3 ms·
To increase performance, you need bigger foundation models, like LLaMA 65B. You will always be RAM-bound for that (~40GB for LLaMA 65B INT4). So the next step i
by espadrine 4y ago
To increase performance, you need bigger foundation models, like LLaMA 65B. You will always be RAM-bound for that (~40GB for LLaMA 65B INT4). So the next step is to upgrade your motherboard to have multiple GPU slots. If you really wanted to stick to a single-GPU setup, you could upgrade to RTX A6000, but it is more expensive than two RTX 3090 while holding the same amount of RAM.
- yieldcrv 4y agoIts a dual slot gpu What benefit does that offer over 2 GPUs?
- espadrine 4y agoThe two GPUs will need to exchange information in order to complete inference, by having one GPU hold half of the network weights, and the other holding the other half. That transmission of information will be limited by the PCIe bandwidth; for instance, 30 GB/s with v4x16. Meanwhile the A6000 VRAM bandwidth is 768 GB/s (= 16 Gb/s (GDDR6) × 384 bit-width ÷ 8 bits per byte).
- coolspot 4y agoTwo 3090 can be connected using NVLink bridge which is much faster than PCI-E.
- espadrine 4y agoI couldn’t find precise information on what bandwidth you’d get with NVLink on the 3090. To be fair, though, if all we do is inference, using Huggingface pipeline parallelism, the amount of data transferred is pretty small: 8192×2×n_tokens bytes; for most uses, with a recent PCIe setup, that bottleneck will take less than 0.1 ms per token generated, which may not be the dominant latency. (Also, the RTX 3090 has faster VRAM, >900 GB/s, than the A6000, because it is GDDR6X.)