Memory Bandwidth is the best thing ever. That's why when I had my Radeon VII, and I had over ONE TERABYTE per SECOND of Memory Bandwidth, I was a very happy Catgirlboy.
I made fun of the GTX 1660 SUPER when I first heard about what it was. That was because I was not Enlightened at that point to the Memory Bandwidth.
GTX 1660 SUPER is a GTX 1660 Non-Ti with fast, 14 Gbps GDDR6 Memory. That means it has more Memory Bandwidth. That is exactly what it is. That means it is simply, a fantastic thing. How much Memory Bandwidth, do you ask?
Well.
This is how much Memory Bandwidth it has. Actually, this is the theoretical peak Memory Bandwidth and this is worked out by doing a calculation on the memory sub-system of this TU116 graphics processor.
(192-bit Interface Width / 8 (8 bits = 1 byte)) * 14 Gbps Memory Data Rate
192 / 8 = 24 * 14 = 336 GB/s of Memory Bandwidth.
However, this is the peak theoretical. And although this little TU116 Graphics Processor has a pretty large amount of Memory Bandwidth for itself, it might not achieve that 100% of the time.
Memory Bandwidth. It is such a great thing to have. This is the actual, measurable Memory Bandwidth in AIDA64's GPGPU test. That said, it is only one test and different software running has different overheads. This is still a large amount of Memory Bandwidth for this little Graphics Processor.
I think its little GPU Cores are very well fed with Memory Bandwidth. And the Pixel Pipelines too. TU116 is already a highly Memory Bandwidth-efficient GPU.
That just means it has EVEN MORE MEMORY BANDWIDTH.
Comments