>tfw fell for the 1060 3gb meme WHAT THE FUCK NVIDIA BROS

>>58844307
I have a feeling even 8GB is going to feel the pinch in the coming years.

File: 012-P3-1575-ER_XL_4.jpg (474KB, 1200x1200px) Image search: [Google]

474KB, 1200x1200px

>>58844307
>tfw based 1.25GB
There's a reason why they make game consoles, you know.

>>58844307
Really makes you ponder, huh, OP?

https://www.youtube.com/watch?v=Ay_NWfaRiPc&t=8s

My 8GB 480 uses over 6GB of VRAM in several games, and I am running at 1600*900.

How are 970 owners right now?

>>58844860
Crying

Bought the 6gb 1060 and it isn't much better. The performance isn't worth the extra 50 dollars. You're better off just keeping the 3gb and upgrading in the next few generations

>>58844307
yeah, must blow when you get memed into thinking that you'll never need a certain amount of vram.

>>58844760
It's all about that space/time trade-off
I'd rather have a game pre render a lot of things then load them into RAM rather then having to do that shit on the fly and lose frames because of it

>>58844307
What's wrong? My 780ti is just fine at 1440p.

>>58844307
It's not just the VRAM, the 3GB version also has fewer CUDA cores. You got double cucked.

/g/ never said that was enough, forgot the 970 3.5gb meme?

>>58844552
Now I kinda want to see what happens when they're clocked the same

>>58845351
The 480 wins by a longshot. Nvidia has a clock advantage that AMD can't touch (for right now) but AMD has performance per core that Nvidia doesn't even hold a candle to.

>>58844307
ayy i just sold a shitty 980 for $300 to some tard. i can live with my even shittier 670 until Vega drops.

>>58844427
I think the point of Vega's High Bandwidth Cache Controller is to negate these concerns.

If they can treat the HBM pool as a true cache (i.e., using CAMs and state bits to make an n-way writeback cache kept coherent with host memory), the pool only needs to be a little bit bigger than the working set for a frame, which is in turn limited by the bandwidth and max frame rate. (e.g., 512 GB/s for 60 Hz = 8.5 GB max, 120 Hz = 4.3 GB, ... The real limits would be substantially lower thanks to imperfect memory bus saturation, part of the bandwidth being used for writes and not just reading assets, etc.)

Of course this would require either fancy scheduling to defer shader work on stalled reads, or just punting on the rendering for the current frame a la MegaTexturing.

>>58845281
Stop larping, dumb cuck

>>58844427
Which is why RX 490 will have 16 GB. You're set for (the card's) life.

>>58847426
Where did that come from? Last I heard that Vega is 8gb.

File: ??????????????.jpg (784KB, 4288x2848px) Image search: [Google]

784KB, 4288x2848px

>>58844307
I wonder who is behind this post.