FREE TOOL · COMPARE
Compare AI Computers side by side
Memory, speed, the largest model each can run and what it costs in taka — pick the GPUs you are deciding between.
| NVIDIA L4 | NVIDIA A100 | NVIDIA H100 | |
|---|---|---|---|
| Price per hour | ৳147 | ৳638 | ৳1,047 |
| GPU memory (VRAM) | 24 GB | 80 GB | 80 GB |
| Architecture | Ada | Ampere | Hopper |
| AI compute (BF16)dense, spec sheet | 121 TFLOPS | 312 TFLOPS | 989 TFLOPS |
| Memory bandwidth | 300 GB/s | 2,039 GB/s | 3,350 GB/s |
| Largest LLM it runssingle GPU, 4k context | 10B at FP16 · 35B at 4-bit | 36B at FP16 · 129B at 4-bit | 36B at FP16 · 129B at 4-bit |
| Llama 3.1 8B speed4-bit, one user | ≈ 40 tokens/s | ≈ 271 tokens/s | ≈ 445 tokens/s |
| 10 hours | ৳1,470 | ৳6,380 | ৳10,470 |
| 40 hours / month | ৳5,880 | ৳25,520 | ৳41,880 |
| Good for | Fast inference, video AI, efficient fine-tuning | LLM fine-tuning, research, high-throughput training | Large model training, transformer workloads, HPC |
Speeds and “largest model” are estimates from public spec sheets (see the VRAM calculator for the method). Prices are ComputeBD’s current hourly rates.
Try it on a real AI Computer
JupyterLab opens in about a minute. Pay in taka with bKash, billed by the minute — stop whenever you like.