Urgent.News

What's breaking now, across thousands of outlets.

Tech

Solid-state cooling could be the secret to faster AI laptops

Memory bandwidth, not processing power, is becoming the real bottleneck for AI laptops. A new white paper explores why fan based cooling is struggling to keep up, and how solid state cooling could replace it.

Solid-state cooling could be the secret to faster AI laptops

The next generation of laptops may face their biggest obstacle not in their processors, but in how effectively they dissipate heat. A white paper from Ventiva, in collaboration with analyst firm Moor Insights and Strategy, suggests that thermal design is set to become the key determinant of an AI laptop's performance capabilities.

Local AI performance is quantified in tokens per second, which is primarily determined by memory bandwidth. Contemporary laptops typically employ 128-bit memory buses coupled with LPDDR5 memory, limiting bandwidth to approximately 150 GB per second. This bandwidth is insufficient to support the operation of modern AI models at practical speeds, prompting chip manufacturers to accelerate their efforts in enhancing memory bus width.

Qualcomm has upgraded its Snapdragon X2 Elite chips to 192-bit buses, while AMD and NVIDIA now provide 256-bit options in their client silicon. Apple has achieved the most significant advancements with its M5 Max, utilizing a massive 512-bit bus capable of executing large open-source models such as GPT-OSS 120 B entirely on the device.

The report anticipates a rapid acceleration in the transition towards wider memory buses with the introduction of LPDDR6 memory, which promises faster data rates and improved efficiency. Mainstream adoption of this technology is projected to occur by 2027 and 2028. However, a significant challenge arises as wider memory buses necessitate memory chips to be positioned extremely close to the processor, within traces shorter than 25 millimeters.

Traditionally, fans and heat pipes occupied this space in thin and light laptops. Apple circumvents this issue by integrating memory directly onto its chip package, but other laptop manufacturers must navigate the same spatial constraints within traditional chassis designs.

Conventional fan-based cooling systems are unable to maintain their effectiveness as memory layouts become denser, particularly as chip manufacturers strive for 256-bit and wider buses. This predicament is driving solid-state cooling into the spotlight as a viable solution. The report highlights Ventiva's ionic cooling platform, which utilizes charged particles to transport air silently without any moving parts.

It also references Frore's AirJet technology, which has successfully demonstrated its cooling capabilities in an Intel reference laptop that is merely 11.3mm thin, sustaining 15 watts of fanless cooling while remaining noiseless. Meanwhile, YPlasma presented a plasma-powered laptop cooler at CES, employing ionized gas instead of a spinning fan to move air, claiming operation at just 17 decibels.

The paper frames 2027 and 2028 as a critical turning point, when wider memory buses and LPDDR6 will become the industry standard. If this timeline materializes, laptop manufacturers will be compelled to reevaluate their cooling strategies instead of treating it as a secondary consideration.

Written by urgent.news from Digital Trends's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.

Read the original at digitaltrends.com →

More in Tech

DaDaScribe API: YouTube Multi-Language Transcript + SRT in One Request

Most transcription APIs still force you to do this: Download the YouTube video Extract the audio with yt-dlp + ffmpeg Upload the file Call the transcription endpoint Call a separate translation API…

  • API simplifies transcription by eliminating steps like audio extraction and translation.
  • Pre-processing improves transcript quality for music videos and real-world recordings.

How the Robot Games are going in Beijing

The World Humanoid Robot Games are underway in Beijing, China, in a stadium repurposed from the 2022 Winter Olympics. Many of the machines on display can now outrun and out-jump even the best…

More from Monday 24 August →