Jonas Chen

Introducing AMD EPYC 4005 Series: Efficiency Meets Value. If you’re in #cloud, #datacenter, #SMB or building high-density servers on a budget, this is big news. ✅ Why it matters: • Zen 5 architecture → 6-16 cores with up to 5.7 GHz boost. ✅ • Top SKU (EPYC 4585PX) packs a full 128 MB L3 cache via 3D V-Cache – huge for in-memory and latency-sensitive workloads. ✅ • AMD benchmarks vs Intel Xeon 6369P, nearly 2× or more performance per CPU dollar in SPEC, MySQL OLTP & FFmpeg. ✅ 💸 Value you can’t ignore: 16-core chip for ~$589 (1,000-unit pricing) → that’s dramatically better cores / dollar for many workloads. If you care about throughput, not just peak MHz, this is a serious contender. 🔧 Ideal for: • Hosting, virtualization, application servers • Cost-sensitive deployments • Businesses wanting modern performance without breaking the bank If you're evaluating server refreshes, tight on budget, or want serious bang for buck — EPYC 4005 deserves a spot at the table. Full details in the blog by Raghu Nambiar, CVP and CTO @ AMD : https://lnkd.in/eZ6rTEk6 #AMD #EPYC4005 #ServerProcessors #Efficiency #Value #TechLeadership #AMDBrandAmbassador

124

35 則回應

The MediaTek Dimensity 9500 features an all new extraordinary architecture for the next generation of flagship smartphones. Let’s break it down: ⚡ CPU: All Big Core design with computing power at PC level 🎮 GPU: Rendering performance at gaming console level 🤖 Dual NPU: Super performance + Super Efficient dual NPU rchitecture The outcome is breakthrough AI, console-class gaming, and power-efficient performance, raising the bar for what’s possible in mobile. #MediaTek #MediaTekDimensity9500 #AllBigCoreCPU #Flagship #5G

123

1 則回應

At FMS: the Future of Memory and Storage 2025, SANDISK showed the newest form factors for 256TB class drives E2, U.2, and E3.L! This is just the beginning of a journey to SSDs with 1PB of storage. Mrinal Kochar detailed the economic and technical drivers of the AI data center and how flash has become bifurcated into performance and capacity more than ever. IOPS/TB/W has become the metric to rule the data center when it comes to storage, showing performance and power efficiency. SANDISK Enterprise SSDs are purpose-built for the AI data center of today and tomorrow. Watch more of Mrinal Kochar's keynote here: https://lnkd.in/ggZAdRDT

637

12 則回應

Rebellions has partnered with Marvell Technology to provide custom hardware for AI deployments in APAC and the Middle East. Rebellions will design AI chips using Marvell's technology to create energy-efficient systems for sovereign AI needs. Sunghyun Park, CEO of Rebellions, said the collaboration aims to meet the specific needs of sovereign entities. Will Chu from Marvell mentioned that custom AI infrastructure is crucial for future data center innovation. Read more: https://lnkd.in/ep3uBUFC 📰 Subscribe to the weekly Silicon Brief Newsletter: https://lnkd.in/ejfzg92J #ai #artificialintelligence #ainews #aistartup #aichips #datacenters

64

1 則回應

As 3D NAND scales higher, SK hynix is tackling key challenges with Charge Trap Flash and new process, design, and algorithm solutions—boosting program efficiency and device performance. In Episode 01 of our new 🎧 TECH TALK series, we share highlights from unsang lee’s #FMS2025 presentation on “PGM Efficiency Improvement Method for QLC Device.” Catch the first TECH TALK and see how we’re driving NAND innovation forward. #SKhynix #TECHTALK #FMS2025 #NAND #QLC

68

High-Bandwidth Flash (HBF) utilizes SanDisk's proprietary BiCS NAND and CBA technologies to tightly stack high-layer-count 3D NAND storage cells with high-speed logic layers, achieving high-bandwidth, low-latency, and high-density interconnects. While access latency is slightly lower than pure DRAM, HBF offers significant advantages in AI inference and edge computing scenarios, where large model data must be maintained for extended periods of time. This significantly reduces energy consumption and heat dissipation, demonstrating its potential. While HBM remains the core of the current AI market, HBF's advantages in energy saving and reducing HBM make it a key future NAND development direction. It has the potential to complement or even transform the existing memory configuration landscape in specific applications. Sandisk and SK hynix announced the signing of a memorandum of understanding (MOU) to jointly develop technical specifications for HBF and promote standardization, injecting new variables into the memory market. HBF memory samples are expected to be available in the second half of 2026, with the first AI inference hardware powered by this technology scheduled for release in early 2027. The industry anticipates that, thanks to SK Hynix's close partnerships with multiple AI chip manufacturers, HBF will be rapidly adopted by the market once standards are finalized. https://lnkd.in/gSceb2iq

24

4 則回應