AMD VEGA confirmed for 2017 H1

[shankly1985]

AMD need to be delivery at the very least 30% faster than the 1070 across the board never mind the FX or they might as well not bother at this point.

FuryX in most titles isn't far off a 1070 tbh
I think most of you guys read/look into the "GPU War" Too much.. AMD Doesn't need to release a GPU faster than a 1080 they just need to release a GPU matching it at better pricing to be competitive.

The PC gamers will then make there own choice what branded they prefer.

The Key for AMD with VEGA is Pricing

 

[Toss3]

I know it is a daft question but if Vega is as quick as some people are making out why are AMD using Pascal Titans to demo their Ryzen CPU v a 6900k on BF1 ?

I can understand using the fastest GPU available to expose a CPU bottleneck on the 6900k which runs @3.2ghz compared to Ryzen @3.4ghz+.

AMD doing the above though means they are saying they have nothing close to the Pascal Titan ATM to do the job with.

http://techbuyersguru.com/ces-2017-amds-ryzen-and-vega-revealed

This could all change though very quickly if AMD have faster GPUs in the pipeline that are not quite ready yet.

They sre using Titans because it is a fair comparison; nvidia drivers aren't optimized for a specific platform. It also allows the tech press to do their own testing to verify amd's claims. It's all about being objective.

 

[Baboonanza]

They sre using Titans because it is a fair comparison; nvidia drivers aren't optimized for a specific platform. It also allows the tech press to do their own testing to verify amd's claims. It's all about being objective.

It's PR, the one thing you can guarantee it's not about is being objective, though in this case it could be seen as an unintended consequence

 

[DavidS14]

I can understand them using the fastest available card but with Vega being sooo close you'd think they'd use there own card.

 

[Jono8]

Just saw a post on Zen thread and read very interesting link.

http://techbuyersguru.com/ces-2017-amds-ryzen-and-vega-revealed?page=0

Interesting stuff. Zen is looking very promising which the industry really needs. Jury still seems to be out on Vega - they don't think it will even best a 1080 based on their numbers.

I personally think it will beat a 1080 and will be closer to Titan Pascal, but we will have to see. If it comes out in the middle of the year and can;t even beat a year old 1080 it will be rubbish (unless it is very cheap).

Also, AMD doing what they have always done (for better or for worse) by saying their cards will be better in the future and including features that aren't even out yet or can't be utilised. I think they really should concentrate on getting the performance right for the "here and now".

 

[Toss3]

It's PR, the one thing you can guarantee it's not about is being objective, though in this case it could be seen as an unintended consequence

Whatever the reason, it certainly isn't because Vega is lacking in the performance department. I'd still say they used them because the benchmark was all about the CPU, and because they wanted to truly cut out everything that could make people doubt its performance.

 

[Besty]

530mm2-560mm2
that is a huge die

Yes its a big chip, lots of grunt, features and a kick-ass memory setup. Set to be wasted on a near useless game dev community no doubt

E3 is in June and Fulcrum will get its reveal so I am hoping we will see a full reveal in April of Vega to avoid the E3 noise train in May.

Too many peeps trolling this thread at the moment, its nice that Nvidia die-hard's post what they think but its beyond funny now.

 

[AthlonXP1800]

This is the operations they are talking about. Game devs will need to code for either 16bit or 32bit. I doubt 8bit will be used much if at all because of how unpredictable it will be. I can remember Sony fanboys crying over their console being 8Tflops which is only in 16bit operations. This is the same thing.

Ah I see, I remembered read Mark Curry talked about PS4 Pro's 16bit operation or FP16 to get 4K games to run at full 60fps but it would lowered image quality because Nvidia had done it before with NV30 Geforce FX which games devs had to coded 16 bit or FP16 to achieved lower image quality while ATi R300 Radeon 9700 Pro had 24 bit or FP24 with higher image quality better than FP16 at same peformance.

Is that how I think so I don't know why either Mark Curry and RTG decided to go back to 2002 to borrowed FP16 from Nvidia's NV30 to the future and put into PS4 Pro Polaris and Vega. Wonder whether or not PS4 Pro and Vega will meet same fate as NV30 and Microsoft will use same FP32 for Xbox Scorpio with more powerful next gen GPU to run 4K games at 60fps.

 

[Toss3]

Yes its a big chip, lots of grunt, features and a kick-ass memory setup. Set to be wasted on a near useless game dev community no doubt

E3 is in June and Fulcrum will get its reveal so I am hoping we will see a full reveal in April of Vega to avoid the E3 noise train in May.

Too many peeps trolling this thread at the moment, its nice that Nvidia die-hard's post what they think but its beyond funny now.

TBH game development is extremely expensive so coding for an extremely small subset of the market isn't really a priority for any game developer. AMD should be the one doing the optimizations for their architectures (or making it worthwhile for the devs to do so).

 

[JediFragger]

FuryX in most titles isn't far off a 1070 tbh
I think most of you guys read/look into the "GPU War" Too much.. AMD Doesn't need to release a GPU faster than a 1080 they just need to release a GPU matching it at better pricing to be competitive.

The PC gamers will then make there own choice what branded they prefer.

The Key for AMD with VEGA is Pricing

True, and I suppose their's a good reason that they have Vega 20 planned for a 2H release this year....

 

[r7slayer]

Ah I see, I remembered read Mark Curry talked about PS4 Pro's 16bit operation or FP16 to get 4K games to run at full 60fps but it would lowered image quality because Nvidia had done it before with NV30 Geforce FX which games devs had to coded 16 bit or FP16 to achieved lower image quality while ATi R300 Radeon 9700 Pro had 24 bit or FP24 with higher image quality better than FP16 at same peformance.

Is that how I think so I don't know why either Mark Curry and RTG decided to go back to 2002 to borrowed FP16 from Nvidia's NV30 to the future and put into PS4 Pro Polaris and Vega. Wonder whether or not PS4 Pro and Vega will meet same fate as NV30 and Microsoft will use same FP32 for Xbox Scorpio with more powerful next gen GPU to run 4K games at 60fps.

I can't remember the whole reason why but i remember them saying that 16bit operations have improved a lot over time making it viable now for some gfx orientated tasks while 8bit is still unpredictable. So perhaps thats why?

 

[h4rm0ny]

Also, AMD doing what they have always done (for better or for worse) by saying their cards will be better in the future and including features that aren't even out yet or can't be utilised. I think they really should concentrate on getting the performance right for the "here and now".

I disagree on this last one. Firstly for all the people who buy a card to last (which I think is most), it's great to see their card keep up and even surpass rival cards that may even have been better than it at launch. Secondly, if AMD don't introduce advanced features then developers are less inclined to actually put the work in and develop for them which would in turn mean they would always remain "future features". It's a chicken and egg situation, so AMD put their money where their mouth is and say: "Okay, we'll go first." AMD's practices help keep driving the market forward. It benefits the community as a whole.

 

[Mtom]

Interesting stuff. Zen is looking very promising which the industry really needs. Jury still seems to be out on Vega - they don't think it will even best a 1080 based on their numbers.

I personally think it will beat a 1080 and will be closer to Titan Pascal, but we will have to see. If it comes out in the middle of the year and can;t even beat a year old 1080 it will be rubbish (unless it is very cheap).

Also, AMD doing what they have always done (for better or for worse) by saying their cards will be better in the future and including features that aren't even out yet or can't be utilised. I think they really should concentrate on getting the performance right for the "here and now".

Actually it is already around 1080 performance, with unfinished drivers, an early ES chip, posdibly low clock.
Also AMD is doing the opposite what you are saying.
Actually many of the Vega design is taking it closer to NV Pascal like Tiled Rendering, higher clock speeds, etc.

 

[nashathedog]

Depends what they charge for it, 1070 performance at a cheaper price will still sell well enough.

It would but if they want to stay in the game it needs to be faster.

 

[Jono8]

I disagree on this last one. Firstly for all the people who buy a card to last (which I think is most), it's great to see their card keep up and even surpass rival cards that may even have been better than it at launch. Secondly, if AMD don't introduce advanced features then developers are less inclined to actually put the work in and develop for them which would in turn mean they would always remain "future features". It's a chicken and egg situation, so AMD put their money where their mouth is and say: "Okay, we'll go first." AMD's practices help keep driving the market forward. It benefits the community as a whole.

I agree, but really they need to be doing both! By the time the fruits of these labours become apparent, new cards and a whole new generation of cards are out usually. It works well for people who want to buy a card once and then get by without upgrading for 4/5 years but for enthusiasts, "the here and now" is the most important (it is for me anyway).

 

[nashathedog]

I can understand them using the fastest available card but with Vega being sooo close you'd think they'd use there own card.

Doing that would have given too much info away and allowed Nvidia to gauge Vega's performance and give them time to formulate how to respond or not respond, Basically they want to keep them clueless and on their toes.

 

[harney]

I know it is a daft question but if Vega is as quick as some people are making out why are AMD using Pascal Titans to demo their Ryzen CPU v a 6900k on BF1 ?

I can understand using the fastest GPU available to expose a CPU bottleneck on the 6900k which runs @3.2ghz compared to Ryzen @3.4ghz+.

AMD doing the above though means they are saying they have nothing close to the Pascal Titan ATM to do the job with.

http://techbuyersguru.com/ces-2017-amds-ryzen-and-vega-revealed

This could all change though very quickly if AMD have faster GPUs in the pipeline that are not quite ready yet.

interesting

 

[RavenXXX2]

I know it is a daft question but if Vega is as quick as some people are making out why are AMD using Pascal Titans to demo their Ryzen CPU v a 6900k on BF1 ?

Nvidia have 80% of the discrete market, that's a lot of potential Zen customers. They be showing the Nvidia brigade that Zen will run that Nvidia GPU as well as Intel will.

 

[fs123]

Doing that would have given too much info away and allowed Nvidia to gauge Vega's performance and give them time to formulate how to respond or not respond, Basically they want to keep them clueless and on their toes.

Exactly. If we see a 6900K running Vega then we can gauge the performance by comparing it with other cards which gives info away to Nvidia. Not the best idea when launch is upto 6 months away.

 

[Rroff]

Doing that would have given too much info away and allowed Nvidia to gauge Vega's performance and give them time to formulate how to respond or not respond, Basically they want to keep them clueless and on their toes.

I'd be more surprised if nVidia didn't know pretty much what Vega's performance will be than if they did know. Companies like that tend to have people dedicated to tearing down and analysing what the competition is doing, etc. when I was a lab tech I spent a lot of time deconstructing competitor's products into component parts for another department to analyse (though largely for legal reasons to check they weren't using our patents, etc.).