Update README.md

README.md

@@ -54,32 +54,32 @@ For commercial licensing, cluster deployment, or enterprise use of the JiRack Co
 
 ## Hardware Recommendations for AMD Systems
 
-Here's a clear and practical hardware recommendation for running **JiRack Coder 7B** (INT8 / INT4) on AMD platforms:
+### Recommended Hardware for JiRack Coder 7B INT8
 
-### Best AMD Hardware Recommendations
+| Use Case              | CPU                              | GPU (ROCm)                        | VRAM / RAM     | Expected Speed      | Recommendation     |
+|-----------------------|----------------------------------|-----------------------------------|----------------|---------------------|--------------------|
+| **Recommended**       | Ryzen 7 7700 / 9700X            | RX 7900 XTX / 7900 XT            | 24GB VRAM      | 50-75 tokens/s      | Best choice        |
+| **High Performance**  | Ryzen 9 7950X / 9950X           | RX 7900 XTX                      | 24GB+ VRAM     | 65-90 tokens/s      | Excellent          |
+| **Enterprise**        | EPYC 7003/9004 series           | MI300X or 2x RX 7900 XTX         | 48GB+ VRAM     | 90-140 tokens/s     | For 32B model      |
+| **Budget Option**     | Ryzen 5 7600 / 9600X            | RX 7800 XT (16GB)                | 16GB VRAM      | 35-50 tokens/s      | Acceptable         |
 
-| Use Case              | Recommended CPU                  | Recommended GPU (ROCm)               | RAM      | Expected Speed       | Notes                     |
-|-----------------------|----------------------------------|--------------------------------------|----------|----------------------|---------------------------|
-| **Good Performance**  | Ryzen 7 7700 / 9700X            | RX 7900 XTX / 7900 XT (24GB)        | 32GB+    | 45-70 tokens/s       | Best balance              |
-| **Very Good**         | Ryzen 9 7950X / 9950X           | RX 7900 XTX (24GB)                   | 64GB     | 55-85 tokens/s       | Strong choice             |
-| **Enterprise / Fast** | EPYC 7003/9004 series           | Instinct MI300X or 2x RX 7900 XTX   | 128GB+   | 80-120+ tokens/s     | For 32B model too         |
-| **Budget / Decent**   | Ryzen 5 7600 / 9600X            | RX 7800 XT (16GB)                    | 32GB     | 35-50 tokens/s       | Acceptable                |
+### Important Memory Notes
 
-### Key Recommendations (AMD)
+Even though the 7B INT8 model itself takes approximately **8–9 GB**, we recommend **at least 24GB VRAM** for the following reasons:
 
-**Top Pick (Best Price/Performance):**
-- **CPU**: Ryzen 7 9700X or Ryzen 9 7950X
-- **GPU**: Radeon RX 7900 XTX (24GB VRAM) — strongly recommended
-- **RAM**: 64GB DDR5
+- KV-cache consumption during generation (especially with long context)
+- ONNX Runtime overhead and temporary buffers
+- System stability and to avoid Out of Memory errors
+- Room for larger context windows
 
-**Important Notes:**
-- ROCm support is decent on RX 7000 series, but less stable than NVIDIA.
-- For pure CPU inference, Ryzen 9 7950X / 9950X with 64GB+ RAM performs excellently.
-- The **INT4** version runs noticeably faster than INT8 on AMD CPUs.
+**Minimum recommended:** 24GB VRAM (RX 7900 series)  
+**Ideal:** 24–32GB VRAM
+
+For pure CPU inference (no GPU), we recommend at least **64GB system RAM** (Ryzen 9 7950X/9950X).
 
 ---
 
-Would you like me to add recommendations based on your budget (e.g. under $1500, mid-range, high-end)?
+
 
 ## 📧 Contact & Licensing
 For joint venture opportunities, hardware integration, or licensing inquiries: