AMD’s 3D V-Cache Boosted EPYC 7V73X Milan-X CPU Delivers Up To 12.5% Performance Increase Over Standard Milan
The previous benchmarks from tech outlet, Chips and Cheese focused entirely on the latency performance of the upcoming Milan-X CPUs such as the flagship EPYC 7V73X. This time, they have published more latency benchmarks alongside overall bandwidth and workload-specific benchmarks against the EPYC 7763 CPU. One of the more interesting comparisons added to their test spree is Intel’s Ice Lake and Cascade Lake Xeon chips. Now how are these interesting you might ask? Well, the Intel CPUs are based on a monolithic design that utilizes a Mesh architecture for interconnect while AMD uses a Ring bus architecture. The Ring bus architecture is designed for lower latency and higher-bandwidth design while Mesh designs trade these in favor of enabling a more scalable chip design such as what Intel’s been on over the past few years. As such, despite featuring a smaller L2 cache, AMD’s Milan-X and Milan CPUs are lower latency and also dish out higher bandwidth thanks to increased L3 cache over their competitors from Intel. But let’s talk about the workload-specific benchmarks. Once again, the AMD EPYC 7V73X Milan-X CPU is used. The flagship AMD EPYC 7V73X CPU will rock 64 cores, 128 threads and feature a maximum TDP of 280W. The clock speeds will be maintained at 2.2 GHz base and 3.5 GHz boost while the cache amount will drive up to an insane 768 MB. This includes the standard 256 MB of L3 cache that the chip features so essentially, we are looking at 512 MB coming from the stacked L3 SRAM which means that each Zen 3 CCD will feature 64 MB of L3 cache. That’s an insane 3x increase over the existing EPYC Milan CPUs. In four of the five benchmarks published by Chips and Cheese, the Milan-X EPYC 7V73X CPU saw a favorable win in four benchmarks. It only loses out to the EPYC 7763 in OpenSSL since the workload doesn’t stress the cache at all and because Milan-X CPU runs at a slightly lower but sustained clock when pushing all CCD’s, the performance loss is to be expected. In Gem5, the Milan-X chip sees a performance increase of 7.6% while running at 5% lower clocks compared to the standard Milan CPU. This shows an iso-clock V-Cache performance increase of 12.5% which is really good from the added V-Cache. AMD Milan-X ‘EPYC 7V73X’ vs Milan ‘EPYC 77633’ CPU Performance Benchmarks (Image Credits: Chips and Cheese): Other benchmarks also show similar performance increases. For Y-Cruncher, the AMD EPYC 7V73X Milan-X CPU delivers a 1.5% performance increase since this is a very FPU and memory-dependent test but it is explained that while the Milan-X CPU dropped down to 1-CCD boost speeds while standard Milan didn’t, the 3D V-Cache made up for the clock loss. With all this said Chips and Cheese state that the 3D V-Cache benefits are real for AMD’s EPYC Milan-X CPUs and they can’t wait to see what comes next for V-Cache technologies. Their conclusion is provided below but I would recommend you head over to their page to read the detailed report on Milan-X.
AMD EPYC Milan-X 7003X Server CPU (Preliminary) Specs:
V-Cache is also a large technical achievement for AMD considering that the latency increase of 3 to 4 cycles is trifling considering the tripling of the L3. On the bandwidth front, while AMD did sacrifice Milan’s lack of variation of L3 bandwidth for around 25% more single threaded bytes per cycle bandwidth but 10% less whole CCD bandwidth with Milan-X, that is similar behavior to previous generation server CPUs from AMD and even with the slight bandwidth reduction, it blows Intel’s L3 bandwidth out of the water. All in all, V-Cache is both a very interesting technology and a decent performance enhancer which makes me very excited for what AMD does next with the technology. via Chips and Cheese
