Futuremark Releases 3DMark v2.3.3663 - Adds Vulkan Support

[bug]

Care to explain that?

Explain what?

Well, I've bolded it for you, but it seems that was too subtle. So let me try again: please explain why Vulkan "Offers better fps, visuals and dev friendly".

 

[Aquinus]

People fail not understand one extremely important thing: coding for Vulkan/D3D is like going from Java to assembler: enormous complexity, very difficult to debug and often zero or negative gains over D3D 11/OpenGL 4.5 unless you're very talented.

That's a bit of an exaggeration. The difference really is command buffers versus a global draw call. If you map everything over in the sense of one buffer for every draw call or one buffer for all of the draw calls, you're missing out on what both DX12 and Vulkan have to offer. It's not hard to use it, it's hard to use it properly when what you're porting is doing stuff the old way. The learning curve from say OpenGL to Vulkan isn't like going from Java to Assembly, it's more like learning a library you've never used before, not learning a completely different language.

JAVA is the worst invention of humanity after Politics and Religion. Common! Lazy programmers are using Java.

If you're saying that but, aren't a software engineer, I have some words for you that would earn me an infraction. People bash the tool but, it's not the tool, it's how people use it. Without the JVM, I wouldn't currently have a job, and I don't write Java...

 

[Super XP]

Well, I've bolded it for you, but it seems that was too subtle. So let me try again: please explain why Vulkan "Offers better fps, visuals and dev friendly".

I'm not a software engineer, ain't going to get into details, but based on many hardware reviews, in a nut shell Vulcan blows DX12 out of the water.

 

[bug]

I'm not a software engineer, ain't going to get into details, but based on many hardware reviews, in a nut shell Vulcan blows DX12 out of the water.

Can you at least link those "hardware reviews" that tell us how a software API "Offers better fps, visuals and dev friendly"?

 

[Super XP]

Can you at least link those "hardware reviews" that tell us how a software API "Offers better fps, visuals and dev friendly"?

You need to do your own DD on the matter. The net has a lot of info. Based on my personal DOOM playing, everything at Ultra High Settings, I gain a lot more FPS when Vulcan is enabled.

 

[bug]

You need to do your own DD on the matter. The net has a lot of info. Based on my personal DOOM playing, everything at Ultra High Settings, I gain a lot more FPS when Vulcan is enabled.

The rules of a polite discussion are that when you make a statement, you have to back it up. I have done my DD on the subject and what you say is pure BS, that's why I was asking for additional info.

 

[Super XP]

The rules of a polite discussion are that when you make a statement, you have to back it up. I have done my DD on the subject and what you say is pure BS, that's why I was asking for additional info.

Your posts started from "Mar 24, 2017 at 2:35 AM"
I already knew where you were going with this. Claiming BS is a figment of your imagination. Facts are Facts, doesn't matter how you view or interpret them, in the end, they remain as fact.

 

[bug]

Your posts started from "Mar 24, 2017 at 2:35 AM"
I already knew where you were going with this. Claiming BS is a figment of your imagination. Facts are Facts, doesn't matter how you view or interpret them, in the end, they remain as fact.

Only, you can't show/link any of these facts.

 

[Super XP]

Only, you can't show/link any of these facts.

WCCfTECH and Guru3D have tests done comparing Vulcan and DX12. I've formalized info based on the actual reviewers and user comments that are familiar with API's. Would like to see TechPowerUp Testing too.

 

[bug]

WCCfTECH and Guru3D have tests done comparing Vulcan and DX12. I've formalized info based on the actual reviewers and user comments that are familiar with API's. Would like to see TechPowerUp Testing too.

I can't find Vulkan vs DX12 on Guru3D (probably because there's no title that supports both APIs, but what do I know?). And wtftech? Lol.
And you've also claimed Vulkan is more developer friendly. Has Guru3D tested that, too?

 

[Aquinus]

The rules of a polite discussion are that when you make a statement, you have to back it up. I have done my DD on the subject and what you say is pure BS, that's why I was asking for additional info.

Before I ditched Windows, I found that Doom ran significantly better on my 390 with Vulkan than with OpenGL. I've also found DOTA 2 to run smoother with Vulkan than OpenGL in Linux with AMDGPU-Pro.

With respect to Doom, AMD cards do get a significant bump in performance as well as more consistent frame latency. This isn't a disputed fact and there are benchmarks to back it up. Before you tell other people to do research, do your own. As far as DX12 versus Vulkan, you're not going to find anything because there are no common code-bases that handle both as far as I know. All I can tell you is that I can run Vulkan in Linux, I can't run DX12.

 

[bug]

Before I ditched Windows, I found that Doom ran significantly better on my 390 with Vulkan than with OpenGL. I've also found DOTA 2 to run smoother with Vulkan than OpenGL in Linux with AMDGPU-Pro.

With respect to Doom, AMD cards do get a significant bump in performance as well as more consistent frame latency. This isn't a disputed fact and there are benchmarks to back it up. Before you tell other people to do research, do your own. As far as DX12 versus Vulkan, you're not going to find anything because there are no common code-bases that handle both as far as I know. All I can tell you is that I can run Vulkan in Linux, I can't run DX12.

Well, that was my whole point. The guy unequivocally decreed Vulkan is better then DX12. While that may be the case, as you have noticed, there's no conclusive evidence so far. Whatever he posted (and also what you posted) is besides the point.

 

[efikkan]

Vulken is being addopted worryingly slow since announced. DXs death is not in thw horizon atm

This is exactly as expected. Vulkan is not going to challenge Direct3D 12 for the desktop anytime soon, especially with many lacking features.

Futuremark is becoming a joke with this second advertisement of Nvidia cards. Time Spy was the first one.

Why? Because it shows which cards are the best?

No, this benchmark is useless because it has almost no relation to real world performance. Even though this benchmark makes it seem like the difference between Direct3D 11 and 12 is huge, the real world difference is less than 2% except for edge cases, so I would completely ignore such benchmarks. This benchmark was made to gain some attantion in the recent "overhead" mania.

Plenty of devs went DX12 over Vulkan as DX12 was more fully featured initially. I feel this will start to drop as Vulkan has started adding missing features, as what self-serving studio would limit its audience when it can more easily target other platforms?

(For anyone about to mention OpenGL - OGL was a joke).

Well, for starters, Direct3D 12 was in planning for years before Vulkan.

Vulkan will eventually rule the mobile world like OpenGL ES does now, but the desktop is not certain.

Anyone thinking OpenGL was a joke is utterly clueless. There are some messy parts, but there is nothing wrong with the API, and it's way less bloated than Direct3D.

Explain what?
Vulcan is pretty much Mantle.
* Vulkan is derived from and built upon components of AMD' s Mantle API, which was donated by AMD to Khronos!!!

You should start by getting the facts right. Vulkan is nothing like Mantle, Vulkan is based upon SPIR-V, which was designed for OpenCL. Vulkan adapted similar syntax for the frontend like Mantle and Direct3D, but it's completely different "underneath the hood".

 

[bug]

No, this benchmark is useless because it has almost no relation to real world performance. Even though this benchmark makes it seem like the difference between Direct3D 11 and 12 is huge, the real world difference is less than 2% except for edge cases, so I would completely ignore such benchmarks. This benchmark was made to gain some attention in the recent "overhead" mania.

Well, one of the selling points of DX12 (and Vulkan) is lower API overhead. And the (lo an behold) "API overhead test" shows that the APIs deliver as expected from the point of view. Now, if some people think that performance of an API in a particular area reflects the overall performance of said API, that's hardly Futuremark's problem.

Me, I just take 3DMark as measure of the muscle a video card can flex. At the same time, I know games will behave differently. Because games also mean AI, physics, asset loading, user input, stuff beyond the scope of a focused benchmark.

 

[Aquinus]

Well, that was my whole point. The guy unequivocally decreed Vulkan is better then DX12. While that may be the case, as you have noticed, there's no conclusive evidence so far. Whatever he posted (and also what you posted) is besides the point.

Depends on what you consider "better." Being able to run Vulkan on Windows, Linux, and a significant number of newer android devices is a lot of reach, a lot more than Windows alone so, in that respect, it's definitely better. Its widespread support on android devices should also be an indicator of performance that can be attended on low power platforms which would indicate that it can efficiently utilize resources on a low power multi-core ARM device. There have been examples of how Vulkan has had gains compared to OpenGL. The examples from DX11 to DX12 don't seem to indicate as much improvement on average but, that can be due to a multitude of things. Either way, even if we considered them both to be equal in terms of performance benefit, the fact remains that the reach DX12 has doesn't even begin to compare to Vulkan given the support in the mobile market and in Linux. Just saying.

 

[bug]

Depends on what you consider "better." Being able to run Vulkan on Windows, Linux, and a significant number of newer android devices is a lot of reach, a lot more than Windows alone so, in that respect, it's definitely better. Its widespread support on android devices should also be an indicator of performance that can be attended on low power platforms which would indicate that it can efficiently utilize resources on a low power multi-core ARM device. There have been examples of how Vulkan has had gains compared to OpenGL. The examples from DX11 to DX12 don't seem to indicate as much improvement on average but, that can be due to a multitude of things. Either way, even if we considered them both to be equal in terms of performance benefit, the fact remains that the reach DX12 has doesn't even begin to compare to Vulkan given the support in the mobile market and in Linux. Just saying.

Ah, kids these days... won't read anything longer than jingle.

It's not about what I consider better, it's about what the guy who posted claim considers better. And he literally said Vulkan "offers better fps, visuals and dev friendly".
But hey, let's ignore actual contents and post whatever we like, right?

NB I agree with all that you said and I'll take a cross platform API any day, but that was never the issue here.

 

[efikkan]

Well, one of the selling points of DX12 (and Vulkan) is lower API overhead. And the (lo an behold) "API overhead test" shows that the APIs deliver as expected from the point of view. Now, if some people think that performance of an API in a particular area reflects the overall performance of said API, that's hardly Futuremark's problem.

I know very well these APIs offer somewhat lower overhead. But as you can see in tests like this, the API overhead test have no real value for end users making a choise to buy hardware. As you can see, the best AMD card performs the worst in those tests. So to have a benchmark focusing on a very narrow "feature" will make more confusion than enlightenment. This is even less useful than a benchmark focusing only on let's say memory bandwidth, etc.

Depends on what you consider "better." Being able to run Vulkan on Windows, Linux, and a significant number of newer android devices is a lot of reach, a lot more than Windows alone so, in that respect, it's definitely better. Its widespread support on android devices should also be an indicator of performance that can be attended on low power platforms which would indicate that it can efficiently utilize resources on a low power multi-core ARM device. There have been examples of how Vulkan has had gains compared to OpenGL. The examples from DX11 to DX12 don't seem to indicate as much improvement on average but, that can be due to a multitude of things. Either way, even if we considered them both to be equal in terms of performance benefit, the fact remains that the reach DX12 has doesn't even begin to compare to Vulkan given the support in the mobile market and in Linux. Just saying.

None of the games supporting two or more APIs is a true benchmark of the APIs, since "all" of them use some kind of abstraction/translation layer. E.g. Valve games and Croteam games makes all APIs behave like Direct3D 11, so any gains will be due to generic optimizations rather than the API itself.

One of the reasons why you see some OpenGL games get bigger gains is not the API itself either. If the games were written in OpenGL 4.5 with AZDO API calls there would be nearly no gains at all, but most OpenGL games are using features older than OpenGL 4.3, some even still OpenGL 3.x. That is after all the advantage and disadvantage of OpenGL; you get to choose between old and new features, which often leads to new games still using outdated API features. Even though I've experienced these new features in action, I've never seen them used in games.

At the moment Vulkan has one disadvantage over OpenGL; no support from Apple. I hope this will lead to developers abandoning their crappy platforms, and that Apple will live to regret their decision. But my greatest concern is whether Vulkan will get enough foothold on the desktop. As we know, OpenGL development has primarily been driven by the needs of professional graphics, which happens to align well with the needs of desktop gaming. If Vulkan ends up being an Android-"only" API, I'm afraid the direction will focus on embedded low-end devices, and Vulkan being dead on the desktop.

 

[Aquinus]

Ah, kids these days... won't read anything longer than jingle.

It's not about what I consider better, it's about what the guy who posted claim considers better. And he literally said Vulkan "offers better fps, visuals and dev friendly".
But hey, let's ignore actual contents and post whatever we like, right?

NB I agree with all that you said and I'll take a cross platform API any day, but that was never the issue here.

I think you missed my point. Even if you can consider the two equal in performance, everything else would indicate, (at least to me,) that Vulkan would be a better API given its reach. I'm giving DX12 the benefit of the doubt on that one but, all of the benchmarks I've seen haven't shown anything amazing going from DX11 to DX12 but, going from OpenGL to Vulkan seems to have more occurrences where performance improves by more than just a couple FPS. Also, you keep using the term "better" which doesn't really just mean performance. If you're going to talk strictly performance, don't just say "better", because that's far to generalized of a statement to derive just performance from such a statement.

None of the games supporting two or more APIs is a true benchmark of the APIs, since "all" of them use some kind of abstraction/translation layer. E.g. Valve games and Croteam games makes all APIs behave like Direct3D 11, so any gains will be due to generic optimizations rather than the API itself.

Ehh, that's only because Valve wrote a translation layer on top of the already existing use of D3D APIs. You can get the same kind of benefit in OpenGL and DX11 by doing some simple thing (that nVidia cards actually probably do under the hood,) and that's accepting the draw calls and just putting them on a queue (just like command buffers,) and executing it later when there is time or when a command is run that needs to flush the queue. This is actually the primary reason why games like Doom really see no benefit using Vulkan with nVidia cards. There is enough black-boxed optimization that is essentially doing the same thing.

One of the reasons why you see some OpenGL games get bigger gains is not the API itself either. If the games were written in OpenGL 4.5 with AZDO API calls there would be nearly no gains at all, but most OpenGL games are using features older than OpenGL 4.3, some even still OpenGL 3.x. That is after all the advantage and disadvantage of OpenGL; you get to choose between old and new features, which often leads to new games still using outdated API features. Even though I've experienced these new features in action, I've never seen them used in games.

That's not how OpenGL works. Every revision tends to add new features the OpenGL as a whole, not move already existing functionality. This is super important for backwards compatibility. If you're using version 4.3, you can basically use everything from 4.3 and earlier because 4.3 isn't going to have simple things like, making a draw call, you need to use OpenGL 1.x for that but, that doesn't change that your maximum supported version is 4.3.

At the moment Vulkan has one disadvantage over OpenGL; no support from Apple. I hope this will lead to developers abandoning their crappy platforms, and that Apple will live to regret their decision. But my greatest concern is whether Vulkan will get enough foothold on the desktop. As we know, OpenGL development has primarily been driven by the needs of professional graphics, which happens to align well with the needs of desktop gaming. If Vulkan ends up being an Android-"only" API, I'm afraid the direction will focus on embedded low-end devices, and Vulkan being dead on the desktop.

Definitely a valid concern. Apple doesn't really care because they have Metal however, that's not a huge concern to me because, Khronos made is clear that Vulkan was not intended to replace OpenGL. Vulkan was intended to be another option for applications that required high efficency and better low-level control at the expense of complexity and greater control of just about everything. OpenGL is still expected to be used in cases where the workload is light enough where using Vulkan for optimizations wouldn't yield much benefit but, would still take added time away from dev. Simply put, the right tool for the right job and Vulkan isn't always the right tool. I don't think that it will become a mobile-only API for that reason. Mobile devs would prefer OpenGL if it's going to fit the bill because, it's easier to use.

 

[efikkan]

Ehh, that's only because Valve wrote a translation layer on top of the already existing use of D3D APIs. You can get the same kind of benefit in OpenGL and DX11 by doing some simple thing (that nVidia cards actually probably do under the hood,) and that's accepting the draw calls and just putting them on a queue (just like command buffers,) and executing it later when there is time or when a command is run that needs to flush the queue. This is actually the primary reason why games like Doom really see no benefit using Vulkan with nVidia cards. There is enough black-boxed optimization that is essentially doing the same thing.

You are mixing two things here. First of all, I was talking about the game engines abstracting the APIs, which is a bad thing; since the "low level" features will no longer be beneficial when everything is implemented using an abstraction layer simulating the old API.

Secondly, you are referring to something important, but slightly off point. Nvidia did bring all the Direct3D 12 driver side optimizations to all APIs, including OpenGL. So yes, there is a reason why Nvidia get "less gains" from the new APIs.

That's not how OpenGL works. Every revision tends to add new features the OpenGL as a whole, not move already existing functionality. This is super important for backwards compatibility. If you're using version 4.3, you can basically use everything from 4.3 and earlier because 4.3 isn't going to have simple things like, making a draw call, you need to use OpenGL 1.x for that but, that doesn't change that your maximum supported version is 4.3.

Try reading my post again. OpenGL 4.4 and 4.5 focused on new features to improve overhead and more efficient batching. The older rendering methods from 4.3 and older are still available, but less efficient. Games still use the older ones for compatibility.

Definitely a valid concern. Apple doesn't really care because they have Metal however, that's not a huge concern to me because, Khronos made is clear that Vulkan was not intended to replace OpenGL. Vulkan was intended to be another option for applications that required high efficency and better low-level control at the expense of complexity and greater control of just about everything. OpenGL is still expected to be used in cases where the workload is light enough where using Vulkan for optimizations wouldn't yield much benefit but, would still take added time away from dev. Simply put, the right tool for the right job and Vulkan isn't always the right tool. I don't think that it will become a mobile-only API for that reason. Mobile devs would prefer OpenGL if it's going to fit the bill because, it's easier to use.

OpenGL will continue to live on for many years for professional graphics. Most of such applications have very long life cycles, typically 15-20 years. None of these are going to be rewritten in Vulkan, so any transition to Vulkan will take very long. These needs of professional graphics is the reason why OpenGL is so slow to deprecate stuff, even back with 3.0 the deprecation recieved a lot of protests, resulting some of the features being added back in in the following revisions.

 

[Prima.Vera]

If you're saying that but, aren't a software engineer, I have some words for you that would earn me an infraction. People bash the tool but, it's not the tool, it's how people use it. Without the JVM, I wouldn't currently have a job, and I don't write Java...

You think I'm taking you seriously after even Oracle stopped supporting Java, not to mention Apple did this decades ago, while Google is thinking to do the same thing...Common.
Apologies for the derailment.

 

[Aquinus]

You think I'm taking you seriously after even Oracle stopped supporting Java, not to mention Apple did this decades ago, while Google is thinking to do the same thing...Common.
Apologies for the derailment.

Oracle hasn't stopped supporting Java. Java 8 is slated to stop receiving updates near the end of this year but, that's only because Java 9 will probably become GA, in which case, it will receive updates. Nothing I've read indicates that Java is nearing EOL. Where do you get your information from because, I expect sources if you're going to make such grandiose claims.

 

[FordGT90Concept]

Well this doesn't make any sense. NVIDIA cards do better with Vulkan than D3D12. AMD does worse with Vulkan than D3D12. First of all, Futuremark again gives a very strong sense of favoritism. Second, AMD and/or Furmark likely needs to do some optimization for GCN.

 

[efikkan]

Well this doesn't make any sense. NVIDIA cards do better with Vulkan than D3D12. AMD does worse with Vulkan than D3D12. First of all, Futuremark again gives a very strong sense of favoritism. Second, AMD and/or Furmark likely needs to do some optimization for GCN.

No, that's exactly what they shouldn't do. A benchmark should reflect how various hardware actually performs, not how they perform with their own optimization.

 

[FordGT90Concept]

Except Futuremark likely consulted with NVIDIA and NVIDIA already did optimizations for the benchmark. AMD, as is usually the case, is left to fend for itself.

 

[efikkan]

Except Futuremark likely consulted with NVIDIA and NVIDIA already did optimizations for the benchmark. AMD, as is usually the case, is left to fend for itself.

And where is the evidence for this?