AMD 3rd Gen EPYC Milan Review: A Peak vs Per Core Performance Balance

The arrival of AMD’s 3rd Generation EPYC processor family, using the new Zen 3 core, has been hotly anticipated. The promise of a new processor core microarchitecture, updates to the connectivity and new security options while still retaining pl…

The arrival of AMD’s 3rd Generation EPYC processor family, using the new Zen 3 core, has been hotly anticipated. The promise of a new processor core microarchitecture, updates to the connectivity and new security options while still retaining platform compatibility are a good measure of an enterprise platform update, but the One True Metric is platform performance. Seeing Zen 3 score ultimate per-core performance leadership in the consumer market back in November rose expectations for a similar slam-dunk in the enterprise market, and today we get to see those results.

The Ampere Altra Review: 2x 80 Cores Arm Server Performance Monster

Earlier in the year we have a good look at the first ever Arm Neoverse-N1 CPU in the form s of the Graviton2 from Amazon. Whilst that chip was great – it wasn’t really a true competitor in the market as it was availably only to Amazon&rsq…

Earlier in the year we have a good look at the first ever Arm Neoverse-N1 CPU in the form s of the Graviton2 from Amazon. Whilst that chip was great – it wasn’t really a true competitor in the market as it was availably only to Amazon’s cloud business.

Today’s release of the new Ampere Computing Altra processors is bound to change that. The Altra is aiming for nothing less than top performance, going all in against AMD and Intel. 80 Neoverse-N1 cores at 3.3GHz and 250W, 8x DDR4-3200, 128 PCIe 4.0 lanes for I/O – the new Altra Q80-33 is the best the Arm world has ever seen, and we’ve put the new chip through its paces.

SiPearl Lets Rhea Design Leak: 72x Zeus Cores, 4x HBM2E, 4-6 DDR5

In what seems to be a major blunder by the SiPearl PR team, a recent visit by a local French politician resulted in the public Twitter posting in what looks like the floor plan of the company’s first-generation server chip project “Rhea”.

During a visit of Alexandra Dublanche, a local Île-de-France politician and vice president of economic development and other responsibilities for the region, the PR teams had made some photo-op captures of a tour of the office. Amongst the published pictures includes a shot of the company’s Rhea server chip project with some new, until now unreleased details of the European Processor Initiative-backed project.

Accélérer le développement européen de microprocesseurs à haute performance est un enjeu de souveraineté 🇪🇺
C’est le défi auquel s’attaque la pépite technologique @SIPEARL_SAS basée à #MaisonsLaffitte en @iledefrance
La Région sera à ses côtés face à cet enjeu majeur! pic.twitter.com/kcBwghlfsP

— Alexandra Dublanche (@ADublanche) September 8, 2020

In a close-up and up-scaling of the image, we can see that it’s a detailed floor-plan of the server SoC Rhea, labelled for a target TSMC 7nm process.

We can detail 72 CPU cores and 68 mesh network L3 cache slices in the floor plan, surrounded by various IP whose labels are too small to be legible. SiPearl had previously confirmed that the project uses Arm’s upcoming Neoverse “Zeus” cores which succeed the Neoverse N1 Ares cores that are being used in current generation Arm server SoC designs such as Amazon’s Graviton2 or Ampere’s Altra.

Beyond the confirmation of a core-count, we also see that the Rhea design sports a high-end memory subsystem, with the floor plan labelled as having 4x HBM2E controllers and 4-6 DDR5 controllers. Such a hybrid memory system would allow for extremely high bandwidth to be able to feed such a large number of cores, while still falling back to regular DIMMs to be able to scale in memory capacity.

 

The Rhea family of processors is roadmapped to come to market in 2021. The only curious divergence here is that SiPearl previously stated that this was an N6 project, whilst the recent Twitter picture states it being N7. Given that both processes are design compatible, it might just be a recent shift in the project, or the company still plans to productise it in the N6 node when it comes to market.

The design’s aggressive memory subsystem with the inclusion of HBM2E points out that the company is aiming for quite high performance targets, joining the ranks of Fujitsu in terms of designing a CPU with advanced HBM memory.

Related Reading:

 

In what seems to be a major blunder by the SiPearl PR team, a recent visit by a local French politician resulted in the public Twitter posting in what looks like the floor plan of the company’s first-generation server chip project “Rhea”.

During a visit of Alexandra Dublanche, a local Île-de-France politician and vice president of economic development and other responsibilities for the region, the PR teams had made some photo-op captures of a tour of the office. Amongst the published pictures includes a shot of the company’s Rhea server chip project with some new, until now unreleased details of the European Processor Initiative-backed project.

In a close-up and up-scaling of the image, we can see that it’s a detailed floor-plan of the server SoC Rhea, labelled for a target TSMC 7nm process.

We can detail 72 CPU cores and 68 mesh network L3 cache slices in the floor plan, surrounded by various IP whose labels are too small to be legible. SiPearl had previously confirmed that the project uses Arm’s upcoming Neoverse “Zeus” cores which succeed the Neoverse N1 Ares cores that are being used in current generation Arm server SoC designs such as Amazon’s Graviton2 or Ampere’s Altra.

Beyond the confirmation of a core-count, we also see that the Rhea design sports a high-end memory subsystem, with the floor plan labelled as having 4x HBM2E controllers and 4-6 DDR5 controllers. Such a hybrid memory system would allow for extremely high bandwidth to be able to feed such a large number of cores, while still falling back to regular DIMMs to be able to scale in memory capacity.

 

The Rhea family of processors is roadmapped to come to market in 2021. The only curious divergence here is that SiPearl previously stated that this was an N6 project, whilst the recent Twitter picture states it being N7. Given that both processes are design compatible, it might just be a recent shift in the project, or the company still plans to productise it in the N6 node when it comes to market.

The design’s aggressive memory subsystem with the inclusion of HBM2E points out that the company is aiming for quite high performance targets, joining the ranks of Fujitsu in terms of designing a CPU with advanced HBM memory.

Related Reading:

 

Marvell Refocuses Thunder Server Platforms Towards Custom Silicon Business

Yesterday during Marvell’s quarterly earnings call, the company had made a surprise announcement that they are planning to restructure their server processor development team towards fully custom solutions, abandoning plans for “off-the-shelf” product designs.

The relevant earning call statements are as follows:

“Very much aligned with our growing emphasis on custom solutions, we are evolving our ARM-based server processor efforts toward a custom engagement model.

[…]

Having worked with them for multiple generations, it has become apparent that the long-term opportunity is for ARM server processors customized to their specific use cases rather than the standard off-the-shelf products. The power of the ARM architecture has always been in its ability to be integrated into highly customized designs optimized for specific use cases, and we see hyperscale data center applications is no different. With our breadth of processor know-how and now our custom ASIC capability, Marvell is uniquely positioned to address this opportunity. The significant amount of unique ARM server processor IP and technology we have developed over the last few years is ideal to create the custom processors hyperscalers are requesting.

Therefore, we have decided to target future investments in the ARM server market exclusively on custom solutions. The business model will be similar to our ASIC and custom programs where customers contribute engineering and mask expenses through NRE for us to develop and produce products specifically for them. We believe that this is the best way for us to continue to drive the growing adoption of ARM-based compute within the server market.”

We’ve had the opportunity to make a follow-up call with the teams at Marvell to get a little more background on the reasoning for such a move, given that only 6 months ago during the launch of the ThunderX3, the company had stated they were planning to ship products by the end of this year.

Effectively, as we’ve come to understand it, is that Marvell views the Arm server market at this moment in time to be purely concentrated around the big hyperscaler customers which have specific requirements in terms of their workloads, which require specific architecture optimisations.

Marvell sees the market beyond these hyperscaler customers to not be significant enough to be of sufficient value to engage in, and thus the company prefers to refocus their efforts in towards closer collaborations with hyperscaler costumers and fulfilling their needs.

The statement paints a relatively bleak view of the open Arm server market right now; in Marvell’s words, they do not rule out off-the-shelf products and designs in several years’ time when and if Arm servers become ubiquitous, but that currently is not the best financial strategy for the current ecoysystem. That’s quite a harsh view of the market and puts into question the ambitions of other Arm server vendors such as Ampere.

The company seemed very upbeat about the custom semicon design business, and they’re seemingly seeing large amounts of interest in the latest generation 5nm custom solutions they’re able to offer.

The company stated during the earning call that it still plans to ship the ThunderX3 by the end of this year, but says that beyond that it’s looking for custom opportunities for their hyperscaler customers.

We don’t yet have confirmation what this means for the dual-die ThunderX3 part that was planned for 2021, but it seems that we won’t be seeing a publicly available product with the in-design ThunderX4 anymore, which is unfortunate given that the company had presented their chip roadmap through 2022 only a few weeks ago at HotChips.

Although the company states that it’ll continue to leverage its custom IP for the future, I do wonder if the move has anything to do with Arm’s recent rise in the datacentre, and their very competitive Neoverse CPU microarchitectures and custom interconnects, essentially allowing anybody to design highly customizable products in-house, creating significant competition in the market.

From Marvell’s perspective, this all seems to make perfect sense as the company is simply readjusting towards where the money and maximum revenue growth opportunities lie. Having a hyperscaler win and keeping it is already a significant pie of the total market, and I think that’s what Marvell’s goal is here in the next several years.

Related Reading:

Yesterday during Marvell’s quarterly earnings call, the company had made a surprise announcement that they are planning to restructure their server processor development team towards fully custom solutions, abandoning plans for “off-the-shelf” product designs.

The relevant earning call statements are as follows:

"Very much aligned with our growing emphasis on custom solutions, we are evolving our ARM-based server processor efforts toward a custom engagement model.

[…]

Having worked with them for multiple generations, it has become apparent that the long-term opportunity is for ARM server processors customized to their specific use cases rather than the standard off-the-shelf products. The power of the ARM architecture has always been in its ability to be integrated into highly customized designs optimized for specific use cases, and we see hyperscale data center applications is no different. With our breadth of processor know-how and now our custom ASIC capability, Marvell is uniquely positioned to address this opportunity. The significant amount of unique ARM server processor IP and technology we have developed over the last few years is ideal to create the custom processors hyperscalers are requesting.

Therefore, we have decided to target future investments in the ARM server market exclusively on custom solutions. The business model will be similar to our ASIC and custom programs where customers contribute engineering and mask expenses through NRE for us to develop and produce products specifically for them. We believe that this is the best way for us to continue to drive the growing adoption of ARM-based compute within the server market."

We’ve had the opportunity to make a follow-up call with the teams at Marvell to get a little more background on the reasoning for such a move, given that only 6 months ago during the launch of the ThunderX3, the company had stated they were planning to ship products by the end of this year.

Effectively, as we’ve come to understand it, is that Marvell views the Arm server market at this moment in time to be purely concentrated around the big hyperscaler customers which have specific requirements in terms of their workloads, which require specific architecture optimisations.

Marvell sees the market beyond these hyperscaler customers to not be significant enough to be of sufficient value to engage in, and thus the company prefers to refocus their efforts in towards closer collaborations with hyperscaler costumers and fulfilling their needs.

The statement paints a relatively bleak view of the open Arm server market right now; in Marvell’s words, they do not rule out off-the-shelf products and designs in several years’ time when and if Arm servers become ubiquitous, but that currently is not the best financial strategy for the current ecoysystem. That’s quite a harsh view of the market and puts into question the ambitions of other Arm server vendors such as Ampere.

The company seemed very upbeat about the custom semicon design business, and they’re seemingly seeing large amounts of interest in the latest generation 5nm custom solutions they’re able to offer.

The company stated during the earning call that it still plans to ship the ThunderX3 by the end of this year, but says that beyond that it’s looking for custom opportunities for their hyperscaler customers.

We don’t yet have confirmation what this means for the dual-die ThunderX3 part that was planned for 2021, but it seems that we won’t be seeing a publicly available product with the in-design ThunderX4 anymore, which is unfortunate given that the company had presented their chip roadmap through 2022 only a few weeks ago at HotChips.

Although the company states that it’ll continue to leverage its custom IP for the future, I do wonder if the move has anything to do with Arm’s recent rise in the datacentre, and their very competitive Neoverse CPU microarchitectures and custom interconnects, essentially allowing anybody to design highly customizable products in-house, creating significant competition in the market.

From Marvell’s perspective, this all seems to make perfect sense as the company is simply readjusting towards where the money and maximum revenue growth opportunities lie. Having a hyperscaler win and keeping it is already a significant pie of the total market, and I think that’s what Marvell’s goal is here in the next several years.

Related Reading:

Hot Chips 2020: Marvell Details ThunderX3 CPUs – Up to 60 Cores Per Die, 96 Dual-Die in 2021

Today as part of HotChips 2020 we saw Marvell finally reveal some details on the microarchitecture of their new ThunderX3 server CPUs and core microarchitectures. The company had announced the existence of the new server and infrastructure processor b…

Today as part of HotChips 2020 we saw Marvell finally reveal some details on the microarchitecture of their new ThunderX3 server CPUs and core microarchitectures. The company had announced the existence of the new server and infrastructure processor back in March, and is now able to share more concrete specifications about how the in-house CPU design team promises to distinguish itself from the quickly growing competition that is the Arm server market.

We had reviewed the ThunderX2 back in 2018 – at the time still a Cavium product before the designs and teams were acquired by Marvell only a few months later that year. Ever since, the Arm server ecosystem has been jump-started by Arm’s Neoverse N1 CPU core and partner designs such as from Amazon (Graviton2) and Ampere (Altra), a quite different set of circumstances and alongside AMD’s successful return in the market, a very different landscape.

Chenbro Announces RB13804, a Dual Socket 1U Xeon 4-Bay HPC Barebones Server

For nearly four decades, Chinese manufacturer Chenbro has been one of the major players in the OEM and ODM rackmount server market. Now the company is releasing a new 1U rack-mount server for the Dual Xeon market. The RB13804 is the latest 1U rack-mountable barebones system to come out of the company and is based on Intel’s C624 platform, offering support for up to two Intel second-generation Xeon Scalable processors. 

One of the notable features of the Chenbro RB13804 is storage options: it includes support for up to four hot-swappable 3.5″ drives, space for up to two 2.5″ drives, and optional mounting for a Slim optical drive. Touching on the core specifications, the RB13804 is using an Intel S2600WFTR server board with two Intel LGA3647 sockets, supporting Intel’s second-generation Cascade Lake SP Xeon Scalable processors up to a maximum of 165 W per socket. For memory, the 1U rackmount barebones kit has twenty-four memory slots with support for up to 7.5 TB of DDR4-2933 UDIMM ECC memory.

It also has high-level networking capabilities with dual 10 GbE Base-T Ethernet ports, a single Ethernet management port, three USB 3.0 Type-A ports, and an integrated RAID module for high-level storage deployment. Other features include six 40×56 mm cooling fans mounted in the middle, with a maximum speed of 22.5 K rpm. Included with the barebones kit is an 800 W CRPS 80 PLUS Platinum power supply, two empty FHHL PCIe 3.0 x16 slots, and one I/O module. Overall the server weighs a hefty 11.6 KG, and that is without other hardware installed.

Otherwise, the Chenbro RB13804 uses a tool-less top cover for easier maintenance and upgrade path-ways. Chenbro is offering the RB13804 for ODMs who want to customize and add logo branding designed for specific customers needs, with a focus on flexibility to allow the server to be used in multiple roles. The RB13804 is available globally through Chenbro direct and its partners, though the company isn’t publishing official pricing at this time.

Related Reading

For nearly four decades, Chinese manufacturer Chenbro has been one of the major players in the OEM and ODM rackmount server market. Now the company is releasing a new 1U rack-mount server for the Dual Xeon market. The RB13804 is the latest 1U rack-mountable barebones system to come out of the company and is based on Intel's C624 platform, offering support for up to two Intel second-generation Xeon Scalable processors. 

One of the notable features of the Chenbro RB13804 is storage options: it includes support for up to four hot-swappable 3.5" drives, space for up to two 2.5" drives, and optional mounting for a Slim optical drive. Touching on the core specifications, the RB13804 is using an Intel S2600WFTR server board with two Intel LGA3647 sockets, supporting Intel's second-generation Cascade Lake SP Xeon Scalable processors up to a maximum of 165 W per socket. For memory, the 1U rackmount barebones kit has twenty-four memory slots with support for up to 7.5 TB of DDR4-2933 UDIMM ECC memory.

It also has high-level networking capabilities with dual 10 GbE Base-T Ethernet ports, a single Ethernet management port, three USB 3.0 Type-A ports, and an integrated RAID module for high-level storage deployment. Other features include six 40x56 mm cooling fans mounted in the middle, with a maximum speed of 22.5 K rpm. Included with the barebones kit is an 800 W CRPS 80 PLUS Platinum power supply, two empty FHHL PCIe 3.0 x16 slots, and one I/O module. Overall the server weighs a hefty 11.6 KG, and that is without other hardware installed.

Otherwise, the Chenbro RB13804 uses a tool-less top cover for easier maintenance and upgrade path-ways. Chenbro is offering the RB13804 for ODMs who want to customize and add logo branding designed for specific customers needs, with a focus on flexibility to allow the server to be used in multiple roles. The RB13804 is available globally through Chenbro direct and its partners, though the company isn't publishing official pricing at this time.

Related Reading