ShogunDarius
Member
I think i'll go 1080 factory OC'd from my 670s I want a huge boost and that would give the most and I have no intention to SLI again, its more hassle than its worth
Should I expect this card to manage 1440@60fps for most games? I'm sick of playing at sub-native, and I need my monitor for work.
970 to 1070 baby
That's a good decision. (Not doing SLI that is)I think i'll go 1080 factory OC'd from my 670s I want a huge boost and that would give the most and I have no intention to SLI again, its more hassle than its worth
970 to 1070 baby
I'm sure there will be ITX sized cards sooner or later.This would be perfect upgrade for my X51 but doubt it's gonna fit
is this true? maybe i shouldn't regret my 970 purchase a couple months ago after all
Why would they? I mean, cards released a year from now will naturally be faster than cards released right now, and chips with a larger die area will naturally be faster than chips with a smaller one.HBM2 cards are going to kick the ever-loving shit out of anything released this summer--regardless of brand.
Why would they? I mean, cards released a year from now will naturally be faster than cards released right now, and chips with a larger die area will naturally be faster than chips with a smaller one.
But "kicking the ever-loving shit out of anything" sounds more extreme to me, and I just can't fathom why that would happen. If you have 4 times the memory bandwidth that is great and all, but if in practice your GPU is only bandwidth limited by 10-30% in the vast majority of gaming use cases then this will only manifest in a 10-30% performance increase. E.g. a Fury has almost 60% higher bandwidth than a 980ti, and yet it's still significantly slower in most games.
I feel like many people are forgetting in their euphoria over HBM that memory bandwidth is just a means to an end, and not an end in and of itself.
. E.g. a Fury has almost 60% higher bandwidth than a 980ti, and yet it's still significantly slower in most games.
This is ridiculous. HBM is just a memory standard which provides more bandwidth, it's not anymore "real" than GDDR5X or DDR4 or HMC, and if some other memory standard (like GDDR5X for example) is able to provide the same effective bandwidth with less costs then it's very easy to spot a better option - unless your brain was eaten by AMD's hype for HBM a year ago.HBM2 cards are going to kick the ever-loving shit out of anything released this summer--regardless of brand.
If you're building a new PC right now and need a card, of course the 1070 and 1080 (and perhaps Polaris) are the way to go. There should be no buyer's remorse.
That said, for folks that own 970s/980s or 290x/390x and higher, why? Knowing that the real standard (HBM2) is not that far off, I just don't see how dropping serious cash right now is a great idea. Of course, if money is no option then who cares.
GDDR5x is a stop-gap solution, period. Nothing more.
Edit: remember the first HDTVs that were only 1080i and featured a 4:3 format and were still CRT? That's what GDDR5x is.
Even though it's all we really have to go by, I'm not sure how comparing two very different GPUs is a great example of GDDR5 vs HBM. The real test, of course, is a 1080 with GDDR5x against one with HBM2. Apples to apples.
This would be perfect upgrade for my X51 but doubt it's gonna fit
This is ridiculous. HBM is just a memory standard which provides more bandwidth, it's not anymore "real" than GDDR5X or DDR4 or HMC, and if some other memory standard (like GDDR5X for example) is able to provide the same effective bandwidth with less costs then it's very easy to spot a better option - unless your brain was eaten by AMD's hype for HBM a year ago.
There will never be a 1080 (or any chip for that matter) with support for both HBM and GDDR5.
Well, sure, it might, when the games will actually require that much bandwidth which isn't the case right now and thus at this very moment HBM is just money down the drain. The whole hype around it has everything to do with AMD as there are no physical evidence which would prove that HBM is any better than GDDR5 let alone GDDR5X for gaming uses currently.HBM2 will become the premier industry standard. The only reason GDDR5x is being utilized is because it's available and it managed to be better than even Micron expected. This has nothing to do with AMD because Nvidia will make the switch too.
Trying to illustrate how an apples to apples comparison is the real way to gauge performance differences between the two standards. Hence why I put apples to apples?
That's a good decision. (Not doing SLI that is)
What about the room HBM leaves for larger chips? Will that make significant difference eventually?Well, sure, it might, when the games will actually require that much bandwidth which isn't the case right now and thus at this very moment HBM is just money down the drain. The whole hype around it has everything to do with AMD as there are no physical evidence which would prove that HBM is any better than GDDR5 let alone GDDR5X for gaming uses currently.
You can put whatever you want but the simple fact is that Fury with HBM _and_ with DCC improvements of GCN3 isn't that much faster than Hawaii with GDDR5 (not even 5X) and a previous version of AMD's DCC. This should already tell you a lot.
Modern games are optimized for 1080p console b/w figures and while this will last gobs of b/w provided by HBM are just not necessary and GDDR5X will do fine for the time being. If you expect some big jumps from future HBM GPUs over their GDDR5X counterparts then you're likely to be disappointed. It's just as much of an incremental increase as GDDR5X is.
What about the room HBM leaves for larger chips? Will that make significant difference eventually?
R9 290X to the 1070. Worth it??