Tech News - AMD vs Intel
By Zahir, an apprentice software developer here at Epix
At the start of March we added a new server to our server farm to provide additional power and redundancy for you and your engineers. This new server is powered by AMD’s new Epyc CPUs that have taken the world by storm thanks to their unmatched performance.
You may not have heard of AMD but you probably know of Intel due to their market dominance for over a decade. However in February of 2017, AMD released a new line of processors in the consumer space called Ryzen powered by the new ZEN architecture. Ryzen helped redeem AMD’s reputation since they were able to deliver high core count CPUs for a fraction of the price that Intel were selling them for. This forced Intel to drop their prices on their consumer grade chips to compete however they weren’t able to keep their costs down like AMD thanks to the new chiplet design in their ZEN architecture.
Even though 1st and 2nd generation Ryzen did not look to be using chiplet design they used CCX’s which are a group of cores and their caches, which on Ryzen 1st and 2nd generation were placed on one monolithic dye.
AMD also released a enthusiast level CPU called Ryzen Threadripper which, as the name suggests, had a massive amount of cores and threads which Intel could not compete with. The Threadripper CPUs were a great showcase of the ZEN’s chiplet design since they had 4 CCX’s each with 4 cores on each to give a total of 16 cores. At this point Intel may still have had the upper hand in terms of performance when comparing it to a 16 core Intel variant. However the cost of the Threadripper 16 core processor was around half the price.
(Image Credit : Der8auer)
The reason AMD’s chiplet design is so cost effective is because of how it reduces their manufacturing costs due to fact that the manufacturing process is bound to produce CPUs that are faulty and cant be sold. Making one large CPU means that if there is one fault with it then the whole CPU has to be written off which reduces AMD’s yield. Using chiplets means that when there is a defective chiplet only that chiplet has to be discarded rather than the whole CPU, saving AMD a heap of money by driving down the cost of production.
Intel have also been facing problems of their own such as supply shortages for their 14nm CPUs and struggling to reduce the manufacturing process to 7nm which AMD has already achieved. Reducing the size of the manufacturing process can increase efficiency which allows for higher performance. This has helped AMD gain a lead over Intel.
Due to the smaller manufacturing process and AMD’s refinement of the architecture they have been able to catch up to Intel’s Instructions per Clock (IPC), which compares the single core performance of a CPU that is clocked to the same clock speed to compare the efficiency of the CPUs architecture.
In 2019 AMD released a 64 core EPYC server processor that was based on an updated ZEN 2 architecture, while Intel's highest core count CPU is 56 cores that’s estimated to cost around £20,000 compared to AMD’s £7,999. Along with Intel's shortages, that could drive prices up which means that AMD may just have taken the lead back from Intel.
(Image Credit : Forbes)
There are many other ‘good points’ for both platforms however the price to performance of the Epyc line-up is a major selling point and now with the latest generation they are as compettitive or even beating Intel's offering. Other reasons like how all of AMD’s line-up supports 2TB of RAM whereas Intel's forces you to pay more if you want more RAM; or that PCIE 4.0 which is supported on AMD CPUs which offer around double the bandwith and speed as PCIE 3.0 allowing expansion cards such as SSD’s to perform far better than their PCIE 3.0 counterparts.
AMD’s recent innovations have had a major benefit to many sectors from the consumer desktop and laptop market, to the next generation of consoles such as the PS5 and XBOX Series X that will both feature their CPUs and GPUs. Their server line has also had a major impact, such as increasing the limits of supercomputing, cloud computing, and many more. That’s why we chose Epyc, to increase the efficiency and productivity of our software and so that we could innovate along with them. We didn't chose Epyc just because it sounds similar to Epix.