Review Index:
Feedback

Intel Xeon Scalable Processor Launch - New Architecture, New Platform for Data Center

Author: Ryan Shrout
Subject: Processors
Manufacturer: Intel

Die Configs, Pricing and Initial Thoughts

Three Unique Die Configurations

These features, performance specifications and SKUs are built around three individual die configurations. The XCC die is the largest, supporting up to 28 cores, uses a 6x6 mesh configuration, has three x16 PCIe 3.0 stacks, one x16 PCIe 3.0 connection for MCP use, and up to 3x UPI links.

View Full Size

The HCC die is up to 18 cores and the LCC die only expands to 10 cores.

View Full Size

Because of the flexibility of the mesh design, latency differences between the cores designs should be minimal, with any core accessing any component on the mesh in at most two hops.

The Lewisburg Chipset

With this new processor technology comes a new chipset to power it.

View Full Size

Compared to the previous generation chipset, the C620 upgrades to DMI 3.0, has 20 lanes of PCIe 3.0 (compared to just 8 lanes of PCIe 2.0 previously), includes up to four 10GigE connections, and integrates the Intel QuickAssist Technology to improve performance on compression and encryption.

View Full Size

QuickAssist is an integrated processor on the chipset that accelerates cryptography and compression, all on-the-fly. The goal is to create the ability for data security to be seamless and not interfere with the performance of the rest of the system and hardware capabilities. Being able to deflate data at 100 Gb/s means that you can accelerate high speed networking even further while encrypting with AES256 at 150 Gbps allows enterprise to enable full time security without slowing network speeds.

View Full Size

There are of course several different variants of the C620 chipsets with support for different speeds of QuickAssist, PCIe uplink support to the CPU, and even different TDPs.

Performance Estimates

This portion of our write up is based purely on Intel provided data, so I am going to be short and sweet with the results.

View Full Size

Intel is quick to point out its advantage in single threaded performance and the generational improvements it has made going back as far as Woodcrest in 2006. This is clearly an advantage that Intel will hold over AMD’s EPYC product family for this generation so it makes sense that Intel would want that to be emphasized.

View Full Size

Intel did show us generational improvements with the Xeon Scalable Processor family, comparing a Broadwell-EP based Xeon E5-2699 v4 2-socket configuration against the new Xeon Platinum 8180 2-socket configuration. With an average improvement of 1.65x, thanks to the combination of core count increase (22 vs 28) and memory bandwidth, on workloads ranging from options pricing to molecular dynamics to Java operations per second, Intel customers should see notable performance gains with a platform upgrade.

But at what cost?

Comparing server pricing is a tricky business. Depending on the server configuration and the workload it is running, the CPU can be a large or small part of the total budget, even if we are talking about several thousand dollars. A machine with dozens of NVMe drives in it might run you more than $300k, so a $5,000 CPU is an easy expenditure to make, but another machine with bare storage focused on compute will be more directly impacted by the pricing on the processors we are discussing today.

With that in mind, the processors that Intel is comparing throughout its press-provided documentation (and the slide above) are the Xeon E5-2699 v4 at ~$4100 and the Xeon Platinum 8180 at ~$10,000. Yup, you read that correctly, there is a more than 2x increase in cost to go from the 22-core Broadwell-EP to the 28-core Skylake-SP.

Here is a look at the pricing of the many different Xeon Scalable Processors.





Intel Xeon Scalable Processors Price Sockets Memory  Capacity
8180M (38.5M cache, 28 Cores, 56 Threads, 2.50GHz (205W) 10.40 GT/sec Intel® QPI, 14nm) $13,011 8 1.5TB
8176M (38.5M cache, 28 Cores, 56 Threads, 2.10GHz (165W) 10.40 GT/sec Intel® QPI, 14nm) $11,722 8 1.5TB
8170M (35.75M cache, 26 Cores, 52 Threads, 2.10GHz (165W) 10.40 GT/sec Intel® QPI, 14nm) $10,409 8 1.5TB
8160M (33M cache, 24 Cores, 48 Threads, 2.10GHz (150W) 10.40 GT/sec Intel® QPI, 14nm) $7,704 8 1.5TB
6142M (22M cache, 16 Cores, 32 Threads, 2.60GHz (150W) 10.40 GT/sec Intel® QPI, 14nm) $5,949 4 1.5TB
6140M (24.75M cache, 18 Cores, 36 Threads, 2.30GHz (140W) 10.40 GT/sec Intel® QPI, 14nm) $5,448 4 1.5TB
6134M (24.75M cache, 8 Cores, 32 Threads, 3.20GHz (130W) 10.40 GT/sec Intel® QPI, 14nm) $5,217 4 1.5TB
8180 (38.5M cache, 28 Cores, 56 Threads, 2.50GHz (205W) 10.40 GT/sec Intel® QPI, 14nm) $10,009 8 768GB
8176 (38.5M cache, 28 Cores, 56 Threads, 2.10GHz (165W) 10.40 GT/sec Intel® QPI, 14nm) $8,719 8 768GB
8170 (35.75M cache, 26 Cores, 52 Threads, 2.10GHz (165W) 10.40 GT/sec Intel® QPI, 14nm) $7,405 8 768GB
8164 (35.75M cache, 26 Cores, 52 Threads, 2.00GHz (150W) 10.40 GT/sec Intel® QPI, 14nm) $6,114 8 768GB
8168 (33M cache, 24 Cores, 48 Threads, 2.70GHz (205W) 10.40 GT/sec Intel® QPI, 14nm) $5,890 8 768GB
8160 (33M cache, 24 Cores, 48 Threads, 2.10GHz (150W) 10.40 GT/sec Intel® QPI, 14nm) $4,702 8 768GB
8158 (24.75M cache, 12 Cores, 24 Threads, 3.00GHz (150W) 10.40 GT/sec Intel® QPI, 14nm) $7,007 8 768GB
8156 (16.5M cache, 4 Cores, 8 Threads, 3.60GHz (105W) 10.40 GT/sec Intel® QPI, 14nm) $7,007 8 768GB
8153 (22M cache, 16 Cores, 32 Threads, 2.00GHz (125W) 10.40 GT/sec Intel® QPI, 14nm) $3,115 8 768GB
6154 (24.75M cache, 18 Cores, 36 Threads, 3.00GHz (200W) 10.40 GT/sec Intel® QPI, 14nm) $3,543 4 768GB
6152 (30.25M cache, 22 Cores, 44 Threads, 2.10GHz (140W) 10.40 GT/sec Intel® QPI, 14nm) $3,655 4 768GB
6150 (24.75M cache, 18 Cores, 36 Threads, 2.70GHz (165W) 10.40 GT/sec Intel® QPI, 14nm) $3,358 4 768GB
6148 (27.5M cache, 20 Cores, 40 Threads, 2.40GHz (150W) 10.40 GT/sec Intel® QPI, 14nm) $3,072 4 768GB
6146 (24.75M cache, 12 Cores, 24 Threads, 3.20GHz (165W) 10.40 GT/sec Intel® QPI, 14nm) $3,286 4 768GB
6144 (24.75M cache, 8 Cores, 16 Threads, 3.50GHz (150W) 10.40 GT/sec Intel® QPI, 14nm) $2,925 4 768GB
6142 (22M cache, 16 Cores, 32 Threads, 2.60GHz (150W) 10.40 GT/sec Intel® QPI, 14nm) $2,946 4 768GB
6140 (24.75M cache, 18 Cores, 36 Threads, 2.30GHz (140W) 10.40 GT/sec Intel® QPI, 14nm) $2,445 4 768GB
6138 (27.5M cache, 20 Cores, 40 Threads, 2.00GHz (125W) 10.40 GT/sec Intel® QPI, 14nm) $2,612 4 768GB
6132 (19.25M cache, 14 Cores, 28 Threads, 2.60GHz (140W) 10.40 GT/sec Intel® QPI, 14nm) $2,111 4 768GB
6136 (24.75M cache, 12 Cores, 24 Threads, 3.00GHz (150W) 10.40 GT/sec Intel® QPI, 14nm) $2,460 4 768GB
6134 (24.75M cache, 8 Cores, 16 Threads, 3.20GHz (130W) 10.40 GT/sec Intel® QPI, 14nm) $2,214 4 768GB
6130 (22M cache, 16 Cores, 32 Threads, 2.10GHz (125W) 10.40 GT/sec Intel® QPI, 14nm) $1,894 4 768GB
6128 (19.25M cache, 6 Cores, 12 Threads, 3.40GHz (115W) 10.40 GT/sec Intel® QPI, 14nm) $1,691 4 768GB
6126 (19.25M cache, 12 Cores, 24 Threads, 2.60GHz (125W) 10.40 GT/sec Intel® QPI, 14nm) $1,776 4 768GB
5122 (16.5M cache, 4 Cores, 8 Threads, 3.60GHz (105W) 10.40 GT/sec Intel® QPI, 14nm) $1,221 4 768GB
5120 (19.25M cache, 14 Cores, 28 Threads, 2.20GHz (105W) 10.40 GT/sec Intel® QPI, 14nm) $1,555 4 768GB
5118 (16.5M cache, 12 Cores, 24 Threads, 2.30GHz (105W) 10.40 GT/sec Intel® QPI, 14nm) $1,273 4 768GB
5115 (13.75M cache, 10 Cores, 20 Threads, 2.40GHz (85W) 10.40 GT/sec Intel® QPI, 14nm) $1,221 4 768GB
4116 (16.5M cache, 12 Cores, 24 Threads, 2.10GHz (85W) 9.60 GT/sec Intel® QPI, 14nm) $1,002 2 768GB
4114 (13.75M cache, 10 Cores, 20 Threads, 2.20GHz (85W) 9.60 GT/sec Intel® QPI, 14nm) $694 2 768GB
4112 (8.25M cache, 4 Cores, 8 Threads, 2.60GHz (85W) 9.60 GT/sec Intel® QPI, 14nm) $473 2 768GB
4110 (11M cache, 8 Cores, 16 Threads, 2.10GHz (85W) 9.60 GT/sec Intel® QPI, 14nm) $501 2 768GB
4108 (11M cache, 8 Cores, 16 Threads, 1.80GHz (85W) 9.60 GT/sec Intel® QPI, 14nm) $417 2 768GB
3106 (11M cache, 8 Cores, 16 Threads, 1.70GHz (85W) 9.60 GT/sec Intel® QPI, 14nm) $306 2 768GB
3104 (8.25M cache, 6 Cores, 12 Threads, 1.70GHz (85W) 9.60 GT/sec Intel® QPI, 14nm) $213 2 768GB
8176F (38.5M cache, 28 Cores, 56 Threads, 2.10GHz (173W) 10.40 GT/sec Intel® QPI, 14nm) $8,874 2 768GB
8160F (33M cache, 24 Cores, 48 Threads, 2.10GHz (160W) 10.40 GT/sec Intel® QPI, 14nm) $4,856 2 768GB
6148F (27.5M cache, 20 Cores, 40 Threads, 2.40GHz (160W) 10.40 GT/sec Intel® QPI, 14nm) $3,227 2 768GB
6142F (22M cache, 16 Cores, 32 Threads, 2.60GHz (160W) 10.40 GT/sec Intel® QPI, 14nm) $3,101 2 768GB
6138F (27.5M cache, 20 Cores, 40 Threads, 2.00GHz (135W) 10.40 GT/sec Intel® QPI, 14nm) $2,767 2 768GB
6130F (22M cache, 16 Cores, 32 Threads, 2.10GHz (135W) 10.40 GT/sec Intel® QPI, 14nm) $2,049 2 768GB
6126F (19.25M cache, 12 Cores, 24 Threads, 2.60GHz (105W) 10.40 GT/sec Intel® QPI, 14nm) $1,931 2 768GB
8160T (33M cache, 24 Cores, 48 Threads, 2.10GHz (150W) 10.40 GT/sec Intel® QPI, 14nm) $4,936 8 768GB
6138T (27.5M cache, 20 Cores, 40 Threads, 2.00GHz (125W) 10.40 GT/sec Intel® QPI, 14nm) $2,742 4 768GB
6130T (22M cache, 16 Cores, 32 Threads, 2.10GHz (125W) 10.40 GT/sec Intel® QPI, 14nm) $1,988 4 768GB
6126T (19.25M cache, 12 Cores, 24 Threads, 2.60GHz (125W) 10.40 GT/sec Intel® QPI, 14nm) $1,865 4 768GB
5120T (19.25M cache, 14 Cores, 28 Threads, 2.20GHz (105W) 10.40 GT/sec Intel® QPI, 14nm) $1,727 4 768GB
5119T (19.25M cache, 14 Cores, 28 Threads, 1.90GHz (85W) 10.40 GT/sec Intel® QPI, 14nm) $1,555 4 768GB
4116T (16.5M cache, 12 Cores, 24 Threads, 2.10GHz (85W) 9.60 GT/sec Intel® QPI, 14nm) $1,112 2 768GB
4114T (13.75M cache, 10 Cores, 20 Threads, 2.20GHz (85W) 9.60 GT/sec Intel® QPI, 14nm) $773 2 768GB
4109T (11M cache, 8 Cores, 16 Threads, 2.00GHz (70W) 9.60 GT/sec Intel® QPI, 14nm) $501 2 768GB

The pricing table is so complicated, it is difficult to wrap your head around even starting at it for long periods. You can pick up a Xeon Bronze 3104 CPU with 6-cores, 12-threads, and a 1.70 GHz clock speed for just $213. Or maybe step it up a bit for a Xeon Platinum 8180M with 28-cores, 56-threads, 2.50 GHz clock speed for $13,001. Or just consider one of the other 56 options available to you.

Despite all the options provided, some areas seem left out to me. Intel’s highest core count processor targeted at the “up to 2-socket” market comes in at $1000 but only has 12-cores and a 2.1 GHz clock speed. Similarly, the highest “up to 4-socket” processor is the 6154 with 18-cores and a 3.0 GHz clock speed, priced at $3500. Neither of these seems to be particularly well positioned to take on the AMD EPYC processors launched last month that will offer 32-cores/64-threads and clock speeds up to 3.2 GHz for around $4200.

Initial Thoughts

Honestly, there is still a lot to try and wrap my brain around with this release, as the data center and enterprise markets are very different than the consumer field. Services, platform reliability and longevity are coveted as much if not more than performance and value, and Intel clearly has the advantage in these areas. It has dominated the market for a decade, and rightfully so, but at the expense of slower advancement and development than we might have seen otherwise. Intel is hoping that technologies like storage, custom FPGAs, and interconnects like Omni-Path will help maintain the Xeon brand as the pinnacle in the space for many years.

View Full Size

But Intel no longer works in a vacuum, and though a small competitor today, AMD and EPYC are going to make waves. Because of that, the price increases we see on the Xeon Scalable platform over the previous generation seem out of place. We know from conversations with partners that the pricing information we see on Intel Ark and other locales is always in flux, so the biggest names will never shell out $10k for a processor. But we also understand that negotiations must start at the listed price, which is higher today than it was previously.

In totality, the Xeon Scalable Processor Family looks like a competent upgrade over the previous generation with higher core counts, added performance capabilities, new features, and platform extensions to migrate the ecosystem forward. It’s a difficult task to make 58 launched products make sense but the Platinum, Gold, Silver and Bronze categorization are a solid attempt, though the complexity of the product stack still comes through. We are hoping to get in a couple of test units from Intel for the new Xeon Scalable family soon, along with a collection of AMD EPYC processors, for a more detailed analysis soon.


July 11, 2017 | 05:50 PM - Posted by Hishnash (not verified)

that is one complex lineup.

I know it's targeted at data centers who probably employ a few people full time to understand this but still since amd are able to provide a single line up with all CPUs having the same Ram and PCIe its a shame Intel can't do offer the same and just make these a little less confusing.

July 11, 2017 | 07:15 PM - Posted by Xebec

Interesting IPC chart.. they're saying -

IPC is up 70% from Merom (Core 2) to today.. (remove the multiplier at the beginning as they're comparing Merom to Yonah, then remultiply everything), and 35% from Sandy Bridge to today.

I'm guessing the 3.8x per socket is more tied with the power efficiency improvements over top of IPC - performance/watt over the decade..

July 11, 2017 | 08:18 PM - Posted by Clmentoz (not verified)

It looks like Anandtech had some server SKUs to look at for a few weeks. And it's too bad that Anandtech did not ask AMD any questions about AMD's Infinity Fabric on Epyc and how it relates to any Infinity Fabric based Direct Attatched GPU acceleration via the Infinity Fabric EPYC CPU to any Vega/Vega micro-arch GPUs(That Support the Infinity Fabric) accelerators for FP/AI workloads.

AMD did cover this as a possible Epyc to Vega "NVLink" style coherent interface option for Epyc/Vega and GPU accelerated compute workloads with fully coherent Infinity Fabric communication Epyc to Vega in there presentation a few months back with those seismic benchmarks on some Epyc engineering samples. But as of yet there have been no demonstrations for AMD and maybe for SIGGRAPH there will be more information forthcoming.

From Anandtech:

"First of all, we have to emphasize that we were only able to spend about a week on the AMD server, and about two weeks on the Intel system. With the complexity of both server hardware and especially server software, that is very little time. There is still a lot to test and tune, but the general picture is clear." [see: closing thoughts (1)]

(1)

"Sizing Up Servers: Intel's Skylake-SP Xeon versus AMD's EPYC 7000 - The Server CPU Battle of the Decade?"

http://www.anandtech.com/show/11544/intel-skylake-ep-vs-amd-epyc-7000-cp...

July 13, 2017 | 04:00 PM - Posted by Anonymouse (not verified)

Tons of Info about both Fabrics and Chips is at ServeTheHome .

Beating both (in performance) is the POWER9 which has been completely sold-out (to Google and Facebook) - next year they'll be more CPUs available for us poor folks.

July 11, 2017 | 08:56 PM - Posted by RealExascale (not verified)

No Optane support for those prices? Come on! Whats all that about OLTP? Anyone really going to choose x86 over SPARC or a real mainframe for OLTP?

The only nice thing about this release is AVX-512 and the potential for the highest end CPUs to scale to 8 sockets gluelessly. Thats not new or unique though. E7s have done that for years.

SPARC or ARM would be better for scale up anyway. ARM and Epyc both have 48bit physical addressing and Intel is still using 46bit. 46bit PA on a $13,000 CPU!!!

Epyc also has 2TB per socket vs 1.5, 128 PCI-e lanes vs 48, and fully encrypted memory(so does Ryzen Pro). Intel gas nothing to compete with that at all. Epyc is a way better value.

AMDs fully encrypted memory really destroys these new Xeons. Intel better implement it for Optane DIMM supporting platforms or people wont need cryogenics for cold boot attacks anymore!

July 12, 2017 | 06:58 AM - Posted by psuedonymous

"Despite all the options provided, some areas seem left out to me. Intel’s highest core count processor targeted at the “up to 2-socket” market comes in at $1000 but only has 12-cores and a 2.1 GHz clock speed. Similarly, the highest “up to 4-socket” processor is the 6154 with 18-cores and a 3.0 GHz clock speed, priced at $3500. Neither of these seems to be particularly well positioned to take on the AMD EPYC processors launched last month that will offer 32-cores/64-threads and clock speeds up to 3.2 GHz for around $4200."

That's because unlike the previous E5/E7 system, you can grab one of those crazy 28-core chips and stick it in a single-socket board if you wanted. Single-socket-across-the-range is a big change from the 1/2/many socket splits previously implemented.

It's; also likely why Intel have no announced any LGA2066 Xeons: there probably aren't any. The presence of 'small' Xeon Gold dies (that's you'd normally expect to see sharing the small prosumer socket) on LGA3647 almost confirms this.

July 12, 2017 | 11:27 AM - Posted by Ryan Shrout

Interesting thought. My thinking had been that Intel was making LGA2066 Xeons for the iMac Pro offerings.

July 13, 2017 | 08:47 AM - Posted by Hishnash (not verified)

Apple might be getting slictly custom chips
a version of https://ark.intel.com/products/124943/Intel-Xeon-Gold-6144-Processor-24_... with the Tubo Boost v3 Enabled (to give the 4.5Ghz boost Apple have listed on the website). Given the i9 parts are basically these chips with ECC disabled it possible Intel could create some.

Apple has in the past been given custom Xeons for their Mac Pro (deelided ones so that Apple could put good thermal contact directly onto the heatsink) given the chatter about heat issues with the i9 line Apple may be required this time round as well.

July 12, 2017 | 06:02 PM - Posted by Toby Broom (not verified)

Traditionally there was the 16xx xeons that were the counter parts of the i7's so you would imagine that these would be created?

It will be intresting to see how HP develops the Z460, Z660 & Z860 workstations in comparision to the Apple xPro's

July 13, 2017 | 08:48 AM - Posted by Hishnash (not verified)

Unless HP can get the same deal as aplle and get custom https://ark.intel.com/products/124943/Intel-Xeon-Gold-6144-Processor-24_... parts with Tubo Boost enabled. the iMac Pro will deliver more power for lower core count tasks were the turbo boost plays a part.

Post new comment

The content of this field is kept private and will not be shown publicly.
  • Lines and paragraphs break automatically.
  • Allowed HTML tags: <a> <em> <strong> <cite> <code> <ul> <ol> <li> <dl> <dt> <dd> <blockquote><p><br>
  • Web page addresses and e-mail addresses turn into links automatically.

More information about formatting options

By submitting this form, you accept the Mollom privacy policy.