๊ฐ€์ž… ํ›„ ์ดˆ๋Œ€ ๋งํฌ๋ฅผ ๊ณต์œ ํ•˜๋ฉด ๋™์˜์ƒ ์žฌ์ƒ ๋ฐ ์ดˆ๋Œ€ ๋ณด์ƒ์„ ๋ฐ›์„ ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค.

David Hendrickson
@TeksEdge
CEO & Founder | PhD | Startup Advisor | @Columbia | Author Generative Software Engineering | ๐Ÿ”” Follow for AI & Vibe Coding Tips ๐Ÿ‘‡
๊ฐ€์ž… July 2023
552 ํŒ”๋กœ์ž‰ ์ค‘    12.5K ํŒฌ
AMD Max+ Pro 495 benchmarks are starting to trickle out. This user got their single 192GB AMD Minisforum mini PC running gpt-oss 120B MXFP4 with llama.cpp at 55.7 tps. Setup ๐Ÿง  gpt-oss 120B ๐Ÿ’พ 192GB LPDDR5X-8533 unified memory ๐ŸŽฎ Radeon 8065S โšก 55.7 tps decode ๐Ÿš€ 1,135 tps prefill And deeper into context, at the source's ~32K test depth โšก 42.3 tps decode ๐Ÿš€ 385 tps prefill This is why these 192GB unified-memory machines could be a big deal for Local AI. Link in ALT
๋” ๋ณด๊ธฐ