Differences in performance with GPU VRAM amount?

[Demmers]

Just started using a GPU for the first time that was given to me by a friend (GTX 1650), and i'm currently researching for a new build in the second half of the year. The RTX 5060 Ti has my interest the most, but i'm curious if getting the 16GB version will give any benefits over the 8GB model?
Following on from that, i'm also considering getting a dual pcie slot motherboard that I can continue using the 1650 on as well. Do I need to bear anything in mind when it comes the pcie bandwidth, or does running two GPU's simultaneously reduce performance regardless?

[muziqaz]

No benefit in vram speeds or amount.
You need minimum pcie x4 of any generation to be able to fold without significant performance drop

[arisu]

As long as the only difference between the two is VRAM amount then it won't matter. FAH doesn't use much VRAM. Even big projects use just a few hundred MB. But be careful because Nvidia has lied in the past and sold a model with less VRAM that also had fewer CUDA cores. But if it only differs in VRAM amount then it is fine.

The card has to have access to enough PCIe lanes to run efficiently. A chipset will have a number of PCIe lanes that it can allocate to different slots. Most slots are physically x16 but might not actually be connected to all 16. Consult your motherboard documentation. It will probably say it has one slot that is a full x16 and one that operates at x4 (even though they are physically x16). Put the slower card in the slot with fewer PCIe lanes because it doesn't need as much bandwidth.

The reason for this limit is that chipsets only have so many lanes. You usually need a higher-end workstation or server processor (like an AMD Threadripper) to get two full x16 PCIe slots. Since the older card will be totally fine with x4 lanes, you can get really any motherboard and chipset that has two physical slots with one x16 and one with less.

[Demmers]

As long as the only difference between the two is VRAM amount then it won't matter……. if it only differs in VRAM amount then it is fine.

Thanks for all that, but can you explain what you mean with the above please? If I can use two different cards (both will be Nvidia), surely there’s more in difference than just the VRAM? This makes it sound like you mean I can run two 5060 Ti’s, one being 16 and the other 8 in VRAM, but if one of the other cards is say a 1650, then that “won’t work”?
I understand what you mean with everything else.

[muziqaz]

As long as the only difference between the two is VRAM amount then it won't matter……. if it only differs in VRAM amount then it is fine.

Thanks for all that, but can you explain what you mean with the above please? If I can use two different cards (both will be Nvidia), surely there’s more in difference than just the VRAM? This makes it sound like you mean I can run two 5060 Ti’s, one being 16 and the other 8 in VRAM, but if one of the other cards is say a 1650, then that “won’t work”?
I understand what you mean with everything else.

You can use whatever cards you want.
What was meant to be said is nVidia from time to time releases two different spec GPUs under same name. Like let's say (example not real, as I can't be bothered to find actual models): GTX1030 6GB and GTX1030 3GB. nVidia calls them that, and one would think, hey it is same GPUs except different amount of VRAM. Wrong, 3GB version might have shader cores cut. So when buying lower VRAM version of two cards, compare their specs before buying

[Demmers]

As long as the only difference between the two is VRAM amount then it won't matter……. if it only differs in VRAM amount then it is fine.

Thanks for all that, but can you explain what you mean with the above please? If I can use two different cards (both will be Nvidia), surely there’s more in difference than just the VRAM? This makes it sound like you mean I can run two 5060 Ti’s, one being 16 and the other 8 in VRAM, but if one of the other cards is say a 1650, then that “won’t work”?
I understand what you mean with everything else.

You can use whatever cards you want.
What was meant to be said is nVidia from time to time releases two different spec GPUs under same name. Like let's say (example not real, as I can't be bothered to find actual models): GTX1030 6GB and GTX1030 3GB. nVidia calls them that, and one would think, hey it is same GPUs except different amount of VRAM. Wrong, 3GB version might have shader cores cut. So when buying lower VRAM version of two cards, compare their specs before buying

Got it, yes I know what you mean, thanks.