| ▲ | striking a day ago |
| Is GPU memory size really changing that quickly? For that matter, is model size? |
|
| ▲ | kadushka 21 hours ago | parent | next [-] |
| What's rapidly changing are quantization algorithms, and hardware features to support those algorithms. For example, Blackwell GPUs support dynamic FP4 quantization with group size 16. At that group size it's close to lossless (in terms of accuracy metrics). |
|
| ▲ | latchkey 20 hours ago | parent | prev | next [-] |
| Both AMD and Nvidia are dumping more and more memory into their GPUs. MI300x is 192GB HMB3, MI325x is 256 HMB3e, MI355x should be 288 HBM3e (and support FP4/6). |
| |
| ▲ | NBJack 19 hours ago | parent [-] | | The professional side of things, yes. For consumer grade GPUs, despite the trends in gaming markets otherwise needing such, the values have stagnated a bit. | | |
| ▲ | latchkey 19 hours ago | parent [-] | | I'm NDA with AMD and sadly can't mention details, but I can say the future is promising. | | |
|
|
|
| ▲ | danielmarkbruce 21 hours ago | parent | prev [-] |
| Yes, yes. Nvidia about to release blackwell ultra with 288GB. Go back to maybe 2018 and max was 16gb if memory serves. DeepSeek recently release a 670 gb model. A couple years ago Falcon's 180gb seemed huge. |
| |
| ▲ | spoaceman7777 20 hours ago | parent [-] | | I'd assume that, in the context of LLM inference, "recent" generally refers to the Ampere generation and later of GPUs, when the demand for on board memory went through the roof (as, the first truly usable LLMs were trained on A100s). We've been stuck with the same general caps on standard GPU memory since then though. Perhaps limited in part because of the generational upgrades happening in the bandwidth of the memory, rather than the capacity. | | |
| ▲ | danielmarkbruce 20 hours ago | parent [-] | | Bandwidth is going up too. "It's not doubling every 18 months and hence it's not moving" isn't a sensible way to view change. A one time effective 30% reduction in model size simply isn't going to be some massive unlocker, in theory or in practice. |
|
|
