• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

AMD Fiji HBM limited to 4GB stacked memory

This has been known from the get go. HBM2 will be needed for larger than 4GB packages, and they'll most likely move to a single chip (no interposer) design when they move to GF's 14nm FINFET process next year. Though if it's cheap enough, I guess an interposer arrangment might be retained.

There's also no way that Fiji will be H2 '15. It's already in production. Bermuda will be H2.

Volta has been cancelled, at least for consumer products ... supposedly it still exists for enterprise. It was replaced with Pascal. Volta was / is an HMC product, not an HBM product. For whatever reason (timing and cost likely primary motivators), NVIDIA abandoned it and went for the AMD / SK developed HBM. Pascal is very, very unlikely to appear before H2 '16. Should give AMD an entire year with what should amount to an impregnible performance lead in any game that values memory bandwidth, regardless of what chips NVIDIA has.
 
Dunno I've never really been VRAM bandwidth limited when playing with 4K - clocking the core gave the most significant gains and it was easy to hit VRAM amount limits - 4GB will be less than ideal at 4K.
 
Really naive question here. If your VRAM has twice the bandwidth is it functionally similar to having twice the VRAM? Or is the amount of VRAM still going to be a limiting factor no matter how high the bandwidth is?
 
I've not used 4K myself so excuse me if this is wrong but I was under the impression that memory size as well as bandwidth were both rather important at these resolutions?

When games use up 4GB vram at 1440 and in some extreme cases 1080 I would have thought that 6GB and 8GB would be 'needed' for 4k?
 
Really naive question here. If your VRAM has twice the bandwidth is it functionally similar to having twice the VRAM? Or is the amount of VRAM still going to be a limiting factor no matter how high the bandwidth is?

Not as I understand it, it will allow the card to process information quicker. So while it will still suffer from hitting the vram limit it wont be effected as bad, it can just can swap data in and out of it memory at a faster rate.



*fully admit I may be wrong on this.
 
Really naive question here. If your VRAM has twice the bandwidth is it functionally similar to having twice the VRAM? Or is the amount of VRAM still going to be a limiting factor no matter how high the bandwidth is?

When you hit the VRAM amount wall you still have to swap memory out to slower storage (System RAM or if you really load it up paged memory on disc) which the speed of won't have changed so you will still see the same performance penalty once you hit the amount limit.
 
Really naive question here. If your VRAM has twice the bandwidth is it functionally similar to having twice the VRAM? Or is the amount of VRAM still going to be a limiting factor no matter how high the bandwidth is?
Think of the VRAM like a teapot. Bandwidth is the speed you can pour the tea out, capacity (4GB) is the amount of tea it holds. If you make the spout bigger (more bandwidth) you aren't changing the amount of tea it can hold, just the speed it comes out at.
 
In reality then these first gen HBM cards are still going to be 4K limited in single gpu use, while they will offer great performance below 4k and will blow pretty much all other gpu's away.

Once they make the 8gb variants then you should see decent 4k performance from 1 gpu?
 
Well either AMD don't care about the new cards being 4K capable or they've found a way to make 4GB VRAM perfectly fine for 4K.

I'm going with the latter as this generation of upcoming cards are when more and more people will start going 4K.

Once they make the 8gb variants then you should see decent 4k performance from 1 gpu?

The article seems to suggest the Fiji chip can only manage 4GB of HBM.
 
In reality then these first gen HBM cards are still going to be 4K limited in single gpu use, while they will offer great performance below 4k and will blow pretty much all other gpu's away.

Once they make the 8gb variants then you should see decent 4k performance from 1 gpu?

Depends on the speed of their core too. It doesn't matter what much or little vram the card has if the core hasn't got the grunt to make use of it.
 
Well either AMD don't care about the new cards being 4K capable or they've found a way to make 4GB VRAM perfectly fine for 4K.

It may just be that this is all they can do with the technology on offer at the moment, with 8gb models down the line as the technology is developed. All they need to do is establish a performance lead over nvidia, though it will be disappointing if there are no 8gb models out for a while.
 
Back
Top Bottom