Why bother with NDAs?

Courtesy of database detective @Leakbench on Twitter, we now have our first decent look at how Intel's next-gen Core i9-11900K Rocket Lake CPU will perform in our CPU Benchmark hierarchy. This test is the first clear result from Geekbench 4 for the 11900K, which is nice to see as it can be a more accurate gauge of raw CPU performance than the other benchmark results we've seen, like Passmark or Geekbench 5. The latest test results show that Rocket Lake will assuredly climb the gaming ranks, and if the price is right, the new chips could upset our list of Best CPUs. 

In a nutshell, you shouldn't trust Geekbench 5's overall scores as an accurate measure of Rocket Lake's performance, and there's a technical reason why. We've encountered strange phenomenons with Geekbench 5, where its use of AVX-512 can widely skew the results in the encryption subtest. In turn, this inflates Rocket Lake's overall Geekbench 5 scores against all other processors that don't support AVX-512. This can lead to an inaccurate picture that makes Rocket Lake appear better in relation to AMD's competing chips, not to mention Intel's previous-gen models. 

Geekbench 4 isn't perfect either, but its lack of AVX-512 support makes the test much more accurate when gauging per-core performance without using an exotic SIMD instruction (AVX-512) that has no meaningful uptake in mainstream desktop PC software. In fact, Geekbench's developer has stated that the AVX-512 testing disparity will be addressed in the Geekbench 6 benchmark that's due out later this year. The big takeaway here — don't look too deeply into the overall Geekbench 5 test results.

This particular test submission seems about as close as we'll get to something solid before the launch, but as usual, we have to take the results with a grain of salt. However, the 11900K boosts to 5.3 GHz throughout this test sequence, signaling it's running at stock core clocks, and the memory appears to be running at the stock DDR4-3200 for Rocket Lake. 

This is important because Geekbench 4 is sensitive to memory frequency, especially when it comes to multi-core tests. To compare, we're using test results that we generated in our own labs for the Core i9-10900K and Ryzen 7 5800X, with both operating at stock memory clocks (DDR4-2933 and DDR4-3200, respectively). 

Intel claims a 19% increase in IPC for the Rocket Lake chips, and that appears to be roughly accurate in this test. The Core i9-11900K was ~15% faster than its predecessor, the 10900K, in the single-core tests. 

However, looking at the multi-core results, the inverse happens and the 10900K is 6.5% faster due to its higher core count. That's actually pretty impressive, though: The ten-core Core i9-10900K has two more cores than the eight-core Core i9-11900K, so we expected a much larger advantage in favor of the chip with two extra cores. Increased IPC truly floats all boats. 

But against the 5800X, the single-core results are much closer, naturally, with Zen 3's much higher IPC performance. Here the 11900K pulls ahead of the 5800X by a mere 4.4%. Strangely though, the 5800X pulls ahead of the 11900K in the multi-core department by 17%, which is a larger delta than we expected because these are both eight-core chips.

This is but one benchmark, though, and several factors could influence the score, including early firmware with the Core i9-11900K. We expect more mature BIOS revisions will be headed out before launch. In either case, these results paint a competitive picture for the desktop PC space soon, one in which price (and supply in light of the shortages) will be exceedingly important.