Realistic performance difference between memory bus sizes (3080 vs 4070)

[officiallyasimulator]

hey guys. lots of chatter about the new 4070 being a cut down card and what not. but purely from a performance standpoint it seems to be a reasonable option when compared to the similarly priced (or more expensive) 3080. $100-150 difference in price in favour of the 4070 (in AUD)

my main question is there realistically a noticeable difference in performance at higher resolution thanks to the 320 bit bus on the 3080. or does he higher clocked memory on the 4070 mean its a more level playing field at high res?

current 1080p 165hz gamer here running a mighty 1070 still looking to upgrade to 1440 165 hz in the future

[Eigenvektor]

It generally makes little sense to compare raw specs like bus width between cards. If anything look at memory bus bandwidth, which is a combination of bus width x clock speed. However that's just one aspect of a cards performance. If you want to compare performance, look at performance (i.e. benchmarks). You can already see for which games it is faster and which it isn't.

 

Screen resolution is largely irrelevant in terms on memory usage, what does matter is texture resolution. The 4070 should handle games that require more VRAM a bit better than the 10 GB variant of the 3080. Having higher bandwidth should play a bigger role the more data needs to be processed each frame, but this depends a lot on how a game's engine works, so comparing benchmarks is the best option.

[starsmine]

The realistic performance difference between the 3080 and 4070 is what you see in benchmarks at those higher resolutions

What is important in 99% of situations is bandwidth and latency. Bandwidth = Bus width * Clock speed.

You drop buswidth, and raise clockspeed by the same ratio, bandwith remains static.

Not to say bus width does not mater, but its ONE factor that only means anything when taking into account the other factors.  

So long as the ram is able to feed data to the core with minimal interruptions, the bandwidth is fast enough, and going faster no longer does increases frames per second as the internal bottleneck elsewhere become more relevant to the overall average frame times. Given that the 4070 does not compute faster then the 3080, there is little reason to spend the money increasing bandwidth over the 3080 by keeping the same buswidth when faster clocked ram is cheaper way to get to the desired bandwidth 

[officiallyasimulator]

This was the impression i was under. Thanks for your better explanations guys