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
Inference2.0GB
Serving: weights + KV cache + runtime overhead. Estimate — not a guarantee.
- Model weights1.5 GB
- KV cache0.30 GB
- Runtime overhead0.15 GB
Memory headroom
Free after estimated load
Utilization
Target ≤ 90% of 256.0 GB
Decode ceiling
Theoretical, batch 1, bandwidth-bound
VRAM usage
2.0 GB / 256.0 GB
Pinned configuration
1 × AMD Instinct MI325X 256GB
- Total VRAM
- 256.0 GB
- Topology
- Single GPU
- Interconnect
- Infinity Fabric
- Board power
- 1,000 W
Compatibility / 1 × MI325X
- Ready
Inference
2.0 GB · 1% of 1 × MI325X
- Ready
Fine-tuning
9.4 GB · 4% of 1 × MI325X
- Ready
Training
61.8 GB · 24% of 1 × MI325X
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.