根本就是「TITAN」嘛!!ASUSROGSTRIXGeForceRTX3090绘图卡( 二 )


82 (1st Gen)
72 (1st Gen)
GPU Boost Clock (MHz)
1710
1695
1770
Peak FP32 TFLOPS (non-Tensor)1
29.8
35.6
16.3
Peak FP16 TFLOPS (non-Tensor)1
29.8
35.6
32.6
Peak BF16 TFLOPS (non-Tensor)1
29.8
35.6
NA
Peak INT32 TOPS (non-Tensor)1,3
14.9
16.3
17.8
Peak FP16 Tensor TFLOPS
with FP16 Accumulate1
119/238
142/284
130.5
Peak FP16 Tensor TFLOPS with FP32 Accumulate1
59.5/119
71/142
65.2
Peak BF16 Tensor TFLOPS
with FP32 Accumulate1
59.5/119
71/142
NA
Peak TF32 Tensor TFLOPS1
29.8/59.5
35.6/71
NA
Peak INT8 Tensor TOPS1
238/476
284/568
261
Peak INT4 Tensor TOPS1
476/952
568/1136
522
Frame Buffer Memory Size and Type
10240MB GDDR6X
24576MB GDDR6X
24576MB GDDR6
Memory Interface
320-bit
384-bit
384-bit
Memory Clock (Data Rate)
19 Gbps
19.5 Gbps
14 Gbps
Memory Bandwidth
760 GB/sec
936 GB/sec
672 GB/sec
ROPs
96
112
96
Pixel Fill-rate (Gigapixels/sec)
164.2
193
169.9
Texture Units
272
328
288
Texel Fill-rate (Gigatexels/sec)
465
566
509.8
L1 Data Cache/SharedMemory
8704 KB
6912 KB
10496 KB
Next: GA102 : 经改良的 Ampere SM 设计 更大、更具弹性的 L1 CachePage1: 10,496 CUDA Cores、350W TDP 定价 US$1,499、取代 TITAN RTXPage2: GA102 : 经改良的 Ampere SM 设计 更大、更具弹性的 L1 CachePage3: GA102︰第 2 代 RT Cores、第 3 代 Tensor Cores RT、DLSS 性能再提升Page4: GA102︰GDDR6X、更大 Register Files首次加入 RTX I/O 传输技术Page5: 升级 HDMI 2.1、支援 8K @ 60Hz HDR 第 5 代 NVDEC、新增 AV1 硬体解码Page6: RTX 3090 卡皇降生 ASUS ROG STRIX GeForce RTX 3090 绘图卡Page7: 效能测试︰GeForce RTX 3090 vs RTX 3080 / RTX 2080 Ti

推荐阅读