It's been two weeks since Intel launched its 4th Gen Xeon Scalable processor—code-named and further referenced here as Sapphire Rapids. Since then, the major server vendors have announced refreshed portfolios that give insights into how these players view Sapphire Rapids and where it will play in the market.
While some would argue that compute has become commoditized in this cloudified world, I don't necessarily agree. This is evidenced by how each server vendor designs and positions platforms. In this piece, I will quickly cover each server vendor and provide thoughts on the significance of their portfolio refreshes.
Before diving into the details, please note that this review is considering only the product portfolios of Cisco, Dell, HPE, Lenovo and Supermicro. There’s no slight intended on any of the other players in the market; these five server vendors are simply the ones to analyze first.
–Cisco emphasizes flexibility and performance
It’s important to note up front that Cisco is a different kind of server company. Sure, we can say that to some extent about all of the vendors. But Cisco is genuinely unique. Whereas the other major vendors built businesses on servers, Cisco came into this market only after building a dominant presence in infrastructure through its networking equipment.
I point this out because it explains how the company approaches designing and delivering compute platforms. For example, when considering Cisco’s UCS X-Series platform, we are, in many ways, seeing how I believe the future of servers will look—disaggregation of technologies that allows platforms to be configured and upgraded in the most bespoke and economical fashion. And all of it connected by fabric that itself can be upgraded or changed as workloads and requirements evolve.
When it comes to Sapphire Rapids, Cisco has focused its efforts on what it considers the two most important priorities for Enterprise IT—flexibility and performance. At the rack series level, Cisco has updated its UCS C-Series one- and two-socket platforms to address the dynamic nature of modern workloads. Cisco has also updated its UCS X-Series platform to the M7 series, supporting up to four sockets and 16TB of DDR5 RAM, all of which are configured easily through Cisco’s SaaS-based systems management offering, Intersight.
–Dell upgrades PowerEdge performance
Because Dell is a close partner of Intel, one would expect it to fully embrace Sapphire Rapids in the 16th generation of its PowerEdge servers. And it has done that with its mainstream platforms—the PowerEdge R760 (2U) and R660 (1U) servers launched with new management capabilities and security protections.
What's of greater interest to me are two other platform families. The first is the new HS series server (HS stands for hyperscale). The two new platforms, HS5610 (1U) and HS5620 (2U), are designed and built on open standards, allowing cloud service providers and other hyperscalers to deploy, provision and manage these servers with the tools and services of their choice.
I like what Dell is doing with its HS lineup, as it enables the company to more competitively play in the hyperscaler space against an ODM community that has captured this space with bare-bones platforms at razor-thin margins. But those savings can come at the cost of quality. Naturally, the cloud providers have figured out their own economics of deploying at scale, but I would be willing to bet that Dell has identified a need and opportunity in the market with its HS portfolio.
My last thought on Dell's HS lineup is to point out the noticeable absence of AMD EPYC processors. Given the popularity of EPYC in the cloud and the richness of its cores, it seems logical that there would be an AMD alternative to the Intel offerings already mentioned.
The other notable element of Dell's announcement is the refresh of its high-end server offerings—the XE9640, XE8640 and XE9680. These four-socket servers are aimed squarely at the data-rich applications that drive the digital enterprise. Machine learning (ML), big data analytics, ERP platforms and the like that need every cycle of computational capability generate the workloads that put the “critical” in “business-critical”. In this context, it’s no hyperbole to say that deriving actionable intelligence as quickly as possible (i.e., shortening time to value) isn't a luxury—it's a requirement.
These XE servers did not see a refresh when Dell launched its 15th generation of PowerEdge servers because Copper Lake (the previous Intel platform) didn't deliver enough of a performance bump or feature upgrade to warrant it. With the 16th generation, this has changed, and Dell has put a big spotlight on these three platforms. While I haven't seen the performance numbers yet, I expect the rich GPU support combined with Intel's acceleration engines will deliver a considerable performance boost.
Also noted in Dell's announcement were significant upgrades to PowerEdge security and management platforms. I believe Dell has stepped up its protections in the security realm to deliver equal footing relative to its competitors.
On the management front, I believe Dell has also delivered significant improvements. In some ways, the bigger story is that Dell emphasizes the capabilities it delivers through its iDRAC BMC and the OpenManage platform. The company has always had a strong systems management position; it's nice to see this story gaining proper promotion. I’m anxious to see the maturation of Dell’s AIOps strategy.
HPE goes low and high with Xeon
HPE always seems ahead of the market in the server space. While there are a lot of examples to cite, a recent one is the June 2022 launch of the Ampere-based ProLiant RL300 server targeting the hyperscale market. You can read my coverage of that here.
With Intel’s launch of its 4th Gen processors a couple of weeks back, HPE has continued this trend by adding a couple of new SKUs to ensure it is addressing the full range of customer and workload requirements. At the lower end is the ProLiant DL320, a single-socket server that HPE is positioning as a "cost-optimized" virtualization and software-defined platform. I see a lot of play here, especially in some edge environments (formerly known as remote office/branch office or ROBO). With 32 cores, up to 2TB of RAM and support for up to two GPUs, there's a lot of capability to deliver VDI and some back-office operations.
Before getting into the high end of the portfolio, it's worth mentioning that Intel also updated its tower server (ProLiant ML350) to support Sapphire Rapids. While this two-socket server is aimed at the small-to-midsize (SMB) market, I think it has a lot of utility beyond that. Like the DL320, the ML350 is a nice platform for the branch office environment where no racks are available for mounting a smaller form factor. And the richness of features can nicely support a small operation—perhaps a local pharmacy or gas station.
In addition to updating its ProLiant portfolio with Sapphire Rapids, HPE added a couple of new SKUs to its Cray lineup. The EX2500, also based on Sapphire Rapids, is the latest addition to the family. This direct liquid-cooled (DLC) system will power the Crossroads supercomputer at the Los Alamos National Laboratory. It is essential to understand the significance of this. While there are many conversations underway around whether Sapphire Rapids will compete on performance (and I've seen many "in your face" comparisons), organizations like Los Alamos don't choose a platform for tracking the national nuclear stockpile on the basis of politics. Or bias. Or “legacy.” Sapphire Rapids has important performance characteristics of its own, and the way HPE exploits those capabilities has relevance.
Since we’re talking about using the right chip for the right task, it’s worth noting that Nvidia has switched from EPYC to Xeon (Sapphire Rapids) to power its DGX platform. This is also a pretty big deal. The DGX platform, powering AI workloads, is highly performant. I suspect the selection comes down to two factors: Intel's huge software ecosystem and Sapphire Rapids' acceleration engines.
Also of note is that HPE has pulled its Apollo line under the Cray umbrella. The Apollo 2000 and 6500 are now called the XD2000 and XD6500. (Note that these also have EPYC variants.)
While we talk about HPE addressing the high and low ends of the computing spectrum in its announcement, the company also did the obligatory enterprise positioning by publishing a VMmark world record for the DL380. VMmark is a VMware benchmark that measures how a platform handles the virtualized workloads supporting the enterprise.
Lenovo + Sapphire Rapids = 25
Lenovo has experienced robust growth in the datacenter over the last couple of years under the leadership of president Kirk Skaugen and CMO Susan Blocher. In addition to its substantial presence in supercomputing and the cloud, Lenovo’s ThinkSystem servers and ThinkAgile solutions have become more common in enterprise datacenters.
I think a good reason for the team's success is simple: when this team goes in on a concept, it goes all in. This is undoubtedly the case with its support for Sapphire Rapids, given that it announced a refresh to 25 different platforms spanning the entirety of the compute needs in the datacenter. No half measures, no workloads off-limits.
In its positioning, Lenovo talks about its Infrastructure Solutions V3 portfolio across four focus areas—mission critical, AI, HPC and containerized (i.e., hybrid multi-cloud). ThinkSystem V3 is the server portfolio supporting mission critical, AI and HPC through platforms like the four-socket ThinkSystem SR850 and the SD650 V3 Neptune DWC (direct water cooling) and SD650-I V3 Neptune DWC.
On the cloud and containerized front, Lenovo has updated its ThinkAgile HX, MX and VX solutions (for Nutanix, Microsoft and VMware, respectively). These V3 platforms drive integrated solutions that enable organizations to deploy hybrid, multi-cloud environments—along with containerized workloads that shift constantly from on-prem to off-prem—quickly and easily.
And, as we’ve come to expect, all of these servers and solutions are available through traditional purchase or as-a-service via Lenovo’s TruScale model.
A couple of thoughts on Lenovo's announcement. First, I like it that the company is not trying to narrowly define where its Intel platforms stand relative to its AMD servers. Build the platforms and let customers choose. With that said, there are areas where Intel operates that AMD has decided to sit out, such as the four-socket platforms that are pretty narrowly focused on data-heavy workloads like ERP.
Second, I like it that Lenovo, like Dell, has stepped up its game in positioning its security and manageability capabilities. ThinkShield is a comprehensive security platform that is foundational for a zero-trust environment. The company has baked security into the design and manufacturing of its infrastructure. And Lenovo System Guard ensures the integrity of platforms from manufacturing to shipping, deployment and provisioning.
As with security, Lenovo got a bit bolder with its V3 announcement, touting the benefits of the XClarity controller. The company claims that IT organizations using it can realize a 40% reduction in time-to-operation when deploying infrastructure. While that top-line number is impressive, it also means fewer IT resources are required to perform the mundane tasks that hamper IT from transforming operations.
With Lenovo Open Cloud Automation (LOC-A), the company is claiming an 81% reduction in time-to-value for cloud deployments by automating the provisioning of bare metal, virtualized and containerized environments.
I like how Lenovo quietly went about the transformation of its datacenter infrastructure portfolio and managed its position in the market. As the company's portfolio and capabilities have grown, so too has its positioning—to the point where it now has a portfolio of products and services that can compete with any server vendor, not to mention a supply chain and manufacturing story that may offer it some separation from its competitors.
Last but certainly not least, Supermicro
We can’t review the big server manufacturers without talking about Supermicro. While the company may not match the brand awareness of its competitors, it has a server portfolio that is perhaps the largest in the industry. And while many tier-one server vendors have pulled back from the cloud and hyperscaler markets, Supermicro has gone all in.
With the rollout of its X13 Server Solutions, Supermicro launched an incredible 50 Sapphire Rapids platforms across 15 solution “families.” Essentially, any use case you can think of, any deployment scenario—Supermicro has a server platform for you. AI? Digital twins? Virtualized infrastructure? Edge? Yup, yup, yup, yup—from a single node up to eight nodes.
As with Lenovo, I appreciate that Supermicro doesn't discriminate regarding platform positioning. Build platforms targeting use cases and deployment scenarios, then let the customer choose which platform fits its needs best.
Supermicro's aggressiveness in the server market has paid off as the company continues to grow its market share. While this has primarily been in the more commoditized cloud and hyperscaler markets, I would expect the company to start to see gains in the enterprise space as well, especially in those markets where integrated solutions are sold to address specific workloads.
As an analyst in the server market, choosing a favorite vendor is much like choosing a favorite child. Even if you have one, you'd never admit it. However, each vendor discussed here does have a unique value proposition it brings to the equation; each is a solid choice for any IT organization.
It's good to see so much innovation taking place in the server community. For those that think a server is simply a server, it’s time to think again. Each one of these vendors brings value to the market. And they all look to exploit the unique features of new CPUs such as Sapphire Rapids that come to market.
Keep in mind, there are plenty of server vendors I didn't cover simply in the interest of space—Inspur, Quanta, . . . the list goes on. No slight intended toward any of these vendors, and please let me know if there is a vendor you'd like to see covered.