In an interview with CRN, Dan McNamara, general manager of AMD's server business unit, talks about how he's 'trying to do more and more' with channel partners to grow EPYC CPU sales in the enterprise and, increasingly, the midmarket and SMB segments.
The leader of AMD's EPYC server CPU business said he's "trying to do more and more" with channel partners to fight Intel's dominance in data centers across the enterprise and, increasingly, the midmarket and SMB segments.
"We've expanded our reach with partners, like the VARs and the entire channel because their reach and their influence are at the top banks all the way down to the smallest SMBs," said Dan McNamara, general manager of AMD's server business unit, in an October interview with CRN.
[Related: Forrest Norrod On How AMD Is Fighting Nvidia With 'Significant' AI Investments]
"I'm trying to do more and more with them because there's so many influence points at the enterprise," added McNamara.
While AMD launched its latest challenge to Intel's Xeon brand with the fifth-generation EPYC "Turin" processors two months ago, McNamara said the chip designer is "still ramping" with channel partners the previous generation between the general-purpose "Genoa" chips, cloud-focused "Bergamo" chips and edge-focused "Sienna" chips.
"I think Genoa has upped the ante. We've expanded our share pretty dramatically with Genoa," said McNamara, who was an Intel executive prior to joining AMD in 2020.
"And interestingly enough, Bergamo has been a bigger winner in enterprise than I expected, because we built Bergamo for cloud-native [applications]. But there's a bunch of workloads that people are using Bergamo in the enterprise now too, which is exciting," he said.
Since AMD launched Genoa a little more than two years ago, the company's x86 server CPU market share has grown 6.7 points to 24.2 percent against Intel, according to CPU-tracking firm Mercury Research. While the company's share movements have slowed down in the past few quarters, they have nevertheless brought AMD closer to the record 26.2 percent share it hit in 2006 during a previous era of server CPU efforts.
Combined with a "strong ramp" of AMD's Instinct data center GPUs, EPYC processor sales helped the company grow data center revenue to a record $3.5 billion in the third quarter, eclipsing Intel's revenue for the same segment by 4.5 percent.
"We're seeing a significant amount of EPYC interest," said Dominic Daninger, vice president of engineering at Burnsville, Minn.-based AMD system integration partner Nor-Tech.
By working more closely with channel partners, McNamara said, AMD is "breaking habits" of enterprise customers who previously relied on Intel CPUs for their data centers.
"I always tell the team, I've been saying to them for two years: We're breaking habits in the enterprise, because you got age-old IT guys getting serviced by long-standing VAR salespeople, and it just happens," he said.
McNamara said his strategy to continue growing EPYC's market share has been to "go win the tough workloads" -- like electronic design automation, which is used to design chips, along with other "vertical workloads" -- and "then spread it into general IT."
The strategy is "going quite well," according to the executive.
"Our field team completely redeployed on the top what I would say Forbes 2,000 [companies], so we're very focused on top accounts, and we're working with our OEM partners, like [Hewlett Packard Enterprise]," McNamara said.
"When you think of the top banks, think of the Verizons, AT&Ts. Think of Ford, [General Motors], Honda. So breaking in across all the different verticals very, very, very well. So we're excited about the enterprise," he added.
While it takes longer for AMD to break into enterprise accounts than it is with cloud service providers -- where its EPYC CPUs are supported by more than 950 public instances -- McNamara said one way the company makes headway is with proofs of concept.
"Our POCs have just exploded. And what we found is, if we can get a POC and get an IT group to actually look at us in detail, we win," he said.
McNamara credited his business unit's field application engineer team, which is led by Corporate Vice President of Engineering Solutions Laura Smith.
"She's been driving that very, very aggressively in the POC front," he said.
But AMD's server business unit is also putting more resources into winning deals with customers in the midmarket and SMB segments.
"There's a lot of money [in the enterprise], but the midmarket and SMB: there's also a lot of units too, yeah, so we want to go after that," McNamara said.
McNamara said his business is doing so "with a more clear channel strategy."
"I'm going to spend more of my time there too, just to make sure that people are aware of what we're trying to accomplish and things like that," he said.
To help AMD win server deals across the enterprise, midmarket and SMB segments, McNamara said he has built a channel development team in his business unit to support the chip designer's salespeople who are working with partners ranging from GSIs to VARs.
"Every region has the top VARs. And then every OEM has the top VARs. They're the top channel partners. So it's really just putting that matrix together and creating real plays that our sales guys can continue to drive. These aren't silver bullets. It's almost like blocking and tackling, but you have to do it," he said.
Two of the AMD server platforms McNamara sees as big plays with channel partners in the midmarket and SMB segments are HPE's Sienna-powered ProLiant DL145 edge server and Cisco's Genoa-powered UCS X215c blade server.
While AMD is starting to push its fifth-gen EPYC Turin chips in new servers, McNamara sees fourth-gen chip families like Sienna and Genoa as important for the foreseeable future.
"Those are high-value products that we're going to sell for a long, long time," he said.
When AMD launched the fifth-gen EPYC Turin processors in October, it called the lineup the "world's best server CPU for enterprise, AI and cloud."
Based on the performance-focused Zen 5 and density-focused Zen 5c architectures, the processors sport up to 192 cores and, in separate models, a 5GHz frequency.
Compared to the flagship, 64-core CPU in Intel's fifth-gen Xeon lineup that launched in late 2023, AMD's top 192-core EPYC CPU is up to four times faster performance for video transcoding, as much as 3.9 times faster for science and high-performance computing workloads, and 60 percent faster in virtualized infrastructure, according to AMD.
These processors are compatible with the same SP5 socket that is supported by the Genoa and Bergamo chips from the fourth generation.
When CRN interviewed McNamara, it was before Intel's Xeon 6 processors with E-cores and P-cores were widely available, but upon Turin's launch, the executive said at the time that early third-party testing indicated that AMD will remain competitive.
"We think with Turin we still maintain a very healthy lead in terms of overall performance and energy efficiency and [total cost of ownership]," he said.
McNamara said he sees Turin being a big deal for enterprise customers looking to scale up their data center architecture.
"I think as the enterprise expands core counts, Turin is going to be there [for what we call] the scale-up play. [As for] the classic Turin, 48-, 64-, 96-core at some of these top verticals is really where I think we'll play," he said.
Executives at AMD system integration partners said they are seeing "significant" interest in EPYC for different kinds of server applications. But while two of them said AMD's channel support has been solid, another noted that there is room for improvement.
Alexey Stolyar, CTO of Northbrook, Ill.-based International Computer Concepts, said he's been "seeing traction" around AMD's new Turin processors, particularly the ones with higher clock speeds for high-frequency trading customers.
"They added some interesting technologies into it that make it interesting for a lot of customers. So we've been seeing it in the financial segment specifically. So that's one that we focus on. And there's a lot of activity around Turin right now," he said on Wednesday.
For example, AMD has improved the core-to-core latency with the Turin generation, which makes them better for high-frequency traders, according to Stolyar.
While Stolyar said high demand for AI computing infrastructure has taken away interest for traditional, CPU-based servers for the past year or so, Turin has started to change that.
"No one's really been focusing on CPU or really talking, 'Hey, let's update our general CPU infrastructure.' Turin, there's noise and momentum around that," he said.
The leader of an HPC system integrator said on Thursday that he has seen AMD continue to win on the CPU side when it comes to AI and HPC cluster deals.
"The demand is very good for AMD CPUs in as a supporting role, so to speak, in an Nvidia GPU system. And of course, in HPC, it is extremely strong because of the lead AMD established when Intel faltered bringing Ice Lake to the market," said the executive, who asked to not be named to speak candidly.
The system integrator leader is referring to the multi-year delay Intel faced in releasing its third-generation Xeon processors, which started to hit the market in late 2020.
While customers were initially reticent about using EPYC processors following previous missteps by AMD in the server CPU market, the fact that EPYC supported PCIe Gen 4 before Xeon did forced customers to make the switch, according to the executive.
Customers were pleased with the results.
"People basically said, 'Wow, this gives me a much better price performance, or any metric: performance per watt, performance per dollar, or in many cases, simply because Intel was not an alternative [due to not supporting PCIe Gen 4 at the time]," he said.
The system integration leader said Intel may have closed the gap for the most part with its Xeon 6 processors, but he said the company's current struggles, including the abrupt retirement of CEO Pat Gelsinger, have not helped with customer perception.
"AMD's channel strategy is very good compared to Intel in terms of the various programs if you're a covered account and such, so I suspect that the momentum, at least in the market where we serve which is HPC and AI, will continue," he said.
Daninger, the engineering executive at Nor-Tech, said on Wednesday the high interest in EPYC is reflected in two customer projects for Bergamo-based HPC clusters, both of which are worth more than $1 million each. One of the projects is with a building materials manufacturer while the other is with an aerospace company.
"It's usually a pretty good value. That's some of the things [customers] look for is the dollars per core, that kind of thing. And then we've had a lot of them jump in and run evaluations, and they've been impressed with what they see," he said.
But despite the traction Daninger has seen with EPYC-based systems, he said AMD's support for Nor-Tech hasn't been as great as it used to when it comes to having access to field application engineers or market development funds.
"Even with Intel now, we still have direct contact with field application engineers, and that kind of thing is helpful. And AMD just seems to run a slimmer operation," he said.
An AMD spokesperson told CRN on Thursday that the company is "always open to feedback from our channel partners as we are continuously evaluating what is working to provide the best impact for sales and marketing of our high-performance computing products in the channel."