3 ms·

It is theoretically possible. Each token only needs 37B parameters and if the same experts are chosen often, it would behave closer to a 37B model than a 671B m

by ryao 2y ago

It is theoretically possible. Each token only needs 37B parameters and if the same experts are chosen often, it would behave closer to a 37B model than a 671B model, since reusing experts can skip loads from system RAM. You might still be right since I have not confirmed that the selected experts change infrequently doing prompt processing / token generation, and someone could have botched the headline. However, treating Deepseek like llama 3 when reasoning about VRAM requirements is not necessarily correct.

superkuh 2y ago
MoE is pretty enabling after you've spent all the extra $$$$ to stuff your server CPU memory channels with ram so it's possible to run at all. But it's still spending a lot of money which makes this a lot less novel or interesting than "just on 1~2 Arc A770" implies. Especially for the marginal performance that even 8-12 channels of CPU memory bandwidth gets you.
- utopcell 2y ago
  Is this amount of RAM really that expensive? 6x 64GiB DDR4 DIMMs are < $1,000.
- utopcell 2y ago
  Actually, 384GiB is already <$400 [1]. [1] https://www.amazon.com/NEMIX-RAM-DDR4-2666MHz-PC4-21300-Reduced/dp/B0BJFN1JYS https://www.amazon.com/NEMIX-RAM-DDR4-2666MHz-PC4-21300-Redu...
  - superkuh 2y ago
    A slow-end DDR4 speed older generation Xeon system is unlikely to be used by Intel for this benchmark. It's far more likely they used an expensive DDR5 modern Xeon with as many memory channels as they could. Single user LLM inference is memory bandwidth bottlenecked. I just can't see Intel using old/deprecated hardware. And if someone not Intel were to build a Xeon DDR4 system it wouldn't reach the DDR5 tokens/s speeds reported here. The reason they used a Xeon is memory channels. Non-server CPUs only have 2 but modern Xeons have 8 to 12 depending on generation/type. And the Xeons with the most are the most $$$$ and it ends up cheaper to just get a GPU or dedicated accelerator.
hmottestad 2y ago
If the same expert is chosen for two consecutive tokens then it’ll act like a 37B model running on the GPU for the second token since it doesn’t need to load that expert from the main RAM again.

by ryao 2y ago

superkuh 2y ago

MoE is pretty enabling after you've spent all the extra $$$$ to stuff your server CPU memory channels with ram so it's possible to run at all. But it's still spending a lot of money which makes this a lot less novel or interesting than "just on 1~2 Arc A770" implies. Especially for the marginal performance that even 8-12 channels of CPU memory bandwidth gets you.

utopcell 2y ago
Is this amount of RAM really that expensive? 6x 64GiB DDR4 DIMMs are < $1,000.
utopcell 2y ago
Actually, 384GiB is already <$400 [1]. [1] https://www.amazon.com/NEMIX-RAM-DDR4-2666MHz-PC4-21300-Reduced/dp/B0BJFN1JYS https://www.amazon.com/NEMIX-RAM-DDR4-2666MHz-PC4-21300-Redu...
- superkuh 2y ago
  A slow-end DDR4 speed older generation Xeon system is unlikely to be used by Intel for this benchmark. It's far more likely they used an expensive DDR5 modern Xeon with as many memory channels as they could. Single user LLM inference is memory bandwidth bottlenecked. I just can't see Intel using old/deprecated hardware. And if someone not Intel were to build a Xeon DDR4 system it wouldn't reach the DDR5 tokens/s speeds reported here. The reason they used a Xeon is memory channels. Non-server CPUs only have 2 but modern Xeons have 8 to 12 depending on generation/type. And the Xeons with the most are the most $$$$ and it ends up cheaper to just get a GPU or dedicated accelerator.

hmottestad 2y ago

If the same expert is chosen for two consecutive tokens then it’ll act like a 37B model running on the GPU for the second token since it doesn’t need to load that expert from the main RAM again.