AMD Vega MegaThread! FAQ and Resources

[MrN1ce9uy]

So now no June or even July but August or later? By the time something comes out nvidia will have already surpassed themselves again won't they? Starting to think red team will never be competitive again...

Yes it's quite disappointing if I must say. They did manage to release the RX 480 before the GTX 1060 though; even though it was priced much higher than MSRP due to limited availability. They could be selling RX 570 and 580 like hotcakes right now, but where are they? They are nowhere to be found. RX Vega should have been released by now, those would sell like hotcakes also. The only good thing they could possibly be doing right now by pushing back the RX Vega launch is to make that many more RX 570/580 units for all the GPU miners to buy up. However many they make, miners will buy them ALL. I know that doesn't seem good for gamers, but if it makes AMD money to further their competitiveness, then it is good for gamers.

 

[jaymc]

Official AMD Vega Benchmark Claims It Is 21% More Powerful Than The Nvidia Titan Xp:

http://segmentnext.com/2017/06/05/official-amd-vega-benchmark/

 

[cdrkf]

Official AMD Vega Benchmark Claims It Is 21% More Powerful Than The Nvidia Titan Xp:

http://segmentnext.com/2017/06/05/official-amd-vega-benchmark/

I could see that being legitimate- a few things to note though- it's at a very high resolution (4K) and is pushing high levels of AA... (x8).

That makes sense as based on the specs we've seen Vega is going to be memory and shader heavy (so high FP 32) but is likely weaker than nVidia on the back end in terms of rop and texture throughput.

What that likely means is it will be slower than nVidia at lower resolutions as the back end will limit the max fps it can hit, and as you increase resolution and add things like AA that shift the load onto the HBM memory and powerful shader engine the benchmarks will swing towards AMD. Note I'll bet there will be DOOM benches at launch 'proving' Titan XP is faster simply by changing the AA settings for example (the point being, pay careful attention to the settings used when comparing benchmarks as given the differences in architectures it'll be easy to push the results either way depending on choice of settings). This is definitely a best case benchmark although as they've given the settings I'd say it's valid. I just wouldn't view this as indicative of the overall performance picture.

 

[renz496]

lol i first saw that exact chart being posted a few weeks ago on VCZ comment section. if look at the foot note there is mentioned about NDA being lifted on May 8th. this is back when there is speculation Vega will launch together with Bethesda Prey. overclock.net forum did have a thread about that chart in early april:

http://www.overclock.net/t/1629599/various-full-vega-faster-than-titan-xp-in-4k-release-date-may-9th

 

[goldstone77]

All we can do it hope that Vega is on point. I am hoping that they learned their lessons from the R7 release to get all their ducks in a row before they release this time!

 

[varis]

All we can do it hope that Vega is on point. I am hoping that they learned their lessons from the R7 release to get all their ducks in a row before they release this time!

So what happened with R7? Not sure even which year it was, was just happily gaming on my old setup

 

[cdrkf]

All we can do it hope that Vega is on point. I am hoping that they learned their lessons from the R7 release to get all their ducks in a row before they release this time!

So what happened with R7? Not sure even which year it was, was just happily gaming on my old setup

I believe he is referring to 'Ryzen 7'- and imo Ryzen launch wasn't bad. The main issue was the early bios on motherboards making some of the machines a bit of a pain to get up and running. Still it was hardly a disaster- everything else (e.g. the 'poor' gaming results) AMD could do little about before hand.

 

[goldstone77]

All we can do it hope that Vega is on point. I am hoping that they learned their lessons from the R7 release to get all their ducks in a row before they release this time!

So what happened with R7? Not sure even which year it was, was just happily gaming on my old setup

cdrkf nailed it. The R7 Ryzen launch this year. I disagree cdrkf it was bad from a PR point of view for sure. It was rushed, and motherboard manufactures complained. Huge problems with many well known YouTube personalities spending hours trying to get the systems stable. RAM incompatibility. Unoptimized Ryzen gaming benchmarks vs. Intel every where you looked. I hope Vega will launch without much issue!

 

[jaymc]

Vega's Shader Engines: Teraflops Unleashed for Gaming:

https://youtu.be/owL_KY9sIx8

 

[goldstone77]

Vega's Shader Engines: Teraflops Unleashed for Gaming:

https://youtu.be/owL_KY9sIx8

Great news! Vega will be good! Winners? EVERYONE!!!!!!!

 

[varis]

Some more leaks and announcements in the meanwhile. http://wccftech.com/amd-threadripper-1920-12-core-cpu-vega-16gb8gb-cards-leaked/

Vega Gaming could be essentially the same as Vega Pro 56? 1080 competitor probably? So it would seem that 1080ti and 1080 competitors launch first and the lower end cards come later, late August/early September perhaps?

 

[Frank Walsh]

Those bloody miners, can't get a hold of an RX580 , ebuyer had some on their site, hit the add to basker button nothing happened, spoke to CS and was told by the agent those showing on the site had already been pre ordered and they were no longer taking pre orders as they still had 300 to go. I ended up just going for a Nvidia Card, thought feck it and splashed out on a MSI NVIDIA GeForce GTX 1080 8GB GAMING X PLUS 11Gbps Graphics Card

Ebuyer could not say when they would have the card again for "gamers to buy" , Scan last I looked said Aug, the reason I just went with Nvidia

 

[Frank Walsh]

As for the DX12 agruement , I seen one online vloger giving price comparisons and saying that with MS still having Win7-8 support up until 2020 before they drop it , DX12 games will be few and far between.

 

[jaymc]

Vega available for pre-order.. Frontiers Edition... big money but hey at last right.

We will have benchmarks soon happy day's.
http://wccftech.com/amd-radeon-vega-frontier-edition-air-liquid-cooled-pre-order-1800-usd/

 

[JamesSneed]

Vega available for pre-order.. Frontiers Edition... big money but hey at last right.

We will have benchmarks soon happy day's.
http://wccftech.com/amd-radeon-vega-frontier-edition-air-liquid-cooled-pre-order-1800-usd/

So it shouldn't be to long until the non-pro cards land which is nice. I do hope they hit this one out of t he park.

 

[MrN1ce9uy]

I hope there will be an RX Vega GPU for me.

 

[Rogue Leader]

On the Radeon Facebook page they have been posting from E3 with stuff like "what do you want to see today?"

Every day I post:

"How about a pre-order program so gamers can get their hands on RX Vega gaming GPUs before all the miners buy them up and jack their price through the roof."

Hopefully someone hears me, but somehow I doubt it.

 

[-Fran-]

On the Radeon Facebook page they have been posting from E3 with stuff like "what do you want to see today?"

Every day I post:

"How about a pre-order program so gamers can get their hands on RX Vega gaming GPUs before all the miners buy them up and jack their price through the roof."

Hopefully someone hears me, but somehow I doubt it.

Haha! RL, if I were from the sells team in AMD I would do the absolute opposite given that information: up the MSRP before release for all the miners to pay the adopters fee, then 3 months later reduce it by 30% or something.

Am I being too evil? haha.

Cheers!

 

[mitch074]

Haha! RL, if I were from the sells team in AMD I would do the absolute opposite given that information: up the MSRP before release for all the miners to pay the adopters fee, then 3 months later reduce it by 30% or something.

Am I being too evil? haha.

Cheers!

Thing is, miners buy the CHEAPEST card they can buy that can still perform cluster computes (and the RX570/580 is cheaper than the GTX1060) while gamers want the best performing card - so if AMD jacked up the price on Polaris, miners wouldn't buy them anymore.

And, in that vein, AMD don't care: their chips ARE selling like hot cakes because they ARE the best in class (high performance for cheap, otherwise miners wouldn't buy them all) - it's just gamers that lose big time here, because they are now forced to buy the overpriced, less efficient competitor's product.

As for me, I still rock my reference RX480 8 Gb.

 

[renz496]

As for the DX12 agruement , I seen one online vloger giving price comparisons and saying that with MS still having Win7-8 support up until 2020 before they drop it , DX12 games will be few and far between.

even if win 8 and win 7 did get DX12 support the issue is more with game developer. because DX12 are not what many people think it was. Vulkan while we see more positive result for it compared to DX12 some developer probably will not going to use it for the same reason as why they are not using OpenGL in the first place.

 

[mitch074]

even if win 8 and win 7 did get DX12 support the issue is more with game developer. because DX12 are not what many people think it was. Vulkan while we see more positive result for it compared to DX12 some developer probably will not going to use it for the same reason as why they are not using OpenGL in the first place.

I'm not so sure - the main thing against OpenGL is that it's HARD to program for while DX11 is pretty much easy-peasy. Vulkan and DX12 are pretty much equal on that regard, and while Xbox consoles now use DX12, the Nintendo Switch uses Vulkan - as do more modern Android devices. so we might just see more Vulkan games in the future, which does cover Win7/8.
The main problem with DX12/Vulkan is that current 3D engines are geared towards DX11, and actually making use of DX12 or Vulkan pretty much requires a ground-up redesign of the engine. Even the latest revisions of Unreal Engine, Unity, or Serious Engine, or even id tech 6, while they DO support Vulkan (and DX12 for some of them) as a render target, by the words of their authors, don't use half of the advantages these APIs offer them.

Yet.

Note: AMD's async extension allows for a HUGE boost for GCN 2, 3 and 4 under Vulkan, but by the admission of id developers, it was a low-hanging fruit - let the hardware handle the dispatching of shadow rendering jobs to idle compute units instead of doing it through the driver thread. Most developers agree that we could see 30-45% more rendering performance on any hardware on tasks similar to today's renders when using engines optimized from the ground-up for APIs such as Metal, Mantle, Vulkan and DX12, while 70-80% of the job needed to target any of those APIs would be common with the others.

 

[cdrkf]

Haha! RL, if I were from the sells team in AMD I would do the absolute opposite given that information: up the MSRP before release for all the miners to pay the adopters fee, then 3 months later reduce it by 30% or something.

Am I being too evil? haha.

Cheers!

Thing is, miners buy the CHEAPEST card they can buy that can still perform cluster computes (and the RX570/580 is cheaper than the GTX1060) while gamers want the best performing card - so if AMD jacked up the price on Polaris, miners wouldn't buy them anymore.

And, in that vein, AMD don't care: their chips ARE selling like hot cakes because they ARE the best in class (high performance for cheap, otherwise miners wouldn't buy them all) - it's just gamers that lose big time here, because they are now forced to buy the overpriced, less efficient competitor's product.

As for me, I still rock my reference RX480 8 Gb.

The reason the miners are buying AMD is that AMD designs are *shader heavy*-which means very high performance in mining tasks compared to their price point. Case in point an RX 580 can mine data as fast as a GTX 1070- it totally eclipses the 1060. AMD have an outright performance (and perf / w) advantage in this area- which is why they are so popular and why miners will still by the cards even at inflated prices.

It unfortunately puts the AMD boards totally out of position vis a vis their gaming performance segment though.

 

[renz496]

even if win 8 and win 7 did get DX12 support the issue is more with game developer. because DX12 are not what many people think it was. Vulkan while we see more positive result for it compared to DX12 some developer probably will not going to use it for the same reason as why they are not using OpenGL in the first place.

I'm not so sure - the main thing against OpenGL is that it's HARD to program for while DX11 is pretty much easy-peasy. Vulkan and DX12 are pretty much equal on that regard, and while Xbox consoles now use DX12, the Nintendo Switch uses Vulkan - as do more modern Android devices. so we might just see more Vulkan games in the future, which does cover Win7/8.
The main problem with DX12/Vulkan is that current 3D engines are geared towards DX11, and actually making use of DX12 or Vulkan pretty much requires a ground-up redesign of the engine. Even the latest revisions of Unreal Engine, Unity, or Serious Engine, or even id tech 6, while they DO support Vulkan (and DX12 for some of them) as a render target, by the words of their authors, don't use half of the advantages these APIs offer them.

Yet.

Note: AMD's async extension allows for a HUGE boost for GCN 2, 3 and 4 under Vulkan, but by the admission of id developers, it was a low-hanging fruit - let the hardware handle the dispatching of shadow rendering jobs to idle compute units instead of doing it through the driver thread. Most developers agree that we could see 30-45% more rendering performance on any hardware on tasks similar to today's renders when using engines optimized from the ground-up for APIs such as Metal, Mantle, Vulkan and DX12, while 70-80% of the job needed to target any of those APIs would be common with the others.

AFAIK nvidia develop NVN for the Switch. while nintendo did support Vulkan we still to this date did not know if nintendo switch use Vulkan or nvidia NVN for the Switch. but looking how the performance on the Tegra soc can be limiting factor using NVN most likely better for game developer to maximize hardware potential on the switch.

extension is part of the reason why vulkan is better DX12 in regards to low level API. but that also one factor that made OpenGL hard in the first place because developer for most part did not want to deal vendor specific extension. so while Vulkan become much more easier to code for compared to OpenGL the extension part is still the same as it was before with OpenGL. we see very positive result with Doom because the developer have been dealing with OpenGL as their primary 3D API for more than a decade. we might not see similar result from other developer. this is one such example where D3D 11 still end up being faster than Vulkan:

https://www.youtube.com/watch?v=kMesHSxomEg&t=198s

 

[mitch074]

AFAIK nvidia develop NVN for the Switch. while nintendo did support Vulkan we still to this date did not know if nintendo switch use Vulkan or nvidia NVN for the Switch. but looking how the performance on the Tegra soc can be limiting factor using NVN most likely better for game developer to maximize hardware potential on the switch.

extension is part of the reason why vulkan is better DX12 in regards to low level API. but that also one factor that made OpenGL hard in the first place because developer for most part did not want to deal vendor specific extension. so while Vulkan become much more easier to code for compared to OpenGL the extension part is still the same as it was before with OpenGL. we see very positive result with Doom because the developer have been dealing with OpenGL as their primary 3D API for more than a decade. we might not see similar result from other developer. this is one such example where D3D 11 still end up being faster than Vulkan:

https://www.youtube.com/watch?v=kMesHSxomEg&t=198s

Yeah, well, this test runs using a DX9-optimized engine (Serious Engine, which supports Vulkan with a "wrapper" around their engine) - because yes, Serious Engine is originally a DX9/OGL2.1 engine with several backends (some still in use, others not) - amongst which, DX9, DX11, OpenGL2.1, OpenGL 4, and Vulkan. Vulkan is the least mature and the most complex of the 4, and Croteam developers did say that their engine needs TONS of changes to properly exploit Vulkan's possibilities : https://www.khronos.org/assets/uploads/developers/library/2017-khronos-uk-vulkanised/005-Vulkanised-Dean-Sekulic-Getting-Serious-with-Vulkan_May17.pdf

 

[renz496]

AFAIK nvidia develop NVN for the Switch. while nintendo did support Vulkan we still to this date did not know if nintendo switch use Vulkan or nvidia NVN for the Switch. but looking how the performance on the Tegra soc can be limiting factor using NVN most likely better for game developer to maximize hardware potential on the switch.

extension is part of the reason why vulkan is better DX12 in regards to low level API. but that also one factor that made OpenGL hard in the first place because developer for most part did not want to deal vendor specific extension. so while Vulkan become much more easier to code for compared to OpenGL the extension part is still the same as it was before with OpenGL. we see very positive result with Doom because the developer have been dealing with OpenGL as their primary 3D API for more than a decade. we might not see similar result from other developer. this is one such example where D3D 11 still end up being faster than Vulkan:

https://www.youtube.com/watch?v=kMesHSxomEg&t=198s

Yeah, well, this test runs using a DX9-optimized engine (Serious Engine, which supports Vulkan with a "wrapper" around their engine) - because yes, Serious Engine is originally a DX9/OGL2.1 engine with several backends (some still in use, others not) - amongst which, DX9, DX11, OpenGL2.1, OpenGL 4, and Vulkan. Vulkan is the least mature and the most complex of the 4, and Croteam developers did say that their engine needs TONS of changes to properly exploit Vulkan's possibilities : https://www.khronos.org/assets/uploads/developers/library/2017-khronos-uk-vulkanised/005-Vulkanised-Dean-Sekulic-Getting-Serious-with-Vulkan_May17.pdf

what happen with Serious Sam might as well happen to majority of current game engine. as you said the game engine needs to be built from the ground up for Vulkan. but we will not going to see such engine very soon. in fact i'm not surprise if such engine will not exist until 2020 and beyond time frame. game engine development can take very long time especially if they were written from the ground up. take unreal engine 4 for example. the development was started in 2003. but the first game using unreal engine 4 did not ship until 2014.