2nd Gen Intel Xeon Scalable Refresh Market Impact
Due to the size and scope of the refresh SKU list, Intel is clearly changing the competitive landscape of its refresh v. pre-refresh parts. The other obvious comparison is to the AMD EPYC 7002 series. At the same time, we believe this refresh can make inroads far beyond the two obvious competitors. It will impact the Xeon W-3200 and Xeon W-2200 series as well as the Xeon E-2200 series which is performance competitive with the Xeon Silver line. Finally, for every future Arm competitor, their point of comparison will need to change.
2nd Gen Intel Xeon Scalable Refresh v. Pre-Refresh Xeon Scalable
If you need 4-socket or 8-socket servers, the pre-refresh parts are for you. Consider these higher scale-up server parts much like the Xeon E7 series was previously. it is easiest to see these as Intel launched the Xeon E7 generation in 2019, now we get the analogous Xeon E5 generation at a lower cost in 2020.
In the two-socket markets, it will be extraordinarily difficult to purchase non-R SKUs outside of a few segments such as the Xeon Bronze 3204. The Xeon Silver performance bumps are great but perhaps the 100W v. 85W TDP will cause a segment of the market to still buy the 2019 generation parts.
For the mainstream dual-socket market, not impacted by high-memory parts, the new Gen 2R Intel Xeon Scalable (our term, not Intel’s) parts are clearly the way to go. Unless you need an L SKU for over 1TB per socket or you need 4/8 socket capabilities, the new Intel Xeon “Refresh” SKU stack is just the way to go. An example of the Xeon Gold 6230R to the previous generation Gold 6230 and Gold 6130 shows why. Intel is drastically increasing performance per dollar in this generation.
Again, given the feature differentiation, we think that Intel could have called this more than a “refresh” or additional SKUs and instead called it a new product line akin to the Xeon E5-2600 V1-V4 generations for mainstream servers versus the previous generation being closer to the Xeon E5-4600/ Xeon E7 lines.
For STH readers, the Xeon Gold R SKUs are a much better value than the older Xeon Gold and Platinum SKUs. It is awesome that Intel is doing this.
2nd Gen Intel Xeon Scalable Refresh v. AMD EPYC 7002
For AMD, the road just got tougher. Contrary to many beliefs out there, Intel does not have to meet or beat AMD’s performance per dollar in markets primarily focused on compute. Intel just needs to be close.
There are a few costs that one overlooks when looking at CPU prices only. For example, a ~$50 “Lewisburg” C62x PCH goes in every dual-socket Intel Xeon Scalable server. Intel also sells a ton of NICs and SSDs that go to big OEMs and end customers. Intel has FPGAs and other accelerators as well as Optane DCPMM.
Intel’s strategy is to create a giant silicon portfolio, then be able to bundle across. For AMD, it can have a better CPU, but if you have $2000 of CPUs in a server and $20,000 of SSD storage, Intel can make up CPU discounting on the bundle. It can also do this at a higher-level giving performance incentives across its portfolios to OEMs and channel partners.
Intel still does not have a great answer for the 64-core AMD EPYC parts since on a per-node basis Intel needs to get to 4-sockets to hit the same levels of performance. Still, for the mainstream 8-32 core markets, Intel now has a “good enough” answer, and that is going to force AMD to respond.
While Intel has AVX-512 and VNNI, it also expanded Optane DCPMM support down to a Silver-level SKU. That will further put pressure on AMD as Intel has a very good feature that is now more accessible even if it is simply as low latency and fast persistent storage.
AMD’s strengths are in PCIe Gen4 and having more memory bandwidth and capacity for mainstream SKUs. It also has an architecture that utilizes (relatively) a ton of cache which can make up for a lot of the clock speed delta Intel is pushing in this generation. AMD is the better value looking at CPUs, but one needs to really look at post-discounting system-level costs to see the benefit.
2nd Gen Intel Xeon Scalable Refresh v. Intel Xeon W-3200
The Intel Xeon W-3200 series is now in a very strange predicament. Here is why. With the new price changes the SKUs in this series are now going to see a lot of pricing pressure. For example, the Intel Xeon W-3275 can have 64 PCIe lanes if the motherboard supports it which is great for a 28-core part that can hit 4.5GHz. Still, one can save around $500 and get the Xeon Gold 6258R and get a lower top turbo chip with other benefits. The Gold 6258R can be used in dual-socket servers which means more scalable performance and lower costs from using fewer systems. It can also support Optane DCPMM. If you wanted a high-end 28 core CPU with Optane DC Persistent Memory, the Gold 6258R is now a considerably better choice than the Xeon W-3275.
There is a chance this market will not care, or simply wants the higher low-core count utilization turbo numbers of the Xeon W-3275. On the other hand, Intel once again has very viable dual Xeon workstation models and pricing with the R parts.
2nd Gen Intel Xeon Scalable Refresh v. Intel Xeon W-2295
That discussion of the R parts moves down to the Xeon W-2200 series. Our review of the Xeon W-2295 is going to be online in the next few weeks, but this new “refresh” launch puts pressure on these chips as well.
The Intel Xeon W-2265 is a $944 part, with 12 cores a 3.5Ghz base clock and a 4.6GHz turbo clock. How does that stack up compared to the 16 cores 2.9GHz to 3.9GHz Xeon Gold 6208U with 50% more memory channels and bandwidth and DCPMM support?
Likewise, we will explore how the Xeon W-2295 at 18 cores compares to the 20 and 24 core Xeon 5218R and Xeon 5220R that sacrifice clock speed for cores and platform features.
The question Intel needs to answer for the Xeon W-2200 series is why should one invest in this platform versus just using LGA3647 which will be more robust.
2nd Gen Intel Xeon Scalable Refresh v. Intel Xeon E-2200
This may seem strange, but there is a good reason for this one. The Xeon E-2200 series like the Intel Xeon E-2288G can now hit 8-cores for $539 and at very high clock speeds. Still, the platform is limited to ECC UDIMMs instead of RDIMMs, and one must buy a node for every 8 cores and 95W TDP of CPUs. Now, one can purchase a 26 core Intel Xeon 6230R for $1,894 with a 150W TDP or a 24-core Xeon Gold 5220R for $1,555. While one loses per-core performance, it is going to be immensely less expensive to consolidate to dual Xeon Gold systems and virtualize than using Xeon E-2200 series servers.
Even the Xeon Gold 4215R with its 8 cores and Xeon Silver 4214R with 12 cores in the sub $750 market are going to make sense for users who want lower core counts and TDP, but also want more memory and platform I/O. A Xeon Scalable single-socket server costs more than a Xeon E-2200 platform, but it also has a lot more I/O and consolidation benefits that can reduce the impact in some segments.
2nd Gen Intel Xeon Scalable Refresh v. Future Arm Competitors
For any Arm server vendor that is using the Platinum 8280 as a “we are 60% less than Intel” their marketing message much change. The new appropriate 28-core Intel comparison point is the Xeon Gold 6258R which is over 60% less than the Xeon Platinum 8280.
Putting this another way, if an Arm vendor was claiming 70% better price/ performance than an Intel Xeon Platinum 8280, they now have a sub-33% benefit. What is more, like AMD the Arm vendors are going to have to contend with Intel’s bundling girth plus some switching costs, even given how much as the Arm ecosystem has matured recently. Is 33% enough if AMD without as much switching cost gained 5% market share in a few years with a better figure?
To the Arm ecosystem, the “refresh” is going to ravage comparison talking points using the Gold 6258R. Intel has needed to get that Platinum 8280 differentiated and seen as an 8-socket part to prevent list price comparisons since the first generation Platinum 8180 launched.
On the flip side, Intel has a solid chance to sell more “refresh” Gold SKUs before many of the next-generation Arm competitors are even available in the market than those Arm competitors will sell in 2020 combined. As they launch, they now have new marketing messages even before Intel transitions to Ice Lake Xeons.
This launch of “refresh” CPUs is really a launch of an interim line between the 2019 2nd Generation Intel Xeon Scalable line and the allegedly 2H 2020 Ice Lake Xeon line. Let us call it what it is. Intel is facing increased market competition in the CPU segment. It needed to respond. The way it responded was by effectively re-launching the Xeon E5-2600 with dual-socket capable only parts at a lower price than the previously launched quad and eight-socket capable parts like the Xeon E7 family had.
Of course, in the world of microprocessors and manufacturing, the fact that these are still “Cascade Lake” generation microarchitecture parts means that we will call them 2nd Generation Intel Xeon Scalable, but let us be clear: this is a new sub-series of processors. 1-10% more performance per dollar has been a refresh. Hitting over 50% more performance per dollar is more than we saw generationally for almost a decade. This is effectively a new line.
By implementing DCPMM support further down the stack, Intel is signaling something obvious, that it is focusing on bundling silicon products. The more SKUs that support DCPMM the more opportunities to bundle that technology along with other technologies. Intel does not need to be the lowest cost on CPUs if it can provide the best value at the server and rack levels.
With this “refresh” or “Intel Xeon Scalable Gen 2R” launch, it is now up to the rest of the industry to respond. Make no mistake, this is a fierce and formidable competitive move by Intel and other players now need to respond in turn.