Beepo 22B llama.cpp performance comparison on Vulkan between thine 780M/MI50/TeslaM40

Channel:
Why can't I have some original thoughts? Mod SM30
Subscribers:
42
Published on ● Video Link: https://www.youtube.com/watch?v=KH410LRaXr4


Duration: 0:00
0 views
0

780M has 12 Navi 3 CUs and 128bits DDR5.
MI50 has 60 Vega CUs and 4096bits HBM2.
Tesla M40 has 384bit GDDR5 and Idk SMs (3072 cores)

Newish llama.cpp from upload date running on Vulkan (Slackware current/6.12 kernel + 570 nvidia) 570.



Other Videos By Why can't I have some original thoughts? Mod SM30


2025-07-23Linux Phone - Poco X3 NFC (SD732G) - A few benchmarks
2025-07-23Beepo 22B llama.cpp performance comparison on Vulkan between thine 780M/MI50/TeslaM40
2025-07-14Speedometer G-forces calculation of the fastest Armored Cores (#1 ACFA Blader)
2025-06-03Time Series -- Vega 20 Number Crunching Curve (No Half)
2025-06-03Time Series -- Temps (C) per second of a 9700X with a Contac 9 Heatsink
2025-06-03Time Series -- Vega 20 Peak Temperatures per PPT
2025-06-03Time Series -- Vega 20 Number Crunching Curve
2025-06-03Time Series -- Vega 20 HBM2 Memory Bandwidth Curve
2025-05-31Piano played by Wolf bots #1
2025-05-30CM4S ClockworkPi Devterm viewing Armored Core TAS - Full twenty minutes!
2025-05-30CM4S ClockworkPi Devterm viewing Armored Core TAS - Kill Struggle Leader
2025-05-30CM4S ClockworkPi Devterm playing Akatsuki (Akiko Shikata)
2025-05-21Just booting Armbian (Tinker Board S RK3288) image on a Geniatech XPI-3288 for no real reason.......
2025-05-19Gini and GDP Graphs of the Netherlands
2025-05-18Cooling airflow drawing for auxiliary GPU case (MI50s)
2025-05-07Failed attempt to put a CPU tower cooler on a MI50
2025-04-18Window stream...
2025-04-18Penta-GPU PC Stream Part 4
2025-04-18Penta-GPU PC Stream Part 3
2025-04-17アスカの泣き声 (日本語)
2025-04-17アスカの血が凍る叫び (日本語) (クリーン版)