Intel’s Arrow Lake-S processors are looking less impressive than previously anticipated. The latest benchmarks of a Core Ultra 9 285K/Core i9-15900K (via Jaykihn) indicate a 7% performance deficit compared to AMD’s upcoming Ryzen 9 9950X. The Zen 5-based “Granite Ridge” flagship scores 46090 points in the Cinebench R23 multi-core benchmark, versus 43118 points for the Core i9-15900K/Core Ultra 9 285K. Cinebench numbers may be thrown around casually, but it’s a reliable indicator of content creation performance, notably 2D/3D rendering.
The Ryzen 9 9950X was tested with a PPT (boost power limit) of 230W, while the 15900K/285K had a PL2 limit of 253W. Given that this is a qualification sample (QS), performance will likely improve as we get closer to the official October launch, but the same can perhaps be said for the 9950X.
Compute-intensive workloads like Cinebench tend to benefit from hyper-threading so that is the likely culprit. Regardless, if Intel can close the gap to a low single-digit percentage, that would still be acceptable as gaming and other mainstream workloads are largely unaffected by hyper-threading.
Previously, a Cinebench 2024 nT leak placed the Ryzen 9 9950X 8-9% ahead of the Core Ultra 9 285K/15900K (ES2), with an 11% lead over the Core i9-14900K. Apart from lacking hyper-threading, the Arrow Lake-S processors will feature lower boost clocks than the mature Raptor Lake offerings (but not stable).
The jump to a new node grants better performance/power efficiency, but resets the frequency gains achieved by the preceding node (same as production yields). In contrast, AMD’s Ryzen 9000 processors will reuse TSMC’s 5nm (N5) process, as 4nm (N4) is an enhancement of the same. The Ryzen 9 9950X has demonstrated ample overclocking potential, with five world records broken a month before the official launch.
