For customers there's no difference. What matters is the final number: multiplication of bus width and frequency, that'll give you the transfer rate. Higher memory width is more expensive to manufacture (each bit width is an actual physical lane and is the total amount of pins in the memory chips dedicated to transfer data), and higher memory clock is harder to achieve.
The frequency determines how many transfers can be done per second (2 per cycle, actually), and the bus width how many bits are transferred simultaneously.
Power requirements and heat are important limiting factors. Also, having too much transfer rate becomes useless. Just look at AMD cards with HBM2, they have much higher transfer rates but don't really perform better than the equivalent nVidia with GDDR5. In other words: it's useless transferring more data than the GPU is going to be able to process.
edit:
I'm going to add something that I'm not sure to what extend is a problem in this case: transferring data over parallel lanes creates interference due to electromagnetic induction (this is why ethernet cables are twisted, to minimize interference). So if you have many parallel lanes sending and receiving data, there are interferences affecting each other. This interference increase as you increase the frecuency of the signal, so in theory, it's harder to have higher clocks with more lanes.
Comment has been collapsed.
1,497 Comments - Last post 2 minutes ago by SeefKroy
14 Comments - Last post 3 minutes ago by PoeticKatana
19 Comments - Last post 44 minutes ago by Comiclove
79 Comments - Last post 1 hour ago by freshduke
12 Comments - Last post 2 hours ago by Carenard
15,539 Comments - Last post 3 hours ago by MarvashMagalli
22 Comments - Last post 3 hours ago by CelticBatman
92 Comments - Last post 1 minute ago by crez3088
93 Comments - Last post 9 minutes ago by fernandopa
6 Comments - Last post 13 minutes ago by ShannonA81
1,233 Comments - Last post 53 minutes ago by Pyrocious
10 Comments - Last post 1 hour ago by Mayanaise
65 Comments - Last post 1 hour ago by ThePonz
2,787 Comments - Last post 1 hour ago by Mhol1071
From what I know Nvidia create cards with high memory clock and low memory bus and AMD creates cards with low memory clock and high memory bus.
What is the difference for customers like us (gamers)?
And what is the difference overall? (something like you can write whole page and than send it or write and send 1 word every few seconds?)
And why no one creates cards with both high memory clock and high memory bus?
Comment has been collapsed.