Originally Posted by DampMonkey
Im wondering if capacity is comparable between HBM and gddr5. Considering the tech is so much different, Im thinking it will be beneficial to see how memory management is handled through benchmarks before passing judgement on "only 4gb" of memory.
Not really, aside from the GEN1 vs GEN2 of HBM.
HBM is smaller and faster and uses less power than GDDR5.
The limitations come on the number of stacks you can have, which theoretically isn't an issue, but beyond the first stack, you aren't increasing the bandwidth, but rather only the capacity.
GEN1 has a limitation of 4GB on 1 stack of 4.
GEN2 has a limitation of 64GB on 1 stack of 4.
So ultimately the question boils down to this ... Do you think that AMD will introduce a brand new memory system on their card and instead of just dealing with the initial limitations of using just 1 stack until they work out any other complications of dealing with multiple stacks and in say 3-6 months down the road introduce another card that has additional stacks (you know, like what they did with the R9 290X ... where they introduced it with 4GB than a year later come out with an 8GB version)?
Given what AMD's Chief Technology Officer said publicly, I tend to think they will release the first cards with only 4GB of HBM, then sometime later (like right before nVidia drops Pascal, which will be using HBM GEN2 and has a 64GB limitation and twice the bandwidth) come out with an 8GB version.
But beyond that, the whole issue is HBM fast enough doesn't really matter in the whole 4GB video memory debate. AMD for years has said that 3GB isn't enough, you need 4GB, and has been making a huge deal about 4K and multiple monitors, which takes up more and more memory. We have had tons of discussions and slides showing what games at what resolution takes up how much memory. So suddenly, that doesn't matter anymore? I'm sorry, I don't buy that. Having faster memory doesn't make up for having less memory.
While I agree with the AMD CTO in that programmers and developers are lazy and aren't utilizing the resources they do have to their fullest, the fact remains that the programs are already out there, and upcoming games aren't suddenly going to be stopped and totally rewritten just to appease AMD because of HBM. Larger and larger VRAM is going to be needed when you start going down the 4K path, especially with multiple 4K monitors and the inevitable 8K monitors.
I just personally think that AMD is trying to minimize a limitation of the initial technology they can't easily (or quickly) get around. Eventually they will get around it, and come up with multiple stacks ... and eventually go to GEN2 memory, which is twice as fast and has sixteen times the capacity ... but for now, I just don't see it happening out of the gate.
I guess we will all find out in a couple of weeks.Edited by 47 Knucklehead - 5/21/15 at 8:56am