Common HotHardware readers and normal {hardware} fanatics in all probability already know this, however SiSoftware is the creator and maintainer of the SANDRA utility for Home windows PCs. Initially created as a device to easily seize and show system info (as Home windows’ personal capabilities in that regard had been fairly primitive when SANDRA was new), SANDRA is now primarily used as a benchmark suite.
Normally, assessments carried out utilizing SANDRA get uploaded to SiSoftware’s database the place of us can peruse the outcomes and examine them in opposition to different assessments. What this in the end means is that the developer has entry to an amazing database of benchmark outcomes for {hardware} previous and new. Generally, these outcomes embrace assessments carried out on pre-release {hardware}, as you’ve got little doubt seen reported many instances on this very web site.
Many of the leaked specs match the specs we have seen in prior leaks. The preliminary CPUs have eight Raptor Cove P-cores with 2MB of L2 cache, in addition to 4 clusters of “Gracemont” E-cores that share 4MB of L2 cache per cluster. That is a important improve in L2 cache (230%) in comparison with Alder Lake.
SiSoftware marks down a peak clock charge of three.7 GHz for the P-cores on the purported Core i9-13900, whereas the E-cores prime out at 2.76 GHz. The previous is probably going down to those being pre-release processors, however the latter lower-clocked E-cores is perhaps meant to enhance the little cores’ effectivity.
As we talked about, SANDRA is basically used for benchmarking as of late. When testing CPUs, it performs a sequence of tightly-optimized arithmetic operations and gauges their efficiency at these duties. Whereas artificial benchmarks like these aren’t essentially indicative of real-world efficiency, they nonetheless give us an attention-grabbing peek into the potential capabilities of those processors.
The “Arithmetic Native” chart right here measures integer and floating level math efficiency with particular information varieties. These check do not make use of SIMD, so there is no SSE or AVX magic happening right here; these are as fundamental because it will get. You is perhaps shocked to be taught that loads of functions nonetheless use these sorts of directions, so this efficiency is not irrelevant in any respect.
Notably, these assessments are extraordinarily multi-threaded, and that is why we see the Ryzen 9 5900X and its 12 huge cores competing with the faster-on-a-single-thread Intel chips. The Core i9-11900K additionally will get to flex its high-TDP muscle, however not one of the challengers can stand as much as Raptor Lake. Even with what’s presumably diminished clock charges because of being an early pattern, the unreleased CPU dominates in each class, beating the 65W Core i9-12900 by double within the Whetstone double-precision check.
This chart is basically the identical check, however utilizing every chip’s quickest SIMD directions. Which means AVX2 on all the things aside from the Core i9-11900K, which may make the most of unlocked AVX-512 assist to permit it to hold with the twelve-, sixteen-, and 24-core chips within the comparability.
Unsurprisingly, AMD’s chip wins many of the benchmarks right here with its twelve full-power cores. RPL usually beats Alder Lake by a handful of proportion factors, though it loses to the Twelfth-gen chip in INT32, and each thirteenth and Twelfth-gen components fall behind Rocket Lake in INT32 and INT128 testing. In fact, the truth that RPL is usually successful in opposition to previous-generation components regardless of a big clock charge benefit bodes nicely for the efficiency of ultimate silicon.
