The first concrete benchmarks of Intel’s 13th Gen Raptor Lake flagship have surfaced. Featuring a total of 24 cores, including 16 Gracemont (E) and 8 Golden Cove (P) cores, it should be a treat to behold. The higher thread count should result in a substantial performance boost in multi-threaded workloads such as content creation, file compression, and encoding. The leaked data is part of a review posted over at the Chinese site, Bilibili (via: @HXL).
The chip in question is an engineering sample (ES) running at 5.5GHz on an ASUS ROG Maximus Z690 Extreme along with 32GB of DDR5-4000 memory. It features 36MB of L3 cache 2MB of L2 cache per P core and 16MB per E core. The competing Core i9-12900KF was running at 5.1GHz.
In terms of single-threaded workloads simulated by Geekbench, CPU-Z, Cinebench, and WinRAR, the Core i9-13900K is 10-13% faster than its predecessor. While a part of these gains comes from the beefed-up frequency (+400MHz vs the 12900KF), it’s impressive how Intel was able to extract a double-digit ST gain by merely tuning the same architecture (Raptor Lake is essentially a refresh with more cache and cores).
Moving to the multi-threaded segment, here’s where the Core i9-13900K gets to stretch its muscles. In Cinebench R20 nT, it’s a whopping 236% faster than the Core i9-12900KF. While this result is an outlier, it does indicate what Intel’s hybrid core architecture is capable of when the application in question is fully optimized for it.
Other multi-core benchmarks such as CB R23, 7-Zip, and 3DMark get more moderate uplifts, gaining an additional 30-50% performance. It’s worth noting that this isn’t the final retail chip and we should see additional optimizations in action as the release draws near. The boost clocks in particular will be improved by at least 100-200MHz.
