• 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.

Do you think we will get better frametimes and smoother gaming with HBM ?

Kaapstad

Kaapstad

Kaapstad

Man of Honour
Joined
21 May 2012
Posts
31,953
Location
Dalek flagship
A lot has been said about the extra bandwidth that comes with HBM, my personal view is it will have very little benefit as once you have enough bandwidth you don't need anymore.

What gets discussed a lot less is the improved latency available with HBM, this is something that I think will bring real benefits resulting in far smoother gameplay and much better frametimes.

I think when we see some frametime graphs from the new Fiji cards they are going to be very impressive indeed.

What do you guys think ?
 
Yes this was one of my first conceptions when I read about Hbm the lower latencey compared to gddr5 so I assumed lower frame times may be a benefit but dont want to get my hopes up too much but if it is a result of Hbm then AMD cards will reap the rewards of lower latency and better frame times. Also I get what your saying about bandwith but you need to remember devs only have certain specs to play with meaning certain bandwith consumptions when it comes to games. Now they have more to play with this will enable them to do more. However will AMD users see this is a different story as only a minute amount of ppl will have this bandwith available. We will have to wait till next gen along with pascal to reap those rewards.
 
No ****** clue. Higher bandwidth memory should allow textures to load faster perhaps, but that's all I can think of as someone who isn't a graphics card expert.

These Fiji cards better be good. I think at the very least the HBM will have some sort of improvement over GDDR5, but no clue how that will translate into performance for the end user. Looking at the whole 970 3.5GB fiasco, usage of the more terrible 0.5GB memory causes a framerates drop. So there's a difference even now. Perhaps there would be an increase in raw performance from just using HBM alone.
 
Also bandwith needs are relative..for instance the 290X matches or overcomes the 980 at higher resolutions..

It will be interesting to see how many ROPs AMD put in the Fiji..if its indeed 128 as the rumours said, then it needs the bandwith to feed them (and if it really has 128ROPs then the card will be an absolute monster)..but this is not confirmed..it could be 64 or 96 ROPs as well.

Edit: they might not go for 128ROPs as it has the enchantments Tonga has, and the little 32ROP R9 285 blows the 64ROP 290X out of thewater in Pixel fillrate.
 
Last edited:
Problem is in gaming you top off at a certain point. Such as the jump to ddr2 To ddr3 in ram saw a good improvment but 1600 -1866Mhz is the sweet spot as anything faster is little improvement to allmost nothing noticable. I think thats what kaap is getting at at a certain point will most of that bandwith get saturated but not all of it meaning what benefit will there be with all that bandwith.
 
To be honest I'm not sure there is anybody here with the expertise to give a thorough answer to the question.

I would think that like all hardware it's a balancing act and removing one bottleneck (in this case bandwidth) can allow for the rebalancing of other areas, say adding more shader units or removing cache now that there is more bandwidth and less latency. But I'm a programmer not a chip engineer.

One would certainly expect to see improvements with heavily memory intesive operations like AA and other post-processing effects though.
 
I have no idea but if it is good I will get one and if it isn't then I won't :)
 
Kaapstad said:
A lot has been said about the extra bandwidth that comes with HBM, my personal view is it will have very little benefit as once you have enough bandwidth you don't need anymore.

What gets discussed a lot less is the improved latency available with HBM, this is something that I think will bring real benefits resulting in far smoother gameplay and much better frametimes.

I think when we see some frametime graphs from the new Fiji cards they are going to be very impressive indeed.

What do you guys think ?
Click to expand...

HBM has other benefits besides more bandwidth. By requiring less power it means you can use the power savings on having a faster GPU.
 
Whilst I don't have a clue to the question. What I have been thinking is. Are game engines and dev's able to take advantage of the increase in processing power and will it actually be visible to the user???

Or have AMD been holding out for Dx12 all along as they know this will be the game changer for this new tech?
 
Strudders said:
Whilst I don't have a clue to the question. What I have been thinking is. Are game engines and dev's able to take advantage of the increase in processing power and will it actually be visible to the user???

Or have AMD been holding out for Dx12 all along as they know this will be the game changer for this new tech?
Click to expand...

Surely it would be better to get it out earlier so the people who are working on dx12 now could use it?
 
LambChop said:
If HBM works the way I think it does, there will almost guaranteed be a bottleneck somewhere
Click to expand...

I agree with this, at some point the bottleneck will move and it will cause another completely different problem, but I do think HBM will help a lot and as the fury will be built with HBM in mind.
 
I got a feeling that HBM probably won't do much for 1920 res, but will increase the performance of 4K gaming by fair bit. However the concern is still the question of "Will 4GB of HBM be up to scratch, as oppose to 8GB GDDR5 for 4K gaming?"

We know that when we run out of memory on GDDR5 in games, we could potentially get stuttering and frame rate dropping to 1-2fps hell...but would the same still apply for HBM? And will HBM efficient in the sense of being able to getting the data in and out in a efficient manner, that it might not saturate as much vram space as GDDR5 do?

We should know this for sure once the official reviews are available.
 
Last edited:
D1craig said:
Surely it would be better to get it out earlier so the people who are working on dx12 now could use it?
Click to expand...

They were given Mantle and failed to see the benefits or commercially chose not to. So the API's have been out there. With DX12 being announced, mantel was dead in the water and the number of games it could open up a lead on we're not there. Delay 18 months, get on the band wagon a couple of months before win 10 is released and you can then demonstrate the GPU's inherent benefits. Anything else before that and it's just another Graphics card...
 
Latency times should improve but I doubt it will be enough for the naked eye to see. Any improvement is good though, so bring it on I say.
 
Kaapstad said:
A lot has been said about the extra bandwidth that comes with HBM, my personal view is it will have very little benefit as once you have enough bandwidth you don't need anymore.

What gets discussed a lot less is the improved latency available with HBM, this is something that I think will bring real benefits resulting in far smoother gameplay and much better frametimes.

I think when we see some frametime graphs from the new Fiji cards they are going to be very impressive indeed.

What do you guys think ?
Click to expand...

You always need more bandwidth.
DX12 will do more for framepacing as Mantle already shown that than what HBM will do but it will aid as you can move information faster with HBM.
Mantle has shown we have a new era of PC gameplay that will be shown with Fury and Dx12. The newer HBM/DX12 technology allowing better interaction between hardware and software.
 
As has been said I think the bottleneck will move to a different area of the pipeline and all of a sudden we will be looking at a whole new problem.
HBM is the future of graphics cards no doubt about that, the first few iterations might not be too good but it will get there and it has to start somewhere, so kudos to AMD for getting the ball rolling.
 
You must log in or register to reply here.
Back
Top Bottom