Intel xe up to 500w
Intel released a new set of benchmarks to emphasize that its Xeon Class Cascade LakeAP processor is more powerful than AMD's equivalent EPYC Rome CPU. Test
places Intel's dual-socket Xeon Platinum 9282 and dual-socket AMD EPYC 7742 side by side. According to Intel, the main conclusion of the test is that the Xeon Platinum 9282 is a staggering 84 level higher than AMD's EPYC 7742. Intel also claims that the benchmark test successfully replicated real-world performance.
There are some bold statements, a little deeper, that suggest that Intel has taken a very dishonest approach.
First, this comparison is not entirely fair. Intel Xeon Platinum 9282 has 56 cores, 112 threads, 14 nm compute nodes, 77 MB cache, 2.6 GHz base clock speed, 3.8 GHz acceleration, 12 memory lines, 400 W TDP, and sells for up to $50,000.
On the other hand, AMD EPYC Rome has 64 cores, 128 threads, 256 MB cache, 7 nm compute node, 2.25 GHz base clock, 3.4 GHz Boost, 225 W TDP, 128 channel PCIe Gen 4, 8 memory channels, It sells for only US$7,000.
Just looking at price and TDP, we are dealing with two very different beasts.
Secondly, the GROMACS version used for benchmarking is actually an outdated version, not optimized for the Zen 2 architecture, especially due to the lack of the 256-bit wide AVX2 SIMD instruction set. The latest version 2019.4 clearly supports AMD EPYC Rome CPU.
Here, we see that Intel uses a version that benefits its products and penalizes AMD chips outside the starting line, so the comparison is not completely comparable to real-world performance as claimed by the microchip giant.
Third, the claim that the Xeon Platinum 9282 is 84 is a bit suspicious, because it represents top speed, and EPYC Roma has increased by an average of 31%, again using outdated utilities.
Intel also explained that despite having a higher TDP, the Xeon chip is more efficient because it requires fewer nodes to achieve the same level of performance, and states that the total cost of ownership is lower as well.
Finally, the benchmark also focused on memory bandwidth, identified it as a key performance element, and praised the advantages of Xeon chips in this process. Interestingly, real-world applications show that AMD chips provide better bandwidth, and companies like Netflix have chosen to switch to AMD due to the significant increase in bandwidth, and the total cost of ownership is similar to that of the CPU. from Intel.
Both companies will launch new flagship products in 2020, especially AMD 7nm CPU + EPYC Milan, and Intel Cooper Lake 14nm and Ice Lake 10nm chips.The competition will not fade anytime soon.