Gunnir Quietly Reveals Intel Arc A380 in MXM Form-Factor
Gunnir Quietly Reveals Intel Arc A380 in MXM Form-Factor
China-based Gunnir is the only company that exclusively produces Intel-based graphics cards. It now transpires (opens in new tab) (thanks to @Zed_Wang) that the company also has an MXM module based on Intel’s Arc A380 graphics processor equipped with 6GB of GDDR6 memory. Such modules can be used for notebooks, desktops in proprietary form-factors, and embedded applications.
Gunnir’s LM-A380-6G MXM 3.1 Type-A modules seemingly use a desktop-grade Intel Arc A380 (ACM-G11 GPU with eight Xe cores or 1024 stream processors) graphics processor with 6GB of memory connected to the chip via a 96-bit interface, which is highly likely to be faster than laptop-oriented Arc A370M — ACM-G11 with eight Xe cores, 4GB VRAM, and a 64-bit memory interface.
MXM modules are not exactly popular these days. Even high-end gaming notebooks often come with soldered-down GPUs and VRAM. The only exceptions are some Alienware / Dell laptops that use proprietary GPU modules. Making MXM modules based on Intel’s desktop GPUs is something unexpected from Gunnir, yet keeping in mind the company’s focus on Intel-based graphics cards, it is possible that the company wants to address desktops and embedded machines with its MXM 3.1 Type-A cards.
Meanwhile, the LM-A380-6G module is rated for up to 75W thermal board power, which is pretty hot for mainstream laptops, though it’s suitable for desktops and embedded applications.
You can expect similar performance to the regular Arc A380, but perhaps more important than raw GPU performance will be the video capabilities. Having a compact setup powering digital signage might be one use.
Gunnir’s Arc A380 6GB MXM 3.1 Type-A modules are currently available from JD.com (opens in new tab) and Taobao for ¥1999 CNY – ¥2199 CNY ($260 – $286 without VAT). That’s considerably higher than MSRP of the desktop Arc A380. However, keeping in mind that the module is likely aimed at ultra-compact desktops and embedded applications, such a price does not seem to be surprisingly high.