Workbench / Online
BUILD YOUR COMPUTE STACK
Start with the model. MuseBoard maps the memory and hardware. Settings are mirrored to the URL, so any configuration can be shared.
Compute workbench
B / Estimated VRAM
Inference5.3GB
Serving: weights + KV cache + runtime overhead. Estimate — not a guarantee.
- Model weights1.9 GB
- KV cache3.2 GB
- Runtime overhead0.19 GB
Memory headroom
Free after estimated load
Utilization
Target ≤ 90% of 40.0 GB
Decode ceiling
Theoretical, batch 1, bandwidth-bound
VRAM usage
5.3 GB / 40.0 GB
Pinned configuration
1 × NVIDIA A100 40GB
- Total VRAM
- 40.0 GB
- Topology
- Single GPU
- Interconnect
- NVLink 3 · 600 GB/s
- Board power
- 400 W
Compatibility / 1 × A100 40GB
- Ready
Inference
5.3 GB · 13% of 1 × A100 40GB
- Ready
Fine-tuning
6.3 GB · 16% of 1 × A100 40GB
- Limited
Training
71.1 GB · Needs 2 × A100 40GB
Alternative configurations
- 01SERVE / READY
Inference
Estimate memory requirements for serving models.
weights + KV cache + overhead
Open in workbench - 02ADAPT / LORA
Fine-tuning
Explore memory requirements for adapting existing models.
frozen base + adapters + activations
Open in workbench - 03TRAIN / ADAM
Training
Understand large-scale compute requirements.
weights + grads + optimizer + activations
Open in workbench
MuseBoard gives first-order estimates for planning conversations. It does not replace profiling, capacity testing or detailed ML infrastructure planning.