At its European Technology Symposium last week TSMC revealed some of the details about its Global Gigafab Manufacturing program, the company's strategy to replicate its manufacturing processes across its multiple gigafab sites.
The need for large-scale multi-national fabs to have a process in place to replicate their facilities is well-documented at this point. As scaling-up at at the gigafab size means scaling-out instead, chip makers need to be able to quickly get new and updated manufacturing processes ported to other facilities in order to hit their necessary throughput – and to avoid a multi-quarter bottlenecks that come from having to freshly-tune a fab.
Intel, for their part, has a well-known Copy Exactly program, which is one of the company's major competitive advantages, allowing it to share process recipes across its fabs around the world to maximize yields and reduce performance variability. Meanwhile, as Taiwan Semiconductor Manufacturing Co. is building additional capacity in different parts of the world, it has reached the point where it needs a similar program in order to quickly maximize its yields and productivity at its new fabs in Japan and the U.S. And in some respects, TSMC's program goes even further than Intel's, with an additional focus on sustainability and social responsibility.
"As mentioned at last year's symposium, [Global Gigafab manufacturing] is a powerful global manufacturing and management platform," said Y.L. Wang, Vice President of Fab Operations TSMC. "We realise one fab management to ensure our Gigafab to achieve consistent operation efficiency as well as production quality on a global scale. Moreover, we also pursue sustainability across our global footprint covering green manufacturing, global talent development, supply chain localization, as well as social responsibility."
TSMC's Global GigaFab Manufacturing Data by TSMC (Compiled by AnandTech) |
|||
Manufacturing Excellence | Sustainability | ||
Global One Fab Manufacturing | Green Manufacturing | ||
ML-based Process Control | Global Talent Development | ||
Manufacturing Agility and Quality | Supply Chain Localization | ||
Maximum Productivity | Social Responsibility |
When it comes to improvements of process technology, there are two main mechanisms: the continuous process improvements (CPI) to improve yields, as well as statistical process control (SPC) reduce performance variations. To do so, the company has multiple internal techniques that rely on machine learning-based process control, constant quality measuring, and various productivity improving methods. With Global Gigafab manufacturing TSMC can use CPI and SPC to improve yields and performance on the global scale by sharing knowledge between different sites.
"When we port a technology from Taiwan to Arizona, the fab set up, the process control system, everything is actually a copy from Taiwan," said Kevin Zhang, Senior Vice President, Business Development and Overseas Operations Office, and Deputy Co-COO at TSMC.
TSMC yet has to start making chips at its fabs in Germany, Japan, and the United States, so it remains to be seen how fast the foundry will increase yields to Taiwanese levels at its Fab 23 (in Kumamoto, Japan) and Fab 21 (in Arizona) when they begin operations in 2024 and 2025, but with Global Gigafab Manufacturing program in place, this is likely set to happen rather sooner than later.
SemiconductorsThe Arctic Cooling Freezer 36 ARGB CPU Cooler Review: Budget Cooling Done Well As modern high-performance CPUs generate more heat, there's been a noticeable increase in the demand for powerful air coolers capable of managing these thermal challenges. Traditional stock air coolers, while sufficient for regular use, are typically designed to be cheap and relatively compact, leaving further improvements to noise control and peak cooling efficiency on the table. This gap has long prompted advanced users and system builders to opt for high-quality aftermarket coolers that designed to better handle the heat output from top-tier processors. Known for their innovative approach to PC hardware, Arctic Cooling has stepped into this competitive market with a product aimed at delivering effective cooling at a very low retail price. The Freezer 36 A-RGB, a dual fan tower cooler, is designed to support the cooling demands of the latest CPUs while also offering customizable RGB lighting for visual flair. This review will explore the features, performance, and value of the Arctic Cooling Freezer 36 A-RGB, comparing it with other leading products in the market to see how it stacks up in providing efficient and effective cooling for modern CPUs. Cases/Cooling/PSUs
Intel Core i9-14900KS Review: The Swan Song of Raptor Lake With A Super Fast 6.2 GHz Turbo For numerous generations of their desktop processor releases, Intel has made available a selection of high-performance special edition "KS" CPUs that add a little extra compared to their flagship chip. With a lot of interest, primarily from the enthusiasts looking for the fastest processors, Intel's latest Core i9-14900KS represents a super-fast addition to its 14th Generation Core lineup with out-of-the-box turbo clock speeds of up to 6.2 GHz and represents the last processor to end an era as Intel is removing the 'i' from its legendary nomenclature for future desktop chip releases. Reaching speeds of up to 6.2 GHz, this sets up the Core i9-14900KS as the fastest desktop CPU in the world right now, at least in terms of frequencies out of the box. Building on their 'regular' flagship chip, the Core i9-14900, the Core i9-14900KS is also using their refreshed Raptor Lake (RPL-R) 8P+16E core chip design with a 200 MHz higher boost clock speed and also has a 100 MHz bump on P-Core base frequency. This new KS series SKU shows Intel's drive to offer an even faster alternative to their desktop regular K series offerings, and with the Core i9-14900KS, they look to provide the best silicon from their Raptor Lake Refresh series with more performance available to unlock to those who can. The caveat is that achieving these ridiculously fast clock speeds of 6.2 GHz on the P-Core comes at the cost of power and heat; keeping a processor pulling upwards of 350 W is a challenge in its own right, and users need to factor this in if even contemplating a KS series SKU. In our previous KS series review, the Core i9-13900KS reached 360 W at its peak, considerably more than the Core i9-13900K. The Core i9-14900KS, built on the same core architecture, is expected to surpass that even further than the Core i9-14900K. We aim to compare Intel's final Core i series processor to the best of what both Intel and AMD have available, and it will be interesting to see how much performance can be extrapolated from the KS compared to the regular K series SKU. CPUs
G.Skill on Tuesday introduced its ultra-low-latency DDR5-6400 memory modules that feature a CAS latency of 30 clocks, which appears to be the industry's most aggressive timings yet for DDR5-6400 sticks. The modules will be available for both AMD and Intel CPU-based systems.
With every new generation of DDR memory comes an increase in data transfer rates and an extension of relative latencies. While for the vast majority of applications, the increased bandwidth offsets the performance impact of higher timings, there are applications that favor low latencies. However, shrinking latencies is sometimes harder than increasing data transfer rates, which is why low-latency modules are rare.
Nonetheless, G.Skill has apparently managed to cherry-pick enough DDR5 memory chips and build appropriate printed circuit boards to produce DDR5-6400 modules with CL30 timings, which are substantially lower than the CL46 timings recommended by JEDEC for this speed bin. This means that while JEDEC-standard modules have an absolute latency of 14.375 ns, G.Skill's modules can boast a latency of just 9.375 ns – an approximately 35% decrease.
G.Skill's DDR5-6400 CL30 39-39-102 modules have a capacity of 16 GB and will be available in 32 GB dual-channel kits, though the company does not disclose voltages, which are likely considerably higher than those standardized by JEDEC.
The company plans to make its DDR5-6400 modules available both for AMD systems with EXPO profiles (Trident Z5 Neo RGB and Trident Z5 Royal Neo) and for Intel-powered PCs with XMP 3.0 profiles (Trident Z5 RGB and Trident Z5 Royal). For AMD AM5 systems that have a practical limitation of 6000 MT/s – 6400 MT/s for DDR5 memory (as this is roughly as fast as AMD's Infinity Fabric can operate at with a 1:1 ratio), the new modules will be particularly beneficial for AMD's Ryzen 7000 and Ryzen 9000-series processors.
G.Skill notes that since its modules are non-standard, they will not work with all systems but will operate on high-end motherboards with properly cooled CPUs.
The new ultra-low-latency memory kits will be available worldwide from G.Skill's partners starting in late August 2024. The company did not disclose the pricing of these modules, but since we are talking about premium products that boast unique specifications, they are likely to be priced accordingly.
MemoryAt 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.
StorageThe 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
0 Comments