- Graphic processors on artificial intelligence servers are already promoting 1 kW; Cooling – narrow place
- Corintis wants to process graphic processors of 10 kW using microfluid technology
- Microsoft confirmed a breakthrough breakthrough tested on live servers
The growth of artificial intelligence sharply increased the demand for more powerful computer chips, but this impetus is accompanied by a constant obstacle.
Since processors are scaled in performance, heat generation becomes a limiting factor, forcing researchers and companies to look for new methods to prevent overheating.
Semiconductor startup Corinthian Currently, it attracts attention with the help of a microfluid cooling system, which, according to, can achieve performance far beyond the borders of today's ordinary approaches.
Cooling as the next border for AI chips
“Our mission is to unlock the 10x of the best cooling to ensure the future calculations during the short cycle, and using existing investments in the infrastructure in the data center today,” said Remco Van ERP, co-founder and Corintis Director General.
Current high -performance graphic processors in AI Data processing centers Usually scattered between 400 watts to 1 kW.
NvidiaFor example, H100 reaches about 700-800 watts, while the next generation accelerators, such as the GB200, should exceed 1 kW.
Thus, the assertion of the coronant about achieving 10 times better cooling means that its goal is to process graphic processors of 10 kW, a jump far beyond the limits of today's levels.
Nevertheless, the startup, by the same, achieves progress in achieving this goal as Microsoft The recently disclosed results of his partnership with Corintis, confirming off -road microfluid cooling on servers working on servers.
“The thermal margin is translated at the software level to get a greater potential for productivity and acceleration,” said Husam Alissa, director of systemic technologies in cloud operations and innovation in Microsoft.
“It also provides new three-dimensional architectures for chips, which cannot today be impossible due to thermal restrictions on high-power SOC laying without internal cooling of the layer.”
Although such claims imply broad consequences for both Cloud computing And the future design of the processor, independent validation will be necessary before wider acceptance may occur.
Unlike blocks that dominate the market, the startup says that it develops cooling systems that direct the liquid exactly where it is most necessary.
The startup also provided the round of the A 24 million dollars series, led by Blueyard Capital, while participating in how Lip-Bu Tan, Intel Chief Executive Director, and Jeff Lyon, earlier from Coolit.
Corintis claims that this financing in combination with its recent cooperation positions it for the rapid scale of production.
He claims that at the age of 2026, it can expand to more than one million units of microfluid cold plates.
Nevertheless, industry leaders, such as NVIDIA, have already shifted towards liquid cooling for a powerful data processing center Graphic processorsAssuming that market demand is real.
Nevertheless, the path from a laboratory demonstration to mass deployment is rarely simple.
Watch Techradar in Google News And Add us as a preferred source To get our expert news, reviews and opinions in your channels. Be sure to click the “following” button!
And, of course, you can also Follow Techradar on Tiktok For news, reviews, unpacking in the form of video and regular updates from us on WhatsApp too much.
