AMD Radeon 890M faster than Intel Arc A140V, but uses more power

The first side-by-side comparison on two latest GPU architectures.

First, let’s get something straight. AMD Ryzen 300 AI (Strix Point) and Intel Core Ultra 200V (Lunar Lake) are not strictly comparable. Sure, they are both aiming for Microsoft Copilot+PC certification with fast built-in NPU units, feature the latest CPU and GPU architectures, and are primarily aimed at laptops. However, the power settings, which are officially much higher for Strix Point, play a significant role in performance differences.

Intel has not disclosed the full specs of Lunar Lake, but it is expected to launch with a 17W (PL1) and 30W (PL2) power configuration. The Core Ultra 7 258V, which was tested in Geekbench, appears to be using these settings. In contrast, the AMD Ryzen AI HX 370 platform, used in the ASUS ProArt P16 laptop, can use up to 70W TDP. The default spec for this APU is 28W, but it can be configured to 54W.

In other words, this isn’t a fair comparison, as power limits are the main contributors to performance bottlenecks. While graphics performance doesn’t seem to increase significantly beyond a certain threshold, the exact limits for these architectures are yet to be determined by reviewers.

Radeon 890M vs. Arc 140V, Source: Geekbench

The Radeon 890M appears to be 35.4% faster than the Arc A140V. The exact result will vary if we look at the sub-benchmarks, some of which show Lunar Lake’s advantages (Feature Matching or Edge Detection). The Radeon 890M uses 16 RDNA3.5 Compute Units, whereas the Arc A140V has 8 Xe2 Cores. Both GPUs are fully configured, though the Core Ultra 7 258V’s GPU is 100 MHz slower than the flagship Core Ultra 9 288V of the Lunar Lake lineup.

The Geekbench Vulkan test is a good synthetic benchmark, but it doesn’t tell the full story and may not reflect gaming performance. There’s much more to evaluate, including improved ray tracing performance on the Xe2 architecture, boosting capabilities, and how both architectures maintain clock speeds at various TDP levels.

Source: Geekbench