Close Menu
VernoNews
  • Home
  • World
  • National
  • Science
  • Business
  • Health
  • Education
  • Lifestyle
  • Entertainment
  • Sports
  • Technology
  • Gossip
Trending

Luke Shaw Welcomes Fourth Child Paisley with Anouska Santos

March 4, 2026

UHart Names $2.4M Athletics Complex After NBA Star Vin Baker

March 4, 2026

Best Supermarket Scones Ranked: Morrisons Beats M&S, Aldi

March 4, 2026

Coffee Prices Surge 20-30% in India Amid Iran-Israel Conflict

March 4, 2026

Bitget Lists Opinion (OPN) for Spot Trading in Innovation Zone

March 4, 2026

Park Ji-yoon Reaches 56kg Lowest Weight with Daily Twists

March 4, 2026

India Secures Fuel Stability with Russia Oil Boost Amid Gulf Crisis

March 4, 2026
Facebook X (Twitter) Instagram
VernoNews
  • Home
  • World
  • National
  • Science
  • Business
  • Health
  • Education
  • Lifestyle
  • Entertainment
  • Sports
  • Technology
  • Gossip
VernoNews
Home»Technology»Subsequent-generation HBF reminiscence will feed AI accelerators sooner than ever, altering how GPUs deal with huge datasets effectively
Technology

Subsequent-generation HBF reminiscence will feed AI accelerators sooner than ever, altering how GPUs deal with huge datasets effectively

VernoNewsBy VernoNewsJanuary 24, 2026No Comments4 Mins Read
Facebook Twitter Pinterest LinkedIn Tumblr Reddit WhatsApp Email
Subsequent-generation HBF reminiscence will feed AI accelerators sooner than ever, altering how GPUs deal with huge datasets effectively
Share
Facebook Twitter LinkedIn Pinterest WhatsApp Email

[ad_1]


  • HBF provides ten occasions HBM capability whereas remaining slower than DRAM
  • GPUs will entry bigger knowledge units via tiered HBM-HBF reminiscence
  • Writes on HBF are restricted, requiring software program to concentrate on reads

The explosion of AI workloads has positioned unprecedented stress on reminiscence techniques, forcing firms to rethink how they ship knowledge to accelerators.

Excessive-bandwidth reminiscence (HBM) has served as a quick cache for GPUs, permitting AI instruments to learn and course of key-value (KV) knowledge effectively.

Nevertheless, HBM is pricey, quick, and restricted in capability, whereas high-bandwidth flash (HBF) provides a lot bigger quantity at slower speeds.


Chances are you’ll like

  • Imec 3D HBM-on-GPU technology HBM-on-GPU set to energy the subsequent revolution in AI accelerators – and simply to verify, there isn’t any means this can come to your video card anytime quickly

  • JEDEC Solid State Technology Association SPHBM4 New ‘serial’ tech will considerably cut back the price of reminiscence — HBM reminiscence, that’s, the type of RAM solely AI hyperscalers can use, however hey, not less than they will not go after shopper RAM, or would they?

  • Phison CEO Pua Khein Seng ‘In AI fashions, the actual bottleneck isn’t computing energy — it’s reminiscence’: Phison CEO on 244TB SSDs, PLC NAND, why high-bandwidth flash isn’t a good suggestion, and why CSP revenue goes hand in hand with storage capability

How HBF enhances HBM

HBF’s design permits GPUs to entry a wider knowledge set whereas limiting the variety of writes, roughly 100,000 per module, which requires software program to prioritize reads over writes.

HBF will combine alongside HBM close to AI accelerators, forming a tiered reminiscence structure.

Professor Kim Joungho of KAIST compares HBM to a bookshelf at dwelling for quick examine, whereas HBF features like a library with way more content material however slower entry.

“For a GPU to carry out AI inference, it should learn variable knowledge known as the KV cache from the HBM. Then, it interprets this and spits out phrase by phrase, and I feel it’ll make the most of the HBF for this activity,” mentioned Professor Kim.

Signal as much as the TechRadar Professional publication to get all the highest information, opinion, options and steering your enterprise must succeed!

“HBM is quick, HBF is gradual, however its capability is about 10 occasions bigger. Nevertheless, whereas HBF has no restrict on the variety of reads, it has a restrict on the variety of writes, about 100,000. Due to this fact, when OpenAI or Google write packages, they should construction their software program in order that it focuses on reads.”

HBF is anticipated to debut with HBM6, the place a number of HBM stacks interconnect in a community, rising each bandwidth and capability.

The idea envisions future iterations like HBM7 functioning as a “reminiscence manufacturing facility,” the place knowledge will be processed immediately from HBF with out detouring via conventional storage networks.


Chances are you’ll like

  • Imec 3D HBM-on-GPU technology HBM-on-GPU set to energy the subsequent revolution in AI accelerators – and simply to verify, there isn’t any means this can come to your video card anytime quickly

  • JEDEC Solid State Technology Association SPHBM4 New ‘serial’ tech will considerably cut back the price of reminiscence — HBM reminiscence, that’s, the type of RAM solely AI hyperscalers can use, however hey, not less than they will not go after shopper RAM, or would they?

  • Phison CEO Pua Khein Seng ‘In AI fashions, the actual bottleneck isn’t computing energy — it’s reminiscence’: Phison CEO on 244TB SSDs, PLC NAND, why high-bandwidth flash isn’t a good suggestion, and why CSP revenue goes hand in hand with storage capability

HBF stacks a number of 3D NAND dies vertically, just like HBM stacking DRAM, and connects them with through-silicon vias (TSVs).

A single HBF unit can attain 512GB capability and obtain as much as 1.638TBps bandwidth, far exceeding customary SSD NVMe PCIe 4.0 speeds.

SK Hynix and Sandisk have demonstrated diagrams displaying higher NAND layers related via TSVs to a base logic die, forming a purposeful stack.

Prototype HBF chips require cautious fabrication to keep away from warping within the decrease layers, and extra NAND stacks would additional improve the complexity of the TSV connections.

Samsung Electronics and Sandisk plan to connect HBF to Nvidia, AMD, and Google AI merchandise inside the subsequent 24 months.

SK Hynix will launch a prototype later this month, whereas the businesses are additionally engaged on standardization via a consortium.

HBF adoption is anticipated to speed up within the HBM6 period, and Kioxia has already prototyped a 5TB HBF module utilizing PCIe Gen 6 x8 at 64Gbps. Professor Kim predicts that the HBF market might surpass HBM by 2038.

Through Sisajournal (initially in Korean)


Comply with TechRadar on Google Information and add us as a most well-liked supply to get our professional information, critiques, and opinion in your feeds. Make sure that to click on the Comply with button!

And naturally you too can observe TechRadar on TikTok for information, critiques, unboxings in video kind, and get common updates from us on WhatsApp too.



[ad_2]

Avatar photo
VernoNews

    Related Posts

    Ingalls Shipbuilding Marks Keel for USS Philadelphia (LPD 32)

    March 4, 2026

    Rakuten Mobile Picks Samsung for Japan-Wide 5G Open RAN Radios

    March 3, 2026

    Supreme Court Blocks Redraw of Sole GOP-Held NYC District for 2026

    March 3, 2026

    Comments are closed.

    Don't Miss
    World

    Luke Shaw Welcomes Fourth Child Paisley with Anouska Santos

    By VernoNewsMarch 4, 20260

    Manchester United defender Luke Shaw and his long-term partner Anouska Santos celebrate the arrival of…

    UHart Names $2.4M Athletics Complex After NBA Star Vin Baker

    March 4, 2026

    Best Supermarket Scones Ranked: Morrisons Beats M&S, Aldi

    March 4, 2026

    Coffee Prices Surge 20-30% in India Amid Iran-Israel Conflict

    March 4, 2026

    Bitget Lists Opinion (OPN) for Spot Trading in Innovation Zone

    March 4, 2026

    Park Ji-yoon Reaches 56kg Lowest Weight with Daily Twists

    March 4, 2026

    India Secures Fuel Stability with Russia Oil Boost Amid Gulf Crisis

    March 4, 2026
    About Us
    About Us

    VernoNews delivers fast, fearless coverage of the stories that matter — from breaking news and politics to pop culture and tech. Stay informed, stay sharp, stay ahead with VernoNews.

    Our Picks

    Luke Shaw Welcomes Fourth Child Paisley with Anouska Santos

    March 4, 2026

    UHart Names $2.4M Athletics Complex After NBA Star Vin Baker

    March 4, 2026

    Best Supermarket Scones Ranked: Morrisons Beats M&S, Aldi

    March 4, 2026
    Trending

    Coffee Prices Surge 20-30% in India Amid Iran-Israel Conflict

    March 4, 2026

    Bitget Lists Opinion (OPN) for Spot Trading in Innovation Zone

    March 4, 2026

    Park Ji-yoon Reaches 56kg Lowest Weight with Daily Twists

    March 4, 2026
    • Contact Us
    • Privacy Policy
    • Terms of Service
    2025 Copyright © VernoNews. All rights reserved

    Type above and press Enter to search. Press Esc to cancel.