Intel’s upcoming Arrow Lake-S and Arrow Lake-HX Refresh CPUs are turning out to be more of a light tune-up than a genuine upgrade. Contrary to earlier rumors suggesting a significantly improved NPU (Neural Processing Unit), the refreshed chips will stick with the same 13 TOPS NPU performance found in their predecessors.
No architectural enhancements, no clock bumps for the NPU – just a modest boost in overall CPU clocks.
This means users hoping for AI acceleration improvements are out of luck. Intel appears to be holding off on any serious NPU evolution until the next-gen Nova Lake platform. The decision seems to stem from a desire to avoid increasing die size and thermal output – two things that plagued the Raptor Lake refresh.
Arrow Lake Refresh is still built on the TSMC N3B process and retains the Lion Cove (P-core) and Skymont (E-core) architectures paired with Xe1 graphics. Performance gains from this refresh are likely to be in the 2-5% range at best, largely due to increased power budgets rather than any deep architectural change.
Intel has been making some platform-level improvements like Core Ultra 200S boost, better memory tuning, and refinements to the D2D interconnect, but these are under-the-hood optimizations. From a gaming or everyday performance standpoint, don’t expect any major difference.
This refresh feels more like a way to keep the LGA 1851 socket alive and help motherboard partners move stock before Intel transitions to the Nova Lake CPUs with the all-new LGA 1954 socket. With desktop CPU sales slowing and retail interest cooling, Intel seems to be buying time – not breaking new ground.
The Arrow Lake Refresh lineup is expected in the second half of this year. But unless you’re locked into LGA 1851 and desperately need a few extra frames or benchmarks, you may want to wait for something more revolutionary.