Three years after the launch of its first Xeon D processors, Intel is following-up with its second-gen take. These SKUs come to us in the form of the D-2100 series, and all are based on the Skylake-SP microarchitecture. A lot of polish has been provided with this latest gen, with quad-channel memory support on tap for the entire line. For those with very heavy network encryption workloads, special QuickAssist CPUs can fill any void.
If anyone’s wondering why Intel’s (NASDAQ: INTC) trying to give you D when E and W chips already exist, I can relate to the confusion. Before getting into more second-gen Xeon D details, here’s a quick breakdown that should clear some questions up.
|
Intel Xeon Series Breakdown |
Xeon Scalable |
For mission-critical operations. Suitable for real-time analytics; boasts the highest reliability and uptime promises. Supports 2, 4, and 8 socket systems, up to 1.5TB memory, and 28 cores per chip. |
Xeon E7 |
For data-hungry workloads easily handicapped by top-end desktop systems. Supports 4 and 8 CPU configurations. Supports 4, and 8 socket systems, up to 3.07TB memory, and 24 cores per chip. |
Xeon E5 |
Versatile mid-range Xeons for the high-end workstation crowd. Also suitable for software defined infrastructures. Supports 1, 2, and 4 socket systems, up to 1.54TB memory, and 22 cores per chip. |
Xeon W |
For serious workstation users. Think Core X for workstation, with enterprise features and ECC memory support added. Supports 1 socket systems, up to 512GB memory, and 18 cores per chip. |
Xeon E3 |
An “essential” choice for small businesses and entry-level workstation users. A good choice for portable workstations. Supports 1 socket systems, up to 64GB memory, and 4 cores per chip. |
Xeon D |
Designed for dense, low-power computing. These SoCs bundle networking, security, and other accelerators on-chip. Supports 1 socket systems, up to 512GB memory, and 18 cores per chip. |
If W stands for workstation, and E, enterprise, then I’m going to assume the D here stands for “dense”. These chips are designed for smaller enclosures where it isn’t OK to have subpar performance. As SoCs, the Xeon D series includes networking and security features under-the-hood, but notably, they don’t include onboard graphics like the Xeon E3 series does.
Built on the latest 14nm process, the D-2100 series brings a lot of refinement to the table, such as with the introduction of a quad-channel memory controller, and a house cleaning with DDR3 and PCIe 2.0 kicked to the curb (PCIe 2.0 devices will still function, of course, but dedicated lanes are not present).
|
Intel Xeon D-series Processors |
|
Cores |
Clock (Turbo) |
Cache |
Memory |
TDP |
Price |
|
Server & Cloud |
Xeon D-2191 |
18 (36T) |
1.60 GHz (3.00) |
24.75MB |
Quad |
86W |
$2,406 |
Xeon D-2161I |
12 (24T) |
2.20 GHz (3.00) |
16.5MB |
Quad |
90W |
$962 |
Xeon D-2141I |
8 (16T) |
2.20 GHz (3.00) |
11MB |
Quad |
65W |
$555 |
|
Network & Enterprise Storage |
Xeon D-2183IT |
16 (32T) |
2.20 GHz (3.00) |
22MB |
Quad |
100W |
$1,764 |
Xeon D-2173IT |
14 (28T) |
1.70 GHz (3.00) |
19MB |
Quad |
70W |
$1,229 |
Xeon D-2163IT |
12 (24T) |
2.10 GHz (3.00) |
17MB |
Quad |
75W |
$930 |
Xeon D-2143IT |
8 (16T) |
2.20 GHz (3.00) |
11MB |
Quad |
65W |
$566 |
Xeon D-2142IT |
8 (16T) |
1.90 GHz (3.00) |
11MB |
Quad |
65W |
$438 |
Xeon D-2123IT |
4 (8T) |
2.20 GHz (3.00) |
8MB |
Quad |
60W |
$213 |
|
Integrated QuickAssist |
Xeon D-2187NT |
16 (32T) |
2.00 GHz (3.00) |
22MB |
Quad |
110W |
$1,989 |
Xeon D-2177NT |
14 (28T) |
1.90 GHz (3.00) |
19MB |
Quad |
105W |
$1,443 |
Xeon D-2166NT |
12 (24T) |
2.00 GHz (3.00) |
17MB |
Quad |
85W |
$1,005 |
Xeon D-2146NT |
8 (16T) |
2.30 GHz (3.00) |
11MB |
Quad |
80W |
$641 |
Xeon D-2145NT |
8 (16T) |
1.90 GHz (3.00) |
11MB |
Quad |
65W |
$502 |
Intel’s catering to three different scenarios with these latest Xeon D chips. Three models at the top of this table are destined for server and cloud services, while the middle batch are optimized for networking and enterprise (inc. cloud) storage. Finally, the bottom chips feature Intel’s QuickAssist, a name that surprisingly doesn’t tell you anything about what it does (and if you say you guessed, I’ll assume you already knew).
QuickAssist is hardware acceleration for cryptography that offloads the work from the CPU to an external engine found on the chipset. On the biggest QuickAssist SKUs, up to 100Gbps of offline acceleration performance is availed. That’s a gain of 2.5x over the previous generation.
Intel has 5G and edge computing platforms locked in Xeon D’s sights
It’s interesting to note that every single one of the D-2100 SKUs sport the same Turbo frequency of 3.0GHz, though the overall stock hovers around 2GHz base. To that end, the biggest chip in this lineup is the 18-core D-2191, which has the most cores, as well as the lowest base clock speed. It also has the most cache, matching the 18-core Xeon Gold 6154 and Core i9-7980XE.
As of the time of writing, I couldn’t find the D-2100 series in stock anywhere, but I’d suspect it’ll only take a few days before SKUs begin popping-up. In the meantime, I hope some NAS vendors are eying the D-2123IT. 4x10GbE NASes that won’t break the bank could actually happen this year thanks to this chip.