Next-gen API benches for RX 480 & 1060.

[tential]

Dont bring up "we work with..." BS here, it has nothing to do with what they do on this patch.



They said that async was only enabled on GCN cards, and that they where doing this to improve perf on old cards(that seems not to include Kepler as it performs worse than in OGL). Its on the release post.

https://community.bethesda.net/thread/54585?start=0&tstart=0

Pascal has some basic Async support, but they never bother to write the arch specific code for it, neither they bother to do a proper kepler code, even trought the focus of this patch was to improve perf on "old cards", as a result Kepler perf on vulkan is broken.

But for using Shaders Intrinsic Funtions, via GCN-only Vulkan extensions there is time.


Anyway, i rather AMD to lead DX12 perf for a gen or two, than see a AMD vs Nvidia title fight, its not gona be fun to be on either side if that happens.

Kepler + Async...
So many lols.....

 

[bononos]

Dont bring up "we work with..." BS here, it has nothing to do with what they do on this patch.

They said that async was only enabled on GCN cards, and that they where doing this to improve perf on old cards(that seems not to include Kepler as it performs worse than in OGL). Its on the release post.

https://community.bethesda.net/thread/54585?start=0&tstart=0

Pascal has some basic Async support, but they never bother to write the arch specific code for it, neither they bother to do a proper kepler code, even trought the focus of this patch was to improve perf on "old cards", as a result Kepler perf on vulkan is broken.
........

From what I'm reading here, its Nvidia who is being slow and not providing an updated async compute capable driver to Bethesda.

 

[Yakk]

Can't wait for DX12 Deus

 

[Bacon1]

Dont bring up "we work with..." BS here, it has nothing to do with what they do on this patch.



They said that async was only enabled on GCN cards, and that they where doing this to improve perf on old cards(that seems not to include Kepler as it performs worse than in OGL). Its on the release post.

https://community.bethesda.net/thread/54585?start=0&tstart=0

Pascal has some basic Async support, but they never bother to write the arch specific code for it, neither they bother to do a proper kepler code, even trought the focus of this patch was to improve perf on "old cards", as a result Kepler perf on vulkan is broken.

But for using Shaders Intrinsic Funtions, via GCN-only Vulkan extensions there is time.


Anyway, i rather AMD to lead DX12 perf for a gen or two, than see a AMD vs Nvidia title fight, its not gona be fun to be on either side if that happens.

My favorite part is the quote:

Currently asynchronous compute is only supported on AMD GPUs and requires DOOM Vulkan supported drivers to run. We are working with NVIDIA to enable asynchronous compute in Vulkan on NVIDIA GPUs. We hope to have an update soon.

Which sounds almost identical to one provided from Oxide when talking about Ashes of the Singularity, Async Compute, and Maxwell.

Lets see if Nvidia ever manages to get those drivers this time. Still waiting for Maxwell ones.

Odd that no one that bought Maxwell is upset they never got their promised async compute support. :\

 

[AnandThenMan]

Odd that no one that bought Maxwell is upset they never got their promised async compute support. :\

Odd is an understatement. And why have no review sites brought this up, Maxwell users were completely shafted.

 

[sirmo]

Odd is an understatement. And why have no review sites brought this up, Maxwell users were completely shafted.

So were Kepler users, and it won't be much different for Pascal users either by the looks of it.

 

[SPBHM]

I would think the Microsoft titles like Forza, QB, GeOW and so on are more neutral DX12 titles, also I haven't seen much testing done on them, so it nice to see, and the 480 is doing well there,

also, hopefully Nvidia drivers have stopped with the random crashes on QB

but I would like to see the tests being repeated with slower CPUs, to avoid surprise results like the Doom Vulkan test...

 

[frozentundra123456]

TR has PureHair, why is NV Biased? NV performaing best on DX11 than AMD should not be a suprise.

Actually i DONT WANT a fight of titles, damn it, if we start having a "Doom w/Vulkan" scenario on games for both brands its just gona destroy PC gaming.

These speculations about the projected (hyped??) performance uplift for GCN with the new APIs have been beaten to death already in the 480 thread and again in the 1060 thread.

Will be nice to have benchmarks when new games actually come out, but until then, all I see is a bunch of partisan speculation, cherry picking of benchmarks, and bickering.

 

[R0H1T]

Odd is an understatement. And why have no review sites brought this up, Maxwell users were completely shafted.

Unless you're promising something, anything to your shareholders, such (vague) statements are basically meaningless. This applies equally to Nvidia, AMD, Intel, Apple, Samsung et al.

 

[AnandThenMan]

I think it is safe to say the vast majority of NVDA customers are more apt to follow them on Twitter versus combing through financial statements and conference calls. So when Nvidia suggests something as fundamental as async compute will be enabled with a driver update people accept it to be true. And base their purchasing decisions in part on the info.

tldr; Nvidia lied.

 

[IllogicalGlory]

These speculations about the projected (hyped??) performance uplift for GCN with the new APIs have been beaten to death already in the 480 thread and again in the 1060 thread.

Will be nice to have benchmarks when new games actually come out, but until then, all I see is a bunch of partisan speculation, cherry picking of benchmarks, and bickering.

Don't try and hand me any of this. You know exactly what you'd be saying to someone defending the RX 480 with that line. I'm fairly certain that almost every available DX12 game was benched in this review, so cherrypicking...?

Anyway... Apparently iD is close enough to NV that they got access to the Vulkan patch well in advance of its release, as it was demo'ed on the 1080 during their event - plenty of time for NV's legendary driver team to work their magic, right? But, no it's iD's/AMD's fault. NV's drivers are blameless. If the situation were flipped, I wonder what the rhetoric would be? Fact is, the 1060 and comparable cards are getting taken to cleaners in DOOM by the RX 480. I think it's just time to accept that.

Do you really think that NV wouldn't be speaking out about this if iD were jerking them around? As far as I can see, they've been silent.

 

[AnandThenMan]

Doom under Vulkan is good for all GPUs not just AMD why can't people see this? AMD gains more which is why some are all up in arms but that misses the bigger picture. More Vulkan support means not being held hostage by Windows who doesn't want that? And it gives people that are not running Windows 10 (millions) access to a low level API. And maybe Linux gaming might actually be a thing, probably not but at least now there is a chance.

Win win and more win.

 

[R0H1T]

Doom under Vulkan is good for all GPUs not just AMD why can't people see this? AMD gains more which is why some are all up in arms but that misses the bigger picture. More Vulkan support means not being held hostage by Windows who doesn't want that? And it gives people that are not running Windows 10 (millions) access to a low level API. And maybe Linux gaming might actually be a thing, probably not but at least now there is a chance.

Win win and more win.

That's true unless you're stuck on Maxwell or older Nvidia cards, where at times we've seen performance regression. The problem going forward, is that as more games release with DX12/Vulkan & unless Nvidia is fairly close in performance to AMD they'll either make the game GW sponsored or delay running it under DX12/Vulkan as was the case with Doom. This all to make sure that Pascal is not shown in a bad light.

 

[Silverforce11]

That's true unless you're stuck on Maxwell or older Nvidia cards, where at times we've seen performance regression. The problem going forward, is that as more games release with DX12/Vulkan & unless Nvidia is fairly close in performance to AMD they'll either make the game GW sponsored or delay running it under DX12/Vulkan as was the case with Doom. This all to make sure that Pascal is not shown in a bad light.

You would think so, NV has all that profits from maximizing margins of Pascal, they can throw wads of cash at devs & publishers to get the sponsorship.

If they wanted to cripple AMD GPU in DX12/Vulkan, there's absolutely nothing AMD can do about it either, can't fix gimping via drivers when they use low-level APIs. lol

 

[AnandThenMan]

That's true unless you're stuck on Maxwell or older Nvidia cards, where at times we've seen performance regression.

True but Nvidia not coming up with a more robust architecture should not be holding the industry back, plus Kepler and Maxwell are not doing so good in newer DX11 titles either.

The problem going forward, is that as more games release with DX12/Vulkan & unless Nvidia is fairly close in performance to AMD they'll either make the game GW sponsored or delay running it under DX12/Vulkan as was the case with Doom. This all to make sure that Pascal is not shown in a bad light.

We all know Nvidia is going to pull everything string and do whatever else they can to harm adoption of true next gen game engines and games. This goes way way back remember DX10.1 for example?

 

[AtenRa]

NVIDIA has 30 games in the oven, i dont know how many will be DX-12/Vulkan but those will start to hit the market some time in 2017. Im sure most if not all will be optimized for Pascal and the Consoles. The oxymoron is that games will run great in GCN Consoles but not in GCN Desktops.
NVIDIA doesnt care if you have a bad PC Game port as long as their hardware if faster than the competition. They also dont care for older architectures, Kepler is dead for them and soon will Maxwell. The same will happen with Pascal the moment VOLTA hit the deck.
For 2016 DX-12 games AMD has the upper hand, lets see if they will continue to support new DX-12/Vulkan Game development.

 

[Silverforce11]

NVIDIA has 30 games in the oven, i dont know how many will be DX-12/Vulkan but those will start to hit the market some time in 2017. Im sure most if not all will be optimized for Pascal and the Consoles. The oxymoron is that games will run great in GCN Consoles but not in GCN Desktops.
NVIDIA doesnt care if you have a bad PC Game port as long as their hardware if faster than the competition. They also dont care for older architectures, Kepler is dead for them and soon will Maxwell. The same will happen with Pascal the moment VOLTA hit the deck.
For 2016 DX-12 games AMD has the upper hand, lets see if they will continue to support new DX-12/Vulkan Game development.

Is there a list somewhere of GameWorks titles coming out soon?

 

[sirmo]

NVIDIA has 30 games in the oven, i dont know how many will be DX-12/Vulkan but those will start to hit the market some time in 2017. Im sure most if not all will be optimized for Pascal and the Consoles. The oxymoron is that games will run great in GCN Consoles but not in GCN Desktops.
NVIDIA doesnt care if you have a bad PC Game port as long as their hardware if faster than the competition. They also dont care for older architectures, Kepler is dead for them and soon will Maxwell. The same will happen with Pascal the moment VOLTA hit the deck.
For 2016 DX-12 games AMD has the upper hand, lets see if they will continue to support new DX-12/Vulkan Game development.

The funny thing however is that there are more Maxwell and Kepler users than AMD users to poach. So they very well might be pushing Vulkan for this reason. So they will be stuck between a rock and a hard place on how to plan obsolescence this time without giving AMD too much of a gain.. they will likely pull a TimeSpy.

 

[AtenRa]

Is there a list somewhere of GameWorks titles coming out soon?

No, i heard it from a event or conference call dont remember where.

 

[AtenRa]

The funny thing however is that there are more Maxwell and Kepler users than AMD users to poach. So they very well might be pushing Vulkan for this reason. So they will be stuck between a rock and a hard place on how to plan obsolescence this time without giving AMD too much of a gain.. they will likely pull a TimeSpy.

For Pascal, TimeSpy is what they will most probable do. As for Kepler/Maxwell, they dont give a damn. Those are already dead for them, they now need to sell Pascal GPUs.

 

[Erenhardt]

The funny thing however is that there are more Maxwell and Kepler users than AMD users to poach. So they very well might be pushing Vulkan for this reason. So they will be stuck between a rock and a hard place on how to plan obsolescence this time without giving AMD too much of a gain.. they will likely pull a TimeSpy.

I've seen in 1060 reviews that nvidia is ready to bite their finger off if that means amd looses an arm. They can sacrifice a bit of performance going from vulkan to OGL, just to make amd cards suffer more performance loss.

They will kill amd with tesselation and render previous nv gens obsolete with preemption.

 

[3DVagabond]

They did work only for GCN, they only enabled async for AMD, no pascal, they said they where doing this patch to improve perf on older cards, but Kepler performas much, much worse, makes no sence.

Thats OK, but at least take the time and do some work for the other brand, its likely they did a Maxwell codepath and use it for Pascal and Kepler, that whould explain why there is no async on pascal and why Kepler performs so bad.


Now imagine if tomorrow a few games come out the oposite way, we dont need that.

They said they've been working directly with nVidia and they need to release a driver for async. Which they've promised (again).

 

[3DVagabond]

Doom under Vulkan is good for all GPUs not just AMD why can't people see this? AMD gains more which is why some are all up in arms but that misses the bigger picture. More Vulkan support means not being held hostage by Windows who doesn't want that? And it gives people that are not running Windows 10 (millions) access to a low level API. And maybe Linux gaming might actually be a thing, probably not but at least now there is a chance.

Win win and more win.

People just get upset when their nVidia cards don't win and claim conspiracy. If they can't do that it either doesn't matter/not important, or they ignore it.

 

[3DVagabond]

You would think so, NV has all that profits from maximizing margins of Pascal, they can throw wads of cash at devs & publishers to get the sponsorship.

If they wanted to cripple AMD GPU in DX12/Vulkan, there's absolutely nothing AMD can do about it either, can't fix gimping via drivers when they use low-level APIs. lol

I don't think they know how yet. They simply write the code as specified in Intel's and Khronos' programming guide and it works on AMD. That's exactly what Oxide said they did. They activated it on nVidia too but it created performance issues. So, they were forced to turn it off.

 

[Bacon1]

Unless you're promising something, anything to your shareholders, such (vague) statements are basically meaningless. This applies equally to Nvidia, AMD, Intel, Apple, Samsung et al.

What do you mean? Nvidia promised async compute for Maxwell during its launch.

Over year later and still no support for it. People would rather throw more money at Nvidia for new hardware than get promised features and performance for their old hardware. :hmm: