum) TBD 32 GB TBD 96 GB 64 GB 64 GB
Thunderbolt ports TBD TBD TBD 4 (TB4) 4 (TB4) 4 (TB4)
Wi-Fi capability TBD Wi-Fi 7 TBD Wi-Fi 6E Wi-Fi 6E Wi-Fi 6E
TDP TBD 17-30W TBD 7W-45W 15-55W 15-55W
Launch. 2 AH 2025 2 AH 2024 2 AH 2024 2 AH 2023 1H 2023 1 AH 2022

Share this story.

Facebook

Twitter

About the Author

Leave a Reply