Other sources wrong, Pascal not taped out in June, potentially still in testing

[patrickjp93]

25 minutes ago, Humbug said:

Of course it could have been done without AMD, but this is how it did pan out...

 

watch below at 1:50

https://youtu.be/QF7gENO6CI8?t=1m50s

That tells you nothing. Sure, Mantle was a good stepping stone, but until the API is out in the open, there is no way to objectively compare it to Mantle and see how much of it ended up in the final product. Why do you think Vulcan has taken this long to be polished? A small part of it may very well be a Mantle copy, but you have no clue one way or the other yet. None of us do.

[Sakkura]

1 hour ago, patrickjp93 said:

You have no way of determining how much of Mantle is actually in Vulkan, and further you think Intel and Apple couldn't simply make a low-level API from the ground up without Mantle? No, it's not misleading. Mantle doesn't count for AMD, period.

It's an open standard, everyone in the industry knows Mantle makes up a large part of Vulkan.

Of course others could make a low-level API from the ground up without Mantle, it's a ridiculously childish strawman argument you're using. Microsoft made one in DirectX 12, of course others could make one too.

And yes, it is misleading to not count perhaps the most important contribution from either AMD or Nvidia, for no good reason whatsoever.

1 hour ago, patrickjp93 said:

It runs on Linux through wine with a minimal performance hit. And again it's not an open standard in any classical sense.

WINE? That's terrible, it's never going to be as good as actually running it natively.

And it is an open standard, no matter how much it makes you pout.

[patrickjp93]

7 minutes ago, Sakkura said:

It's an open standard, everyone in the industry knows Mantle makes up a large part of Vulkan.

Of course others could make a low-level API from the ground up without Mantle, it's a ridiculously childish strawman argument you're using. Microsoft made one in DirectX 12, of course others could make one too.

And yes, it is misleading to not count perhaps the most important contribution from either AMD or Nvidia, for no good reason whatsoever.

WINE? That's terrible, it's never going to be as good as actually running it natively.

And it is an open standard, no matter how much it makes you pout.

Sources? No, I'm sorry but we don't know much of anything yet other than Mantle was used as a starting point. Do you even know the definition of straw man? Because that's far from it.

 

It's not misleading because we can't yet count it and it never did go open as AMD promised.

 

Not as good as native, yes. Terrible? No.

 

If you try to implement Vulcan without being a registered member of Khronos, you can and will be sued out of house and home. It's not an open standard. It just has enough companies in the mix that you grow complacent. It's the same argument as JEDEC's technologies being open standards. Just no..

[Humbug]

16 minutes ago, patrickjp93 said:

That tells you nothing. Sure, Mantle was a good stepping stone, but until the API is out in the open, there is no way to objectively compare it to Mantle and see how much of it ended up in the final product. Why do you think Vulcan has taken this long to be polished? A small part of it may very well be a Mantle copy, but you have no clue one way or the other yet. None of us do.

none of know exactly how much of mantle is in vulkan. What we do know is what the Khronos group tells us. And they tell us that AMD's contribution to the vulkan development process was gigantic, and give them special thanks for handing over the mantle spec with no strings attached. At this point you are not arguing with me you are arguing with members of the Khronos group that have full access to mantle and Vulkan. This does not demonstrate how much of mantle will remain in vulkan, but it demonstrates how important it was in the process.

[Sakkura]

10 minutes ago, patrickjp93 said:

Sources? No, I'm sorry but we don't know much of anything yet other than Mantle was used as a starting point. Do you even know the definition of straw man? Because that's far from it.

 

It's not misleading because we can't yet count it and it never did go open as AMD promised.

 

Not as good as native, yes. Terrible? No.

 

If you try to implement Vulcan without being a registered member of Khronos, you can and will be sued out of house and home. It's not an open standard. It just has enough companies in the mix that you grow complacent. It's the same argument as JEDEC's technologies being open standards. Just no..

You just got a source. And yes, I know the definition of a strawman, and that was exactly what you just used. You deliberately and grossly misrepresented my position.

 

It is misleading because we can count on it, and it went open as AMD promised, just in a different and more comprehensive way.

 

Not as good as native, therefore definitely terrible. Have you even used WINE?

 

Vulkan is a royalty-free open standard. They will not sue you out of a damn thing, they'll thank you and maybe throw an SDK or two your way. It is definitively an open standard, you can't just claim the sky is green here.

[patrickjp93]

8 minutes ago, Humbug said:

none of know exactly how much of mantle is in vulkan. What we do know is what the Khronos group tells us. And they tell us that AMD's contribution to the vulkan development process was gigantic, and give them special thanks for handing over the mantle spec with no strings attached. At this point you are not arguing with me you are arguing with members of the Khronos group that have full access to mantle and Vulkan. This does not demonstrate how much of mantle will remain in vulkan, but it demonstrates how important it was in the process.

"Gigantic" Oh please. A twitter or Facebook post so short it tells you nothing of substance (and is there purely as a courtesy) is not enough to back your claims. I'm sorry but BS. No, I am arguing with what you are claiming Khronos has said, and you have no sources backing you.

 

5 minutes ago, Sakkura said:

You just got a source. And yes, I know the definition of a strawman, and that was exactly what you just used. You deliberately and grossly misrepresented my position.

 

It is misleading because we can count on it, and it went open as AMD promised, just in a different and more comprehensive way.

 

Not as good as native, therefore definitely terrible. Have you even used WINE?

 

Vulkan is a royalty-free open standard. They will not sue you out of a damn thing, they'll thank you and maybe throw an SDK or two your way. It is definitively an open standard, you can't just claim the sky is green here.

No, I got a piss-poor excuse for one that says less than 1/4 what was claimed. No, you fumbled your language and painted yourself into a corner

 

You can't count on it because the numbers aren't in. No, it would have been comprehensive if the code when straight up on a public Github or Subversion repository. 

 

I have used Wine. Average performance hit is 5%. That's easily tolerable.

 

It's royalty-free for Khronos members such as AMD, Nvidia, and Intel, but not for outsiders. I'm not claiming the sky is green. They sued outsiders over OpenGL implementations in years past too. It's not an open standard. The number of companies which can implement it is just large enough that you can't see the difference. If I started up a GPU company and gave Khronos the finger and implemented Vulkan, my ass would be in court within months.

[Sakkura]

1 minute ago, patrickjp93 said:

No, I got a piss-poor excuse for one that says less than 1/4 what was claimed. No, you fumbled your language and painted yourself into a corner

 

You can't count on it because the numbers aren't in. No, it would have been comprehensive if the code when straight up on a public Github or Subversion repository. 

 

I have used Wine. Average performance hit is 5%. That's easily tolerable.

 

It's royalty-free for Khronos members such as AMD, Nvidia, and Intel, but not for outsiders. I'm not claiming the sky is green. They sued outsiders over OpenGL implementations in years past too. It's not an open standard. The number of companies which can implement it is just large enough that you can't see the difference. If I started up a GPU company and gave Khronos the finger and implemented Vulkan, my ass would be in court within months.

I didn't fumble anything, and you're the one who's painted himself into a corner with demonstrably false claims - as usual. You don't accept other people's sources, and provide none of your own.

 

The numbers are in, Vulkan is completed and Mantle was an important part of it. Releasing Mantle via Vulkan is more comprehensively open, since Vulkan is under the control of an industry standards body, where Mantle would still have been under AMD's control.

 

An average performance loss of 5% isn't representative of all the times when it's a lot more than 5%, nor of the MANY situations where it doesn't work properly.

 

Vulkan is an open standard and a royalty-free API. Stop claiming the sky is green.

[Humbug]

24 minutes ago, patrickjp93 said:

If you try to implement Vulcan without being a registered member of Khronos, you can and will be sued out of house and home. It's not an open standard. It just has enough companies in the mix that you grow complacent. It's the same argument as JEDEC's technologies being open standards. Just no..

any application developer will be able to write vulkan apps without Khronos membership on any of the supported platforms.

and do you know of any major IHVs that are missing out on the action? I think that's what sakkura meant. Maybe a better way to describe it would be a cross-platform API.

[Humbug]

37 minutes ago, patrickjp93 said:

"Gigantic" Oh please. A twitter or Facebook post so short it tells you nothing of substance (and is there purely as a courtesy) is not enough to back your claims. I'm sorry but BS. No, I am arguing with what you are claiming Khronos has said, and you have no sources backing you.

I am claiming nothing regarding how much of mantle spec is identical to vulkan 1.0, I have no access to either spec.

What I said is what the khronos group members tell us regarding AMD's contribution to the Vulkan development process.

 

Since you are asking me to repeat my sources again within a few minutes I will do so for your benefit

Tom Olsen- Director, Graphics Research at ARM

John McDonald- Game Engine Developer, Valve corporation

 

watch below at 1:50

https://youtu.be/QF7gENO6CI8?t=1m50s

 

 

[patrickjp93]

7 hours ago, Humbug said:

any application developer will be able to write vulkan apps without Khronos membership on any of the supported platforms.

and do you know of any major IHVs that are missing out on the action? I think that's what sakkura meant. Maybe a better way to describe it would be a cross-platform API.

Not the same thing. Using an implementation of Vulkan is perfectly fine. Implementing a Vulkan interface for your graphics product (as in you've made it possible for Vulkan code to run on your tech) without being a member company of Khronos will land you in court staring down the wrong end of Intel's shotgun (they basically provide Khronos' lawyers). It is cross-platform as an API, but it is far from being an open standard. OpenMP is an open standard. CilkPlus is an open standard. Vulkan is not.

[Jahramika]

On 2/7/2016 at 7:57 PM, niofalpha said:

You forgot PhysX 

https://developer.nvidia.com/physx-source-gith

On 2/7/2016 at 7:57 PM, niofalpha said:

You forgot PhysX 

https://developer.nvidia.com/physx-source-github

PhysX is a proprietary realtime physics engine middleware SDK. PhysX was authored at NovodeX, an ETH Zurich spin-off. In 2004 NovodeX was acquired by Ageia, and in February 2008 Ageia was acquired by Nvidia.^[1]

[Jahramika]

21 hours ago, patrickjp93 said:

And did Ageia ever open-source it? Did Nvidia never improve upon it? DX 12 is not open you twit. Neither will Vulkan be. Both of them are managed by businesses. One is Microsoft, and the other is the Khronos Consortium. Nvidia did help co-develop HMC which has been deployed for 2.5 years in Sparc Fujitsu servers and will be launching for main system memory for x86 servers in the next 2 years with consumerville to follow later when JEDEC can't replace DDR4 with anything better. FreeSync also isn't open, and Adaptive Sync predated AMD's foray into it. Most of it was in eDP long before DP 1.2 anyway, which was ALREADY developed by VESA.

 

For Pete's sake. Nvidia isn't the devil and AMD is no saint.

 

Nvidia competes perfectly fairly. Just because it decides to MAKE ITS OWN PRODUCTS that work together as a better ecosystem than the open standards can support does not mean it's being unfair. Nvidia is doing nothing wrong. They've also proven Gameworks isn't a hindrance to AMD. Both companies do most of their driver optimization based on the binaries anyway. Just because the raw C/C++ and shader code isn't handed to AMD doesn't mean Nvidia's being any better or worse. You AMD fanchildren need to get over yourselves. Buy the best product for you, but don't go spouting this BS that AMD is somehow THE PREMIER open source company in this mess. They are far from it.

Point is it was never anything Nvidia brain stormed

[Jahramika]

Just look at amd's cpus and gpus geared towards Dx12,mantal,Vulkan, Hsa.  Nvidia and Intel single thread process that should tell you who has been pushing what and where it came from to begin with. Oh look Nvidia hired amd's top Hsa guy why? Because they never thought of it and now trying to learn to walk into the unknown or not even thought of.

[Jahramika]

11 hours ago, Humbug said:

 

Of course it could have been done without AMD, but this is how it did pan out...

 

watch below at 1:50

https://youtu.be/QF7gENO6CI8?t=1m50s

Vulkan, initially referred to as "glNext", is a low-overhead, cross-platform 3D graphics and compute API first announced at GDC 2015 by the Khronos Group.^[7][8][6] The Vulkan API was initially referred to as the "next generation OpenGL initiative" by Khronos, but use of those names were discontinued once the Vulkan name was announced.^[9] Like OpenGL, Vulkan targets high-performance realtime 3D graphics applications such as games and interactive media across all platforms, and offers higher performance and lower CPU usage, much like Direct3D 12 and Mantle. In addition to its lower CPU usage, Vulkan is also able to better distribute work amongst multiple CPU cores.^[10] Vulkan is derived from and built upon components of AMD'sMantle API, which was donated by AMD to Khronos with the intent of giving Khronos a foundation on which to begin developing a low-level API that they could standardize across the industry, much like OpenGL.^{[3][7][11][12][13][14][15]}

 

^{Just read who dinated this to the vulkan mean wher it came from. This wa setup in wikiidia by the kronous group itself make up more shit if you like}

[Jahramika]

So ya the core structure of Vulkan is mantel now kronous group is adding onto it. 

[Humbug]

19 minutes ago, Jahramika said:

snip

 

I don't see what point you are trying to make as you haven't said anything contradictory to my posts in this thread.

[Jahramika]

Ya your right got lost a little in the various posts at time!

[patrickjp93]

12 hours ago, Sakkura said:

I didn't fumble anything, and you're the one who's painted himself into a corner with demonstrably false claims - as usual. You don't accept other people's sources, and provide none of your own.

 

The numbers are in, Vulkan is completed and Mantle was an important part of it. Releasing Mantle via Vulkan is more comprehensively open, since Vulkan is under the control of an industry standards body, where Mantle would still have been under AMD's control.

 

An average performance loss of 5% isn't representative of all the times when it's a lot more than 5%, nor of the MANY situations where it doesn't work properly.

 

Vulkan is an open standard and a royalty-free API. Stop claiming the sky is green.

I proved your source was garbage by informal proof. Just because you're not learned in Jeffersonian debate and informal logic (those fallacies are actually mathematical structures by the way) doesn't mean you didn't lose when I said you did, and you did.

 

Vulkan is a completed API, but no one outside Khronos and a few select game studios has seen it, so the numbers are not yet in. Now you're moving the goal post to "important."

 

And being under control of a consortium means it's only open to members of that consortium (i.e. companies willing to pay to play). You yourself can't create an implementation of it and sell it without Khronos' explicit permission. It's not an open standard in any classical sense. OpenMP is an open standard. Cilk Plus is an open standard. Vulkan is not.

 

2 hours ago, Jahramika said:

Point is it was never anything Nvidia brain stormed

You really think the PhysX of today remotely resembles what Ageia left it as? Come on. Nvidia does not sit on its ass in software or game libraries. I think we can all agree on that.

 

2 hours ago, Jahramika said:

Just look at amd's cpus and gpus geared towards Dx12,mantal,Vulkan, Hsa.  Nvidia and Intel single thread process that should tell you who has been pushing what and where it came from to begin with. Oh look Nvidia hired amd's top Hsa guy why? Because they never thought of it and now trying to learn to walk into the unknown or not even thought of.

^Ignorance confirmed. Intel pushed multithreaded APIs before any other big company did with OpenMP all the way in 2000 when multi-socket boards were in their infancy. Intel also released OpenMP as an open standard to a committee (2006 ish which includes AMD, IBM, Cray, and others) which holds no patent (Intel released them to the open world, and GCC, Clang, and MVCC all have their own implementations (2 of those being totally open-source)) and continues to expand upon it with industry and community input from around the world.

 

Intel hasn't been pushing single-threaded at all. It's been pushing multithreaded hard, but it also knows its markets. HPC was raring to go and jumped right on board. Consumerville never did even though programming in parallel is extremely easy and efficient under OpenMP, far easier than native threading. Also, let's not forget Intel put an iGPU on their CPUs a decent time before AMD did, and Intel has helped expand OpenMP into heterogeneous acceleration of code too (still just as easy). It only requires a driver to use the graphics processors (AMD and Intel both have these already).

 

OpenMP's first released textbook guide (note the authors of the 1.0 guide were a combination of Intel, Cray, and academic engineers , AMD nowhere to be found).

http://www.intel.com/pressroom/archive/releases/2000/cn040600.htm

http://www.amazon.com/Parallel-Programming-OpenMP-Rohit-Chandra/dp/1558606718

 

This is a bit dated of a guide (still the one I recommend for beginners since no one seems to have made a new one for 4.0 and upward, forcing you to rely on openmp.org and Intel software pages to get good documentation and examples), but version 3.0 was already implemented back in 2008, long before Bulldozer even existed and people started claiming the consumer multicore revolution (but still Intel is the villain in all this pushing against multithreaded, consumers, and good-guy AMD, much troll...).

http://bisqwit.iki.fi/story/howto/openmp/

 

Heterogeneous acceleration has been in OpenMP since 2012, but no, AMD was definitely ahead and original with HSA...

https://software.intel.com/en-us/articles/three-progamming-points-to-mention-on-offloaded-code-for-intel-graphics-technology

[Sakkura]

7 hours ago, patrickjp93 said:

I proved your source was garbage by informal proof. Just because you're not learned in Jeffersonian debate and informal logic (those fallacies are actually mathematical structures by the way) doesn't mean you didn't lose when I said you did, and you did.

 

Vulkan is a completed API, but no one outside Khronos and a few select game studios has seen it, so the numbers are not yet in. Now you're moving the goal post to "important."

 

And being under control of a consortium means it's only open to members of that consortium (i.e. companies willing to pay to play). You yourself can't create an implementation of it and sell it without Khronos' explicit permission. It's not an open standard in any classical sense. OpenMP is an open standard. Cilk Plus is an open standard. Vulkan is not.

It wasn't my source, and you didn't prove anything other than your own ignorant bias. And an assertion that "you lose when I say you do" doesn't do much to support your claims of mastering debate or logic, in fact it proves you a liar on those points as so many others (where's Jim Keller again?).

 

Vulkan is completed, so the numbers are in. You're moving the goalposts while accusing me of moving the goalposts. Priceless hypocrisy.

 

It's under the control of an open standards consortium. And the API is still royalty-free. No, it's not fully free and open source, ie. you can't just take their code, add a comma, and market that as "Patrick-Volcano" or whatever. Doesn't mean it's not an open standard.

 

OpenMP is managed by a non-profit organization dedicated to open standards. Vulkan is managed by a non-profit organization dedicated to open standards. Seeing any similarities yet?

[Jahramika]

7 hours ago, patrickjp93 said:

I proved your source was garbage by informal proof. Just because you're not learned in Jeffersonian debate and informal logic (those fallacies are actually mathematical structures by the way) doesn't mean you didn't lose when I said you did, and you did.

 

Vulkan is a completed API, but no one outside Khronos and a few select game studios has seen it, so the numbers are not yet in. Now you're moving the goal post to "important."

 

And being under control of a consortium means it's only open to members of that consortium (i.e. companies willing to pay to play). You yourself can't create an implementation of it and sell it without Khronos' explicit permission. It's not an open standard in any classical sense. OpenMP is an open standard. Cilk Plus is an open standard. Vulkan is not.

 

 

You really think the PhysX of today remotely resembles what Ageia left it as? Come on. Nvidia does not sit on its ass in software or game libraries. I think we can all agree on that.

 

^Ignorance confirmed. Intel pushed multithreaded APIs before any other big company did with OpenMP all the way in 2000 when multi-socket boards were in their infancy. Intel also released OpenMP as an open standard to a committee (2006 ish which includes AMD, IBM, Cray, and others) which holds no patent (Intel released them to the open world, and GCC, Clang, and MVCC all have their own implementations (2 of those being totally open-source)) and continues to expand upon it with industry and community input from around the world.

 

Intel hasn't been pushing single-threaded at all. It's been pushing multithreaded hard, but it also knows its markets. HPC was raring to go and jumped right on board. Consumerville never did even though programming in parallel is extremely easy and efficient under OpenMP, far easier than native threading. Also, let's not forget Intel put an iGPU on their CPUs a decent time before AMD did, and Intel has helped expand OpenMP into heterogeneous acceleration of code too (still just as easy). It only requires a driver to use the graphics processors (AMD and Intel both have these already).

 

OpenMP's first released textbook guide (note the authors of the 1.0 guide were a combination of Intel, Cray, and academic engineers , AMD nowhere to be found).

http://www.intel.com/pressroom/archive/releases/2000/cn040600.htm

http://www.amazon.com/Parallel-Programming-OpenMP-Rohit-Chandra/dp/1558606718

 

This is a bit dated of a guide (still the one I recommend for beginners since no one seems to have made a new one for 4.0 and upward, forcing you to rely on openmp.org and Intel software pages to get good documentation and examples), but version 3.0 was already implemented back in 2008, long before Bulldozer even existed and people started claiming the consumer multicore revolution (but still Intel is the villain in all this pushing against multithreaded, consumers, and good-guy AMD, much troll...).

http://bisqwit.iki.fi/story/howto/openmp/

Did not rember thisconversastion being about A

Heterogeneous acceleration has been in OpenMP since 2012, but no, AMD was definitely ahead and original with HSA...

https://software.intel.com/en-us/articles/three-progamming-points-to-mention-on-offloaded-code-for-intel-graphics-technology

Well first this conversation was about Amd and Nvidia never said Intel did not contribute hardware and software tech they sure do. As far as Physic x okay Nvidia improved what Ageia started let's say it's true, still it was not there leadership in coming up with the idea. This was a conversation about Nvidias lack of developing crap of there own from scratch.  Amd has proven way better than Nvidia at it.

[patrickjp93]

1 hour ago, Sakkura said:

It wasn't my source, and you didn't prove anything other than your own ignorant bias. And an assertion that "you lose when I say you do" doesn't do much to support your claims of mastering debate or logic, in fact it proves you a liar on those points as so many others (where's Jim Keller again?).

 

Vulkan is completed, so the numbers are in. You're moving the goalposts while accusing me of moving the goalposts. Priceless hypocrisy.

 

It's under the control of an open standards consortium. And the API is still royalty-free. No, it's not fully free and open source, ie. you can't just take their code, add a comma, and market that as "Patrick-Volcano" or whatever. Doesn't mean it's not an open standard.

 

OpenMP is managed by a non-profit organization dedicated to open standards. Vulkan is managed by a non-profit organization dedicated to open standards. Seeing any similarities yet?

It's only royalty free if you're a member of the standards consortium. You can freely use their implementations. You cannot make your own technology Vulkan-compliant and have that be a selling feature without being a member of that standards consortium, or do we really have to go back to the point Matrox was being sued over implementing OpenGL?

 

OpenMP is fully open. Vulkan is not.

[patrickjp93]

1 hour ago, Jahramika said:

Well first this conversation was about Amd and Nvidia never said Intel did not contribute hardware and software tech they sure do. As far as Physic x okay Nvidia improved what Ageia started let's say it's true, still it was not there leadership in coming up with the idea. This was a conversation about Nvidias lack of developing crap of there own from scratch.  Amd has proven way better than Nvidia at it.

AMD hasn't proven crap about it yet. And now they're pawning off doing the work to the Open Source community. And I see you failed to respond to the rest of my post which has effectively ripped your claims about Intel's single-threaded conspiracy to shreds...

 

Yet over on wccf you're still spouting this crap.

[Jahramika]

1 hour ago, patrickjp93 said:

AMD hasn't proven crap about it yet. And now they're pawning off doing the work to the Open Source community. And I see you failed to respond to the rest of my post which has effectively ripped your claims about Intel's single-threaded conspiracy to shreds...

 

Yet over on wccf you're still spouting this crap.

Sorry if I did not respond I missed your question.  The whole point of open source is so all can be evolved that's not posing off in my eyes that including all so all pitch in no?

[Jahramika]

1 hour ago, patrickjp93 said:

AMD hasn't proven crap about it yet. And now they're pawning off doing the work to the Open Source community. And I see you failed to respond to the rest of my post which has effectively ripped your claims about Intel's single-threaded conspiracy to shreds...

 

Yet over on wccf you're still spouting this crap.

Hey I know Amd is not everything and maybe at times I do spew some crap but Amd deserves more credit than the beating they get just because there gpus are a tad slower. Ya cpus need work at the sametime my furyx and 9590 do all I need in every game I play or spreadsheet I create. That's what most home users do with there computers.

[Sakkura]

1 hour ago, patrickjp93 said:

It's only royalty free if you're a member of the standards consortium. You can freely use their implementations. You cannot make your own technology Vulkan-compliant and have that be a selling feature without being a member of that standards consortium, or do we really have to go back to the point Matrox was being sued over implementing OpenGL?

 

OpenMP is fully open. Vulkan is not.

What lawsuit? As far as I recall, Matrox initially had no OpenGL support, then crappy support, then ended up in the OpenGL consortium when they were more or less done in the gaming graphics market.