• 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.

Poll: ** The AMD VEGA Thread **

On or off the hype train?

  • (off) Train has derailed

    Votes: 207 39.2%
  • (on) Overcrowding, standing room only

    Votes: 100 18.9%
  • (never ever got on) Chinese escalator

    Votes: 221 41.9%

  • Total voters
    528
Status
Not open for further replies.
It is still not clear to me whether the draw stream binning approach requires games to facilitate it (in order to achieve good utilization) or not.

If it does, then it may be that without games doing X, the improvement is minimal (hence the abysmal performance of Vega FE and by extension it will not improve much in RX Vega). Anyway... We'll soon find out...

From what I've seen there are 3 different pipelines - traditional, some kind of quasi tile based deferred immediate mode which is basically a compatibility hack without full performance benefits and the full tiled based implementation with maximum performance - from the looks of it drivers can force the first 2 but its upto the application developer to support the highest performance variant.

AMD seem to have gone for a more "elegant" solution rather than nVidia's brute force implementation with better ability to flexibly handle different loads but downside is it seems it needs more input from developers to get the same or better levels of performance from it so....

EDIT: From reading up on it the approach AMD is using is more about power (snigger) and bandwidth saving than all out performance while nVidia uses a brute force approach that can also have bigger implications for performance.

I believe the DSBR will have to be implemented by developers ..., because he said this :

"And some of Vega’s features, like our High Bandwidth Cache Controller, HBM2, Rapid-Packed Math, or the new geometry pipeline, have the potential to really break new ground and fundamentally improve game development. These aren’t things that can be mastered overnight. It takes time for developers to adapt and adopt new techniques that make your gaming experience better than ever"

ie. AMD concentrating on potential gains instead of known gains... Just like briging double half precision FP to consumer cards, I bet that will be useless for gaming and I wouldn't be surprised if it hinders the performance of the GPU as a whole when it comes to gaming
 
This thread described in one gif:

More like "Never mind a Ti, will it beat a 1080?" :D :p

giphy.gif
 
I believe the DSBR will have to be implemented by developers ..., because he said this :

"And some of Vega’s features, like our High Bandwidth Cache Controller, HBM2, Rapid-Packed Math, or the new geometry pipeline, have the potential to really break new ground and fundamentally improve game development. These aren’t things that can be mastered overnight. It takes time for developers to adapt and adopt new techniques that make your gaming experience better than ever"

Yeah that has been my understanding that their TBR method relies on DSBR which to get the most out of requires developer support while there is a fallback it won't do much for performance from what I can see mostly power saving.
 
I believe the DSBR will have to be implemented by developers ..., because he said this :

"And some of Vega’s features, like our High Bandwidth Cache Controller, HBM2, Rapid-Packed Math, or the new geometry pipeline, have the potential to really break new ground and fundamentally improve game development. These aren’t things that can be mastered overnight. It takes time for developers to adapt and adopt new techniques that make your gaming experience better than ever"

Hm, once again, AMD building their cards for the future and not for now. Trouble is, for enthusiasts at least, in a few years Vega will be irrelevant anyway.

I just don't think AMD are going down the right path. They need to build an architecture that can compete NOW, not at some point in the future and only when a developer pulls their finger out and uses some bit of tech the card has.
 
ill be extremely surprised if we see Navi in 2018 at all
+1

If I knew Navi was coming 100% in August 2018 I would even gladly chill with my 7970 until then and buy that. By then there will be a build up games worth playing also. It is just highly unlikely. Sad to say but true.
 
Hm, once again, AMD building their cards for the future and not for now. Trouble is, for enthusiasts at least, in a few years Vega will be irrelevant anyway.

I just don't think AMD are going down the right path. They need to build an architecture that can compete NOW, not at some point in the future and only when a developer pulls their finger out and uses some bit of tech the card has.

Yeah that's their problem, building tech for the potential of future usage isn't doing them any good at all. We also have to keep in mind that the architecture as a whole is aging now as well it's still GCN at it's core, I expect them to come out with something half decent when they get on to a new arch from ground up. And I don't believe that that will be Navi, I expect navi to be a modular shrunken Vega , probably the arch after will be new though (well I hope)
 
A question for the knowledgable people, do you think if they limited the FP16 and FP64 perf on the RX vega they would gain a fair amount of performance for gaming ? Coz I can see the point of Sony wanting double FP16 for the PS4 PRO, because they develop exclusives, so they can take advantage of that without worrying about everybody else's performance, but I really don't see the point of having it in a consumer GPU.
 
I sure AMD is also working on navi right now. What's to say this also on schedule for 2018.
So how will they be behind a generation? If Vega is indeed 1080 / 1080ti levels them AMD have matched this generation. So they not behind a generation.

Nicosia? lol! I was there a couple of years ago :D


To be honest the way things seem to be going with AMD, I won't be surprised is Navi comes out 1+ year after Volta.

If AMD are struggling now with Vega drivers, how are they going to handle Navi drivers one must ask themselves?



Lets say Nvidia release Volta May 2018 and AMD Release Navi August 2018 They they will have catched up. Sure if Navi comes 2019 then they are defo way behind.
The Road map shows Navi between 2018 and 2019 with Next Gen 2019 and 2020 Onwards

gpuroadmap_575px.png

It's taken 2 years for AMD to replace the Fiji cards, Vega is releasing at the end of July I can't see Navi being less than a year later, I think it'll more than likely be longer. Pascal's been around for quite a while now so some Volta gpu's appearing early next year wouldn't be a surprise. And if Nav's then a 2019 product that's a long time without competition, as we've seen with Fiji's replacement and Nvidia riding bareback over the consumer base. AMD seem to be struggling with the gpu products, As for the roadmap it shows Vega on 14nm and then on a refreshed 14nm+ so are we going to see what we saw with Polaris where Vega get's a refresh a year later which brings nothing more than a small performance jump via slightly better clocks? Personally I don't mind that if it means we get the Flagship remaining a driver priority for longer. As long as I can get a card that does a good job at 3440x1440 I'll keep it for 3 years if it's still going strong.
 
Yeah I don't get why people are claiming otherwise either, it is a jack of all trades, and so should be tested in all areas.

Game ready according to AMD? well I think it was probably rushed to meet a deadline.

It seems it's either that or they're holding something back driver wise until the RX releases.


Did Raja not say that the RX would be better for gaming? Hard to know what is meant by that, could just mean better price/performance ratio (I.E a cheaper card with the same performance) or it could mean better performance.

He did but it's hard to know what he means exactly, He also stated that it would still be a great card for playing games if you can't wait for the RX.

Personally I think it'll be a mix of better drivers, features not yet running and higher clocks.
 
If Vega is a disappointment it will go to show that AMD can only really focus on one thing at a time. Makes you think that maybe they will just leave Ryzen ticking over and switch full attention to Navi. As they have clearly done with Ryzen.

Fingers cross however that Vega is more then just a catch up job with less impressive thermals and power usage.

However if Vega is 'a little bit rubbish' it will make my choice of going green as usual an easy and comfortable one. If Vega comes out and is awesome, it will mean I will end up with an AMD card and Freesync, which for some reason doesn't feel as comfortable as an Nvidia card and Gsync.
 
Last edited:
It is still not clear to me whether the draw stream binning approach requires games to facilitate it (in order to achieve good utilization) or not.

If it does, then it may be that without games doing X, the improvement is minimal (hence the abysmal performance of Vega FE and by extension it will not improve much in RX Vega). Anyway... We'll soon find out...

It is a hardware feature that is independent to the game, it is something AMD will sort in the driver itself. It is not something that is opened to a game through an API.
 
I believe the DSBR will have to be implemented by developers ..., because he said this :

"And some of Vega’s features, like our High Bandwidth Cache Controller, HBM2, Rapid-Packed Math, or the new geometry pipeline, have the potential to really break new ground and fundamentally improve game development. These aren’t things that can be mastered overnight. It takes time for developers to adapt and adopt new techniques that make your gaming experience better than ever"

The new geometry pipeline is the primitive shaders part of things, nothing to do with the DSBR. The primitive shaders as with the rest of then can be taken advantage of in drivers on a per game basis, but for best performance they need to be directly coded for.
 
Lets say Nvidia release Volta May 2018 and AMD Release Navi August 2018 They they will have catched up. Sure if Navi comes 2019 then they are defo way behind.
The Road map shows Navi between 2018 and 2019 with Next Gen 2019 and 2020 Onwards

gpuroadmap_575px.png

The latest the 1080 successor will be out is April, and plausibly before that. They can put it out whenever really, since they only need 12nm and 14 Gbps GDDR6.

Navi needs 7nm to be ready, and the earliest indications are 2H2018 for volume production. So I'd expect October/November at the very very earliest.

In other words, it's pretty likely Volta and Navi will be 9 months apart.

Only saving grace will be medium Volta will come out first, where the 2070 is 1080 Ti performance, and the 2080 is ~25% faster. Nvidia will hold back the 1080 Ti successor till Q1 2019.
 
Status
Not open for further replies.
Back
Top Bottom