As the deployment of PCIe 5.0 picks up steam in both datacenter and consumer markets, PCI-SIG is not sitting idle, and is already working on getting the ecosystem ready for the updats to the PCIe specifications. At FMS 2024, some vendors were even talking about PCIe 7.0 with its 128 GT/s capabilities despite PCIe 6.0 not even starting to ship yet. We caught up with PCI-SIG to get some updates on its activities and have a discussion on the current state of the PCIe ecosystem.
PCI-SIG has already made the PCIe 7.0 specifications (v 0.5) available to its members, and expects full specifications to be officially released sometime in 2025. The goal is to deliver a 128 GT/s data rate with up to 512 GBps of bidirectional traffic using x16 links. Similar to PCIe 6.0, this specification will also utilize PAM4 signaling and maintain backwards compatibility. Power efficiency as well as silicon die area are also being kept in mind as part of the drafting process.
The move to PAM4 signaling brings higher bit-error rates compared to the previous NRZ scheme. This made it necessary to adopt a different error correction scheme in PCIe 6.0 - instead of operating on variable length packets, PCIe 6.0's Flow Control Unit (FLIT) encoding operates on fixed size packets to aid in forward error correction. PCIe 7.0 retains these aspects.
The integrators list for the PCIe 6.0 compliance program is also expected to come out in 2025, though initial testing is already in progress. This was evident by the FMS 2024 demo involving Cadence's 3nm test chip for its PCIe 6.0 IP offering along with Teledyne Lecroy's PCIe 6.0 analyzer. These timelines track well with the specification completion dates and compliance program availability for previous PCIe generations.
We also received an update on the optical workgroup - while being optical-technology agnostic, the WG also intends to develop technology-specific form-factors including pluggable optical transceivers, on-board optics, co-packaged optics, and optical I/O. The logical and electrical layers of the PCIe 6.0 specifications are being enhanced to accommodate the new optical PCIe standardization and this process will also be done with PCIe 7.0 to coincide with that standard's release next year.
The PCI-SIG also has ongoing cabling initiatives. On the consumer side, we have seen significant traction for Thunderbolt and external GPU enclosures. However, even datacenters and enterprise systems are moving towards cabling solutions as it becomes evident that disaggregation of components such as storage from the CPU and GPU are better for thermal design. Additionally maintaining signal integrity over longer distances becomes difficult for on-board signal traces. Cabling internal to the computing systems can help here.
OCuLink emerged as a good candidate and was adopted fairly widely as an internal link in server systems. It has even made an appearance in mini-PCs from some Chinese manufacturers in its external avatar for the consumer market, albeit with limited traction. As speeds increase, a widely-adopted standard for external PCIe peripherals (or even connecting components within a system) will become imperative.
StorageMicrochip Demonstrates Flashtec 5016 Enterprise SSD Controller Microchip recently announced the availability of their second PCIe Gen 5 enterprise SSD controller - the Flashtec 5016. Like the 4016, this is also a 16-channel controller, but there are some key updates: PCIe 5.0 lane organization: Operation in x4 or dual independent x2 / x2 mode in the 5016, compared to the x8, or x4, or dual independent x4 / x2 mode in the 4016. DRAM support: Four ranks of DDR5-5200 in the 5016, compared to two ranks of DDR4-3200 in the 4016. Extended NAND support: 2400 MT/s NAND in the 4016, compared to the 3200 MT/s NAND support in the 5016. Performance improvements: The 5016 is capable of delivering 3.5M+ random read IOPS compared to the 3M+ of the 4016. Microchip's enterprise SSD controllers provide a high level of flexibility to SSD vendors by providing them with significant horsepower and accelerators. The 5016 includes Cortex-A53 cores for SSD vendors to run custom applications relevant to SSD management. However, compared to the Gen4 controllers, there are two additional cores in the CPU cluster. The DRAM subsystem includes ECC support (both out-of-band and inline, as desired by the SSD vendor). At FMS 2024, the company demonstrated an application of the neural network engines embedded in the Gen5 controllers. Controllers usually employ a 'read-retry' operation with altered read-out voltages for flash reads that do not complete successfully. Microchip implemented a machine learning approach to determine the read-out voltage based on the health history of the NAND block using the NN engines in the controller. This approach delivers tangible benefits for read latency and power consumption (thanks to a smaller number of errors on the first read). The 4016 and 5016 come with a single-chip root of trust implementation for hardware security. A secure boot process with dual-signature authentication ensures that the controller firmware is not maliciously altered in the field. The company also brought out the advantages of their controller's implementation of SR-IOV, flexible data placement, and zoned namespaces along with their 'credit engine' scheme for multi-tenant cloud workloads. These aspects were also brought out in other demonstrations. Microchip's press release included quotes from the usual NAND vendors - Solidigm, Kioxia, and Micron. On the customer front, Longsys has been using Flashtec controllers in their enterprise offerings along with YMTC NAND. It is likely that this collaboration will continue further using the new 5016 controller. Storage
U.S. Signs $1.5B in CHIPS Act Agreements With Amkor and SKhynix for Chip Packaging Plants Under the CHIPS & Science Act, the U.S. government provided tens of billions of dollars in grants and loans to the world's leading maker of chips, such as Intel, Samsung, and TSMC, which will significantly expand the country's semiconductor production industry in the coming years. However, most chips are typically tested, assembled, and packaged in Asia, which has left the American supply chain incomplete. Addressing this last gap in the government's domestic chip production plans, these past couple of weeks the U.S. government signed memorandums of understanding worth about $1.5 billion with Amkor and SK hynix to support their efforts to build chip packaging facilities in the U.S.
Later this year Intel is set to introduce its Xeon 6-branded processors, codenamed Granite Rapids (6x00P) and Sierra Forest (6x00E). And with it will come a new slew of server motherboards and pre-built server platforms to go with it. On the latter note, this will be the first generation where Intel won't be offering any pre-builts of its own, after selling that business off to MiTAC last year.
To that end, MiTAC and its subsidiary Tyan were at this year's event to demonstrate what they've been up to since acquiring Intel's server business unit, as well as to show off the server platforms they're developing for the Xeon 6 family. Altogether, the companies had two server platforms on display – a compact 2S system, and a larger 2S system with significant expansion capabilities – as well as a pair of single-socket designs from Tyan.
The most basic platform that MiTAC had to show is their TX86-E7148 (Katmai Pass), a half-width 1U system that's the successor to Intel's D50DNP platform. Katmai Pass has two CPU sockets, supports up to 2 TB of DDR5-6400 RDIMMs over 16 slots (8 per CPU), and has two low-profile PCIe 5.0 x16 slots. Like its predecessor, this platform is aimed at mainstream servers that do not need a lot of storage or room to house bulky add-in cards like AI accelerators.
The company's other platform is TX77A-E7142 (Deer Creek Pass), a considerably more serious offering that replaces Intel's M50FCP platform. This board can house up to 4 TB of DDR5-6400 RDIMMs over 32 slots (16 per CPU with 2DPC), four PCIe 5.0 x16 slots, one PCIe 5.0 x8 slot, two OCP 3.0 slots, and 24 hot-swap U.2 bays. Deer Creek Pass can be used both for general-purpose workloads, high-performance storage, as well as workloads that require GPUs or other special-purpose accelerators.
Meanwhile Tyan had the single-socket Thunder CX GC73A-B5660 on display. That system supports up to 2 TB of DDR5-6400 memory over 16 RDIMMs and offers two PCIe 5.0 x16 slots, one PCIe 4.0 x4 M.2 slot, two OCP 3.0 slots, and 12 hot-swappable U.2 drive bays.
Finally, Tyan's Thunder HX S5662 is an HPC server board specifically designed to house multiple AI accelerators and other large PCIe cards. This board supports one Xeon 6 6700 processor, up to 1 TB of memory over eight DDR5-6400 RDIMMs, and has five tradiitonal PCIe 5.0 x16 slots as well as two PCIe 5.0 x2 M.2 slots for storage.
MiTAC is expected to start shipments of these new Xeon 6 motherboards in the coming months, as Intel rolls out its next-generation datacenter CPUs. Pricing of these platforms is unknown for now, but expect it to be comparable to... Servers
The CXL consortium has had a regular presence at FMS (which rechristened itself from 'Flash Memory Summit' to the 'Future of Memory and Storage' this year). Back at FMS 2022, the company had announced v3.0 of the CXL specifications. This was followed by CXL 3.1's introduction at Supercomputing 2023. Having started off as a host to device interconnect standard, it had slowly subsumed other competing standards such as OpenCAPI and Gen-Z. As a result, the specifications started to encompass a wide variety of use-cases by building a protocol on top of the the ubiquitous PCIe expansion bus. The CXL consortium comprises of heavyweights such as AMD and Intel, as well as a large number of startup companies attempting to play in different segments on the device side. At FMS 2024, CXL had a prime position in the booth demos of many vendors.
The migration of server platforms from DDR4 to DDR5, along with the rise of workloads demanding large RAM capacity (but not particularly sensitive to either memory bandwidth or latency), has opened up memory expansion modules as one of the first set of widely available CXL devices. Over the last couple of years, we have had product announcements from Samsung and Micron in this area.
At FMS 2024, SK hynix was showing off their DDR5-based CMM-DDR5 CXL memory module with a 128 GB capacity. The company was also detailing their associated Heterogeneous Memory Software Development Kit (HMSDK) - a set of libraries and tools at both the kernel and user levels aimed at increasing the ease of use of CXL memory. This is achieved in part by considering the memory pyramid / hierarchy and relocating the data between the server's main memory (DRAM) and the CXL device based on usage frequency.
The CMM-DDR5 CXL memory module comes in the SDFF form-factor (E3.S 2T) with a PCIe 3.0 x8 host interface. The internal memory is based on 1α technology DRAM, and the device promises DDR5-class bandwidth and latency within a single NUMA hop. As these memory modules are meant to be used in datacenters and enterprises, the firmware includes features for RAS (reliability, availability, and serviceability) along with secure boot and other management features.
SK hynix was also demonstrating Niagara 2.0 - a hardware solution (currently based on FPGAs) to enable memory pooling and sharing - i.e, connecting multiple CXL memories to allow different hosts (CPUs and GPUs) to optimally share their capacity. The previous version only allowed capacity sharing, but the latest version enables sharing of data also. SK hynix had presented these solutions at the CXL DevCon 2024 earlier this year, but some progress seems to have been made in finalizing the specifications of the CMM-DDR5 at FMS 2024.
Micron had unveiled the CZ120 CXL Memory Expansion Module last year based on the Microchip SMC 2000 series CXL memory controller. At FMS 2024, Micron and Microchip had a demonstration of the module on a Granite Rapids server.
Additional insights into the SMC 2000 controller were also provided.
The CXL memory controller also incorporates DRAM die failure handling, and Microchip also provides diagnostics and debug tools to analyze failed modules. The memory controller also supports ECC, which forms part of the enterprise... Storage
At FMS 2024, the technological requirements from the storage and memory subsystem took center stage. Both SSD and controller vendors had various demonstrations touting their suitability for different stages of the AI data pipeline - ingestion, preparation, training, checkpointing, and inference. Vendors like Solidigm have different types of SSDs optimized for different stages of the pipeline. At the same time, controller vendors have taken advantage of one of the features introduced recently in the NVM Express standard - Flexible Data Placement (FDP).
FDP involves the host providing information / hints about the areas where the controller could place the incoming write data in order to reduce the write amplification. These hints are generated based on specific block sizes advertised by the device. The feature is completely backwards-compatible, with non-FDP hosts working just as before with FDP-enabled SSDs, and vice-versa.
Silicon Motion's MonTitan Gen 5 Enterprise SSD Platform was announced back in 2022. Since then, Silicon Motion has been touting the flexibility of the platform, allowing its customers to incorporate their own features as part of the customization process. This approach is common in the enterprise space, as we have seen with Marvell's Bravera SC5 SSD controller in the DapuStor SSDs and Microchip's Flashtec controllers in the Longsys FORESEE enterprise SSDs.
At FMS 2024, the company was demonstrating the advantages of flexible data placement by allowing a single QLC SSD based on their MonTitan platform to take part in different stages of the AI data pipeline while maintaining the required quality of service (minimum bandwidth) for each process. The company even has a trademarked name (PerformaShape) for the firmware feature in the controller that allows the isolation of different concurrent SSD accesses (from different stages in the AI data pipeline) to guarantee this QoS. Silicon Motion claims that this scheme will enable its customers to get the maximum write performance possible from QLC SSDs without negatively impacting the performance of other types of accesses.
Silicon Motion and Phison have market leadership in the client SSD controller market with similar approaches. However, their enterprise SSD controller marketing couldn't be more different. While Phison has gone in for a turnkey solution with their Gen 5 SSD platform (to the extent of not adopting the white label route for this generation, and instead opting to get the SSDs qualified with different cloud service providers themselves), Silicon Motion is opting for a different approach. The flexibility and customization possibilities can make platforms like the MonTitan appeal to flash array vendors.
Storage
0 Comments