AMD Vega GPU launch reportedly moved to October

[NeverBackDown]

Actually they didn't start mass production until January of this year.

You are correct. I was thinking of tsmc 16nm then. Started around September for volume production. 14nm LPP and HBM2 didn't start until January.

 

[shambles1980]

It's a huge focus because it's a huge market. It will surpass $5billion(or 6 if memory is wrong) market within the next couple years. You can bet any company will want some of that market share.

i think you are missing that the majority of those sales will be the phone based ones

 

[Kushiro]

Still a big market.

 

[Tolemac]

Hmmm wait for vega or get 2 polaris cards in crossfire for 1440p.

 

[NeverBackDown]

i think you are missing that the majority of those sales will be the phone based ones

Your missing part of the point. It's a growing market and is growing rapidly in all area's. Outside of gaming there's still many markets it can address. Whether it be education or science, it's something both companies need. They can't rely on PC market, they need to expand into other profitable area's to reduce there PC exposure as it continues to lose profitability at quick rate.

You don't have to like VR, but it's definitely not leaving and since it is a big market, expect both AMD and Nvidia to focus on it. You should be glad to, more money they make means more powerful GPUs

 

[Agost]

I REALLY hope this rumor is true, I'm really looking forward to switching to UHD and Polaris10 could not be enough...

 

[AngryGoldfish]

Hmmm wait for vega or get 2 polaris cards in crossfire for 1440p.

It's stipulated that AMD will be reducing the size of the chips with Navi. If, instead of relying on large chips to perform well, they rely on the "scalability" of multiple smaller GPU's on the same die, Crossfire might be the way to go. This is just a stipulation from AdoredTV, but it makes sense. He believes AMD has a "Master Plan" that involves shrinking chip sizes—to improve yields and reduce costs—and putting two smaller chips on the same interposer. With the new API's out at the moment, and the next generation of consoles likely to powered by AMD again, if AMD can persuade Sony, Microsoft and Nintendo to go with a dual-GPU APU for 980ti-like performance, without the huge costs and efficiency problems of that, the game developers will be forced to develop their games around DX12 and Vulkan that better support multi GPU's. This means it won't be down to drivers from AMD or nVidia; it'll be the game developers themselves that make Crossfire and SLI more attractive.

That's the theory. Also, VR loves multi GPU's, and as NeverBackDown is saying, it's here to stay and is a big potential market. If VR developers want multi GPU's, AMD and nVidia—and also the game developers—might focus on it more.

It ties in with Moore's Law as well.

 

[NeverBackDown]

It's stipulated that AMD will be reducing the size of the chips with Navi. If, instead of relying on large chips to perform well, they rely on the "scalability" of multiple smaller GPU's on the same die, Crossfire might be the way to go. This is just a stipulation from AdoredTV, but it makes sense. He believes AMD has a "Master Plan" that involves shrinking chip sizes—to improve yields and reduce costs—and putting two smaller chips on the same interposer. With the new API's out at the moment, and the next generation of consoles likely to powered by AMD again, if AMD can persuade Sony, Microsoft and Nintendo to go with a dual-GPU APU for 980ti-like performance, without the huge costs and efficiency problems of that, the game developers will be forced to develop their games around DX12 and Vulkan that better support multi GPU's. This means it won't be down to drivers from AMD or nVidia; it'll be the game developers themselves that make Crossfire and SLI more attractive.

That's the theory. Also, VR loves multi GPU's, and as NeverBackDown is saying, it's here to stay and is a big potential market. If VR developers want multi GPU's, AMD and nVidia—and also the game developers—might focus on it more.

It ties in with Moore's Law as well.

Interesting stuff. It's probably still better to go one bigger die rather than 2 smaller connected dies for GPUs considering that should add latency(even if 2ms is to much), when you play VR every ms counts. I'm no GPU engineer so i'll leave it to them if they know it will be better. One thing scalability lets them do is make really small dies for mobile, whether it be smartphones or tablets or even "mini tower" OEMs. It can allow them to get into more markets. Which is good considering they need the money..

I feel like Vulkan is better suited to multiplat games. PS is the only one not able to use DX, so supporting two different APIs like has been the traditional method, may entice Devs to try just one universal one. That in turn could reduce costs on there side and put more resources into things like VR since it's gaining traction. We'll see how everything turns out^_^

 

[AngryGoldfish]

Interesting stuff. It's probably still better to go one bigger die rather than 2 smaller connected dies for GPUs considering that should add latency(even if 2ms is to much), when you play VR every ms counts. I'm no GPU engineer so i'll leave it to them if they know it will be better. One thing scalability lets them do is make really small dies for mobile, whether it be smartphones or tablets or even "mini tower" OEMs. It can allow them to get into more markets. Which is good considering they need the money..

I feel like Vulkan is better suited to multiplat games. PS is the only one not able to use DX, so supporting two different APIs like has been the traditional method, may entice Devs to try just one universal one. That in turn could reduce costs on there side and put more resources into things like VR since it's gaining traction. We'll see how everything turns out^_^

Is latency still an expected problem with multi GPU's in DX12/Vulkan though *edit* when handled by the game developers and not the GPU manufactuers?

 

[NeverBackDown]

Is latency still an expected problem with multi GPU's in DX12/Vulkan though *edit* when handled by the game developers and not the GPU manufactuers?

Yes. Latency is always a factor
If you have two separate GPUs on the same die like you said earlier, you introduce latency just like having two separate GPUs. Defeats the purpose! May as well have a single bigger die! That's the most logical sense. Don't know if there's any tech out there that allows this without penalty.

 

[AngryGoldfish]

Yes. Latency is always a factor
If you have two separate GPUs on the same die like you said earlier, you introduce latency just like having two separate GPUs. Defeats the purpose! May as well have a single bigger die! That's the most logical sense. Don't know if there's any tech out there that allows this without penalty.

I've not actually ever gamed on an SLI or Crossfire system so I can't say what it's like, but it's a really popular setup for enthusiasts who insist on the best, so maybe it's not as bad as paying a premium for a massive chip that's theoretically not as powerful.

 

[Tolemac]

I may well go polaris I really don't know yet i'm gaming at 1440p and have no plans for 4K so I won't speculate I just have to wait for the polaris benchmarks to see if it is worthwhile upgrading my 290X is showing its age in doom but i'm still getting playable framerates. As we all know AMD cards are like a good wine they mature well as they get a little older

 

[NeverBackDown]

I'm excited for Polaris, but I won't be upgrading, I'll wait till Vega, my FX is more than enough at 1440p already