ASUS rejoins Intel discrete GPU market with Arc Pro B70 32GB card after 5 years out of Intel graphics. It's big deal. Because of a discrete GPU option for localized AI inference workloads that doesn't need cloud. If you're seeking budget entry to generative AI, you now have the option to test out single GPU on desktop.
ASUS releases its first Intel Arc discrete GPU for professional AI workloads
ASUS's Arc Pro B70 is aimed at AI inference pro tasks, not gaming. It's powered by Intel's BMG-G31 GPU with 32 Xe2 cores and 256 XMX engines. This card is a nice option for local AI frameworks like Ollama, llama.cpp, vLLM, Pytorch, and OpenVINO.
Specifications
- GPU: Intel BMG-G31
- Memory: 32GB GDDR6 ECC
- Bandwidth: 608 GB/s
- Max GPU Clock: 2800 MHz
- INT8 Performance: 367 TOPS
It's got 32GB of GDDR6 ECC memory running through 256-bit interface for big language models. Memory bandwidth is at 608 GB/s and max clock is 2800 MHz. It can achieve up to 367 INT8 TOPS with 310W TBP.
PCIe 5.0 x16 interface with 12V-2×6 power connector links this card to your system. There are 3 DisplayPort 2.1 and 1 HDMI 2.1b outputs for monitors. Dual slot blower cooler for thermal dissipation in workstations.
Pricing and availability for UGen Arc Pro B70 from ASUS are unknown for now. Although they launched Intel DG1 card in 2021, they missed out on Arc Alchemist and Battlemage gaming. Now they're targeting professional compute tasks, not gaming.
Discussion
0 comments
Log in to join the thread with a thoughtful take, question, or correction.