Caveats to the Butter Test
The growth of AI-powered data centers presents major challenges. These systems use processors that generate substantial heat. This creates economic and environmental hurdles. A fundamental shift is possible with highly efficient, cooler AI chips.
Breaking news
Anthropic Unveils Fable 5.2 and Mythos 5.2 AI Models
Revitalizing Classic PC Gaming Through Advanced Path Tracing Technology
Big Tech Deals: Z Fold 8 Discount, Free Buds, and OLED Savings
Nominal Raises $240M at $3B Valuation to Streamline Hardware ProductionDEEPX, recently covered by CNBC, offers a solution. They demonstrated their DX-M1 processor's low-temperature capability. This unusual thermal test used a dairy product. While competitor chips warmed a piece of butter quickly, the butter on the DEEPX platform remained solid.
This tech evolution matters globally. AI data center expansion raises sustainability concerns. Public sentiment is strong regarding these infrastructures. CNBC's report noted that hot processors and complex cooling systems cause difficulties.
Production and Future Plans
DEEPX claims a simple fix: optimizing their chips for lower operating temperatures. The CNBC video details the energy-efficient DX-M1. A direct comparison shows the DX-M1 NPU's thermal superiority. Kim Nok-won, DEEPX CEO, estimates a 25 to over 40-degree Celsius difference from rivals. He stated both processed the same AI task; all chips melted the butter except the DX-M1.
The dairy-based thermal challenge is striking. However, details in the comparison lack crucial data. Specifically, the test execution speed for each processor is missing. Any chip can stay cool by reducing power and clock speeds. PC enthusiasts often undervolt GPUs for better thermal performance and quieter operation.
Application Niche and Industry Context
The Elec reports the DX-M1 was made on Samsung 5nm technology starting August 2025. This performance secured about $13.3 million in orders. DEEPX plans the next chip, DX-M2. This uses Samsung 2nm and targets mass production in the second half of 2027.
These NPUs suit edge computing and robotics well. They differ from high-power Nvidia Blackwell and Ruben GPU racks mentioned by CNBC. Future systems might replace hot Nvidia GPUs with cooler custom accelerators. This could ease negative public reactions to AI data centers. However, DX-M1 and DX-M2 do not target the high-power niche; the butter test is more attention-grabbing than a direct segment proof.
CNBC also highlighted SK hynix's iHBM progress. This new technology promises lower energy use and cooler operation. An executive told CNBC that 'i' stands for „Integrated Cooling Element.”CoolIT is also exploring similar methods, bringing water cooling closer to heat-generating silicon.
