With the launch of the Intel Xeon D-2100 series, “Skylake-D” series SoCs, we wanted to provide staples at STH, the initial SKU list and value analysis. There are 14 publicly released SKUs in the Intel Xeon D-2100 series at launch. These SKUs are focused into three families, Edge Server and Cloud, Network Edge and Storage, and QuickAssist Technology (QAT) SKUs.
Intel Xeon D-2100 Series SKU List and Comparison
Here is a summary table of the 14 SKUs at launch which ranges from 4 to 18 core parts.
We wanted to point out that the Xeon D-2191 does not have the 4x 10GbE available. Looking back to the previous generation, Facebook utilized Mellanox multi-host adapters along with a custom version of the original Xeon D to lower networking costs and improve performance. We suspect that Intel is keenly aware of this and that is a part of the reason for that de-feature move.
We are also told that the Xeon D-2191 is expected to be a limited availability SKU in the stack. Although it exists, it is not going to be a mainstream part.
On the memory speed, one can see DDR4-2133 to DDR4-2666 being supported. We wish that across Intel’s lines that the company simply used a single maximum memory speed. This would assist OEMs and users with making configurations easier.
One of the biggest items you will notice is that TDPs are significantly higher than with the previous generation, along with clock speeds. The previous generation Intel Xeon D-1587 was the highest-end part of that generation yet had a TDP of only 65W. This generation has the D-2183T which has higher clocks yet moves up to 75W. As a whole, there are three SKUs at 100-110W. Nine of the fourteen SKUs have TDPs greater than the previous generation’s (non-QAT parts) maximum of 65W. There are more features, more cores and higher clocks, but the stack has moved up.
While we had 25W TDP parts in the previous generation, 60W is the lowest TDP in this generation. It seems as though Intel is moving sub 60W TDP SoC’s to be covered by the Intel Atom C3000 line.
Intel Xeon D-2100 Series Value Analysis
With each new processor generation over the last five years we have provided a simple value analysis. This value analysis simply looks at how much one is paying for each core GHz. Here is that table for this release:
Here we can see that Intel is extracting a high premium for the compute capacity of the Xeon D-2191. Although it has the highest core count at 18 cores, it also has the highest price and lowest frequencies. Conversely, the 4 core/ 8 thread Xeon D-2123IT is the lowest priced option at $213 and has some of the highest clock speeds which make it look like a great value.
This value analysis, especially in this processor generation is extremely flawed. The 4x 10GbE ports carry a value that would alternatively be measured in power consumption, lost PCIe connectivity, and an add-on board or chipset NICs. The QuickAssist technology offers faster compression and encryption. Another key point is that consolidating more edge systems into a single box with more cores can mean less power, space, licensing, and component (e.g. motherboard, chassis, PSU, and networking) costs. One reason higher core count CPUs are popular is that consolidation plays can have enormous values.
With any new generation, there are changes. Perhaps the biggest change of this generation is that these CPUs act more like slightly de-featured and more compact Xeon Silver 4100 / Gold 5100 platforms. As a result, there are going to be more discussions about the mainstream Xeon Scalable line and the Xeon D-2100 series than there were around the Xeon E5 V4 generation and the Xeon D-1500 series.
For more information, check out our Intel Xeon D-2100 series launch coverage central.