Not really, they're pretty much independent of each other, with the possible exception that typically picking a target bus width or target bandwidth means sticking to certain multiples of memory modules.
The latest radeon cards all have 16Gb of vRAM but their bus width is smaller than Nvidia at 256bit, vs Nvidias 320bit bus width. And it also has less memory bandwidth as the GDDR6 is slower, at 512 GB/s vs Nvidia counterpart at 760 GB/sec. The actual memory capacity itself maxes out eventually, once you have more than games need then it has no impact on frame rate. You can clearly see that in the benchmarks of modern games from both camps.
Bandwidth has four factors that I know of.
Frequency, latency, pump rate (SDR or DDR etc) and Bus width.
(memory clock in Hz × bus width ÷ 8) × memory clock type multiplier = Bandwidth
https://www.gamersnexus.net/dictionary/5-memory-bandwidth-gpu
GDDR6 has a burst length of 16 bytes (BL16), meaning that each of its two 16-bit channels can deliver 32 bytes per operation. GDDR6X has a burst length of 8 bytes (BL8), but because of PAM4 signaling, each of its 16-bit channels will also deliver 32 bytes per operation. To that end, GDDR6X is not faster than GDDR6 at the same clock. -
https://www.tomshardware.com/uk/new...ls-the-future-of-memory-or-a-proprietary-dram
The way a memory controller works on a GPU, the less RAM chips mean a lower bus bit width. Each chip is two 16 bits channels or 32 bits (x 10 = 320 bits rtx 3080). If the 3080 had 12 chips it would have 384 bits bus width like the 3090. This is why more RAM chips are better. You can pick your bus width by removing chips from the PCB or increase it by adding more. Up to a limit anyway. Right now, Micron's GDDR6X memory chips are only available in 8Gb densities, translating to 1GB of capacity. The 3090 has to use two to get the increased capacity for 24GB but this wont increase the bus width.
With maximum speeds of 21Gbps, GDDR6X memory delivers a bandwidth increase of over 30% when compared to 16Gbps GDDR6 memory modules. This lead extends to 50% after you consider that most GDDR6-based consumer GPUs only use 14 Gbps GDDR6 memory modules.
The 6900xt most likely has 2GB x 8 chips
https://www.extremetech.com/computing/258901-samsung-introduces-new-16gbps-gddr6-2gb-capacities. 8x2 is 256 bit. Memory controller matters as well and sets the maximum bus width.
So working out the 3080's bandwidth. The frequency is 9500 the bus width is 320 bit and the ram is DDR.
Note that this is 1188MHz RAM. 19Gbps effective. 19Gbps /2 = 9500
9500 x 320 /8 *2 = 760,000 MB/s or 760GB/s
https://www.techpowerup.com/gpu-specs/geforce-rtx-3080.c3621
Now for the 3090. 1219MHz RAM. 19.5Gbps effective. 19Gbps /2 = 9750
9750 x 384 /8*2 = 936,000 MB/s or 960GB/s
https://www.techpowerup.com/gpu-specs/geforce-rtx-3090.c3622
For the Radeon 6900xt RAM frequency 2000MHz 16Gbps effective. 16GB/2=8000
8000x 256/8*2 = 512,000MB/s or 512GB/s
https://www.techpowerup.com/gpu-specs/radeon-rx-6900-xt.c3481
Say for interest sake if we lowered the bus bit width to 256 for GDDR6x.
For the 3090. 19.5Gbps effective. 19Gbps /2 = 9750
9750 x 256 /8*2 = 624,000 or 624GB/s
For the 3080. 19Gbps effective. 19Gbps /2 = 9500
9500 x 256 /8 *2 = 608,000 or 608GB/s
So the GDDR6x RAM is faster even at the same bus bit width. Thus, higher effective I/O Data Rate is a factor.
GDDR6 RAM on the 6900xt has a higher clock speed than the GDDR6x RAM on both the 3080 and 3090. Its faster clock wise. The difference is both bus width and a higher effective I/O Data Rate. Even at the same bus width GDDR6x has a higher effective I/O Data Rate. Thus faster.
Sources
https://www.overclock3d.net/news/so...ls_its_gddr6x_memory_and_their_future_plans/1
GDDR6x on the 3080 and 3090 is running with a slower effective I/O Data Rate. GDDR6x was intended to run at 21GB/s effective I/O Data Rate. If this had of happened.
For the 3090. 21Gbps effective. 21Gbps /2 = 10500
10500 x 384 /8*2 = 1,008,000 or 1008GB/s
For the 3080. 21Gbps effective. 21Gbps /2 = 10500
10500 x 320 /8*2 = 840,000 or 840GB/s
In 2021, Micron plans to offer 16Gb (2GB) GDDR6X memory chips, which will allow for the creation of Ampere graphics cards with larger memory densities. So a 20GB RTX 3080 is unlikely in 2020 but not impossible.