Announcement PDF 4DS Memory Limited | ABN 43 145 590 110 Level 2, 50 Kings Park Road, West Perth WA 6005 PO Box 271, West Perth WA 6872 +61 8 6377 8043 | [email protected] | www.4dsmemory.com 24 October 2024 4DS Memory presenting at Semiconductor Australia 2024 4DS Memory Limited (ASX:4DS) advises Chief Strategic Officer Mr Peter Himes will present at Semiconductor Australia 2024 today, Thursday, 24 October 2024, at 11:45am AEDT. Peter’s presentation for the conference is attached to this announcement. The one-day Conference is held at Deloitte, Quay Quarter Tower from 8am – 6pm. This inaugural event brings together Australia’s deep-tech entrepreneurs, industry experts, innovation policy makers, and investor community to explore opportunities and chart the roadmap to secure Australia’s semiconductor future. 4DS Interactive Investor Hub If you have any questions on this announcement or any past 4DS announcements, check out our Investor Hub. Like, comment, ask a question and view video summaries on important announcements. Shareholders can link through to the Investor Hub via: announcements.4dsmemory.com ENDS Authorised for release by the Board. Contact information David McAuliffe Executive Chairman 4DS Memory +61 408 994 313 [email protected] About 4DS 4DS Memory Limited (ASX: 4DS), with facilities located in Silicon Valley, is a semiconductor technology company bringing high bandwidth, high endurance, persistent non-volatile memory to advanced CMOS process nodes. Its technology, known as Interface Switching ReRAM, features tunable persistence and low energy per bit for today’s most challenging compute intensive and AI processor applications. Established in 2007, 4DS owns a patented IP portfolio, comprising 34 USA patents, and is the first company to develop PCMO ReRAM, on an advanced CMOS processing node. 4DS has a development agreement with Belgium based imec – a world leading research and innovation hub in nano electronics and digital technologies, as well as a joint development agreement with Western Digital subsidiary HGST, a global leader in storage solutions. For more information, please visit www.4dsmemory.com Disclaimer This release contains certain forward-looking statements that are based on the Company’s management’s beliefs, assumptions and expectations and on information currently available to management. Such forward looking statements involve known and unknown risks, uncertainties, and other factors which may cause the actual results or performance of 4DS to be materially different from the results or performance expressed or implied by such forward looking statements. Such forward looking statements are based on numerous assumptions regarding the Company’s present and future business strategies and the political and economic environment in which 4DS will operate in the future, which are subject to change without notice. Past performance is not necessarily a guide to future performance and no representation or warranty is made as to the likelihood of achievement or reasonableness of any forward-looking statements or other forecast. To the full extent permitted by law, 4DS and its directors, officers, employees, advisers, agents and intermediaries disclaim any obligation or undertaking to release any updates or revisions to information to reflect any change in any of the information contained in this release (including, but not limited to, any assumptions or expectations set out in the release). You should not place undue reliance on these forward-looking statements. Except as required by law or regulation (including the ASX Listing Rules) we undertake no obligation to update these forward- looking statements. Beyond Barriers: Enabling Future AI PERSISTENT RERAM FROM 4DS MEMORY A high speed, high endurance, low energy memory technology to enable Future AI systems Peter Himes | Chief Strategic Officer OCTOBER 2024 About 4DS Memory Began groundbreaking work on PCMO ReRAM in 2007 Strategic relationship with HGST (now WD) in 2014 Public on ASX (Australia) in 2015 Started advanced process development with imec in 2017 Improved PCMO cell design efforts started in 2020 Latest generation Mb chip using 60nm cell fabricated in 2023 34th US patent issued 2023 20nm Mb chip (with 1.6B elements) to be run at imec in 2024 4DS is the only company to demonstrate PCMO ReRAM on an advanced CMOS node What are the challenges facing AI companies today? SCALE ENERGY SPEED LATENCY Big data compute needs are scaling exponentially The power needed to run increasingly large models constrains use cases As computational needs scale, everything must work at higher speeds to maintain responsiveness As performance demands greater scale the tolerance for system latencies due to memory bandwidths evaporates OPTIMISATION ADAPTATION FAULT TOLERANCE COST Reducing model size without compromising performance and reliability is a key differentiator across players The ability of models to adapt based on changing resources and conditions Increasing cluster sizes of big data systems brings higher system fault risks. Instantaneous recovery is vital Balancing the cost of AI systems with real world economic constraints What are today’s approaches to solve? MORE SRAM MORE DRAM FASTER DRAM The most advanced AI processor chips have up to 280MB of SRAM on board The need to hold larger and larger datasets in memory drives the need for more DRAM banks Compute latency due to data transfer from DRAM – CPU or GPU MORE INTEGRATION BIGGER CLUSTERS HYBRID MEMORIES Compute In Memory to address data transfer cost More clusters for larger datasets and faster processing NV SRAM, PMEM, CXL – all are approaches to address some of the challenges by combining different technologies How should an advanced memory help these challenges? ZERO LATENCY HIGH ENDURANCE HIGH SCALE but most NV technologies are slow but NV technologies suffer from low lifetime but embedded memory options are few LOW ENERGY LOW COST HIGH RETENTION but DRAM requires constant refresh even for static data but some advanced memories require expensive ALD integration but persistence, speed and endurance have long been trade offs What is PCMO and why is it unique? Current is spread across entire interface area Low current densities lead to high cell endurance Top electrode Thin oxide layer FAST Super fast programming Perovskite layer comprised of: Praseodymium Calcium Manganese Oxygen One Shot write Area based current High endurance material Bottom electrode E Field injects or removes oxygen from PCMO vacancies 4DS: Persistent Memory at DRAM Speeds Highly responsive PCMO with extremely fast single-shot write* times of 4.7ns to deliver low energy per bit writing at DRAM speeds High endurance and high retention sectors can be dynamically allocated with data retention from hours to months High bandwidth persistent memory for high performance data protection High scale and density w 20 nm cell (to be demonstrated Q4 2024) High endurance memory with up to 109 endurance demonstrated Area based programming, energy scales with size Pulse programmable for Analog characteristics * Switching in PCMO is thru Electric Pulse Induced Resistance switching, or EPIR 4DS – The Memory that Breathes © Why important? Warm Data Applications Data at DRAM speeds that doesn’t need constant updating Prevalent in Inference (as opposed to Training) AI Engines Lowers energy cost of memory for inference applications Complement to CXL bus memory but at faster access, response times Can be In Memory or Near Memory Compute Jovanovic, Bojan & Brum, Raphael & Torres, Lionel. (2015). MTJ-based hybrid storage cells for “normally-off and instant-on” computing. Facta universitatis - series: Electronics and Energetics. 28. 465-476. 10.2298/FUEE1503465J. 4DS believes that both CXL and Warm Data need a memory that responds at the speed of DRAM Why important? Persistent Memory for Big Data Applications Combines the speed of DRAM with long term data retention Enables faster data access and improved system reboot Lowers energy cost of memory for inference applications Integrates seamlessly into the memory hierarchy, residing between volatile memory and storage devices Can provide extra memory capacity to DRAM for big data analytics, or can be ultrafast storage for in-memory databases Persistent memory combines speed, resilience, and capacity with nonvolatility, making it a great option for applications demanding rapid data access and retention Intel PMEM structure 4DS believes there is still an unmet need for a high bandwidth persistent memory solution and that 4DS could potentially deliver that solution Why important? High Bandwidth, Low Energy Memory for Edge AI Applications Edge AI needs to deliver the same AI performance within highly constrained resources Many edge applications are battery powered meaning power consumption becomes a primary design constraint Low cost and low pincount requirements add to the system design challenges for Edge AI applications A memory technology needs to meet the speed requirements while contributing to power savings and efficiency ISPD 2024, “Computing Architecture for Large Language Models (LLMs) and Large Multimodal Models (LMMs),” Bor-Sung Liang 4DS believes that PCMO ReRAM can be an ideal candidate for this emerging memory need Thank you for your time Beyond Barriers: Enabling Future AI FOR MORE INFORMATION Registered Office Level 2, 50 Kings Park Road West Perth WA 6005 USA Office 3155 Skyway Court Fremont CA 94539 Postal Address PO Box 271 West Perth WA 6872 Phone +61 6377 8043 Email [email protected] Web 4DSmemory.com Chief Strategic Officer