09-17-2020, 11:26 AM
The number of CUDA cores is a little bit more complicated. NVIDIA did a little trick by Ampere. Using Turin you have 1xINT32 and 1xFP32, RTX 2080 TI has in total 4352 INT32 cores and 4352 FP32 cores. Using Ampere it is either 1xINT32 and 1xFP32 or 2xFP32; they can use the INT32 unit as FP32. Thus 3080 can have up to 2x4352=8704 FP32 core but 3080 can only have up to 4352 INT32 cores. The so called CUDA core number for 3080 is under optimal circumstances for FP32.
The base clock of 3080 is only a little bit higher than 2080 TI (1440 MHz vs. 1350 MHz).
Based on AIDA64 pictures I saw 3080 FE is only a little bit faster in SHA1 than 2080 TI FE and a little bit slower than 2080 TI OC. But I do not know how optimized AIDA64 is and if they use OpenCL or CUDA (text mentioned both).
I could not find the article I read, I could only found this: https://babeltechreviews.com/wp-content/...iv3080.jpg
The base clock of 3080 is only a little bit higher than 2080 TI (1440 MHz vs. 1350 MHz).
Based on AIDA64 pictures I saw 3080 FE is only a little bit faster in SHA1 than 2080 TI FE and a little bit slower than 2080 TI OC. But I do not know how optimized AIDA64 is and if they use OpenCL or CUDA (text mentioned both).
I could not find the article I read, I could only found this: https://babeltechreviews.com/wp-content/...iv3080.jpg