Most of the curved screens have a pretty low vertical number of pixels rendering them pretty useless for anything but gaming through a letterbox opening.
Which ones have that? The majority of the wide curved displays have a resolution of 3440x1440 which is about 30% more than the 1080p that majority of gamers (~63%) use and is actually in the 98th percentile of vertical resolution for gaming users.
FLOPS is what will make 1'000+ "old" games i have in my Steam account work faster, i doubt all those old games will suddenly start using those tensor/rt cores, and even new ones will probably start using them in few years not in 3 days...
It seems odd that I have to point this out but i think it's pretty clear that this product is the wrong product for you then isn't it? If what you want is higher FLOPS and an even Price:FLOPS ratio this is not the product for that because what you would be paying for, in part, is RT and tensor cores which you don't seem to want.
The first game to support raytracing, battlefield v, releases a week before the hardware arrives. No need to wait a year, you can use it immediately.
Nice! Well the answer is that if you want to play the latest titles that feature new technology and actually use that technology (Battlefield V doesn't actually require and RTX GPU) then that's how much it costs.
Well nVidia can tout the benefits all they like but these are effectively halo features from their AI/workstation cards and not available at all on any lesser GPUs. So how many games will create unique effects that'll only work on extremely high end 2018+ cards?
Those features are exposed to applications via APIs like DirectX and Vulkan. DirectX's DXR for example has a layer to support using it without hardware support and those features already have application support via DXR in Unity, Unreal and some of EA's engines. Being able to switch from the rasterizer to the raytracer in a game would be pretty awesome though.
Errr...ok, that's a weird way to describe raytracing. What games are doing are raytraced lighting (more to the point for this product, what are doing raytracing with AI inferencing - which is what all those tensor cores are used for)? I think you'll find pretty much all games are rasterized, not raytraced and any raytracing is done in offline steps to build things like lightmaps.
so given that I said I wanted to play games with it, why would you tell me these cards aren't for me?
Because - other than your misunderstanding of raytracing - you said you don't want to pay 54% more for a new card so clearly it's not for you.
It's unclear what the tensor cores would be used for in games aside from antialiasing and raytracing denoising.
They can be used for inferencing, if you take a look at how they've been training neural networks with raytraced images that's a big part of how this generation can do realtime raytracing.
So, out of interest, who would use a card like this?
Gamers and graphics enthusiasts who want the best performance despite paying for things they don't need right now, i.e. more concerned about performance than price value. Developers building the next generation of games and 3d applications.
Well I don't think I would have to say it, isn't it clear from the announcement and all the articles and documentation that yes indeed it is very very different?
and will not lower the price of the 1080?
I'm not saying that, I have no idea what pricing effects will occur.
Tell me, why should gamers care about tensor cores? And I doubt the RT cores will be used much outside of a couple of demos for a long long time.
Because the next generation of games will utilise them, developers need silicon and APIs to build these new games. Very often new products include new features that you can't effectively utilise on day one but you'll pay for anyway because they need to solve the chicken and egg problem, if you think this is supposed to just be a faster 1080 then you've misinterpreted it. That product simply doesn't exist regardless of how much you want it but it does leave the market open for a competitor like AMD to fill that gap.
I want a card that will be capable of any game on the market for the next year or so and can do VR.
Well then a multi-GPU setup with the current generation is probably appropriate. While I'm sure the RTX GPUs will be capable they include things you likely won't need for the next year of games like cores devoted to tensor operations and cores designed for raytracing operations and so by buying one you would be paying for things you don't need. Of course if no products exist on the market then no company is going to produce software to utilise those features so this is how you solve the chicken and egg problem so you either don't buy it at all or you accept that if you do you're going to be paying for things you don't need right now.
Again it's pretty clear this isn't the product for you.
Because I don't want to spend 54% more for a new card? I thought 1080's were already over priced.
So buy an AMD GPU then. The thing you seem to be failing to understand here is this isn't "a 1080, but faster", it's a GPU focussed on raytracing. If all you want is a new card that's like the current crop but faster and priced accordingly then this is not the product for you.
The RTX 2080Ti has 19% more FLOPS than the 1080Ti... and costs 54% more money.
NVIDIA emphasized the new raytracing performance, presumably to deflect that fact.
Or perhaps it's not just about FLOPS? The 1080Ti lacks RT cores and Tensor cores but if raw FLOPS is all you care about then yes there are better value options.
So you admit that nothing has changed as far as OpenGL is concerned between High Sierra and Mojave and thusly that both you and Autodesk are acting deceitful in you claims.
Don't be such a drama queen, they aren't supporting it going forward, probably because updates will utilise modern features of OpenGL that Apple's platforms simply don't have. Moreover in Mojave OpenGL is deprecated, meaning now it could be removed at any time.
What exactly are you getting so bent out of shape for? This 'deceit' is some grand conspiracy for what exactly? Maybe the software will continue to work, maybe it won't but it's nonsense for a vendor to advertise support for a platform where they depend on a deprecated feature.
So if you want to use any modern features of OpenGL (compute shaders, atomics, SSBOs, etc) you can't do that on Apple's platform, though the hardware does actually support it so if you install Windows or Linux on your Mac you can use the later versions or even Vulkan.
And when you use Vulkan, you can use MoltenVK - The End.
Yes if they were to rewrite the renderer in Vulkan it could certainly support macOS via MoltenVK, but what we're talking about here is an OpenGL application, not a Vulkan one.
"Sorry, we are using DirectX" is the #1 answer that developers give on why they won't port their game to macOS or Linux.
Specifically who gives that answer? Porting between platforms by maintaining (among other things) multiple renderers is par for the course in game development, how do you think games get ported between consoles?
But Metal (not even considering Metal 2) is over twice as efficient as Vulkan; so why would Apple Devs. want to give up all that extra performance, just to use a VASTLY inferior API?
I can't find any benchmarks to back up that claim at all, well not even any benchmarks, just any information at all that backs up that claim. Also in that post there seems to be a comparison of some broad term of "efficiency" to something with a more specific term "draw call" (draw calls being just one bit of what you do with a graphics API).
Whatever it is that those terms are actually referring to when you look at what it translates to in terms of performance in the real world it's not that much: Metal vs OpenGL.
Suggesting it's a vastly inferior API based on comparison of two different things with no numbers, evidence, benchmarks or justification at all is a bit of a stretch.
Yeah, I know, it's only available PRACTICALLY EVERYWHERE. Except, now, on Apple.
Except, now, it's actually still on Apple. Are you as stupid or as deceitful as Autodesk?
That's a bit disingenuous, the latest version available on Apple is version 4.1 from 8 years ago. So if you want to use any modern features of OpenGL (compute shaders, atomics, SSBOs, etc) you can't do that on Apple's platform, though the hardware does actually support it so if you install Windows or Linux on your Mac you can use the later versions or even Vulkan.
What's interesting about this to me is that development for platforms that never supported OpenGL continues. The only way they work on windows is because nVidia and AMD implement OpenGL for them.
No, platform vendors don't have to support the API they just have to allow the hardware vendors to supply the API to drive that hardware. In this case the API is OpenGL and Apple actively prevents vendors from providing an API to interface with their hardware on Apple's platforms.
Khronos didn't write entire Vulkan API from scratch, it's built upon AMD's Mantle API. You talked about many things in this topic, but somehow you forgot/didn't know that fact, really?
Why are you talking as though AMD and Khronos are separate things? AMD is a promoter member of the Khronos Group...Apple is also a promoter member of the Khronos Group.
Yeah so what. AutoCAD is a mature product. It just needs to keep working on the existing API.
But the point is it isn't really the existing API, it's the API from 2010. Apple went 5 years with no updates to 3D graphics APIs on their platform before they even introduced Metal on the desktop.
Why would they need to do that? OpenGL still works on macOS. The library didn't get removed. The headers didn't even get removed! The only thing that changed was that they added __attribute__((deprecated("use Metal instead"))) to the headers.
Because OpenGL on macOS hasn't been updated since the version released in 2010. You cannot use any of the features introduced in the past 7 years in OpenGL on a Mac because there is no ICD mechanism and Apple are not updating their implementation. Want to use atomics, compute shaders, SSBOs, etc... in your product? Well it won't run on macOS.
What has been highlighted though is that you can use BootCamp because the hardware is actually capable, macOS just doesn't expose those OpenGL features where Windows and Linux do.
I just pointed out a really significant one to you, OpenGL compute shaders, along with atomics and other buffer types. Vulkan was developed to allow for lower level access to GPU functionality so optimizations that were usually restricted to the driver could be done by application developers rather than the current state of large drivers containing a lot of application-specific optimization code.
They've made some optimizations in recent years but they haven't really advanced it.
Optimization is the whole reason for lower-level APIs like DX12, Vulkan and Metal, I don't know what you mean by "advanced it" but can you explain how Vulkan and/or Metal have done that other than optimization?
Slashdot Mirror
VRrequires a stereo image, so why wouldn't it do better with two separate GPUs rendering left and right eye frames?
Because the Pascal architecture introduced single-pass stereo rendering.
Most of the curved screens have a pretty low vertical number of pixels rendering them pretty useless for anything but gaming through a letterbox opening.
Which ones have that? The majority of the wide curved displays have a resolution of 3440x1440 which is about 30% more than the 1080p that majority of gamers (~63%) use and is actually in the 98th percentile of vertical resolution for gaming users.
FLOPS is what will make 1'000+ "old" games i have in my Steam account work faster, i doubt all those old games will suddenly start using those tensor/rt cores, and even new ones will probably start using them in few years not in 3 days ...
It seems odd that I have to point this out but i think it's pretty clear that this product is the wrong product for you then isn't it? If what you want is higher FLOPS and an even Price:FLOPS ratio this is not the product for that because what you would be paying for, in part, is RT and tensor cores which you don't seem to want.
The first game to support raytracing, battlefield v, releases a week before the hardware arrives. No need to wait a year, you can use it immediately.
Nice! Well the answer is that if you want to play the latest titles that feature new technology and actually use that technology (Battlefield V doesn't actually require and RTX GPU) then that's how much it costs.
Well nVidia can tout the benefits all they like but these are effectively halo features from their AI/workstation cards and not available at all on any lesser GPUs. So how many games will create unique effects that'll only work on extremely high end 2018+ cards?
Those features are exposed to applications via APIs like DirectX and Vulkan. DirectX's DXR for example has a layer to support using it without hardware support and those features already have application support via DXR in Unity, Unreal and some of EA's engines. Being able to switch from the rasterizer to the raytracer in a game would be pretty awesome though.
Ray tracing is for lighting in games
Errr...ok, that's a weird way to describe raytracing. What games are doing are raytraced lighting (more to the point for this product, what are doing raytracing with AI inferencing - which is what all those tensor cores are used for)? I think you'll find pretty much all games are rasterized, not raytraced and any raytracing is done in offline steps to build things like lightmaps.
so given that I said I wanted to play games with it, why would you tell me these cards aren't for me?
Because - other than your misunderstanding of raytracing - you said you don't want to pay 54% more for a new card so clearly it's not for you.
It's unclear what the tensor cores would be used for in games aside from antialiasing and raytracing denoising.
They can be used for inferencing, if you take a look at how they've been training neural networks with raytraced images that's a big part of how this generation can do realtime raytracing.
So, out of interest, who would use a card like this?
Gamers and graphics enthusiasts who want the best performance despite paying for things they don't need right now, i.e. more concerned about performance than price value. Developers building the next generation of games and 3d applications.
You are saying this card is different some how
Well I don't think I would have to say it, isn't it clear from the announcement and all the articles and documentation that yes indeed it is very very different?
and will not lower the price of the 1080?
I'm not saying that, I have no idea what pricing effects will occur.
Tell me, why should gamers care about tensor cores? And I doubt the RT cores will be used much outside of a couple of demos for a long long time.
Because the next generation of games will utilise them, developers need silicon and APIs to build these new games. Very often new products include new features that you can't effectively utilise on day one but you'll pay for anyway because they need to solve the chicken and egg problem, if you think this is supposed to just be a faster 1080 then you've misinterpreted it. That product simply doesn't exist regardless of how much you want it but it does leave the market open for a competitor like AMD to fill that gap.
I want a card that will be capable of any game on the market for the next year or so and can do VR.
Well then a multi-GPU setup with the current generation is probably appropriate. While I'm sure the RTX GPUs will be capable they include things you likely won't need for the next year of games like cores devoted to tensor operations and cores designed for raytracing operations and so by buying one you would be paying for things you don't need. Of course if no products exist on the market then no company is going to produce software to utilise those features so this is how you solve the chicken and egg problem so you either don't buy it at all or you accept that if you do you're going to be paying for things you don't need right now.
Again it's pretty clear this isn't the product for you.
Because I don't want to spend 54% more for a new card? I thought 1080's were already over priced.
So buy an AMD GPU then. The thing you seem to be failing to understand here is this isn't "a 1080, but faster", it's a GPU focussed on raytracing. If all you want is a new card that's like the current crop but faster and priced accordingly then this is not the product for you.
The RTX 2080Ti has 19% more FLOPS than the 1080Ti... and costs 54% more money. NVIDIA emphasized the new raytracing performance, presumably to deflect that fact.
Or perhaps it's not just about FLOPS? The 1080Ti lacks RT cores and Tensor cores but if raw FLOPS is all you care about then yes there are better value options.
So you admit that nothing has changed as far as OpenGL is concerned between High Sierra and Mojave and thusly that both you and Autodesk are acting deceitful in you claims.
Don't be such a drama queen, they aren't supporting it going forward, probably because updates will utilise modern features of OpenGL that Apple's platforms simply don't have. Moreover in Mojave OpenGL is deprecated, meaning now it could be removed at any time.
What exactly are you getting so bent out of shape for? This 'deceit' is some grand conspiracy for what exactly? Maybe the software will continue to work, maybe it won't but it's nonsense for a vendor to advertise support for a platform where they depend on a deprecated feature.
So if you want to use any modern features of OpenGL (compute shaders, atomics, SSBOs, etc) you can't do that on Apple's platform, though the hardware does actually support it so if you install Windows or Linux on your Mac you can use the later versions or even Vulkan.
And when you use Vulkan, you can use MoltenVK - The End.
Yes if they were to rewrite the renderer in Vulkan it could certainly support macOS via MoltenVK, but what we're talking about here is an OpenGL application, not a Vulkan one.
"Sorry, we are using DirectX" is the #1 answer that developers give on why they won't port their game to macOS or Linux.
Specifically who gives that answer? Porting between platforms by maintaining (among other things) multiple renderers is par for the course in game development, how do you think games get ported between consoles?
But Metal (not even considering Metal 2) is over twice as efficient as Vulkan; so why would Apple Devs. want to give up all that extra performance, just to use a VASTLY inferior API?
https://www.reddit.com/r/Andro...
I can't find any benchmarks to back up that claim at all, well not even any benchmarks, just any information at all that backs up that claim. Also in that post there seems to be a comparison of some broad term of "efficiency" to something with a more specific term "draw call" (draw calls being just one bit of what you do with a graphics API).
Whatever it is that those terms are actually referring to when you look at what it translates to in terms of performance in the real world it's not that much: Metal vs OpenGL.
Suggesting it's a vastly inferior API based on comparison of two different things with no numbers, evidence, benchmarks or justification at all is a bit of a stretch.
"It's not like OpenGL is much of a standard"
Yeah, I know, it's only available PRACTICALLY EVERYWHERE. Except, now, on Apple.
Except, now, it's actually still on Apple. Are you as stupid or as deceitful as Autodesk?
That's a bit disingenuous, the latest version available on Apple is version 4.1 from 8 years ago. So if you want to use any modern features of OpenGL (compute shaders, atomics, SSBOs, etc) you can't do that on Apple's platform, though the hardware does actually support it so if you install Windows or Linux on your Mac you can use the later versions or even Vulkan.
What's interesting about this to me is that development for platforms that never supported OpenGL continues. The only way they work on windows is because nVidia and AMD implement OpenGL for them.
No, platform vendors don't have to support the API they just have to allow the hardware vendors to supply the API to drive that hardware. In this case the API is OpenGL and Apple actively prevents vendors from providing an API to interface with their hardware on Apple's platforms.
Khronos didn't write entire Vulkan API from scratch, it's built upon AMD's Mantle API. You talked about many things in this topic, but somehow you forgot/didn't know that fact, really?
Why are you talking as though AMD and Khronos are separate things? AMD is a promoter member of the Khronos Group...Apple is also a promoter member of the Khronos Group.
Yeah so what. AutoCAD is a mature product. It just needs to keep working on the existing API.
But the point is it isn't really the existing API, it's the API from 2010. Apple went 5 years with no updates to 3D graphics APIs on their platform before they even introduced Metal on the desktop.
Why would they need to do that? OpenGL still works on macOS. The library didn't get removed. The headers didn't even get removed! The only thing that changed was that they added __attribute__((deprecated("use Metal instead"))) to the headers.
Because OpenGL on macOS hasn't been updated since the version released in 2010. You cannot use any of the features introduced in the past 7 years in OpenGL on a Mac because there is no ICD mechanism and Apple are not updating their implementation. Want to use atomics, compute shaders, SSBOs, etc... in your product? Well it won't run on macOS.
What has been highlighted though is that you can use BootCamp because the hardware is actually capable, macOS just doesn't expose those OpenGL features where Windows and Linux do.
It's not like OpenGL is much of a standard.
How do you figure?
A large number of apps and games on Windows don't use OpenGL anyway
And a large number of the do, the point is that because OpenGL is a standard even the ones that don't use it can use it.
Autodesk isn't even a member of the Khronos Group...at any level at all, they have absolutely nothing whatsoever to do with it.
Vulkan's main benefit is obviously that it's a cross platform industry standard. However it's not very capable yet
Can you be more specific about that?
They've made some optimizations in recent years but they haven't really advanced it.
Optimization is the whole reason for lower-level APIs like DX12, Vulkan and Metal, I don't know what you mean by "advanced it" but can you explain how Vulkan and/or Metal have done that other than optimization?