Gemma 4 tested on three old GTX-1070 GPUs

Several Gemma 4 models were tested for on a machine with three Nvidia GTX-1070 s. The system used Kubuntu 26.04, an AMD Ryzen 5 3600 6-core CPU, 48GiB of memory, and 8GiB of VRAM per card, for 24GiB total VRAM. Each GPU was power-limited to about 120 watts, which reduced by about 5% but lowered stress on the .

The tests used llama.cpp with Vulkan and GGUF model files. The 31B quantized model was 17.52GiB, took 3 minutes 35 seconds to benchmark, and generated 7.12 . The 12B Q8 model was 12.69GiB, took 1 minute 59 seconds, and generated 13.47 .

The 26B.A4B Q4 model was 15.83GiB, took 1 minute 45 seconds, and generated 41.28 , while the 26B.A4B QAT model was 13.26GiB, took 1 minute 30 seconds, and reached 53.08 . The three cards were installed at different PCIe speeds, with one card connected through a 1x extender.

Key points

  • The setup used three GTX-1070 GPUs with 24GiB total VRAM.
  • Power limits caused about a 5% loss but reduced power stress.
  • The 26B.A4B QAT model produced the fastest generation result at 53.08 .
  • The larger 31B Q4 model generated much more slowly at 7.12 .
  • For , and hardware layout can strongly affect cost and speed.
Read original