Um look at sites that relooked that the matter, its straight up truth. How many people run games at that kinda settings? Running game on a 1440p using high DSR which pretty much runs game at 4k rez, throw on top of that using 4x MSAA anyone that are not idiots will tell you that would kill fps on any card ram limit or not. Most people don't run games at that kinda settings cause fps is gonna make it mostly unplayable to start with. Better use most common use case then some insanely high settings that no one else uses and not say that anything about that fact.
Any Hardware reviewer will speak to that fact that settings they used to get card to do what is the issue is well beyond what a vast majority of users wouldn't run for settings.
Well, the poster wasn't actually stating a fact, just what he felt was right and his assertion was wrong about testing. He'd have to link to something demonstrating what he's saying, or show us his own testing results to call it a fact.
Regardless, I get your point about using so much RAM, resolution and quality settings that the GPU can't handle it anyway and this can often be the case, especially with low end cards that have lots of memory. However, my 780 Ti is a bit less powerful than the 970, yet GPU-Z has shown RAM usage sometimes approaching the full 3GB* and the framerate remained reasonable, so it's not unreasonable that a high end 4GB card can use 99% of its RAM and still maintain performance. Hence, stress testing a 970 to show up its flawed memory subsystem is valid.
If you think about it, there's no reason why a card couldn't have 90%+ of its memory used up and still run a game or test animation at 200fps or even 1000fps. It all depends on how it's using the data and the effective size of the redraw loop. A programmer could easily design a test to allocate 90% of the RAM and just make the card work with the last bit of its RAM buffer without any problems and it wouldn't surprise me if tests like this are going to be designed after this fiasco to check for issues like this in future models.
*To be fair, it was only a casual test when playing CoD: Ghosts**, so I don't clearly remember the resolution. I think it was 1080p with quality settings maxed or nearly maxed out (I never use ambient occlusion for example due to the significant performance hit and turn down shadow quality too). I've also played around a bit with 4K using DSR, so maybe the RAM limit was approaching 3GB in this mode.
**Yes, I'm the one person in the gaming universe who really liked this game, lol.
.
