Today, we have a bunch of Geekbench scores of the upcoming notebook processors from both Intel and AMD. This is the first time we’ve been able to compare the two products side-by-side, and yet the results aren’t really all that surprising. On the blue side, we have the Core i5-11300H and 11370H, both quad-core chips with SMT and boost clocks in the 4.4-4-8GHz range.
These quad-core i5s differ from one another with respect to more than just the clock speeds. The Core i5-11300H has a boost clock of 4.4GHz and an L3 cache of 12MB. On the other hand, the 11370H has a boost clock of 4.8GHz and an L3 cache of 12MB (50% more) across the same four cores. Intel had earlier mentioned octa-core Tiger Lake-H parts with 24MB of L3 cache, but we’re yet to see them in any database.
The AMD Cezanne-H competitor to these quad-core mobile CPUs is the Ryzen 5 5600H (6 cores, 12 threads). It has a base clock of 3.3GHz and a boost of 4.2-4.3GHz. Interestingly, the single-core performance of the Tiger Lake-H CPUs is still quite a bit higher than the Zen 3 part. We’re looking at a delta of roughly 100 points between the 11300H and the 5600H, one that doubles when considering the 11370H. While the higher clocks do contribute to much of this lead, I suspect Willow Cove still has a slightly higher IPC than Zen 3. In the multi-threaded benchmark, Cezanne has a 15-20% advantage on account of the two additional cores. A rather small delta for packing 50% more cores, isn’t it?
It takes the higher clocked Ryzen 7 5800H (4.44GHz) to catch up with the Core i5-11300H, but the 11370H still remains out of reach, maintaining a lead of another 100 points, thanks to the near 5GHz single-core boost.
Even the Ryzen 9 5900HX (clocked at 4.74GHz) fails to beat the 11370H in the single-threaded test, all the while being nearly twice as faster in the multi-threaded benchmark. Interestingly, Geekbenck is affected quite a bit by the operating clocks. Going from 4.4GHz to 4.7GHz (+300MHz) increases the multi-core score by a sizeable ~1,500 points on the Zen 3 core. You won’t see this kind of increase with Cinebench or other heavily threaded workloads, as you’ll be limited by a thermal or power bottleneck.