AMD Claims Businesses Pay 'Intel Tax' With Xeon Server CPUs
AMD is claiming that customers of its largest rival are paying an "Intel tax" when they use Intel's market-dominating Xeon processors in data centers.
The Santa Clara, Calif.-based company recently released a white paper on its website to promote its AMD EPYC server CPUs and criticize its competitor for what it calls the "Intel tax," a set of costs associated with "self-imposed, designed-in performance bottlenecks" that allegedly force Intel's customers to buy more expensive Xeon processors.
[Related: AMD Fares Better Than Intel On Wall Street As It Steals CPU Market Share]
"It is the extra price for Intel processors that you have to pay to get the features and performance you need," the white paper says.
The white paper was released ahead of Intel's Data Centric Innovation Day on April 2, when the semiconductor giant is expected to make new announcements around Xeon and other server products.
An Intel spokesperson said its response will come with next week's event.
Andrew Kretzer, director of sales and marketing at Bold Data Technology, a Fremont, Calif.-based custom system builder, said he agrees with AMD's message, which called out Intel for its “maze of more than 42 different processor SKUs, 4 different metal codes, and 5 different series.”
"Not only is Intel hamstringing its processors in order to create artificial, slotted SKUs (and at extremely high price points), but they’ve added an incredible amount of complexity to the purchasing process for potential customers," he said in an email. "We’ve had to create a tool to help customers with understanding what they are buying and what they need when it comes to Intel Xeon processors and that tool has gotten more complex with each new family of Xeons."
However, he said, the added complexity isn't necessarily a benefit for AMD, with customers instead opting to buy older Xeon products from Intel, such as the Broadwell line.
"The majority of these customers are still clamoring for Broadwell parts in large part to avoid the brain cramp and price gouge brought upon from this newer line," he said. "Unfortunately for AMD, it seems that these customers are looking to older generations of Intel Xeon over an arguably better AMD experience, reinforcing the validity of that old chestnut that 'nobody ever got fired for buying IBM.'"
Dominic Daninger, vice president of engineering at Nor-Tech, Burnsville, Minn.-based high-performance computing systems builder, said AMD's return to the server market with EPYC is good for competition, which will ultimately benefit customers while also pushing Intel
"I know particularly on single socket, AMD has got some attractive competition," he said, referring to AMD's single-socket EPYC solutions, "and they're always competitive on price."
Daninger said while the AMD white paper makes some good points, such as the costliness of Intel's higher-end Xeon CPUs and AMD's strong I/O capabilities, AMD's argument about more cores per processor isn't necessarily relevant for certain markets, such as high-performance computing.
For high-performance computing, faster clock speeds are often more important than higher core counts, Daninger said, and the former is where Intel has done an especially good job. This is relevant in high-performance computing when it comes to commercial software that charge licenses on a per-core basis, meaning that it's important to maximize performance within each core.
"The thing that Intel is consistently done is having higher per-core performance," Daninger said. "In the HPC business, oftentimes a customer will elect to go with fewer cores and higher clock speed."
In AMD's white paper, the company said that businesses can save money on software licenses by choosing its single-socket processor or a processor with fewer cores without having to sacrifice on features like memory capacity and bandwidth.
The white paper makes price and performance comparisons between Intel's Xeon processors and Amd's EPYC processors, showing that AMD's CPUs have better or similar performance while costing less. However, the white paper only makes the comparison using one benchmark, SPECrate2017_fp_base, which is a standard test for compute-intensive performance. The document also doesn't consider Intel's wider selection of Xeon processors for comparison.
AMD has been making gains against Intel in the server market, especially among enterprises, according to a recent survey. But even with the company set to beat Intel to the market this year with processors based on a next-generation manufacturing process, AMD still has a lot of hurdles to clear.
"You're going to have a hard time ripping Intel out of an enterprise," Bob Venero, CEO of Holbrook, N.Y.-based IT service provider Future Tech, told CRN earlier this month.