With the rise of the handheld gaming PC market, we've seen PC vendors and their partners toy with a number of tricks and tweaks to improve improve framerates in games, with some of their latest efforts on display at this year's Computex trade show. Perhaps the most interesting find thus far comes from ADATA sub-brand XPG, who is demoing their prototype "Nia" handheld PC, which uses eye tracking and dynamic foveated rendering to further improve their rendering performance.
For those unfamiliar, dynamic foveated rendering is a graphics technique that is sometimes used to boost performance in virtual reality (VR) and augmented reality (AR) applications by taking advantage of how human vision works. Typically, humans can only perceive detailed imagery in the relatively small central area of our vision called the fovea, while our peripheral vision is much less detailed. Dynamic foveated rendering, in turn, exploits this by using real-time eye tracking to determine where the user is looking, and then rendering just that area in high/full resolution, while rendering the peripheral areas in lower resolution. The net result is that only a fraction of the screen is rendered at full detail, which cuts down on the total amount of rendering work required and boosting framerates on performance-limited devices.
As stated before, this technology is sometimes used in high-end AR/VR headsets, where high resolution displays are placed mere inches from one's face. This ends up being an ideal use case for the technique, since at those distances, only a small fraction of the screen is within the fovea.
Using dynamic foveated rendering for a handheld, on the other hand, is a more novel application. All of the same visual principles apply, but the resolutions at play are lower, and the screen is farther from the users' eyes. This makes a handheld device a less ideal use case, at least on paper, as a larger portion of the screen is going to be in the fovea, and thus will need to be rendered at full resolution. None the less, it will be interesting to see how XPG's efforts pan out, and if dynamic foveated rendering is beneficial enough for handheld PCs. As we sometimes see with trade show demos, not everything makes it out of the prototype stage.
According to a press release put out by ADATA ahead of the trade show, the eye tracking technology is being provided by AMD collaborator Eyeware. Notably, their software-based approach runs on top of standard webcams, rather than requiring IR cameras. So the camera hardware itself should be pretty straight-forward.
Foveated rendering aside, XPG is making sure that the Nia won't be a one-trick pony. The handheld's other major claim to fame is its hardware swappability. The prototype handheld not only features a removable M.2-2230 SSD, but the company is also taking advantage of the recently-introduced LPCAMM2 memory module standard to introduce removable DRAM. Via a hatch in the back of the handheld, device owners would be able to swap out LPCAMM2 LPDDR5X modules for higher capacity versions. This would give the handheld an additional degree of future-proofness over current handhelds, which use non-replaceable soldered-down memory.
Rounding out the package, the current prototype is based on an AMD's Zen 4 Phoenix APU, which is used across both of the company's current mobile lines (Ryzen Mobile 7000/8000 and Ryzen Z1). Meanwhile, the unit's display is adjustable, allowing it to be angled away from the body of the handheld.
Assuming all goes well with the prototype, XPG aims to release a finished product in 2025.
ADATAOptical connectivity – and especially silicon photonics – is expected to become a crucial technology to enable connectivity for next-generation datacenters, particularly those designed HPC applications. With ever-increasing bandwidth requirements needed to keep up with (and keep scaling out) system performance, copper signaling alone won't be enough to keep up. To that end, several companies are developing silicon photonics solutions, including fab providers like TSMC, who this week outlined its 3D Optical Engine roadmap as part of its 2024 North American Technology Symposium, laying out its plan to bring up to 12.8 Tbps optical connectivity to TSMC-fabbed processors.
TSMC's Compact Universal Photonic Engine (COUPE) stacks an electronics integrated circuit on photonic integrated circuit (EIC-on-PIC) using the company's SoIC-X packaging technology. The foundry says that usage of its SoIC-X enables the lowest impedance at the die-to-die interface and therefore the highest energy efficiency. The EIC itself is produced at a 65nm-class process technology.
TSMC's 1st Generation 3D Optical Engine (or COUPE) will be integrated into an OSFP pluggable device running at 1.6 Tbps. That's a transfer rate well ahead of current copper Ethernet standards – which top out at 800 Gbps – underscoring the immediate bandwidth advantage of optical interconnects for heavily-networked compute clusters, never mind the expected power savings.
Looking further ahead, the 2nd Generation of COUPE is designed to integrate into CoWoS packaging as co-packaged optics with a switch, allowing optical interconnections to be brought to the motherboard level. This version COUPE will support data transfer rates of up to 6.40 Tbps with reduced latency compared to the first version.
TSMC's third iteration of COUPE – COUPE running on a CoWoS interposer – is projected to improve on things one step further, increasing transfer rates to 12.8 Tbps while bringing optical connectivity even closer to the processor itself. At present, COUPE-on-CoWoS is in the pathfinding stage of development and TSMC does not have a target date set.
Ultimately, unlike many of its industry peers, TSMC has not participated in the silicon photonics market up until now, leaving this to players like GlobalFoundries. But with its 3D Optical Engine Strategy, the company will enter this important market as it looks to make up for lost time.
Kioxia's booth at FMS 2024 was a busy one with multiple technology demonstrations keeping visitors occupied. A walk-through of the BiCS 8 manufacturing process was the first to grab my attention. Kioxia and Western Digital announced the sampling of BiCS 8 in March 2023. We had touched briefly upon its CMOS Bonded Array (CBA) scheme in our coverage of Kioxial's 2Tb QLC NAND device and coverage of Western Digital's 128 TB QLC enterprise SSD proof-of-concept demonstration. At Kioxia's booth, we got more insights.
Traditionally, fabrication of flash chips involved placement of the associate logic circuitry (CMOS process) around the periphery of the flash array. The process then moved on to putting the CMOS under the cell array, but the wafer development process was serialized with the CMOS logic getting fabricated first followed by the cell array on top. However, this has some challenges because the cell array requires a high-temperature processing step to ensure higher reliability that can be detrimental to the health of the CMOS logic. Thanks to recent advancements in wafer bonding techniques, the new CBA process allows the CMOS wafer and cell array wafer to be processed independently in parallel and then pieced together, as shown in the models above.
The BiCS 8 3D NAND incorporates 218 layers, compared to 112 layers in BiCS 5 and 162 layers in BiCS 6. The company decided to skip over BiCS 7 (or, rather, it was probably a short-lived generation meant as an internal test vehicle). The generation retains the four-plane charge trap structure of BiCS 6. In its TLC avatar, it is available as a 1 Tbit device. The QLC version is available in two capacities - 1 Tbit and 2 Tbit.
Kioxia also noted that while the number of layers (218) doesn't compare favorably with the latest layer counts from the competition, its lateral scaling / cell shrinkage has enabled it to be competitive in terms of bit density as well as operating speeds (3200 MT/s). For reference, the latest shipping NAND from Micron - the G9 - has 276 layers with a bit density in TLC mode of 21 Gbit/mm2, and operates at up to 3600 MT/s. However, its 232L NAND operates only up to 2400 MT/s and has a bit density of 14.6 Gbit/mm2.
It must be noted that the CBA hybrid bonding process has advantages over the current processes used by other vendors - including Micron's CMOS under array (CuA) and SK hynix's 4D PUC (periphery-under-chip) developed in the late 2010s. It is expected that other NAND vendors will also move eventually to some variant of the hybrid bonding scheme used by Kioxia.
StorageOne of the core challenges that Rapidus will face when it kicks off volume production of chips on its 2nm-class process technology in 2027 is lining up customers. With Intel, Samsung, and TSMC all slated to offer their own 2nm-class nodes by that time, Rapidus will need some kind of advantage to attract customers away from its more established rivals. To that end, the company thinks they've found their edge: fully automated packaging that will allow for shorter chip lead times than manned packaging operations.
In an interview with Nikkei, Rapidus' president, Atsuyoshi Koike, outlined the company's vision to use advanced packaging as a competitive edge for the new fab. The Hokkaido facility, which is currently under construction and is expecting to begin equipment installation this December, is already slated to both produce chips and offer advanced packaging services within the same facility, an industry first. But ultimately, Rapidus biggest plan to differentiate itself is by automating the back-end fab processes (chip packaging) to provide significantly faster turnaround times.
Rapidus is targetting back-end production in particular as, compared to front-end (lithography) production, back-end production still heavily relies on human labor. No other advanced packaging fab has fully automated the process thus far, which provides for a degree of flexibility, but slows throughput. But with automation in place to handle this aspect of chip production, Rapidus would be able to increase chip packaging efficiency and speed, which is crucial as chip assembly tasks become more complex. Rapidus is also collaborating with multiple Japanese suppliers to source materials for back-end production.
"In the past, Japanese chipmakers tried to keep their technology development exclusively in-house, which pushed up development costs and made them less competitive," Koike told Nikkei. "[Rapidus plans to] open up technology that should be standardized, bringing down costs, while handling important technology in-house."
Financially, Rapidus faces a significant challenge, needing a total of ¥5 trillion ($35 billion) by the time mass production starts in 2027. The company estimates that ¥2 trillion will be required by 2025 for prototype production. While the Japanese government has provided ¥920 billion in aid, Rapidus still needs to secure substantial funding from private investors.
Due to its lack of track record and experience of chip production as. well as limited visibility for success, Rapidus is finding it difficult to attract private financing. The company is in discussions with the government to make it easier to raise capital, including potential loan guarantees, and is hopeful that new legislation will assist in this effort.
SemiconductorsThe 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
Optical connectivity – and especially silicon photonics – is expected to become a crucial technology to enable connectivity for next-generation datacenters, particularly those designed HPC applications. With ever-increasing bandwidth requirements needed to keep up with (and keep scaling out) system performance, copper signaling alone won't be enough to keep up. To that end, several companies are developing silicon photonics solutions, including fab providers like TSMC, who this week outlined its 3D Optical Engine roadmap as part of its 2024 North American Technology Symposium, laying out its plan to bring up to 12.8 Tbps optical connectivity to TSMC-fabbed processors.
TSMC's Compact Universal Photonic Engine (COUPE) stacks an electronics integrated circuit on photonic integrated circuit (EIC-on-PIC) using the company's SoIC-X packaging technology. The foundry says that usage of its SoIC-X enables the lowest impedance at the die-to-die interface and therefore the highest energy efficiency. The EIC itself is produced at a 65nm-class process technology.
TSMC's 1st Generation 3D Optical Engine (or COUPE) will be integrated into an OSFP pluggable device running at 1.6 Tbps. That's a transfer rate well ahead of current copper Ethernet standards – which top out at 800 Gbps – underscoring the immediate bandwidth advantage of optical interconnects for heavily-networked compute clusters, never mind the expected power savings.
Looking further ahead, the 2nd Generation of COUPE is designed to integrate into CoWoS packaging as co-packaged optics with a switch, allowing optical interconnections to be brought to the motherboard level. This version COUPE will support data transfer rates of up to 6.40 Tbps with reduced latency compared to the first version.
TSMC's third iteration of COUPE – COUPE running on a CoWoS interposer – is projected to improve on things one step further, increasing transfer rates to 12.8 Tbps while bringing optical connectivity even closer to the processor itself. At present, COUPE-on-CoWoS is in the pathfinding stage of development and TSMC does not have a target date set.
Ultimately, unlike many of its industry peers, TSMC has not participated in the silicon photonics market up until now, leaving this to players like GlobalFoundries. But with its 3D Optical Engine Strategy, the company will enter this important market as it looks to make up for lost time.
Kioxia's booth at FMS 2024 was a busy one with multiple technology demonstrations keeping visitors occupied. A walk-through of the BiCS 8 manufacturing process was the first to grab my attention. Kioxia and Western Digital announced the sampling of BiCS 8 in March 2023. We had touched briefly upon its CMOS Bonded Array (CBA) scheme in our coverage of Kioxial's 2Tb QLC NAND device and coverage of Western Digital's 128 TB QLC enterprise SSD proof-of-concept demonstration. At Kioxia's booth, we got more insights.
Traditionally, fabrication of flash chips involved placement of the associate logic circuitry (CMOS process) around the periphery of the flash array. The process then moved on to putting the CMOS under the cell array, but the wafer development process was serialized with the CMOS logic getting fabricated first followed by the cell array on top. However, this has some challenges because the cell array requires a high-temperature processing step to ensure higher reliability that can be detrimental to the health of the CMOS logic. Thanks to recent advancements in wafer bonding techniques, the new CBA process allows the CMOS wafer and cell array wafer to be processed independently in parallel and then pieced together, as shown in the models above.
The BiCS 8 3D NAND incorporates 218 layers, compared to 112 layers in BiCS 5 and 162 layers in BiCS 6. The company decided to skip over BiCS 7 (or, rather, it was probably a short-lived generation meant as an internal test vehicle). The generation retains the four-plane charge trap structure of BiCS 6. In its TLC avatar, it is available as a 1 Tbit device. The QLC version is available in two capacities - 1 Tbit and 2 Tbit.
Kioxia also noted that while the number of layers (218) doesn't compare favorably with the latest layer counts from the competition, its lateral scaling / cell shrinkage has enabled it to be competitive in terms of bit density as well as operating speeds (3200 MT/s). For reference, the latest shipping NAND from Micron - the G9 - has 276 layers with a bit density in TLC mode of 21 Gbit/mm2, and operates at up to 3600 MT/s. However, its 232L NAND operates only up to 2400 MT/s and has a bit density of 14.6 Gbit/mm2.
It must be noted that the CBA hybrid bonding process has advantages over the current processes used by other vendors - including Micron's CMOS under array (CuA) and SK hynix's 4D PUC (periphery-under-chip) developed in the late 2010s. It is expected that other NAND vendors will also move eventually to some variant of the hybrid bonding scheme used by Kioxia.
StorageOne of the core challenges that Rapidus will face when it kicks off volume production of chips on its 2nm-class process technology in 2027 is lining up customers. With Intel, Samsung, and TSMC all slated to offer their own 2nm-class nodes by that time, Rapidus will need some kind of advantage to attract customers away from its more established rivals. To that end, the company thinks they've found their edge: fully automated packaging that will allow for shorter chip lead times than manned packaging operations.
In an interview with Nikkei, Rapidus' president, Atsuyoshi Koike, outlined the company's vision to use advanced packaging as a competitive edge for the new fab. The Hokkaido facility, which is currently under construction and is expecting to begin equipment installation this December, is already slated to both produce chips and offer advanced packaging services within the same facility, an industry first. But ultimately, Rapidus biggest plan to differentiate itself is by automating the back-end fab processes (chip packaging) to provide significantly faster turnaround times.
Rapidus is targetting back-end production in particular as, compared to front-end (lithography) production, back-end production still heavily relies on human labor. No other advanced packaging fab has fully automated the process thus far, which provides for a degree of flexibility, but slows throughput. But with automation in place to handle this aspect of chip production, Rapidus would be able to increase chip packaging efficiency and speed, which is crucial as chip assembly tasks become more complex. Rapidus is also collaborating with multiple Japanese suppliers to source materials for back-end production.
"In the past, Japanese chipmakers tried to keep their technology development exclusively in-house, which pushed up development costs and made them less competitive," Koike told Nikkei. "[Rapidus plans to] open up technology that should be standardized, bringing down costs, while handling important technology in-house."
Financially, Rapidus faces a significant challenge, needing a total of ¥5 trillion ($35 billion) by the time mass production starts in 2027. The company estimates that ¥2 trillion will be required by 2025 for prototype production. While the Japanese government has provided ¥920 billion in aid, Rapidus still needs to secure substantial funding from private investors.
Due to its lack of track record and experience of chip production as. well as limited visibility for success, Rapidus is finding it difficult to attract private financing. The company is in discussions with the government to make it easier to raise capital, including potential loan guarantees, and is hopeful that new legislation will assist in this effort.
SemiconductorsThe 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.
StorageSamsung had quietly launched its BM1743 enterprise QLC SSD last month with a hefty 61.44 TB SKU. At FMS 2024, the company had the even larger 122.88 TB version of that SSD on display, alongside a few recorded benchmarking sessions. Compared to the previous generation, the BM1743 comes with a 4.1x improvement in I/O performance, improvement in data retention, and a 45% improvement in power efficiency for sequential writes.
The 128 TB-class QLC SSD boasts of sequential read speeds of 7.5 GBps and write speeds of 3 GBps. Random reads come in at 1.6 M IOPS, while 16 KB random writes clock in at 45K IOPS. Based on the quoted random write access granularity, it appears that Samsung is using a 16 KB indirection unit (IU) to optimize flash management. This is similar to the strategy adopted by Solidigm with IUs larger than 4K in their high-capacity SSDs.
A recorded benchmark session on the company's PM9D3a 8-channel Gen 5 SSD was also on display.
The SSD family is being promoted as a mainstream option for datacenters, and boasts of sequential reads up to 12 GBps and writes up to 6.8 GBps. Random reads clock in at 2 M IOPS, and random writes at 400 K IOPS.
Available in multiple form-factors up to 32 TB (M.2 tops out at 2 TB), the drive's firmware includes optional support for flexible data placement (FDP) to help address the write amplification aspect.
The PM1753 is the current enterprise SSD flagship in Samsung's lineup. With support for 16 NAND channels and capacities up to 32 TB, this U.2 / E3.S SSD has advertised sequential read and write speeds of 14.8 GBps and 11 GBps respectively. Random reads and writes for 4 KB accesses are listed at 3.4 M and 600 K IOPS.
Samsung claims a 1.7x performance improvement and a 1.7x power efficiency improvement over the previous generation (PM1743), making this TLC SSD suitable for AI servers.
The 9th Gen. V-NAND wafer was also available for viewing, though photography was prohibited. Mass production of this flash memory began in April 2024.
StorageOptical connectivity – and especially silicon photonics – is expected to become a crucial technology to enable connectivity for next-generation datacenters, particularly those designed HPC applications. With ever-increasing bandwidth requirements needed to keep up with (and keep scaling out) system performance, copper signaling alone won't be enough to keep up. To that end, several companies are developing silicon photonics solutions, including fab providers like TSMC, who this week outlined its 3D Optical Engine roadmap as part of its 2024 North American Technology Symposium, laying out its plan to bring up to 12.8 Tbps optical connectivity to TSMC-fabbed processors.
TSMC's Compact Universal Photonic Engine (COUPE) stacks an electronics integrated circuit on photonic integrated circuit (EIC-on-PIC) using the company's SoIC-X packaging technology. The foundry says that usage of its SoIC-X enables the lowest impedance at the die-to-die interface and therefore the highest energy efficiency. The EIC itself is produced at a 65nm-class process technology.
TSMC's 1st Generation 3D Optical Engine (or COUPE) will be integrated into an OSFP pluggable device running at 1.6 Tbps. That's a transfer rate well ahead of current copper Ethernet standards – which top out at 800 Gbps – underscoring the immediate bandwidth advantage of optical interconnects for heavily-networked compute clusters, never mind the expected power savings.
Looking further ahead, the 2nd Generation of COUPE is designed to integrate into CoWoS packaging as co-packaged optics with a switch, allowing optical interconnections to be brought to the motherboard level. This version COUPE will support data transfer rates of up to 6.40 Tbps with reduced latency compared to the first version.
TSMC's third iteration of COUPE – COUPE running on a CoWoS interposer – is projected to improve on things one step further, increasing transfer rates to 12.8 Tbps while bringing optical connectivity even closer to the processor itself. At present, COUPE-on-CoWoS is in the pathfinding stage of development and TSMC does not have a target date set.
Ultimately, unlike many of its industry peers, TSMC has not participated in the silicon photonics market up until now, leaving this to players like GlobalFoundries. But with its 3D Optical Engine Strategy, the company will enter this important market as it looks to make up for lost time.
Kioxia's booth at FMS 2024 was a busy one with multiple technology demonstrations keeping visitors occupied. A walk-through of the BiCS 8 manufacturing process was the first to grab my attention. Kioxia and Western Digital announced the sampling of BiCS 8 in March 2023. We had touched briefly upon its CMOS Bonded Array (CBA) scheme in our coverage of Kioxial's 2Tb QLC NAND device and coverage of Western Digital's 128 TB QLC enterprise SSD proof-of-concept demonstration. At Kioxia's booth, we got more insights.
Traditionally, fabrication of flash chips involved placement of the associate logic circuitry (CMOS process) around the periphery of the flash array. The process then moved on to putting the CMOS under the cell array, but the wafer development process was serialized with the CMOS logic getting fabricated first followed by the cell array on top. However, this has some challenges because the cell array requires a high-temperature processing step to ensure higher reliability that can be detrimental to the health of the CMOS logic. Thanks to recent advancements in wafer bonding techniques, the new CBA process allows the CMOS wafer and cell array wafer to be processed independently in parallel and then pieced together, as shown in the models above.
The BiCS 8 3D NAND incorporates 218 layers, compared to 112 layers in BiCS 5 and 162 layers in BiCS 6. The company decided to skip over BiCS 7 (or, rather, it was probably a short-lived generation meant as an internal test vehicle). The generation retains the four-plane charge trap structure of BiCS 6. In its TLC avatar, it is available as a 1 Tbit device. The QLC version is available in two capacities - 1 Tbit and 2 Tbit.
Kioxia also noted that while the number of layers (218) doesn't compare favorably with the latest layer counts from the competition, its lateral scaling / cell shrinkage has enabled it to be competitive in terms of bit density as well as operating speeds (3200 MT/s). For reference, the latest shipping NAND from Micron - the G9 - has 276 layers with a bit density in TLC mode of 21 Gbit/mm2, and operates at up to 3600 MT/s. However, its 232L NAND operates only up to 2400 MT/s and has a bit density of 14.6 Gbit/mm2.
It must be noted that the CBA hybrid bonding process has advantages over the current processes used by other vendors - including Micron's CMOS under array (CuA) and SK hynix's 4D PUC (periphery-under-chip) developed in the late 2010s. It is expected that other NAND vendors will also move eventually to some variant of the hybrid bonding scheme used by Kioxia.
StorageOne of the core challenges that Rapidus will face when it kicks off volume production of chips on its 2nm-class process technology in 2027 is lining up customers. With Intel, Samsung, and TSMC all slated to offer their own 2nm-class nodes by that time, Rapidus will need some kind of advantage to attract customers away from its more established rivals. To that end, the company thinks they've found their edge: fully automated packaging that will allow for shorter chip lead times than manned packaging operations.
In an interview with Nikkei, Rapidus' president, Atsuyoshi Koike, outlined the company's vision to use advanced packaging as a competitive edge for the new fab. The Hokkaido facility, which is currently under construction and is expecting to begin equipment installation this December, is already slated to both produce chips and offer advanced packaging services within the same facility, an industry first. But ultimately, Rapidus biggest plan to differentiate itself is by automating the back-end fab processes (chip packaging) to provide significantly faster turnaround times.
Rapidus is targetting back-end production in particular as, compared to front-end (lithography) production, back-end production still heavily relies on human labor. No other advanced packaging fab has fully automated the process thus far, which provides for a degree of flexibility, but slows throughput. But with automation in place to handle this aspect of chip production, Rapidus would be able to increase chip packaging efficiency and speed, which is crucial as chip assembly tasks become more complex. Rapidus is also collaborating with multiple Japanese suppliers to source materials for back-end production.
"In the past, Japanese chipmakers tried to keep their technology development exclusively in-house, which pushed up development costs and made them less competitive," Koike told Nikkei. "[Rapidus plans to] open up technology that should be standardized, bringing down costs, while handling important technology in-house."
Financially, Rapidus faces a significant challenge, needing a total of ¥5 trillion ($35 billion) by the time mass production starts in 2027. The company estimates that ¥2 trillion will be required by 2025 for prototype production. While the Japanese government has provided ¥920 billion in aid, Rapidus still needs to secure substantial funding from private investors.
Due to its lack of track record and experience of chip production as. well as limited visibility for success, Rapidus is finding it difficult to attract private financing. The company is in discussions with the government to make it easier to raise capital, including potential loan guarantees, and is hopeful that new legislation will assist in this effort.
Semiconductors
0 Comments