| Year |
2022 |
2020 |
2023 |
2022 |
2022 |
2024 |
| Manufacturing |
7nm |
7nm |
7+nm |
千问 Qwen 教程
4nm |
4nm |
4nm |
| Architecture |
Ampere |
Ampere |
HUAWEI Da Vinci |
Hopper |
Hopper |
Hopper |
| Max Power |
300/400 W |
300/400 W |
400 W |
|
350/700 W |
700W |
| GPU Mem |
80G HBM2e |
80G HBM2e |
64G HBM2e |
80G HBM3 |
80G HBM3 |
141GB HBM3e |
| GPU Mem BW |
|
1935/2039 GB/s |
|
|
2/3.35 TB/s |
4.8 TB/s |
| GPU Interconnect (one-to-one max bw) |
NVLINK 400GB/s |
PCIe Gen4 64GB/s, NVLINK 600GB/s |
HCCS |
NVLINK 400GB/s |
PCIe Gen5 128GB/s, NVLINK |
PCIe Gen5 128GB/s, NVLINK 900 GB/s |
| GPU Interconnect (one-to-many total bw) |
NVLINK 400GB/s |
PCIe Gen4 64GB/s, NVLINK 600GB/s |
HCCS |
NVLINK 400GB/s |
PCIe Gen5 128GB/s, NVLINK |
PCIe Gen5 128GB/s, NVLINK 900 GB/s |
| FP32 |
|
|
|
|
`51 |
67*` |
| TF32 |
|
`156 |
312*` |
|
|
`756 |
| BF16 |
|
`156 |
312*` |
|
|
`1513 |
| FP16 |
|
`312 |
624*` |
|
|
`1513 |
| FP8 |
NOT support |
NOT support |
|
|
`3026 |
3958*` |
| INT8 |
|
`624 |
1248*` |
|
|
`3026 |
版权声明:本文内容由互联网用户自发贡献,该文观点仅代表作者本人。本站仅提供信息存储空间服务,不拥有所有权,不承担相关法律责任。如发现本站有涉嫌侵权/违法违规的内容, 请联系我们举报,一经查实,本站将立刻删除。
发布者:Ai探索者,转载请注明出处:https://javaforall.net/261314.html原文链接:https://javaforall.net