Assuming you've read my posts, this is deeply dishonest. I've posted a lot in both the 8Gb and 10Gb threads, I've taken all debate seriously. I've made extremely high amount of effort to understand the nuance of memory usage in games and the different methods of measuring vRAM and their drawbacks such as:
1) The settings menu (an estimate, can be very wrong/misleading)
2) Memory Allocated (an upper bound of what is request, NOT what is actually used)
<-- Most tools report this
3) The new per process memory allocation in the MSI Afterburner beta (which hooks the Direct3DKMT API and reports memory in use for the current process)
4) Special K modding tool which reports DXGI_Budget (The memory budget response from DirectX, used to check if you're over memory budget or not)
Yes I've made long posts that go into this detail but it's not just "waffle", it's actually a nuanced topic which people are attempting to over simplify into a statement like "8Gb is not enough". My actual arguments are
evidence based, I've made an effort to look at every example people have given me of games using large amounts of vRAM to actually inspect if that claim is true or not. Then linked to evidence of either my own testing results OR where other users have tested and provided evidence. Here are some examples:
1) FS2020 was originally claimed to use 12.5Gb peak, it doesn't it uses about 9.5Gb peak when measured properly, confirmed with in-game dev tools to display real memory usage. And is unplayable in these settings on a 3090
2) Read Dead Redemption 2 was claimed to need 11Gb for Ultra 4k, it doesn't. Plenty of users showed when measured properly this takes about 6.5Gb.
3) There's an entire thread on this at the resetera forums here
https://www.resetera.com/threads/msi-afterburner-can-now-display-per-process-vram.291986/page-2 that show when measured properly:
3a) Gears Tactics Max settings 3440x1440 is 4,767MB of vRAM
3b) Control Max settings 3440x1440 is 4,164MB vRAM
3c) Age of Empires 3 1,680MB of vRAM
3d) Death stranding 3,764MB of vRAM
3e) Metro Exodus Max settings with RTX 3440x1440 5.1Gb of vRAM
3f) There's loads of example in that thread with evidence, just go and read and look at all the screenshots of memory usage
4) Then someone claimed Wolf 2 using "16k textures" was using like 20Gb of vRAM so I downloaded that, tested it. Turns out they were just turning up the engines texture pool size to 16Gb with zero benefit, real memory usage was like 5Gb, I posted screenshots of that for evidence here on the forum.
5) People made similar claims about Doom Eternal I downloaded that and tested and posted evidence that at max it was using 7Gb but that actually just like Wolf 2 (same engine) all it was doing was reserving a massive texture pool that it never used. In "Ultra Nightmare" texture pool it's 4.5Gb reserved just for the pool but at "High" it's only 1.5Gb and that this brought no visual benefit. Real actual memory usage when not reserving a large unused pool tok it from 7Gb to 4.5Gb
6) And now we have more claims about Resident Evil 3, which I had previously pointed out others had given evidence for severely misleading menu estimates. So I've just downloaded and test again.
Here...completely maxed settings including highest 8Gb textures, in 4k, the menu warns it will use 12.5Gb of vRAM (LOL, the game is only a 22Gb install), and here are random screenshots as I played through the start of the game, displaying again
vRAM Allocated as what Afterburner would typically report, and
vRAM Used which is the newer per process memory in use, as reported by the D3D API
Warning these images are very large 7-19Mb uncompressed. Sorry I had to post a bunch but I'm facing accusations of cherry picking now and I'm just not going to stand for that kind of bad faith argument.
For those of you who CBA to download/expand/zoom to see OSD values they are:
8018 Reported 5521 actually used
7099 Reported 5428 actually used
7701 Reported 6071 actually used
7409 Reported 5865 actually used
8018 Reported 5521 actually used
5850 Reported 4367 actually used
That's an average of memory actually used of about 5.4Gb, the idea this is bumping up against 8Gb limits is WRONG, it's demonstrably wrong. And your reliance on reviewers is what is leading you to a bad conclusion. Consider the fact that this new way of measuring vRAM was only incorporated into a beta of a public tool on the 18th Sept this year, prior to that there was no real easy way to do it, certainly no industry standard tool, all the tools except 1 extremely obscure modding tool all reported memory allocated. So any reviewer you rely on who hasn't explicitly switched to this new method is not telling you how much vRAM games actually use. They're telling you what the game engine requests to be allocated, it's not the same thing. All the research on this shows that the difference between these 2 values can often be extremely big sometimes games will just allocate 90% of the vRAM you have available in which case they appear to be using like 22Gb on a 3090 but they're really using like...3Gb or something small.