• 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 VEGA confirmed for 2017 H1

Status
Not open for further replies.
Is anyone else looking at the Vega spec and thinking that it looks very similar to a Fury X? Is this really just a die-shrink and overclock of the Fury X? (I actually want to be proven totally wrong here, just a lot of the specs seem the same).
Itis not too far from a die-shrunk and clocked FuryX. There was some small improvements in Polaris and there should be some more modest improvements with Vega and some additional functionality such as FP16 support. Vega is an evolution of GCN and many of the GCN architectural decision still exist.

That isn;t a bad thing. AMD have abad track record when it comes to ground up decisions. The last thing we need is another Bulldozer or 2900XT!
 
Itis not too far from a die-shrunk and clocked FuryX. There was some small improvements in Polaris and there should be some more modest improvements with Vega and some additional functionality such as FP16 support. Vega is an evolution of GCN and many of the GCN architectural decision still exist.

That isn;t a bad thing. AMD have abad track record when it comes to ground up decisions. The last thing we need is another Bulldozer or 2900XT!
Yea it seems that way. Vega is still GCN under the hood, now ive been putting more research into it when i thought it was now NCU lol. But like you said an evolution of GCN with improvements being the evolution. Seem's they have tried to optimise the architecture to get it closer to it's theoretical performance. And at the same time the die shrink seems to allow for higher clock speeds. But looking at the specs seems they have stuck with a similar configuration to the Furyx. I do think AMD will get slammed about only having 8Gb of Vram though because that seems like what its going to be. Though TBF ive not got close to 8GB yet in any games on my 1080ti so we will see.
 
Yea it seems that way. Vega is still GCN under the hood, now ive been putting more research into it when i thought it was now NCU lol. But like you said an evolution of GCN with improvements being the evolution. Seem's they have tried to optimise the architecture to get it closer to it's theoretical performance. And at the same time the die shrink seems to allow for higher clock speeds. But looking at the specs seems they have stuck with a similar configuration to the Furyx. I do think AMD will get slammed about only having 8Gb of Vram though because that seems like what its going to be. Though TBF ive not got close to 8GB yet in any games on my 1080ti so we will see.

Both you and D.P are right and wrong at the same time, yes the shaders are GCN, yet that is irrelevant, what is relevant is what's around those shaders feeding them and extracting from them, in that just about everything is different.

Its easy to say its the same GPU because its based on GCN, yet in reality thats like saying the engine in the Noble M600 is the same as the engine in the Volvo XC90 because its the same block, of course that is completely not the same engine, because everything that is attached to that block is different in the two cars they are anything but the same, one produces 200 BHP the other 650 BHP

The Engine in Vega is the same as it is in the Fury-X in the same way the engine in the Noble M600 is the same as the engine in the Volvo XC90.
 
Both you and D.P are right and wrong at the same time, yes the shaders are GCN, yet that is irrelevant, what is relevant is what's around those shaders feeding them and extracting from them, in that just about everything is different.

Its easy to say its the same GPU because its based on GCN, yet in reality thats like saying the engine in the Noble M600 is the same as the engine in the Volvo XC90 because its the same block, of course that is completely not the same engine, because everything that is attached to that block is different in the two cars they are anything but the same, one produces 200 BHP the other 650 BHP

The Engine in Vega is the same as it is in the Fury-X in the same way the engine in the Noble M600 is the same as the engine in the Volvo XC90.
my point was they kept the same configuration as in same amount of shaders, texture mapping units, Rops, same amount of NCU as was previously GCN compute units.
Yea fair enough they changed a lot such as the compute units etc but they still kept the furx configuration for top end vega. It just seems like GCN which has been heavily optimised but using the Furyx GPU to apply it to.
 
my point was they kept the same configuration as in same amount of shaders, texture mapping units, Rops, same amount of NCU as was previously GCN compute units.
Yea fair enough they changed a lot such as the compute units etc but they still kept the furx configuration for top end vega. It just seems like GCN which has been heavily optimised but using the Furyx GPU to apply it to.
Fair enough. :)

While this may not then be relevant to anything you say i still make the point 'all be it on my own then' that Vega is nothing like the Fury-X, the only thing it has in common are the number of shades it has.

It according to D.P
is not too far from a die-shrunk and clocked FuryX
could not be further from any truth. they are completely and totally different GPU's
 
Fair enough. :)

While this may not then be relevant to anything you say i still make the point 'all be it on my own then' that Vega is nothing like the Fury-X, the only thing it has in common are the number of shades it has.

It according to D.P could not be further from any truth. they are completely and totally different GPU's

Id be naive if i didn't think Vega is not the Furyx that has been taken as the card to optimise. Optimise the architecture improve it then put it on a node shrink. I said that Pascal is pretty much Maxwell on a die shrink but Vega has more significant optimisations to the arch. It's still the same arch just improved. AMD even said so themselves. It's the most changes to the architecture even since we introduced GCN. But then look at the configuration of the top end vega. Looks exaclty the same on paper as the fury x all be clock speeds ohh and its 8GB of HBM2 not 4GB of HBM.

The fury x is a cracker of a card and still is. It just seems like its getting held back some how. Look at it in DX12 games. Stomps the 980ti and beats the 1070 now and again, which is supposedly faster. This is the fury x lol. How old is this card again? Thats why i think AMD have taken this card as the subject to improve upon.
 
Either way, I am not paying the silly money they want for the new monitor's with HDR when there is hardly any HDR games for it. I would rather wait for OLED and proper HDR in couple of year time, at least the premium paid will be worth it.

That's the problem we'll face with the Freesync 2 specs if HDR is a condition, We need AMD to take control of the cheaper non HDR monitors that are using the Freesync label so we know which ones have the spec needed for a good Freesync experience. It's no good AMD only taking control of the premium monitors.
 
That's the problem we'll face with the Freesync 2 specs if HDR is a condition, We need AMD to take control of the cheaper non HDR monitors that are using the Freesync label so we know which ones have the spec needed for a good Freesync experience. It's no good AMD only taking control of the premium monitors.
They are taking so long to show us some freesync 2 monitors and their specs that I got tired of waiting. End of the day once you take HDR out of the equation, freesync 2 is essentially g-sync. This is why when I went for the agon ag271ug when I saw it at a competitive price. Just hope the panel quality matches my dell 4K when it comes today, if not what's the point. G-sync/freesync is good, but not that good that I would pay money to downgrade the quality my monitor to have it.

To be honest it was not so much tired of waiting, more so knowing any hdr monitor they are going to charge silly price for. Look at that lg monitor that came out recently with its crappy HDR 10 standard and what they charging for it.
 
They are taking so long to show us some freesync 2 monitors and their specs that I got tired of waiting. End of the day once you take HDR out of the equation, freesync 2 is essentially g-sync. This is why when I went for the agon ag271ug when I saw it at a competitive price. Just hope the panel quality matches my dell 4K when it comes today, if not what's the point. G-sync/freesync is good, but not that good that I would pay money to downgrade the quality my monitor to have it.

To be honest it was not so much tired of waiting, more so knowing any hdr monitor they are going to charge silly price for. Look at that lg monitor that came out recently with its crappy HDR 10 standard and what they charging for it.

Oh certainly, I really miss my LG 34UM95. That was a true thing of beauty; especially since it was calibrated and had a hardware LUT.

dnicgRm.jpg

I really regret getting rid of it, but despite the massive downgrade in visual quality, the overall gaming experience is significantly better on both PG278Q and MG278Q.

What I really want is a freesync 2 monitor that has the features of that monitor, but also low response time.
Hopefully computex has Ultrawides and 4K monitors that'll fill that gap we want.
 
Oh certainly, I really miss my LG 34UM95. That was a true thing of beauty; especially since it was calibrated and had a hardware LUT.

dnicgRm.jpg

I really regret getting rid of it, but despite the massive downgrade in visual quality, the overall gaming experience is significantly better on both PG278Q and MG278Q.

What I really want is a freesync 2 monitor that has the features of that monitor, but also low response time.
Hopefully computex has Ultrawides and 4K monitors that'll fill that gap we want.
Hopefully they do without asking four figures for them :p
 
They are taking so long to show us some freesync 2 monitors and their specs that I got tired of waiting. End of the day once you take HDR out of the equation, freesync 2 is essentially g-sync. This is why when I went for the agon ag271ug when I saw it at a competitive price. Just hope the panel quality matches my dell 4K when it comes today, if not what's the point. G-sync/freesync is good, but not that good that I would pay money to downgrade the quality my monitor to have it.

To be honest it was not so much tired of waiting, more so knowing any hdr monitor they are going to charge silly price for. Look at that lg monitor that came out recently with its crappy HDR 10 standard and what they charging for it.

Thankfully the Ultrawide monitor I have now has a 30-75 working range and LFC so I'm in no rush to upgrade again, When I do upgrade next I want it to be another Ultrawide though, preferably a 4k version with a minimum of 100hz and Free2 with HDR, By waiting another 2 or 3 years the premium pricing HDR holds should have dropped away.

What I really want is a freesync 2 monitor that has the features of that monitor, but also low response time.
Hopefully computex has Ultrawides and 4K monitors that'll fill that gap we want.

The sooner the better so the sooner the price premium drops off :)
 
Itis not too far from a die-shrunk and clocked FuryX. There was some small improvements in Polaris and there should be some more modest improvements with Vega and some additional functionality such as FP16 support. Vega is an evolution of GCN and many of the GCN architectural decision still exist.

That isn;t a bad thing. AMD have abad track record when it comes to ground up decisions. The last thing we need is another Bulldozer or 2900XT!

What about K8, Zen and GCN? I would say their track record isn't as bad as some are making out.
 
That isn;t a bad thing. AMD have abad track record when it comes to ground up decisions. The last thing we need is another Bulldozer or 2900XT!

What about K8, Zen and GCN? I would say their track record isn't as bad as some are making out.

The 2900XT was bad, but Terascale itself was great; it just couldn't work at 80nm.
The 4870, 5870 and 6970 were all Terascale, same as the 2900XT and they were fantastic GPUs.

Honestly I think Polaris is more 2900XT than Vega will be. Polaris once you try and make it compete with NVIDIA's upper-midrange would be far too big, hot, and power hungry. Just like what happened with the 2900XT; and it only really shines when using a smaller die, and okay clocks; then it keeps Power and Heat well in check. See Polaris Mobile parts.

The old Computerbase graphs looks like Polaris 10 at the moment; launch drivers were still back back then as well :P

fa23613fae694f3d9ba00d47bdd0a5ac.png
 
Last edited:
The 2900XT was bad, but Terascale itself was great; it just couldn't work at 80nm.
The 4870, 5870 and 6970 were all Terascale, same as the 2900XT and they were fantastic GPUs.

Honestly I think Polaris is more 2900XT than Vega will be. Polaris once you try and make it compete with NVIDIA's upper-midrange would be far too big, hot, and power hungry. Just like what happened with the 2900XT; and it only really shines when using a smaller die, and okay clocks; then it keeps Power and Heat well in check. See Polaris Mobile parts.

The old Computerbase graphs looks like Polaris 10 at the moment; launch drivers were still back back then as well :p

fa23613fae694f3d9ba00d47bdd0a5ac.png

But that is my point, the initial Terrsacale was pretty terrible. After several iterations and die shrinks it was really good but the initial release was terrible. When AMD iterated that design it got better, when AMD iterated the GCN architecture it got much better up to Hawaii, and then some bottlenecks appeared in Fiji that were only partially solved in Polaris. Vega as an iteration of GCN should be a good improvement. Fiji/FuryX can be excused on the fact that AMD were heavily expecting a 20nm process to be available and their backup plan just wasn't as refined as Nvidia's. AMD ploughed R&D into getting HBM on consumer cards which is commendable but ultimately was relatively pointless. Polaris is an odd release wihtout any high-end release. I think this is just a result of massively cut R&D and AMD trying to be efficient with development costs, with Polaris aimed for Apple products and the new consoles. It hit the low-end market which was important.
Vegas is the first GPU since Hawaii where everything should be on track.
 
What about K8, Zen and GCN? I would say their track record isn't as bad as some are making out.


For sure they have made some great ground up designs but there is much more risk. Even architectures like Zen, i think it will really shine after an iteration or 2. I would also have concerns about a brand new NVidia architecture.
 
Status
Not open for further replies.
Back
Top Bottom