M2 Ultra (76 GPU)

15 records across 2 models (quants: F16, Q4_0, Q8_0).

hardwaremodelquantbackendtok/spp tok/sWttft sctx/tgdateclassflagssource
M2 Ultra (76 GPU)Llama 2 7B Q4_0llama.cpp (Metal)125.211488.81–––2023-11-22community–source
M2 Ultra (76 GPU)Llama 2 7B Q4_0llama.cpp (Metal)109.411412.42–––2023-11-22community–source
M2 Ultra (76 GPU)Llama 2 7B Q4_0llama.cpp (Metal)108.81391.78–––2023-11-22community–source
M2 Ultra (76 GPU)Llama 2 7B Q4_0llama.cpp (Metal)94.271238.48–––2023-11-22community–source
M2 Ultra (76 GPU)Llama 2 7B Q8_0llama.cpp (Metal)82.831486.51–––2023-11-22community–source
M2 Ultra (76 GPU)Llama 2 7B Q8_0llama.cpp (Metal)73.351386.97–––2023-11-22community–source
M2 Ultra (76 GPU)Llama 2 7B Q8_0llama.cpp (Metal)73.111368.18–––2023-11-22community–source
M2 Ultra (76 GPU)Llama 2 7B Q8_0llama.cpp (Metal)66.641248.59–––2023-11-22community–source
M2 Ultra (76 GPU)Llama 2 7B F16llama.cpp (Metal)49.551627.82–––2023-11-22community–source
M2 Ultra (76 GPU)Llama 2 7B F16llama.cpp (Metal)43.241561.35–––2023-11-22community–source
M2 Ultra (76 GPU)Llama 2 7B F16llama.cpp (Metal)43.151525.95–––2023-11-22community–source
M2 Ultra (76 GPU)Llama 2 7B F16llama.cpp (Metal)41.021401.85–––2023-11-22community–source
M2 Ultra (76 GPU)llama 7B Q4_0llama.cpp (Metal)94.271238.48–––2023-11-22community–source
M2 Ultra (76 GPU)llama 7B Q8_0llama.cpp (Metal)66.641248.59–––2023-11-22community–source
M2 Ultra (76 GPU)llama 7B F16llama.cpp (Metal)41.021401.85–––2023-11-22community–source

All rows cite their source. Provenance: community 40018 sourced 1085

← all hardware