AMD and Intel Race Towards High Core Count CPU Future
As we prepare for a surprisingly robust summer season of new hardware technologies to be released to the consumer, both Intel and AMD have moved in a direction that both seems inevitable and wildly premature. The announcement and pending introduction of high core count processors, those with many cores that share each company’s most modern architecture and design, brings with it an interesting combination of opportunity and discussion. First and foremost, is there a legitimate need for this type of computing horsepower, in this form factor, and secondly, is this something that consumers will want to purchase?
To be clear, massive core count CPUs have existed for some time but in the server and enterprise markets. Intel’s Xeon line of products have breached the 20-core count in previous generations and if you want to dive into Xeon Phi, a chip that uses older, smaller cores, you will find options with over 70 cores. Important for applications that require a significant amount of multi-threading or virtualization, these were expensive. Very expensive – crossing into the $9000 mark.
What Intel and AMD have begun is a move to bring these high core count products to consumers at more reasonable price points. AMD announced Threadripper as part of its Ryzen brand at its financial analyst day, with core counts as high as 16 and thread counts of 32 thanks to SMT. Then at Computex in Taipei, Intel one-upped AMD with its intent to bring an 18-core/36-thread Skylake-X CPU to the new Core i9 lineup. Both are drastic increases over the current consumer landscape that previously capped out at 10-cores for Intel and 8-cores for AMD.
Let’s first address the need for such a product in the world of computing today. There are many workloads that benefit easily from multi-threading and consumers and prosumers that focus in areas of video production, 3D rendering/modeling, and virtualization will find single socket designs with 16 or 18 cores improve performance and scalability without forcing a move to a rackmount server infrastructure. Video encoding and transcoding has long become the flagship workload to demonstrate the power of many-core processors. AMD used that, along with 3D rendering workloads in applications like Blender, to demonstrate the advantages of its 8-core Ryzen 7 processors in the build up to their release.
Other workloads like general productivity applications, audio development, and even PC gaming, are impacted less by the massive core quantity increases. And in fact, any application that is heavily dependent on single threaded performance may see a decrease in overall performance on these processors as Intel and AMD adjust clock speeds down to fit these new parts into some semblance of a reasonable TDP.
The truth is that hardware and software are constantly in a circular pattern of development – one cannot be fully utilized without the other. For many years, consumer processors were stuck mostly in a quad-core rut, after an accelerated move to it from the single core architecture days. The lack of higher core count processors let software developers get lazy with code and design, letting the operating system handle the majority of threading operations. Once many-core designs are the norm, we should see software evolve to take advantage of it, much as we do in the graphics market with higher performance GPUs pushing designers forward. This will lead to better utilization of the hardware being released this year and pave the road for better optimization for all application types and workloads.
From a production standpoint Intel has the upper hand, should it chose to utilize it. With a library of Xeon parts built for enterprise markets already slated for release this year and in 2018, the company could easily bring those parts to consumers as part of the X299 platform rollout. Pre-built, pre-designed and pre-validated, the Xeon family were already being cannibalized for high-end consumer processors in previous generations, but Intel capped its migration in order to preserve the higher prices and margins of the Xeon portfolio. Even at $1700 for the 10-core 6950X processor, Intel was discounting dramatically compared to the Xeon counterpart.
Similarly, AMD is utilizing its EPYC server product line for the Threadripper processors targeting the high-end consumer market. But, AMD doesn’t have large market share of workstation or server customers to be concerned about cannibalization. To them, a sale is a sale, and any Ryzen or Threadripper or EPYC sold is an improvement to the company’s bottom line. It would surprise no one if AMD again took an aggressive stance on pricing its many-core consumer processors, allowing the workstation and consumer markets to blend at the top. Gaining market share has taken precedent over margins for AMD; it started as the initiative for the Polaris GPU architecture and I think it continues with Threadripper.
These platforms will need to prove their value in the face of dramatic platform requirements. Both processor vendors are going to ship the top performing parts with a 165-watt TDP, nearly double that of the Ryzen and Kaby Lake desktop designs in the mainstream market. This requires added complexity for cooling and power delivery on the motherboard. Intel has muddied the waters on its offering by varying the number of PCI Express lanes available and offering a particular set of processors with just four cores, half the memory channels and 16 lanes of PCIe, forcing platforms into convoluted solutions. AMD announced last week that all Threadripper processors would have the same 64 lanes of PCIe and quad-channel memory support, simplifying the infrastructure.
With that knowledge and assumption in place, is higher core count processing something that the consumer has been asking for? Is it just a solution without a problem? The truth is that desktop computers (and notebooks by association) have been stuck at 4-cores in the mainstream markets for several years, and some would argue artificially so. Intel, without provocation from competent competing hardware from AMD, has seen little reason to lower margins at the expense of added performance and capability in its Core line. Even the HEDT market, commonly referred to as the E-series (Broadwell-E, Ivy Bridge-E and now Skylake-X) was stagnant at 8-cores for longer than was likely necessary. The 10-core option Intel released last year seemed like an empty response, criticized as much for its price ($1700) then praised for its multi-threaded performance.
AMD saw the opportunity and released Ryzen 7 to the market this year, at mainstream prices, with double the core count of Intel Core parts in the sub-$400 category. The result has been a waterfall of an effect that leads to where we are today.
Yes, consumers have been asking for higher core processors at lower prices than they are currently available. Now it seems they will have them, from both Intel and AMD. But pricing and performance will have the final say on which product line garners the most attention.