Intel grows the Xeon D Platform, with 8 new Xeon D CPUs announced today, and 16 core version arriving in Q1 2016

Posted by Paul Braren on Nov 9 2015 (updated on Jan 8 2015) in

Nov 10 2015 Update #1 - Supermicro has already updated some of their product pages, with apparently no changes to SKUs or part #s or components, details below.

Nov 10 2015 Update #2 - ServeTheHome has now published SR-IOV details at THE BROADWELL-DE INTEL XEON D-1500 SERIES AND SR-IOV. More thoughts below.

Nov 11 2015 Update - ServeTheHome has now published Smore details on CPU variants that all fit in the same tiny Mini-ITX footprint New Supermicro Intel Xeon D Motherboards – Even with 12 and 16 core chips!.

Back in early 2015, System on a Chip (SoC) was reaching new levels of efficiency and power, enjoying success in its third generation. When Intel announced the Xeon D-1520 (no 10GbE, 4 cores) and D-1540 (10GbEx2, 8 cores) on March 9, they were the first-ever Xeon processors to use SoC, touting:

First 14nm Intel® Xeon® Product Family Delivers Server-Class Capabilities in a Dense, Low- Power System-on-Chip Optimized for Cloud, Telco Service Providers and Web Hosters
...
provides customers with enhanced intelligence and greater agility to rapidly deliver new services at a lower total cost of ownership (TCO). Also, with server-class reliability, availability and serviceability (RAS) features now available in an ultra-dense, low-power device...

ARK_Menu_Intel_Xeon_Processor_D-1500_Family

So today's announcement clearly indicates ongoing investment in rounding out the platform, with 8 more SKUs at a variety of 4 and 8 core sizes. Given the 8 core current Xeon D-1540 SoC is worth about $850 alone, one can expect next year's 16 core version will be quite pricey. All of these new 4Q15 SKUs are lesser models, with one exception.

Buried in the announcement is very brief mention of the new Xeon D-1541 SoC, which Patrick does a great job summarizing here:

INTEL XEON D-1500 STORAGE ACCELERATED SKUS
Nov 09 2015 by Patrick Kennedy at Serve The Home

along with a more general overview article:

EIGHT NEW INTEL XEON D PROCESSORS LAUNCHED – 16 CORE CHIPS COMING
Nov 09 2015 by Patrick Kennedy at Serve The Home

More to come but for general purpose servers, the Xeon D-1540 and D-1520 will still be highly relevant. For high-density and embedded applications, the new processors will likely be a game changer. We have reported previously that these SoCs were in extremely high demand.

For readers who already own a Xeon D-1540 in their Supermicro SuperServer 5028D-TN4T Bundle, you may be wondering exactly what has changed. From my read of the announcements, very little, it turns out:

the CPU clock multiplier moved from 20 to 21 (less than 5% performance boost), from 2.0GHz/2.6GHz turbo, to 2.1GHz/2.7GHz turbo
the ability to use Storage Performance Development Kit (SPDK)
SR-IOV feature will now be supported
[this new item #3 was added on Nov 10]

See also the nice view of the sameness over at CPU World:

Intel Xeon D-1540 vs D-1541

Xeon_D-1540v1541_from_CPU_World — From "Intel Xeon D01540 s D-1541," click to visit the complete article.

Missing from the Intel press release was any mention of what is happening with the ongoing SR-IOV saga that could affect a subset of enthusiasts, being actively discussed now at always-wonderful ServeTheHome. Interesting that AnandTech also mentions that same thread. See also comments here at TinkerTry. Keeping perspective here, I've never owned anything that didn't have some bugs to work out, this one just doesn't happen to sting me, as I don't currently have a compelling use case for that feature in my simple home lab. But I realize you might, and I'm continuing to track this, with requests for comment and/or documentation clarification already sent to Supermicro.

One can rather safely assume the same components will likely be used for everything but the CPU, including the 10GbE that's now enjoying a shiny new VMware driver VIB. As for what Supermicro will do with the new CPU part, sure seems likely it'll just be a slight speed bump but sold as the existing SKU at resellers, the SYS-5028D-TN4T, and the X10SDV-TLN4F motherboard within. Supermicro will be letting me know.

I do not yet have any estimated ship dates for the Supermicro Xeon D-1541 system from either Intel or Supermicro. All I have is the press release's Q4 2015 estimate. Last time around, they announced in early March, but didn't ship until late June. But that doesn't mean they won't make their Nov 2015 estimate this time around. Whatever happens, I'll be sure to try to test that the Windows 10 (Bundle 1) and ESXi 6.0 Update 1a (user-installed on Bundle 2 or 3) keep working just fine, which seems extremely likely, given how well they're working now, and that only the CPU itself is likely to change slightly (multipler change, and one new instruction set)

Summary

Intel's growing investment in the platform bodes well for a bright future in efficient computing, and a healthy ecosystem surrounding it. In other words, it's likely vast numbers of users and business will be using these very same I-350 1GbE and X552 10GbE chips. For the home lab enthusiast, this means a lot of folks will be joining you with PCIe 3.0 x 4 lanes goodness (PCIe slot and M.2 socket), empowering you to keep having more fun choices joining the world's fastest consumer SSD, which isn't even sold in legacy SATA3 2.5" form factor. The future is here now, and these new announcements help us clearly see into Intel's big future for these tiny and efficient 14nm CPUs.

In the above presentation, the following information about one of the many test configurations is given on page 36, giving a good sense of how much faster this Xeon is, compared to Atom:

Dynamic Web Serving
New Configuration: Intel® Xeon Processor D-based reference platform with with one Xeon Processor D 1540 (8C, 2.0GHz, 45W), Turbo Boost Enabled, Hyper-Threading enabled, 64GB
memory (4x16GB DDR4-2133 RDIMM ECC), 2x10GBase-T X552, 3x S3700 SATA SSD, Ubuntu 14.10 (3.16.0-23 generic), Nginx 1.4.4, Php-fpm 15.4.14, memcached 1.4.14, Simultaneous
users=47152
Base Configuration: Supermicro SuperServer 5018A-TN4 with one Intel Atom Processor C2750 (8C, 2.4GHz,20W), Turbo Boost Enabled, 32GB memory (4x8GB DDR3-1600 SO-DIMM
ECC), 1x10GBase-T X520, 2x S3700 SATA SSD, Ubuntu 14.10(3.16.0-23 generic), Nginx 1.4.4, Php-fpm 15.4.14, memcached* 1.4.14, Simultaneous users=12896
...
Video on Demand
New Configuration: Intel® Xeon Processor D-based reference platform with Xeon-D 1541 (8C, 2.1GHz), Turbo Boost Disabled, Hyper-Threading Enabled, 32 GB memory (2x16GB DDR4-
2400 RDIMM ECC), 20x 4TB WD SATA 64MB cache, 2x LSI 9207 HBA, 2x 10GbE Bonded, Ubuntu 14.04.2 (3.16.0-30-generic x86_64), Swift 2.2.x, COSBench 4.2.x, 512 Workers- 4
Drivers- 1000C- 1000 Objects per container, 32MB 90%Reads 10% Writes= 943 MB/s
Base Configuration: Intel® Atom C2000-based reference platform with one Intel Atom C2750 (8C, 2.4GHz,20W), Turbo Boost Disabled, Hyper-Threading Disabled, 16 GB memory (2x8GB
DDR3-1600 RDIMM ECC), 10x 3TB WD SATA 64MB cache, 1x LSI 9207 HBA, 1x 10GbE, Ubuntu 14.04.2 (3.16.0-30-generic x86_64), Swift 2.2.x, COSBench 4.2.x, 512 Workers- 4 Drivers-
1000C- 1000 Objects per container, 32MB 90%Reads 10% Writes= 484 MB/s
Virtualized Storage
New Configuration: Intel® Xeon Processor D-based reference platform with Xeon-D 1541 (8C, 2.1GHz), Turbo Boost Disabled, Hyper-Threading Enabled, 32 GB memory (2x16GB DDR4-
2400 RDIMM ECC), 20x 4TB WD SATA 64MB cache, 2x LSI 9207 HBA, 2x 10GbE Bonded, Ubuntu 14.04.2 (3.16.0-30-generic x86_64, Ceph hammer , FIO 2.2.9, FIO- 40 RBDs- 4 Clients-
32 Queue Depth per RBD- 1 worker per RBD volume, 4KB 90%Reads 10% Writes= 1372 IO/s
Base Configuration: Intel® Atom C2000-based reference platform with one Intel Atom C2750 (8C, 2.4GHz,20W), Turbo Boost Disabled, Hyper-Threading Disabled, 16 GB memory (2x8GB
DDR3-1600 RDIMM ECC), 10x 3TB WD SATA 64MB cache, 1x LSI 9207 HBA, 1x 10GbE, Ubuntu 14.04.2 (3.16.0-30-generic x86_64), Ceph hammer , FIO 2.2.9, FIO- 40 RBDs- 4 Clients-
32 Queue Depth per RBD- 1 worker per RBD volume, 4KB 90%Reads 10% Writes= 707 IO/s

Nov 10 2015 Update #1

Well there it is, Supermicro has already updated the existing pages for the the X10SDV-TLN4F motherboard, the heart of the SYS-5028D-TN4T SuperServer.

The CPU part is now listed simply as D-1540/1541, with the model numbers of the SuperServer and motherboard remaining unchanged, and the Supermicro list of tested 32GB RDIMMs also unchanged.

Too early to say for be completely sure, but this sure looks promising, that they'll likely be no new qualification testing needed. This is good!

2nd image credit is from the Internet Archive Wayback Machine Sep 14 2015 snapshot here.

Nov 10 2015 Update #2

Re-thinking about this whole tiny speed bump thing, I'll admit, it's a little annoyance to me than anything useful. You can imagine Intel had some manufacturing process improvements these past 4 months of trying to ship these constrained CPUs in volume. So they decided to overclock the original, and sell is at new, giving them something slightly newer to sell. Unfortunately, when you combine that with no solid date in the press release, you only hurt sales a bit in the short term. As far as articles like my recent Samsung 950 EVO disk benchmark tests, it will likely stay valid exactly as-is, since I doubt the <5% CPU boost will have a noticeable effect on those results. But for things like Futuremark PCMark 8, those results would need to be re-run.

Oh well, that's the way it goes with a new and growing platform, and we might want to just get used to it. It's up to you if you feel the slight speed boost, at likely the same price, is worth holding off for your next or first SuperServer order, for an undetermined amount of time.

Nov 10 2015 Update #3

This is preliminary information:
Now it seems like full SR-IOV support will only be in the Xeon D-1541 revision. This SR-IOV shortcomings of the Xeon D-1540 has already been notated in recent TinkerTry Xeon D-1540 articles.

SR-IOV, the one feature missing among dozens of new features in this new platform. Still hands-down the best choice for me, personally, having saved up for 4.5 years and knowing full well I was an early adopter, getting mine the first days they were avaialble to the public, on June 25th.

This one missing feature implementation wouldn't be such an annoyance for folks that need SR-IOV if it weren't for these 5 factors:

Intel's annoucements earlier this year included SR-IOV, according to this article.
(I'm currently searching for the original source material)
Intel's lack of direct public response to the longstanding questions about when SR-IOV support would arrive, with months conjecture left unanswered, until ServeTheHome's article today.
Supermicro's lack of BIOS upgrade release notes or download page warnings to document that the SR_IOV menu option in 1.0a would go away once you flash it to 1.0b.
(I've now added a warning to my BIOS upgrade instructions)
Supermicro's press release back on Mar 09 2015 mentions SR-IOV:

Product Specifications

X10SDV-F/-TLN4F – Mini-ITX Motherboard (6.7” x 6.7”) supports single Intel® Xeon® processor D-1540 SoC (8 core, 45W), VT-d/x, TXT, AES-NI, SR-IOV, Xeon RAS, built in 10GbE. 128GB 2133MHz DDR4 RDIMM or 64GB UDIMM in 4x DIMMs, 6x SATA 3.0, 1x M.2 slot (M key for SSD, 2242/2280, PCIe3.0 x4), 2x USB 3.0, 4x USB 2.0, 1x PCIe 3.0 x16, Quad LAN ports with SoC dual 10GbE and I350-AM2 dual GbE (-F with dual GbE only), IPMI 2.0 with KVM and dedicated port, 0-60°C operating temperature, 4 pin 12V DC and ATX power source
Some companies, and individual people using their own hard-earned money, bought into the Xeon D-1540 product family based on seeing SR-IOV mentioned, and are now upset and disappointed to hear it likely won't ever work in their current motherboard. The Xeon D-1540 is a SoC design, system on a chip, which means that replacing the motherboard/cpu assembly with Xeon D-1541 would be costly (~$1000 USD).
which ends with the usual boilerplate, scroll down to the very bottom:

Information in this document is subject to change without notice.

What is SR-IOV? As usual, Patrick Kennedy does a great job explaining it, here's an excerpt:

Is SR-IOV absolutely necessary for virtualization? No. It is a PCI-SIG standard technology to accelerate virtual machine networking.

What does lack of SR-IOV mean to me? Well, I don't personally need it, and frankly hadn't even thought about it until pinged by a commenter very recently. I'm also not likely to be pushing to my VMs hard enough to see benefits in my single-user home lab environment. Finally, I'm not willing to give up the freedom of vMotion (along with many other drawbacks, see also VMware KB 2038739, which hasn't been updated since March 25 2015, and doesn't even mention ESXi 6.0 yet:

SR-IOV support status FAQ (2038739)

It would seem Intel simply encountered issues with getting SR-IOV working on the Xeon D-1540 chipset, which would not be the fault of the OEMs and resellers. I have no solid confirmation, this assessment is just from what I'm hearing and reading second-hand.