High Bandwidth Memory (HBM) is emerging as the preferred solution for overcoming memory transfer speed restrictions due to the bandwidth limitations of DDR SDRAM in high-speed computation. HBM is recognized for its revolutionary transmission efficiency and plays a pivotal role in allowing core computational components to operate at their maximum capacity. Top-tier AI server GPUs have set a new industry standard by primarily using HBM. TrendForce forecasts that global demand for HBM will experience almost 60% growth annually in 2023, reaching 290 million GB, with a further 30% growth in 2024.

TrendForce's forecast for 2025, taking into account five large-scale AIGC products equivalent to ChatGPT, 25 mid-size AIGC products from Midjourney, and 80 small AIGC products, the minimum computing resources required globally could range from 145,600 to 233,700 Nvidia A100 GPUs. Emerging technologies such as supercomputers, 8K video streaming, and AR/VR, among others, are expected to simultaneously increase the workload on cloud computing systems due to escalating demands for high-speed computing.

AMD's "Vega" graphics architecture powering graphics cards such as the Radeon VII, Radeon PRO VII, sees a discontinuation of maintenance with ROCm GPU programming software stack. The release notes of ROCm 5.6 states that the AMD Instinct MI50 accelerator, Radeon VII client graphics card, and Radeon PRO VII pro-vis graphics card, collectively referred to as "gfx906," will reach EOM (end of maintenance) starting Q3-2023, which aligns with the release of ROCm 5.7. Developer "EwoutH" on GitHub, who discovered this, remarks gfx906 is barely 5 years old, with the Radeon PRO VII and Instinct MI50 accelerator currently being sold in the market. The most recent AMD product powered by "Vega" has to be the "Cezanne" desktop processor, which uses an iGPU based on the architecture. This chip was released in Q2-2021.

AMD has designed a physically smaller version of its 4 nm "Phoenix" mobile processor silicon. The chip could power lower-end mobile SKUs in the Ryzen 3 and Ryzen 5 series, and it's likely that it could make it to Socket AM5, where it will power Ryzen 3 and lower-end versions of Ryzen 5 desktop processors. Built on the same 4 nm foundry process as the standard "Phoenix" silicon, the so-called "Phoenix 2" or "PHX2" die physically features a 6-core/12-thread CPU based on the "Zen 4" microarchitecture, and a physically smaller iGPU with just two WGPs (workgroup processors), or 4 CU (compute units), which work out to 256 stream processors. This iGPU is based on the same RDNA3 graphics architecture as the one powering the regular "Phoenix" silicon. At this point we don't know if the Ryzen AI component gets the axe, but given AMD's enthusiasm with consumer AI acceleration, the die might just retain it.

The PHX2 die likely retains the I/O of the regular "Phoenix," including a dual-channel (4 sub-channel) DDR5 and LPDDR5 memory interface, and a 24-lane PCI-Express Gen 4 root complex. These changes result in a die that appears to be around three-quarters the size of the regular "Phoenix," with an area of around 137 mm², compared to 178 mm² of the regular "Phoenix." The smaller die will save AMD big on costs and yields. At this time, there are at least two processor models reported to be based on this die, the Ryzen 5 7540U and Ryzen 3 7440U. Both are 15 W to 28 W class mobile processors aimed at thin-and-light notebooks.

US Computer component retailer Micro Center has announced that the store has struck an exclusive deal with AMD to be the sole retailer for the Ryzen 5 5600X3D processor. The CPU is apparently a limited edition release, although it's not clear how limited it'll be in terms of available quantities. The new CPU will launch on the 7th of July and has a base clock of 3.3 GHz and a boost clock of 4.4 GHz, each 100 MHz slower than the Ryzen 7 5800X3D. The CPU obviously has six CPU cores, which results in a total cache of 99 MB due to the missing two cores.

The TDP remains at 105 W and it appears that the Ryzen 5 5600X3D might just consist of failed Ryzen 7 5800X3D chips sold with two cores disabled. From what TPU understands, there should be some OEM availability of the Ryzen 5 5600X3D as well, based on what our sources have told us, but we don't have any details on which system integrators might be offering the CPU. Micro Center will be charging US$229.99, which is US$220 less than what the Ryzen 7 5800X3D launched at, although Micro Center is currently selling it at US$279.99. The Ryzen 5 5600X3D will also be offered at a discounted price when bought with eligible motherboard and memory bundles from Micro Center.

New research from Omdia reveals that the semiconductor market declined in revenue for a fifth straight quarter in the first quarter of 2023. This is the longest recorded period of decline since Omdia began tracking the market in 2002. Revenue in 1Q23 settled at $120.5B, down 9% from 4Q22. The semiconductor market is cyclical, and this prolonged decline follows the upsurge as the market grew to record revenues in each quarter between 4Q20 through 4Q21 following increased demand from the global pandemic.

The memory and MPU market are major areas of the semiconductor market that are contributing to the decline. The MPU market in 1Q23 was $13.1B, just 65% of its size in 1Q22 when it was $20B. The memory market fared worse, with 1Q23 coming in at $19.3B, just 44% of the market in 1Q22 when it was $43.6B. The combined MPU and memory markets declined 19% in 1Q23, dragging the market down to the 9% quarter-over-quarter (QoQ) decline.

According to Taiwanese media, TSMC will start production of its first 1.4 nm fab in 2026, with chip production in the fab said to start sometime in 2027 or 2028. The new fab will be located in Longtan Science Park outside of Hsinchu in Taiwan, where many of TSMC's current fabs are located. TSMC is currently constructing a 2 nm and below node R&D facility at a nearby plot of land to where the new fab is expected to be built. This facility is expected to be finished in 2025 and TSMC has been allocated a total area of just over 158 hectares of land for future expansion in the area.

In related news, TSMC is expected to be charging US$25,000 per 2 nm GAA wafer, which is an increase of about a fifth compared to its 3 nm wafers which are going for around US$20,000. This is largely due to the nodes being fully booked and TSMC being able to charge a premium for its cutting edge nodes. TSMC is also expanding in CoWoS packaging facilities due to increased demand from both AMD and NVIDIA for AI related products. Currently TSMC is said to be able to output 12,000 CoWoS wafers per month and this is twice as much as last year, yet TSMC is unable to meet demand from its customers.

AMD and Bethesda have today revealed that Starfield will be best experienced on a Ryzen processor and Radeon graphics card-equipped PC. Team Red has been announced as the giant open world game's official graphics and GPU partner, but its Xbox Series hardware also gets a couple of friendly shout-outs. Todd Howard, director and executive producer at Bethesda Game Studios, stated in the video presentation: "We have AMD engineers in our code base working on FSR (FidelityFX Super Resolution) 2.0 image processing and upscaling and it looks incredible. You're going to get the benefits of that obviously on your PC but also on Xbox. We're super excited and can't wait to show everybody more."

Jack Huynh, Senior Vice President and General Manager of its Computing and Graphics Group at AMD, added: "Making this game even more special, is the close collaboration between Bethesda and AMD to unlock the full potential of Starfield. We have worked hand-in-hand with Bethesda Game Studios to optimize Starfield for both Xbox and PC with Ryzen 7000 series processors and Radeon 7000 series graphics. The optimizations both accelerate performance and enhance the quality of your gameplay using highly multi-threaded code that both Xbox and PC players will get to take advantage of."

AMD today announced the AMD Versal Premium VP1902 adaptive system-on-chip (SoC), the world's largest adaptive SoC. The VP1902 adaptive SoC is an emulation-class, chiplet-based device designed to streamline the verification of increasingly complex semiconductor designs. Offering 2X the capacity over the prior generation, designers can confidently innovate and validate application-specific integrated circuits (ASICs) and SoC designs to help bring next generation technologies to market faster.

AI workloads are driving increased complexity in chipmaking, requiring next-generation solutions to develop the chips of tomorrow. FPGA-based emulation and prototyping provides the highest level of performance, allowing faster silicon verification and enabling developers to shift left in the design cycle and begin software development well before silicon tape-out. AMD, through Xilinx, brings over 17 years of leadership and six generations of the industry's highest capacity emulation devices, which have nearly doubled in capacity each generation.

Noctua today introduced its new NM-DD1 direct die kit. Developed in cooperation with professional overclocker and direct die cooling expert Roman "der8auer" Hartung, the NM-DD1 is a mounting spacer kit that makes it possible to use a wide range of Noctua CPU coolers on delidded AMD AM5 processors. Removing the processor's integrated heat spreader (delidding) and putting the heatsink directly onto the dies allows for much more efficient thermal transfer and can thereby lower CPU temperatures significantly, with typical gains in the range of 10-15°C.

"Delidding and direct die cooling will void your CPU's warranty and bear a certain risk of damaging it, so this certainly isn't for everyone," explains Roland Mossig (Noctua CEO). "However, the performance gains to be had are simply spectacular, typically ranging from 10 to 15°C but in some cases, we've even seen improvements of almost 20°C in combination with our offset mounting bars, so we're confident that this is an attractive option for enthusiast users. Thanks to Roman for teaming up with us in order to enable customers to implement this exciting tuning measure with our CPU coolers!"

Giga Computing, a subsidiary of GIGABYTE and an industry leader in high-performance servers, workstations, and mini-PCs, today announced a newly designed ultra-compact, mainstream mini-PC for the GIGABYTE BRIXs lineup that adopts AMD Ryzen 7030 series processors. With AMD Zen 3 architecture and TSMC 7 nm process, AMD advanced its high-performance, efficient mobile processors for the mobile market to multitask and stay entertained. Even while adopting a compact size, all new mainstream BRIXs still delivers powerful computing that fits perfectly in any IoT deployment, whether for office use, education use, home use, digital signage, medical care, or KIOSK.

The most recent generation of mainstream BRIXs products in 2023 adopt an all-new chassis design. Computing performance has increased, but the design does not require an increase in the size of the chassis. Its appearance has curved, soft, and twisted lines with a liveliness that makes it very unobtrusive. By the way it was assembled and layered, and designing oblique angles around the base, a sense of simplicity is created to achieve a lightweight design. This allows users to enjoy ultimate computing performance while also having a stylish and elegant product.

The AMD Radeon RX 7600 mainstream graphics card slides a little closer to its ideal price, with an online retailer price-cut sending it down to $249, about $20 less than its MSRP of $269. The cheapest RX 7600 graphics card in the market right now is the MSI RX 7600 MECH 2X Classic, going for $249 on Amazon; followed by the XFX RX 7600 SWFT 210 at $258, and the ASRock RX 7600 Challenger at $259.99.

The sliding prices of the RX 7600 should improve its prospects against the upcoming NVIDIA GeForce RTX 4060, which leaked 3DMark benchmarks show to be around 17% faster than the previous-generation RTX 3060 (12 GB) and 30% faster than its 8 GB variant. Our real-world testing puts the RX 7600 about 15% faster than the RTX 3060 (12 GB) at 1080p, which means there could be an interesting square-off between the RTX 4060 and RX 7600. NVIDIA has announced $299 as the baseline price for the RTX 4060, which should put pressure on AMD partners to trim prices of the RX 7600 to below the $250-mark.

Hardware leaker MEGAsizeGPU has uploaded photos of unreleased G.Skill DIMMs—they claim that the leaked hardware "is the world's first 24G*2 DDR5 expo module: F5-6000J4048F24GX2-TZ5NR." The next-gen Trident Z5 memory is said to be rated for a 6000 MT/s data transfer rate, and close-up shots of labels on heatsinks point to the sample units being non-binary 24 GB DDR5 memory modules that can support EXPO profiles for AMD's Ryzen 7000-series CPUs. MEGAsizeGPU claims that "6000 MHz is the sweetspot for Ryzen" (AM5).

Off-screen captures show a PC system booting up in DDR5-6000 mode—within a Windows OS environment, CPU-Z demonstrates that these new Trident Z5 modules are based on SpecTek-made 24Gb DRAM ICs (instead of binary 16Gb)—SpecTek is a division working under Micron Technology. G.SKILL will likely be selling non-binary Z5 memory in pairs, so we expect to see matched 48 GB dual-channel kits popping up on the market soon. MEGAsizeGPU did not mention anything about pricing or availability. Kingston debuted its own non-binary memory offerings at Computex 2023, but presentation material on hand did not mention whether their new models support AMD's Extended Profiles for Overclocking (EXPO).

AMD Radeon RX 7800 XT will be a much-needed performance-segment addition to the company's Radeon RX 7000-series, which has a massive performance gap between the enthusiast-class RX 7900 series, and the mainstream RX 7600. A report by "Moore's Law is Dead" makes a sensational claim that it is based on a whole new ASIC that's neither the "Navi 31" powering the RX 7900 series, nor the "Navi 32" designed for lower performance tiers, but something in between. This GPU will be AMD's answer to the "AD103." Apparently, the GPU features the same exact 350 mm² graphics compute die (GCD) as the "Navi 31," but on a smaller package resembling that of the "Navi 32." This large GCD is surrounded by four MCDs (memory cache dies), which amount to a 256-bit wide GDDR6 memory interface, and 64 MB of 2nd Gen Infinity Cache memory.

The GCD physically features 96 RDNA3 compute units, but AMD's product managers now have the ability to give the RX 7800 XT a much higher CU count than that of the "Navi 32," while being lower than that of the RX 7900 XT (which is configured with 84). It's rumored that the smaller "Navi 32" GCD tops out at 60 CU (3,840 stream processors), so the new ASIC will enable the RX 7800 XT to have a CU count anywhere between 60 to 84. The resulting RX 7800 XT could have an ASIC with a lower manufacturing cost than that of a theoretical Navi 31 with two disabled MCDs (>60 mm² of wasted 6 nm dies), and even if it ends up performing within 10% of the RX 7900 XT (and matching the GeForce RTX 4070 Ti in the process), it would do so with better pricing headroom. The same ASIC could even power mobile RX 7900 series, where the smaller package and narrower memory bus will conserve precious PCB footprint.

Oracle today introduced the latest generation of the Oracle Exadata platforms, the X10M, delivering unrivaled performance and availability for all Oracle Database workloads. Starting at the same price as the previous generation, these platforms support higher levels of database consolidation with more capacity and offer dramatically greater value than previous generations. Thousands of organizations, large and small, run their most critical and demanding workloads on Oracle Exadata including the majority of the largest financial, telecom, and retail businesses in the world.

"Our 12th generation Oracle Exadata X10M continues our strategy to provide customers with extreme scale, performance, and value, and we will make it available everywhere—in the cloud and on-premises," said Juan Loaiza, executive vice president, Mission-Critical Database Technologies, Oracle. "Customers that choose cloud deployments also benefit from running Oracle Autonomous Database, which further lowers costs by delivering true pay-per-use and eliminating database and infrastructure administration."

AMD's flagship graphics card, the Radeon RX 7900 XTX, based on the RDNA3 architecture, is seeing its street pricing in fall, as the cheapest custom-design card can be had for as low as $881. The XFX Radeon RX 7900 XTX Speedster Merc 310 is listed on Amazon for $979, with a $97 checkbox coupon that sends its price down to $881. The next cheapest card is the Sapphire RX 7900 XTX Pulse, going for $899 after a $100 checkbox coupon on its $999 price. Meanwhile, the RX 7900 XT can be had for as low as $719 with an $80 coupon. Considering that the RX 7900 XTX has shown performance at-par or better than the GeForce RTX 4080, with ray tracing performance comparable to the RTX 3090 Ti, this is tremendous value, given that $881 is what some premium RTX 4070 Ti cards are being sold at.

TrendForce reports that explosive growth in generative AI applications like chatbots has spurred significant expansion in AI server development in 2023. Major CSPs including Microsoft, Google, AWS, as well as Chinese enterprises like Baidu and ByteDance, have invested heavily in high-end AI servers to continuously train and optimize their AI models. This reliance on high-end AI servers necessitates the use of high-end AI chips, which in turn will not only drive up demand for HBM during 2023~2024, but is also expected to boost growth in advanced packaging capacity by 30~40% in 2024.

TrendForce highlights that to augment the computational efficiency of AI servers and enhance memory transmission bandwidth, leading AI chip makers such as Nvidia, AMD, and Intel have opted to incorporate HBM. Presently, Nvidia's A100 and H100 chips each boast up to 80 GB of HBM2e and HBM3. In its latest integrated CPU and GPU, the Grace Hopper Superchip, Nvidia expanded a single chip's HBM capacity by 20%, hitting a mark of 96 GB. AMD's MI300 also uses HBM3, with the MI300A capacity remaining at 128 GB like its predecessor, while the more advanced MI300X has ramped up to 192 GB, marking a 50% increase. Google is expected to broaden its partnership with Broadcom in late 2023 to produce the AISC AI accelerator chip TPU, which will also incorporate HBM memory, in order to extend AI infrastructure.

AMD today announced that its AMD EPYC Embedded Series processors are powering Hewlett Packard Enterprise's new modular, multi-protocol storage solution, HPE Alletra Storage MP. AMD EPYC Embedded processors provide the performance and energy efficiency required for enterprise-class storage systems with high availability, resilience, and industry-leading connectivity and longevity.

The HPE Alletra Storage MP supports a disaggregated infrastructure with multiple storage protocols on the same hardware that can scale independently for performance and capacity. Configurable for block and file stores, HPE Alletra Storage MP gives customers the ability to deploy, manage, and orchestrate data and storage services via the HPE GreenLake edge-to-cloud platform, regardless of the workload and storage protocol. This eliminates data silos, reducing cost and complexity while improving performance.

AMD today announced plans for continued growth in Ireland through an investment of up to $135 million over four years. The investment is intended to fund several strategic R&D projects through the addition of up to 290 highly skilled engineering and research positions, as well as a broad range of additional support roles. The development is being formally announced in Dublin today by Minister for Enterprise, Trade and Employment, Simon Coveney TD, and Ruth Cotter, senior vice president, Marketing, Communications and Human Resources at AMD. The new investment is supported by the Irish government through IDA Ireland.

"I warmly welcome the ambitious plans of AMD to expand their advanced R&D and engineering operations in Ireland. This significant investment will not only bolster our thriving technology sector but also create long-term career opportunities for both highly experienced professionals and new graduates from engineering disciplines. The company's plans to add up to 290 new positions and its funding of strategically important R&D projects demonstrate its confidence in Ireland's supportive enterprise environment and infrastructure. The Irish government, through IDA Ireland, is delighted to support this expansion, further solidifying our commitment to nurturing a vibrant ecosystem for research, development, and engineering," said Simon Coveney TD, Minister for Enterprise, Trade and Employment.

Supermicro Inc., a Total IT Solution Provider for Cloud, AI/ML, Storage, and 5G/Edge, is introducing a new server that gives IT and data center owners a high performance and scalable solution to meet the needs for E-commerce, cloud gaming, code development, content creation, and virtual private servers. The new systems are designed to use AMD Ryzen 7000 Series processors optimized for server usage, based on the latest "Zen 4" core architecture, which has a max boost speed of up to 5.7 GHz, including PCIe 5.0 support, DDR5-5200 MHz, and up to 16 cores (32 threads) per CPU. The new Supermicro MicroCloud is designed to use the latest system technology for a wide range of applications, including web hosting, cloud gaming, and virtual desktop applications.

"We are expanding our application optimized server product lines to include the latest AMD Ryzen 7000 Series processors," said Michael McNerney, VP of Marketing and Security, Supermicro. "These new servers from Supermicro will give IT administrators a compact and high-performance option in order to offer more services with lower latencies to their internal or external customers. By working closely with AMD to optimize the Ryzen 7000 Series firmware for server usage, we can bring a range of solutions with new technologies with PCIe 5.0, DDR5 memory, and very high clock rates to market faster, which allows organizations to reduce costs and offer advanced solutions to their clients."

CD Projekt RED has released the new Patch v1.63 for Cyberpunk 2077 which includes several quest, open world, UI, visual, as well as PC-and console-specific improvements and fixes, as it is rolling out on PC, PlayStation 5, and Xbox Series X/S. While there are plenty of changes, PC users will be glad that it improves performance of DLSS Frame Generation on AMD CPUs, which had been hit by stuttering issues. The Cyberpunk 2077 Patch v1.63 also fixes visual color issues with DLSS and Path Tracing.

The new patch also fixes Photo Mode issues, as well as a crash issue when using Razer Chroma. There are plenty of other quest, open world, miscellaneous, console-specific and REDmod fixes, which you can check out it the full release notes.

TrendForce reports that inventory reduction in Q1 fell short of expectations and coincided with the industry's traditional off-season, leading to overall subdued demand. However, due to new product release and a surge in urgent orders for specialized specifications, Q1 revenue of the global top ten IC design houses remained on par with 4Q22, with a modest QoQ increase of 0.1% for a total revenue of US$33.86 billion. Changes in ranking included Cirrus Logic slipping from the top ten as well as the ninth and tenth positions being replaced by WillSemi and MPS, respectively. The rest of the rankings remained unchanged.

The smartphone supply continues to grapple with overstock, but AI applications are entering a period of rapid growth
Qualcomm witnessed an uptick in its revenue, largely attributed to the launch and subsequent shipments of its latest flagship chip, the Snapdragon 8Gen2. The company saw 6.1% in QoQ growth in its smartphone business, which effectively offset the downturn from its automotive and IoT sectors. As a result, Qualcomm's Q1 revenue increased marginally by 0.6%, cementing its position at the top of the pack with a market share of 23.5%.

GIGABYTE unveiled the first Mini-ITX motherboard to feature the entry-level AMD A620 chipset. The UD-A620I-X offers a comprehensive I/O feature-set. The board draws power from a 24-pin ATX and a single 8-pin EPS, and uses a 9-phase VRM to power the SoC. This board is restricted to 65 W TDP processors (7600, 7700, 7900 and possibly Ryzen PRO desktop processors that are 65 W). The processor is wired to two DDR5 DIMM slots, and a PCI-Express 4.0 x16, which is really all you need for this generation of graphics cards.

A downside of the A620 platform is that it doesn't support PCIe Gen 5 on even the CPU-attached M.2 NVMe slots—the one on this board is Gen 4. Display connectivity on the UD-A620I includes an HDMI and DisplayPort. Networking interfaces include Wi-Fi 6 wireless, and 2.5 GbE wired. Storage connectivity, besides the Gen 4 NVMe slot, includes two SATA 6 Gbps ports. You get at least two USB 3.2 ports from the processor, four USB 3.2 type-A ports on the rear I/O, an internal type-E port (for security keys), and an internal USB 3.2 header. The company didn't reveal pricing.

Gamers Nexus has uploaded a video feature dedicated to the history of AMD's Zen CPU architecture—editor-in-chief and founder Stephen Burke ventured to Team Red's Austin, Texas-based test and engineering campus. Longer and more in-depth coverage of his lab tour will be released at a later date, but today's upload included an interesting segment covering unreleased hardware. The Gamers Nexus crew spent some time looking at several examples of current and past generation AMD 3D V-Cache CPUs. Prototype Ryzen 7000-series Zen 4 designs were shown off by principal engineer Amit Mehra and technical team member Bill Alverson. They also brought out older 5000-series Zen 3 units that never reached retail—the 16-core Ryzen 9 5950X3D was demonstrated as having a 3.5 GHz base clock, and it can boost up to 4.1 GHz. The 12-core Ryzen 9 5900X3D had 3.5 GHz base and 4.4 GHz boost clocks.

Team Red only sells one AM4 3D V-Cache model at the moment, in the form of its well received Ryzen 7 5800X3D CPU. It was released over a year ago, but recent price cuts have resulted in increased unit sales—system builders looking to maximize the potential of their older generation Ryzen 5000-series compatible mainboards are snapping up 5800X3Ds. AMD could be readying a cheaper alternative, with previous reports proposing that a "Ryzen 5 5600X3D" is positioned to take on Intel's 13th Gen Core i5 series (with DDR4). The unreleased Ryzen 9 5950X3D and 5900X3D have 3D V-Cache stacks on both of their CCDs (granting 192 MB of L3 cache), which is unique given that all retail 3D V-Cache CPUs (released so far) restrict this to a single CCD stack. Apparently AMD decided to stick with the latter setup due to it offering the best balance of performance and efficiency, plus gaming benchmarks demonstrated that there was not much of a difference between the configurations.

Harukaze5719 has brought attention to a curious registration of unreleased AsRock graphics cards at the Eurasian Economic Commission (ECC) regulatory office. The self-described (South) Korean PC Tech enthusiast has found out that ASRock is likely preparing for an imminent launch of custom design AMD Radeon RX 7800 XT PG 16 GB and RX 7800 XT PGW 16 GB models.

No specifications were found in the ECC registration, so it is too early to confirm whether the leaked RX 7800 XT series is based on AMD's RDNA 3 Navi 31 or Navi 32 GPU. Igor Lab's simulated a hypothetical version via the benchmarking of a workstation Radeon Pro W7800 (Navi 31) 32 GB graphics card. Model codes (registered on May 18 2023) indicate that the two AsRock Radeon RX 7800 XT models could sport the company's Phantom Gaming (PG) triple-fan cooling solution, possibly available in a standard shade or a (PGW) white option.

AMD recently revealed its Instinct MI300X GPU at their Data Center and AI Technology Premiere event on Tuesday (June 15). The keynote presentation did not provide any details about the new accelerator model's power consumption, but that did not stop one tipster - Hoang Anh Phu - from obtaining this information from Team Red's post-event footnotes. A comparative observation was made: "MI300X (192 GB HBM3, OAM Module) TBP is 750 W, compared to last gen, MI250X TBP is only 500-560 W." A leaked Giga Computing roadmap from last month anticipated server-grade GPUs hitting the 700 W mark.

NVIDIA's Hopper H100 took the crown - with its demand for a maximum of 700 W - as the most power-hungry data center enterprise GPU until now. The MI300X's OCP Accelerator Module-based design now surpasses Team Green's flagship with a slightly greater rating. AMD's new "leadership generative AI accelerator" sports 304 CDNA 3 compute units, which is a clear upgrade over the MI250X's 220 (CDNA 2) CUs. Engineers have also introduced new 24G B HBM3 stacks, so the MI300X can be specced with 192 GB of memory (as a maximum), the MI250X is limited to a 128 GB memory capacity with its slower HBM2E stacks. We hope to see sample units producing benchmark results very soon, with the MI300X pitted against H100.