Ashiedu's picture
Sync unified workbench
0490201 verified
# ISSUE T-017C: RX 6700 XT Optimization and Tuning Framework
**Status:** `status:complete`
**Parent:** #139 (GENESIS)
**GitHub ID:** 175
## Summary
Owns the hardware-specific optimization contract for the workstation.
## Completed
- [x] Forced `HSA_OVERRIDE_GFX_VERSION=10.3.0` for RDNA 2 stability.
- [x] Implemented `PYTORCH_HIP_ALLOC_CONF` for memory fragmentation control.
- [x] Defined MatFormer strategy for elastic inference (800M -> 200M).
- [x] Integrated 4-bit (NF4) quantization plan.
## New Sub-tasks / Refinements
- [ ] Implement VRAM spike monitoring in Trackio logs.
- [ ] Create benchmark script for `fp16` vs `int8` vs `nf4` on RX 6700 XT.
- [ ] Tuning: verify `HSA_ENABLE_SDMA=0` impact on large model transfers.