GMKtec EVO T2 Series Supercomputer with Intel 18A Technology
The GMKtec EVO T2 Series Supercomputer which achieves 180 TOPS of AI performance uses Intel 18A technology. Intel unveiled its third generation Core Ultra platform which has the Panther Lake code name during the Shanghai flagship event on March 12 2026. The Intel ecosystem partner GMKtec used this event platform to showcase its EVO T2 Desktop AI Supercomputer Series. The hardware enables consumers to use Intels 18A process node for their first implementation of artificial intelligence models which operate entirely within local computing environments.
The EVO T2 operates with Intel Core Ultra X9 388H and X7 358H processors as its main system components. The chips enter the sub 2nm technological era through their use of 18A manufacturing process which operates at 1.8nm class dimensions. The consumer desktop 12nm process node introduces two essential technological advancements
- RibbonFET A gate all around (GAA) transistor architecture that improves power efficiency and performance density
- PowerVia Backside power delivery technology that separates power routing from signal routing to reduce interference and heat
The Panther Lake architecture features 16 cores which include 4 Performance cores and 8 Efficient cores and 4 Low Power Efficient cores. The performance enhancements now deliver 18% better IPC efficiency compared to earlier CPU designs while multi threaded execution performance improved by 30% or more.
The primary objective of the EVO T2 project establishes direct AI operation without needing cloud resources. The system achieves 180 TOPS of processing power through its combined CPU and NPU and integrated Xe3 based Arc B390 GPU performance capabilities. The desktop system supports local execution of large language models with 70 billion parameters by using its available processing capacity.
The system provides users with immediate access through its GMKtec Claw AI system which comes equipped with all necessary local AI tools. Testing shows that code generation and complex document analysis tasks which used to take several minutes now complete within 40 seconds.
The EVO T2 system uses Phison aiDAPTIV technology to overcome memory constraints which occur during execution of large AI models on small scale hardware. The system employs a NAND based caching system which uses aiDAPTIV Cache as additional memory to expand system DRAM resources. The system achieves more extensive dataset processing capabilities through AI model partitioning and inactive slice storage on AI SSD which ensures that inference speeds remain stable for 70B parameter models.
The Intel Arc B390 GPU uses Xe3 architecture and has 12 Xe cores which enable it to perform real time ray tracing and XeSS 3 AI upscaling functions. This solution enables users to achieve graphics performance that matches mid range discrete desktop graphics cards while using a compact mini PC design.
The EVO T2 system provides multiple high speed I/O features which serve professional users
- Dual Ethernet 10GbE and 5GbE ports for enterprise networking
- OCuLink Support for external GPU (eGPU) enclosures to further expand rendering or AI compute capabilities
- Storage Two M.2 slots supporting PCIe 5.0 and PCIe 4.0 for up to 16TB of total capacity
- Power USB4 connectivity with 100W Power Delivery and 40Gbps data transfer
Intel CEO Lip Bu Tan personally endorsed the EVO T2 highlighting it as a benchmark for local AI workload demonstration. The system is currently available on JD.com in China with global distribution through GMKtecs official website currently in preparation. The device represents a turning point in the mini PC sector which now provides dedicated computing power for developers and researchers and creative professionals.
