home.social

#hbm4 — Public Fediverse posts

Live and recent posts from across the Fediverse tagged #hbm4, aggregated by home.social.

  1. Samsung Strikes $200 Billion AI Chip Deal; SK Extends HBM4 Supply to Microsoft

    Broadcom CEO Hock Tan (left), Samsung Electronics Foundry President Han Jin-man (third from left), and Samsung Electronics Chairman…
    #EuropeSays #Korea #KR #Samsung #AIsemiconductor #Broadcom #foundry #HBM4 #Microsoft #NVIDIA #SamsungElectronics #SamsungGroup #SKhynix
    europesays.com/korea/99270/

  2. Nvidia and SK Group Forge $500 Billion AI Alliance to Build Next-Gen Data Centers and Memory — BigGo Finance

    Nvidia and South Korea’s SK Group announced a massive partnership on Friday, committing more than $500 billion to…
    #EuropeSays #Korea #KR #SK #Brookfield #CheyTae-won #HBM4 #JensenHuang #LeeJaeMyung #Naver #NVIDIA #SKGroup #SKhynix #SKTelecom #SouthKorea #VeraRubin
    europesays.com/korea/98810/

  3. U.S.-South Korea AI Alliance Takes Shape: Samsung, SK Hynix Ink $950 Billion Chip Megadeal — BigGo Finance

    U.S.-South Korea AI Alliance Takes Shape: Samsung, SK Hynix Ink $950 Billion Chip Megadeal The global artificial intelligence…
    #EuropeSays #Korea #KR #SK #Anthropic #Broadcom #HBM4 #LeeJaeMyung #Micron #NVIDIA #OpenAI #SamsungElectronics #SanFrancisco #SKGroup #SKhynix #SKTelecom #TrendForce #VeraRubin
    europesays.com/korea/98668/

  4. RT @SebAaltonen: Die AMD MI4555X scheint großartig zu sein: 40 Petaflops MXFP6 (gemeinsamer Exponent → Qualität nahe an fp8). 432 GB RAM mit ~20 TB/s. MXFP6-quantisiertes Kimi K3 sollte auf 6 GPUs passen. Das sind rund 300.000 US-Dollar für unbegrenzte Frontier-Modell-Token für etwa 10 Entwickler… videocardz.com/newz/amd-inst… Link AMD Instinct MI455X GPU: 432 GB HBM4 und 23,3 TB/s Speicherbandbreite - VideoCardz.com Die neue Instinct-GPU verwendet ein Chiplet-Paket mit zwei Fabric-Dies und zwei I/O-Dies. Die maximale Speicherbandbreite beträgt 23,3 TB/s. videocardz.com

    mehr auf Arint.info

    #AI #AMD #GPU #HBM4 #MI455X #MXFP6 #arint_info

    https://x.com/SebAaltonen/status/2080496707406909455#m

  5. RT @SebAaltonen: Die AMD MI4555X scheint großartig zu sein: 40 Petaflops MXFP6 (gemeinsamer Exponent → Qualität nahe an fp8). 432 GB RAM mit ~20 TB/s. MXFP6-quantisiertes Kimi K3 sollte auf 6 GPUs passen. Das sind rund 300.000 US-Dollar für unbegrenzte Frontier-Modell-Token für etwa 10 Entwickler… videocardz.com/newz/amd-inst… Link AMD Instinct MI455X GPU: 432 GB HBM4 und 23,3 TB/s Speicherbandbreite - VideoCardz.com Die neue Instinct-GPU verwendet ein Chiplet-Paket mit zwei Fabric-Dies und zwei I/O-Dies. Die maximale Speicherbandbreite beträgt 23,3 TB/s. videocardz.com

    mehr auf Arint.info

    #AI #AMD #GPU #HBM4 #MI455X #MXFP6 #arint_info

    https://x.com/SebAaltonen/status/2080496707406909455#m

  6. RT @SebAaltonen: Die AMD MI4555X scheint großartig zu sein: 40 Petaflops MXFP6 (gemeinsamer Exponent → Qualität nahe an fp8). 432 GB RAM mit ~20 TB/s. MXFP6-quantisiertes Kimi K3 sollte auf 6 GPUs passen. Das sind rund 300.000 US-Dollar für unbegrenzte Frontier-Modell-Token für etwa 10 Entwickler… videocardz.com/newz/amd-inst… Link AMD Instinct MI455X GPU: 432 GB HBM4 und 23,3 TB/s Speicherbandbreite - VideoCardz.com Die neue Instinct-GPU verwendet ein Chiplet-Paket mit zwei Fabric-Dies und zwei I/O-Dies. Die maximale Speicherbandbreite beträgt 23,3 TB/s. videocardz.com

    mehr auf Arint.info

    #AI #AMD #GPU #HBM4 #MI455X #MXFP6 #arint_info

    https://x.com/SebAaltonen/status/2080496707406909455#m

  7. Nvidia Establishes Joint Research Labs with KAIST, Seoul National University to Secure South Korean AI Talent — BigGo Finance

    Nvidia is partnering with the Korea Advanced Institute of Science and Technology (KAIST) and Seoul National University to…
    #EuropeSays #Korea #KR #Seoul #AMD #AmkorTechnology #HBM4 #Hims&Hers #KAIST #KimHyun-woo #LisaSu #NVIDIA #peptide #SamsungElectronics #SeoulNationalUniversity #TSMC #U.S.FoodandDrugAdministration
    europesays.com/korea/98402/

  8. SK Signs $500 Billion AI Infrastructure LOI With Nvidia

    SK Chairman Chey Tae-won and Nvidia CEO Jensen Huang pose for a photo after signing a memorandum of…
    #EuropeSays #Korea #KR #SK #AIinfrastructure #HBM4 #K-AISummit #Microsoftmemorysupply #NVIDIA #SKGroup #SKhynix #SKTelecom
    europesays.com/korea/98212/

  9. Lisa Su Says “Almost There” as Samsung Nears AMD HBM4 Deal

    ▲AI PRISM* Customized Economic Briefing *Editor’s Note: ‘AI PRISM’ (Personalized Report & Insight Summarizing Media) is an “AI-based…
    #EuropeSays #Korea #KR #SamsungElectronics #AMD #AmkorNvidiapackaging #CXMTIPO #HBM4 #HeliosAIserver #high-bandwidthmemory #LisaSu #Samsung
    europesays.com/korea/98078/

  10. AMD Unveils AI Server with Samsung HBM4, Challenging Nvidia’s Dominance — BigGo Finance

    AMD unveiled its next-generation AI server rack system “Helios” at the “Advancing AI 2026” event in San Francisco…
    #EuropeSays #Korea #KR #SamsungElectronics #Alphabet #AMD #Anthropic #Cerebras #HBM4 #Helios #LisaSu #MI455X #NVIDIA #OpenAI #Samsung #SKhynix
    europesays.com/korea/97604/

  11. KB Securities Keeps 600,000 Won Target on Samsung, Sees Third Straight Earnings Beat

    The Samsung Group flag flies in the wind at the Samsung Electronics building in Seocho-gu, Seoul. News1 KB…
    #EuropeSays #Korea #KR #Samsung #AIcapitalexpenditure #DRAM #earningssurprise #HBM4 #KBSecurities #SamsungElectronics #SamsungGroup #semiconductorstocks #targetprice
    europesays.com/korea/97446/

  12. RT @wccftech: TRANSLASHT: AMD bringt die Instinct MI455X GPU auf den Markt, einen 320 Milliarden Transistoren umfassenden Riesen, der entwickelt wurde, um NVIDIA's Rubin zu bekämpfen, mit 50 % mehr HBM4-Speicher und bis zu 40 PFLOPs an KI-Computing-Leistung. wccftech.com/amd-instinct-mi… Link AMD veröffentlicht Instinct MI455X GPU, ein 320 Milliarden Transistoren großes Monster, das entwickelt wurde, um... AMDs Instinct MI455X-GPUs erweitern die KI-Roadmap des Unternehmens und bieten führende HBM4-Kapazitäten und über 40 PFLOPs an Rechenleistung für Agentic AI, im Wettbewerb mit NVIDIAs Rubin-Chip. AMD hat eine Antwort auf NVIDIA's... wccftech.com

    mehr auf Arint.info

    #AICompute #AMD #HBM4 #InstinctMI455X #NVIDIA #Rivalry #arint_info

    https://x.com/wccftech/status/2080394171446067351#m

  13. RT @wccftech: TRANSLASHT: AMD bringt die Instinct MI455X GPU auf den Markt, einen 320 Milliarden Transistoren umfassenden Riesen, der entwickelt wurde, um NVIDIA's Rubin zu bekämpfen, mit 50 % mehr HBM4-Speicher und bis zu 40 PFLOPs an KI-Computing-Leistung. wccftech.com/amd-instinct-mi… Link AMD veröffentlicht Instinct MI455X GPU, ein 320 Milliarden Transistoren großes Monster, das entwickelt wurde, um... AMDs Instinct MI455X-GPUs erweitern die KI-Roadmap des Unternehmens und bieten führende HBM4-Kapazitäten und über 40 PFLOPs an Rechenleistung für Agentic AI, im Wettbewerb mit NVIDIAs Rubin-Chip. AMD hat eine Antwort auf NVIDIA's... wccftech.com

    mehr auf Arint.info

    #AICompute #AMD #HBM4 #InstinctMI455X #NVIDIA #Rivalry #arint_info

    https://x.com/wccftech/status/2080394171446067351#m

  14. RT @wccftech: TRANSLASHT: AMD bringt die Instinct MI455X GPU auf den Markt, einen 320 Milliarden Transistoren umfassenden Riesen, der entwickelt wurde, um NVIDIA's Rubin zu bekämpfen, mit 50 % mehr HBM4-Speicher und bis zu 40 PFLOPs an KI-Computing-Leistung. wccftech.com/amd-instinct-mi… Link AMD veröffentlicht Instinct MI455X GPU, ein 320 Milliarden Transistoren großes Monster, das entwickelt wurde, um... AMDs Instinct MI455X-GPUs erweitern die KI-Roadmap des Unternehmens und bieten führende HBM4-Kapazitäten und über 40 PFLOPs an Rechenleistung für Agentic AI, im Wettbewerb mit NVIDIAs Rubin-Chip. AMD hat eine Antwort auf NVIDIA's... wccftech.com

    mehr auf Arint.info

    #AICompute #AMD #HBM4 #InstinctMI455X #NVIDIA #Rivalry #arint_info

    https://x.com/wccftech/status/2080394171446067351#m

  15. Jensen Huang Reportedly to Hold Secret Friday Meeting with Samsung, SK Chiefs; Korea-US AI Giants to Discuss HBM4 and Sovereign AI Plans

    Global memory and GPU giants are about to engage in an unprecedented high-level dialogue in the United States.…
    #EuropeSays #Korea #KR #SK #Anthropic #Apple #HBM4 #JensenHuang #JimCramer #Naver #NVIDIA #OpenAI #SamsungElectronics #SKGroup
    europesays.com/korea/95480/

  16. Samsung, SK, Naver Chiefs to Meet Nvidia’s Jensen Huang in Silicon Valley AI Summit; Altman, Amodei Also Expected — BigGo Finance

    Top executives from South Korea’s and the United States’ leading artificial intelligence and semiconductor companies will gather in…
    #EuropeSays #Korea #KR #SK #AIfactory #Anthropic #CheyTae-won #HBM4 #JensenHuang #LeeHae-jin #LeeJae-yong #Naver #NVIDIA #OpenAI #SamsungElectronics #siliconvalley #SKGroup #SKhynix #VeraRubin
    europesays.com/korea/95100/

  17. HBM4 Prices to Double Next Year as Samsung, SK hynix Keep Upper Hand

    SK Group Chairman Chey Tae-won, SK hynix CEO Kwak Noh-jung and SK hynix Outside Director and Board Chairman…
    #EuropeSays #Korea #KR #SamsungElectronics #AImemorydemand #DRAMprices #HBM3e #HBM4 #Micron #NvidiaVeraRubin #Samsung #SKhynix
    europesays.com/korea/83389/

  18. Nvidia’s ‘Vera’ CPU Expansion Raises Hopes for Samsung, SK Hynix Memory Windfall — BigGo Finance

    Nvidia is accelerating its push into the data center market with its next-generation central processing unit (CPU) “Vera,”…
    #EuropeSays #Korea #KR #SamsungElectronics #Anthropic #HBM4 #JensenHuang #NVIDIA #OpenAI #Perplexity #PM1763 #Samsung #SKhynix #SOCAMM2 #Vera #VeraRubin
    europesays.com/korea/82417/

  19. Samsung Posts Record Earnings, Shares Plunge as Analysts Split

    ▲ AI PRISM* Customized Economic Briefing *Editor’s Note: ‘AI PRISM’ (Personalized Report & Insight Summarizing Media) is an…
    #EuropeSays #Korea #KR #SamsungElectronics #AImemorysupercycle #HBM4 #KOSPIvolatility #marginrequirements #Q2earnings #Samsung #SKhynix #VeraRubin
    europesays.com/korea/79687/

  20. SK Hynix’s Nasdaq Listing Sets Up New AI Memory Battle With MU Stock

    SK Hynix’s Nasdaq listing can not only reset the AI memory trade but also accelerate it. The company…
    #EuropeSays #Korea #KR #SKHynix #AImemorytrade #analystratings #HBM4 #high-bandwidthmemory #MACDmomentum #Micron #Nasdaqlisting #NVIDIA #semiconductorstocks #SK #SKhynix
    europesays.com/korea/77897/

  21. Dentro la gerarchia della memoria delle GPU: come i server AI spostano i dati dagli SSD alla HBM

    Ti sei mai chiesto come i modelli di intelligenza artificiale trasferiscono i dati alla GPU? Questo articolo spiega in modo semplice i diversi livelli di memoria all'interno di un server AI, perché la memoria della GPU è diventata un collo di bottiglia e quali nuove tecnologie potrebbero migliorare le prestazioni dell'intelligenza artificiale.

    buysellram.com/blog/inside-the…

    #HBM #HBM4 #GPUMemory #AIInfrastructure #DataCenter #CXL #HighBandwidthFlash #AIHardware #MemoryHierarchy #AIInference #NVMe #ITAD

  22. Dentro la gerarchia della memoria delle GPU: come i server AI spostano i dati dagli SSD alla HBM

    Ti sei mai chiesto come i modelli di intelligenza artificiale trasferiscono i dati alla GPU? Questo articolo spiega in modo semplice i diversi livelli di memoria all'interno di un server AI, perché la memoria della GPU è diventata un collo di bottiglia e quali nuove tecnologie potrebbero migliorare le prestazioni dell'intelligenza artificiale.

    buysellram.com/blog/inside-the…

    #HBM #HBM4 #GPUMemory #AIInfrastructure #DataCenter #CXL #HighBandwidthFlash #AIHardware #MemoryHierarchy #AIInference #NVMe #ITAD

  23. Dentro la gerarchia della memoria delle GPU: come i server AI spostano i dati dagli SSD alla HBM

    Ti sei mai chiesto come i modelli di intelligenza artificiale trasferiscono i dati alla GPU? Questo articolo spiega in modo semplice i diversi livelli di memoria all'interno di un server AI, perché la memoria della GPU è diventata un collo di bottiglia e quali nuove tecnologie potrebbero migliorare le prestazioni dell'intelligenza artificiale.

    buysellram.com/blog/inside-the…

    #HBM #HBM4 #GPUMemory #AIInfrastructure #DataCenter #CXL #HighBandwidthFlash #AIHardware #MemoryHierarchy #AIInference #NVMe #ITAD

  24. Dentro la gerarchia della memoria delle GPU: come i server AI spostano i dati dagli SSD alla HBM

    Ti sei mai chiesto come i modelli di intelligenza artificiale trasferiscono i dati alla GPU? Questo articolo spiega in modo semplice i diversi livelli di memoria all'interno di un server AI, perché la memoria della GPU è diventata un collo di bottiglia e quali nuove tecnologie potrebbero migliorare le prestazioni dell'intelligenza artificiale.

    buysellram.com/blog/inside-the…

    #HBM #HBM4 #GPUMemory #AIInfrastructure #DataCenter #CXL #HighBandwidthFlash #AIHardware #MemoryHierarchy #AIInference #NVMe #ITAD

  25. Dentro la gerarchia della memoria delle GPU: come i server AI spostano i dati dagli SSD alla HBM

    Ti sei mai chiesto come i modelli di intelligenza artificiale trasferiscono i dati alla GPU? Questo articolo spiega in modo semplice i diversi livelli di memoria all'interno di un server AI, perché la memoria della GPU è diventata un collo di bottiglia e quali nuove tecnologie potrebbero migliorare le prestazioni dell'intelligenza artificiale.

    buysellram.com/blog/inside-the…

    #HBM #HBM4 #GPUMemory #AIInfrastructure #DataCenter #CXL #HighBandwidthFlash #AIHardware #MemoryHierarchy #AIInference #NVMe #ITAD

  26. A modern AI server runs five layers of memory, from nanosecond on-chip SRAM to petabyte-scale SSDs, and keeping the GPU fed is the whole engineering game. This piece walks the full hierarchy: why HBM became the bottleneck, why adding more isn't simple, and where HBF, CXL, and PIM fit into the next generation.

    buysellram.com/blog/inside-the

    #HBM #HBM4 #GPUMemory #AIInfrastructure #DataCenter #CXL #HighBandwidthFlash #AIHardware #MemoryHierarchy #AIInference #NVMe #ITAD #buysellram

  27. A modern AI server runs five layers of memory, from nanosecond on-chip SRAM to petabyte-scale SSDs, and keeping the GPU fed is the whole engineering game. This piece walks the full hierarchy: why HBM became the bottleneck, why adding more isn't simple, and where HBF, CXL, and PIM fit into the next generation.

    buysellram.com/blog/inside-the

    #HBM #HBM4 #GPUMemory #AIInfrastructure #DataCenter #CXL #HighBandwidthFlash #AIHardware #MemoryHierarchy #AIInference #NVMe #ITAD #buysellram

  28. Why can't a GPU just carry more HBM? Interposers max out in size, stacking 12 to 16 DRAM dies compounds yield losses, and every stack sits beside a kilowatt-class package that hates sharing heat. So capacity climbs in careful steps — 80 GB on the H100, 141 on the H200, 192 on the B200, 288 on Blackwell Ultra — while KV caches for long-context inference balloon past 40 GB per request.

    That gap between what models demand and what packaging permits is reshaping server design. NVIDIA's Rubin platform treats CPU memory and HBM as one coherent pool. SanDisk and SK hynix are standardizing High Bandwidth Flash as a capacity tier under HBM, with first samples due this half. CXL 4.0 pooling hardware is landing in racks now.

    This article walks the whole memory hierarchy, from on-chip SRAM to NVMe, and explains what each emerging technology actually solves — and what it doesn't.

    buysellram.com/blog/inside-the

    #HBM #HBM4 #GPUMemory #AIInfrastructure #DataCenter #CXL #HighBandwidthFlash #AIHardware #MemoryHierarchy #AIInference #NVMe #ITAD #technology

  29. Why can't a GPU just carry more HBM? Interposers max out in size, stacking 12 to 16 DRAM dies compounds yield losses, and every stack sits beside a kilowatt-class package that hates sharing heat. So capacity climbs in careful steps — 80 GB on the H100, 141 on the H200, 192 on the B200, 288 on Blackwell Ultra — while KV caches for long-context inference balloon past 40 GB per request.

    That gap between what models demand and what packaging permits is reshaping server design. NVIDIA's Rubin platform treats CPU memory and HBM as one coherent pool. SanDisk and SK hynix are standardizing High Bandwidth Flash as a capacity tier under HBM, with first samples due this half. CXL 4.0 pooling hardware is landing in racks now.

    This article walks the whole memory hierarchy, from on-chip SRAM to NVMe, and explains what each emerging technology actually solves — and what it doesn't.

    buysellram.com/blog/inside-the

  30. Why can't a GPU just carry more HBM? Interposers max out in size, stacking 12 to 16 DRAM dies compounds yield losses, and every stack sits beside a kilowatt-class package that hates sharing heat. So capacity climbs in careful steps — 80 GB on the H100, 141 on the H200, 192 on the B200, 288 on Blackwell Ultra — while KV caches for long-context inference balloon past 40 GB per request.

    That gap between what models demand and what packaging permits is reshaping server design. NVIDIA's Rubin platform treats CPU memory and HBM as one coherent pool. SanDisk and SK hynix are standardizing High Bandwidth Flash as a capacity tier under HBM, with first samples due this half. CXL 4.0 pooling hardware is landing in racks now.

    This article walks the whole memory hierarchy, from on-chip SRAM to NVMe, and explains what each emerging technology actually solves — and what it doesn't.

    buysellram.com/blog/inside-the

    #HBM #HBM4 #GPUMemory #AIInfrastructure #DataCenter #CXL #HighBandwidthFlash #AIHardware #MemoryHierarchy #AIInference #NVMe #ITAD #technology

  31. Why can't a GPU just carry more HBM? Interposers max out in size, stacking 12 to 16 DRAM dies compounds yield losses, and every stack sits beside a kilowatt-class package that hates sharing heat. So capacity climbs in careful steps — 80 GB on the H100, 141 on the H200, 192 on the B200, 288 on Blackwell Ultra — while KV caches for long-context inference balloon past 40 GB per request.

    That gap between what models demand and what packaging permits is reshaping server design. NVIDIA's Rubin platform treats CPU memory and HBM as one coherent pool. SanDisk and SK hynix are standardizing High Bandwidth Flash as a capacity tier under HBM, with first samples due this half. CXL 4.0 pooling hardware is landing in racks now.

    This article walks the whole memory hierarchy, from on-chip SRAM to NVMe, and explains what each emerging technology actually solves — and what it doesn't.

    buysellram.com/blog/inside-the

    #HBM #HBM4 #GPUMemory #AIInfrastructure #DataCenter #CXL #HighBandwidthFlash #AIHardware #MemoryHierarchy #AIInference #NVMe #ITAD #technology

  32. Why can't a GPU just carry more HBM? Interposers max out in size, stacking 12 to 16 DRAM dies compounds yield losses, and every stack sits beside a kilowatt-class package that hates sharing heat. So capacity climbs in careful steps — 80 GB on the H100, 141 on the H200, 192 on the B200, 288 on Blackwell Ultra — while KV caches for long-context inference balloon past 40 GB per request.

    That gap between what models demand and what packaging permits is reshaping server design. NVIDIA's Rubin platform treats CPU memory and HBM as one coherent pool. SanDisk and SK hynix are standardizing High Bandwidth Flash as a capacity tier under HBM, with first samples due this half. CXL 4.0 pooling hardware is landing in racks now.

    This article walks the whole memory hierarchy, from on-chip SRAM to NVMe, and explains what each emerging technology actually solves — and what it doesn't.

    buysellram.com/blog/inside-the

    #HBM #HBM4 #GPUMemory #AIInfrastructure #DataCenter #CXL #HighBandwidthFlash #AIHardware #MemoryHierarchy #AIInference #NVMe #ITAD #technology

  33. A modern AI server runs five layers of memory, from nanosecond on-chip SRAM to petabyte-scale SSDs, and keeping the GPU fed is the whole engineering game. This piece walks the full hierarchy: why HBM became the bottleneck, why adding more isn't simple, and where HBF, CXL, and PIM fit into the next generation.

    buysellram.com/blog/inside-the

    #HBM #HBM4 #GPUMemory #AIInfrastructure #DataCenter #CXL #HighBandwidthFlash #AIHardware #MemoryHierarchy #AIInference #NVMe #tech

  34. A modern AI server runs five layers of memory, from nanosecond on-chip SRAM to petabyte-scale SSDs, and keeping the GPU fed is the whole engineering game. This piece walks the full hierarchy: why HBM became the bottleneck, why adding more isn't simple, and where HBF, CXL, and PIM fit into the next generation.

    buysellram.com/blog/inside-the

    #HBM #HBM4 #GPUMemory #AIInfrastructure #DataCenter #CXL #HighBandwidthFlash #AIHardware #MemoryHierarchy #AIInference #NVMe #tech

  35. A modern AI server runs five layers of memory, from nanosecond on-chip SRAM to petabyte-scale SSDs, and keeping the GPU fed is the whole engineering game. This piece walks the full hierarchy: why HBM became the bottleneck, why adding more isn't simple, and where HBF, CXL, and PIM fit into the next generation.

    buysellram.com/blog/inside-the

    #HBM #HBM4 #GPUMemory #AIInfrastructure #DataCenter #CXL #HighBandwidthFlash #AIHardware #MemoryHierarchy #AIInference #NVMe #tech