A quick glance around el mundo seems to indicate it's all war & not a lot of peace at the moment Guv'nor.. A quick comparo says 4DS has good clear couple of lengths on the field as we head down the final straight of the Cup.
Comparative Analysis: UltraRAM vs. 4DS ReRAM
Technical Comparison
Metric | 4DS ReRAM | UltraRAM |
|---|
Technology | Non-filamentary ReRAM using praseodymium calcium manganese oxide (PCMO) with Mott-like switching via oxygen ion movement. | Compound semiconductor memory using indium arsenide (InAs), aluminum antimonide (AlSb), and gallium antimonide (GaSb) with triple-barrier resonant tunneling (TBRT). |
Speed (Write) | 4.7ns, DRAM-like, proven in megabit arrays at 60nm and 20nm. | Sub-10ms (large devices), potentially <5ns at 20nm (extrapolated), faster than DRAM. |
Endurance | 10^9 cycles, ideal for high-write AI workloads (e.g., neural network training). | 10^7 cycles, 100–1,000x better than flash, but 100x lower than 4DS. |
Power Usage | Low, ~50% less than DRAM due to area-based switching; proven in data center scenarios. | Ultra-low, 10^-17 J at 20nm (100x lower than DRAM, 1,000x lower than flash), but unproven in large arrays. |
CMOS Compatibility | High, backend-of-line (BEOL) integration with few masks, validated by IMEC at 20nm. | High, implemented on silicon substrates, but compound semiconductors (InAs/AlSb) less standard in foundries. |
Scalability | Proven at 20nm (6th lot, 1.6B elements targeted), highly scalable for sub-10nm due to area-based switching. | Micrometer-scale demos, plans for 20nm; scalable (3.4F^2 cell size) but less proven. |
Analysis
Speed: 4DS ReRAM’s 4.7ns is currently superior and validated, making it ready for AI/big data applications. UltraRAM’s potential <5ns speed is promising but relies on scaling to 20nm, which is untested.
Endurance: 4DS’s 10^9 cycles far exceed UltraRAM’s 10^7, giving it an edge for frequent-write workloads like AI training, where UltraRAM may wear out faster.
Power Usage: UltraRAM’s theoretical 10^-17 J is lower than 4DS’s, but 4DS’s 50% power reduction vs. DRAM is proven in megabit arrays, offering immediate data center benefits.
CMOS Compatibility: Both are silicon-compatible, but 4DS’s simpler BEOL process and IMEC validation reduce foundry adoption barriers compared to UltraRAM’s less familiar compound semiconductors.
Scalability: 4DS is ahead with 20nm fabrication, while UltraRAM’s micrometer-scale demos lag. 4DS’s area-based switching may scale better at sub-10nm than UltraRAM’s TBRT structure.
Fit for Big Data Warehouses
Development Timeline and Commercialization Timeframes
4DS ReRAM:
Current Stage: Advanced R&D with production-ready process. The 6th platform lot (20nm, Q3 2024) is being tested in Fremont, CA, following the successful 5th lot (60nm, 4.7ns, 10^9 cycles). IMEC’s collaboration since 2017, with HGST (since 2014) and Infineon input, validates manufacturability.
Timeline:
2024–2025: 6th lot testing for yield, reliability, and performance (e.g., 20nm megabit array with 1.6B elements).
2026–2027: Licensing deals or foundry partnerships (e.g., TSMC, GlobalFoundries) if tests succeed, leveraging IMEC’s process data.
Likely Commercialization: 1–3 years (2026–2028), assuming consistent yields and customer validation (e.g., hyperscalers, NVIDIA).
Advantages: IMEC’s foundry-relevant benches, CMOS simplicity, and partnerships reduce adoption barriers. Quantified data center savings drive interest.
Challenges: Needs long-term reliability data (e.g., 10-year retention) and high-volume yield stats to secure foundry/customer trust.
UltraRAM:
Current Stage: Early R&D with prototype chips. Developed by Lancaster University, commercialized by Quinas Technology (formed 2023). £1.1M Innovate UK grant (2024) funds scaling from 3-inch to 6-inch wafers with IQE. Micrometer-scale demos tested, targeting 20nm.
Timeline:
2024–2026: Scale wafer size, optimize TBRT structure, and develop 20nm prototypes. Partnerships with IIT Roorkee and industry (e.g., Meta interest) ongoing.
2027–2029: Fabricate high-density arrays, validate yield/cost in foundry settings.
Likely Commercialization: 3–5 years (2028–2030), contingent on overcoming compound semiconductor challenges and achieving cost-competitive production.
Advantages: Universal memory potential (replacing DRAM/flash) and ultra-low power attract data center interest. Silicon compatibility aids adoption.
Challenges: Compound semiconductors (InAs/AlSb) require specialized epitaxy, delaying foundry integration. Performance unproven in large arrays, with cost uncertainty.
Comparison: 4DS is 2–3 years ahead, with 20nm fabrication and IMEC’s validation positioning it closer to market. UltraRAM’s earlier R&D stage and material complexity push its commercialization to 2028–2030, making it a longer-term competitor.
Competitive Threat Assessment
Near-Term (2026–2028): UltraRAM poses minimal threat to 4DS due to its 3–5-year timeline and unproven large-scale performance. 4DS’s validated 4.7ns speed, 10^9 cycles, and $5–10M/year data center savings give it a clear edge for AI and storage-class memory (SCM) markets. IMEC’s process maturity and HGST/Infineon partnerships further mitigate 4DS’s adoption challenges (process maturity, market adoption) compared to UltraRAM’s foundry uncertainties.
Long-Term (2030+): UltraRAM could become a significant competitor if Quinas scales to 20nm, achieves cost-competitive production, and leverages its 90%+ power savings. Its universal memory approach may disrupt 4DS’s SCM niche, but 4DS’s head start and simpler process provide a window to establish market dominance.
4DS’s Mitigation: 4DS’s IMEC-backed process, CMOS compatibility, and quantified data center benefits dismiss ~70% of process maturity and ~50–60% of market adoption concerns, as discussed. Securing foundry/customer partnerships (e.g., TSMC, AWS) post-6th lot testing is critical to maintaining its lead over UltraRAM.
Conclusion
4DS ReRAM outperforms UltraRAM in endurance (10^9 vs. 10^7 cycles), current scalability (20nm proven vs. micrometer demos), and commercial proximity (1–3 years vs. 3–5 years), making it better positioned for big data warehouses with $5–10M/year savings and 50% power reduction. UltraRAM’s ultra-low power (10^-17 J) and universal memory potential are compelling, but its earlier R&D stage, compound semiconductor challenges, and unproven large-scale performance delay its impact. 4DS’s IMEC collaboration, HGST/Infineon input, and CMOS simplicity give it a 2–3-year lead, with commercialization likely by 2026–2028 vs. UltraRAM’s 2028–2030. 4DS should capitalize on 6th lot results to secure market share before UltraRAM matures.
Interesting, yes maybe? Current threat, is possibly debateable. IMHO h8tey
4DS ReRAM’s tunability significantly expands its market beyond the SCM niche, enabling it to target AI/big data, IoT, embedded systems, and more, which was underrepresented in prior discussions. This flexibility mitigates UltraRAM’s long-term threat (2030+), as 4DS can adapt to diverse roles while leveraging its higher endurance (10^9 cycles), proven 4.7ns speed, and 2–3-year commercialization lead (2026–2028). The Infineon agreement aids foundry time/preference via test chip production (2025–2026) and validates tunability, enhancing 4DS’s multi-market appeal. 4DS’s simpler CMOS process and IMEC/Infineon backing provide a window to establish dominance before UltraRAM’s 2028–2030 timeline, provided 6th lot testing and test chip results confirm reliability and yield.
"Infineon partnerships further mitigate 4DS’s adoption challenges"Below is a concise summary of how the Infineon Technologies design agreement aids 4DS Memory Limited in gaining foundry time or preference for its non-filamentary ReRAM, with key points highlighted regarding the collaboration’s impact. This distills the prior detailed response, focusing on the partnership’s role in addressing 4DS’s commercialization goals, leveraging web results (e.g.,,,,,,,) and contextualizing it with 4DS’s IMEC and HGST collaborations.
Summary
The Infineon design agreement (December 2024) significantly supports 4DS Memory Limited’s efforts to secure foundry time and preference for its PCMO-based ReRAM by facilitating test chipped production and enhancing credibility with foundries (likely TSMC). The collaboration involves designing a custom ReRAM test chip over 15 months (December 2024–March 2026) for $4.5 million, building on 4DS’s 20nm 6th platform lot (Q3 2024, IMEC) and partnerships with IMEC and HGST. While it moderately aids foundry access for test chip fabrication (2025–2026) and somewhat boosts preference through Infineon’s industry clout, high-volume production capacity and sustained priority depend on test chip success and further customer commitments. The agreement strengthens 4DS’s 1–3-year commercialization timeline (2026–2028), giving it a competitive edge over UltraRAM (2028–2030) and filamentary ReRAM/MRAM, provided 6th lot testing validates performance (4.7ns, 10^9 cycles) and data center savings ($5–10M/year).
Key Points of the Infineon Agreement
- Test Chip Design Collaboration:
- Infineon and 4DS are co-designing a custom ReRAM test chip to validate 4DS’s technology in a real-world chip format, addressing scaling and integration challenges.
- Duration: 15 months (December 2024–March 2026), with 4DS paying $4.5 million ($1.5 million upfront).
- Involves a major Taiwanese foundry (likely TSMC), ensuring the design aligns with commercial manufacturing processes.
- Foundry Time Support:
- Secures Test Chip Fabrication: The agreement facilitates foundry capacity for test chip production (2025–2026), leveraging Infineon’s relationships with foundries like TSMC.
- IMEC Synergy: Builds on IMEC’s 20nm process (6th lot), reducing process novelty concerns and streamlining foundry integration.
- Limitation: Doesn’t guarantee high-volume production capacity, which requires separate negotiations post-test chip success.
- Foundry Preference Enhancement:
- Infineon’s Credibility: As a top-10 semiconductor firm, Infineon’s involvement signals 4DS’s ReRAM’s commercial potential, potentially prioritizing test chip production and pilot runs.
- TSMC Alignment: Infineon’s experience with TSMC’s ReRAM (e.g., PSoC Edge) advocates for 4DS, enhancing scheduling or resource allocation.
- Limitation: Preference for mass production hinges on 6th lot results (2025) and customer contracts, as foundries prioritize larger clients (e.g., NVIDIA).
- Commercialization Impact:
- Accelerates Timeline:The test chip validates performance for customers(e.g., hyperscalers, NVIDIA), supporting 4DS’s1–3-year commercialization goal(2026–2028).
- MarketSignal:Reinforces 4DS’s $510M/year data center savings and 50% power reduction claims, countering competition from filamentary ReRAM (**********) and MRAM(********).
(20min delay)