Why Two 4GB GPUs Isn't Necessarily 8GB
We're trying something new here at PC Perspective. Some topics are fairly difficult to explain cleanly without accompanying images. We also like to go fairly deep into specific topics, so we're hoping that we can provide educational cartoons that explain these issues.
This pilot episode is about load-balancing and memory management in multi-GPU configurations. There seems to be a lot of confusion around what was (and was not) possible with DirectX 11 and OpenGL, and even more confusion about what DirectX 12, Mantle, and Vulkan allow developers to do. It highlights three different load-balancing algorithms, and even briefly mentions what LucidLogix was attempting to accomplish almost ten years ago.
If you like it, and want to see more, please share and support us on Patreon. We're putting this out not knowing if it's popular enough to be sustainable. The best way to see more of this is to share!
Subject: General Tech | February 18, 2016 - 02:16 PM | Ken Addison
Tagged: x16 LTE, vulkan, video, ssd, Samsung, qualcomm, podcast, pb328q, opengl, nvidia, micron, Khronos, gtx 950, asus, apple, 840 evo, 750ti, 750 evo, 3d nand
PC Perspective Podcast #387 - 02/18/2016
Join us this week as we discuss the ASUS PB328Q, Samsung 750 EVO SSD, the release of Vulkan and more!
The URL for the podcast is: http://pcper.com/podcast - Share with your friends!
- iTunes - Subscribe to the podcast directly through the Store (audio only)
- RSS - Subscribe through your regular RSS reader (audio only)
- MP3 - Direct download link to the MP3 file
Hosts: Ryan Shrout, Jeremy Hellstrom, Josh Walrath, and Allyn Malventano
Program length: 1:34:18
Week in Review:
0:35:00 This episode of the PC Perspective Podcast is brought to you by Audible, the world's leading provider of audiobooks with more than 180,000 downloadable titles across all types of literature including fiction, nonfiction, and periodicals. For your free audiobook, go to audible.com/pcper
News items of interest:
1:07:00 This episode of PC Perspective Podcast is brought to you by Braintree. Even the best mobile app won’t work without the right payments API. That’s where the Braintree v.0 SDK comes in. One amazingly simple integration gives you every way to pay. Try out the sandbox and see for yourself at braintreepayments.com/pcper
Hardware/Software Picks of the Week
Subject: Graphics Cards | December 29, 2015 - 07:05 AM | Scott Michaud
Tagged: opengl, mesa, linux, Intel
The open-source driver for Intel is known to be a little behind on Linux. Because Intel does not provide as much support as they should, the driver still does not support OpenGL 4.0, although that is changing. One large chunk of that API is support for tessellation, which comes from DirectX 11, and recent patches are adding it for supported hardware. Proprietary drivers exist, at least for some platforms, but they have their own issues.
According to the Phoronix article, once the driver succeeds in supporting OpenGL 4.0, it will not be too long to open the path to 4.2. Tessellation is a huge hurdle, partially because it involves adding two whole shading stages to the rendering pipeline. Broadwell GPUs were recently added, but a patch that was committed yesterday will expand that to Ivy Bridge and Haswell. On Windows, Intel is far ahead -- pushing OpenGL 4.4 for Skylake-based graphics, although that platform only has proprietary drivers. AMD and NVIDIA are up to OpenGL 4.5, which is the latest version.
While all of this is happening, Valve is working on an open-source Vulkan driver for Intel on Linux. This API will be released adjacent to OpenGL, and is built for high-performance graphics and compute. (Note that OpenCL is more sophisticated than Vulkan "1.0" will be on the compute side of things.) As nice as it would be to get high-end OpenGL support, especially for developers who want a more simplified structure to communicate to GPUs with, Vulkan will probably be the API that matters most for high-end video games. But again, that only applies to games that are developed for it.
Subject: Graphics Cards | December 21, 2015 - 01:04 PM | Jeremy Hellstrom
Tagged: GameWorks VR 1.1, nvidia, Oculus, opengl, vive
If you are blessed with the good fortune of already having a VR headset and happen to be running an NVIDIA GPU then there is a new driver you want to grab as soon as you can. The driver includes a new OpenGL extension that enables NVIDIA SLI support for OpenGL apps that display on an Oculus or Vive. NVIDIA's PR suggests you can expect your performance to improve 1.7 times, not quite doubling but certainly offering a noticeable performance improvement. The update is for both GeForce and Quadro cards.
They describe how the update will generate images in their blog post here. They imply that the changes to your code in order to benefit from this update will be minimal and it will also reduce the CPU overhead required to display the images for the right and left eye. Read on if you are interested in the developer side of this update, otherwise download your new driver and keep an eye out for application updates that enable support for SLI in VR.
Subject: Graphics Cards | October 7, 2015 - 07:01 AM | Scott Michaud
Tagged: opengl, metal, apple
Ars Technica took it upon themselves to benchmark Metal in the latest OSX El Capitan release. Even though OpenGL on Mac OSX is not considered to be on par with its Linux counterparts, which is probably due to the driver situation until recently, it pulls ahead of Metal in many situations.
Image Credit: Ars Technica
Unlike the other graphics APIs, Metal uses the traditional binding model. Basically, you have a GPU object that you attach your data to, then call one of a handful of “draw” functions to signal the driver. DirectX 12, Vulkan, and Mantle, on the other hand, treat work like commands on queues. The latter model works better in multi-core environments, and it aligns with GPU compute APIs, but the former is easier to port OpenGL and DirectX 11 applications to.
Ars Technica notes that faster GPUs, such as the NVIDIA GeForce GTX 680MX, show higher gains than slower ones. Their “best explanation” is that “faster GPUs can offload more work from the CPU”. That is pretty much true, yes. The new APIs are designed to keep GPUs loaded and working as much as possible, because they really do sit around doing nothing a lot. If you are able to keep a GPU loaded, because it can't accept much load in the first place, then there is little benefit to decreasing CPU load or spreading out across multiple cores.
Granted, there are many ways that benchmarks like these could be incorrectly used. I'll assume that Ars Technica and GFXBench are not making any simple mistakes, though, but it's good to be critical just in case.
Subject: Graphics Cards, Processors, Mobile, Shows and Expos | August 10, 2015 - 09:01 AM | Scott Michaud
Tagged: vulkan, spir, siggraph 2015, Siggraph, opengl sc, OpenGL ES, opengl, opencl, Khronos
When the Khronos Group announced Vulkan at GDC, they mentioned that the API is coming this year, and that this date is intended to under promise and over deliver. Recently, fans were hoping that it would be published at SIGGRAPH, which officially begun yesterday. Unfortunately, Vulkan has not released. It does hold a significant chunk of the news, however. Also, it's not like DirectX 12 is holding a commanding lead at the moment. The headers were public only for a few months, and the code samples are less than two weeks old.
The organization made announcements for six products today: OpenGL, OpenGL ES, OpenGL SC, OpenCL, SPIR, and, as mentioned, Vulkan. They wanted to make their commitment clear, to all of their standards. Vulkan is urgent, but some developers will still want the framework of OpenGL. Bind what you need to the context, then issue a draw and, if you do it wrong, the driver will often clean up the mess for you anyway. The briefing was structure to be evident that it is still in their mind, which is likely why they made sure three OpenGL logos greeted me in their slide deck as early as possible. They are also taking and closely examining feedback about who wants to use Vulkan or OpenGL, and why.
As for Vulkan, confirmed platforms have been announced. Vendors have committed to drivers on Windows 7, 8, 10, Linux, including Steam OS, and Tizen (OSX and iOS are absent, though). Beyond all of that, Google will accept Vulkan on Android. This is a big deal, as Google, despite its open nature, has been avoiding several Khronos Group standards. For instance, Nexus phones and tablets do not have OpenCL drivers, although Google isn't stopping third parties from rolling it into their devices, like Samsung and NVIDIA. Direct support of Vulkan should help cross-platform development as well as, and more importantly, target the multi-core, relatively slow threaded processors of those devices. This could even be of significant use for web browsers, especially in sites with a lot of simple 2D effects. Google is also contributing support from their drawElements Quality Program (dEQP), which is a conformance test suite that they bought back in 2014. They are going to expand it to Vulkan, so that developers will have more consistency between devices -- a big win for Android.
While we're not done with Vulkan, one of the biggest announcements is OpenGL ES 3.2 and it fits here nicely. At around the time that OpenGL ES 3.1 brought Compute Shaders to the embedded platform, Google launched the Android Extension Pack (AEP). This absorbed OpenGL ES 3.1 and added Tessellation, Geometry Shaders, and ASTC texture compression to it. It was also more tension between Google and cross-platform developers, feeling like Google was trying to pull its developers away from Khronos Group. Today, OpenGL ES 3.2 was announced and includes each of the AEP features, plus a few more (like “enhanced” blending). Better yet, Google will support it directly.
Next up are the desktop standards, before we finish with a resurrected embedded standard.
OpenGL has a few new extensions added. One interesting one is the ability to assign locations to multi-samples within a pixel. There is a whole list of sub-pixel layouts, such as rotated grid and Poisson disc. Apparently this extension allows developers to choose it, as certain algorithms work better or worse for certain geometries and structures. There were probably vendor-specific extensions for a while, but now it's a ratified one. Another extension allows “streamlined sparse textures”, which helps manage data where the number of unpopulated entries outweighs the number of populated ones.
OpenCL 2.0 was given a refresh, too. It contains a few bug fixes and clarifications that will help it be adopted. C++ headers were also released, although I cannot comment much on it. I do not know the state that OpenCL 2.0 was in before now.
And this is when we make our way back to Vulkan.
SPIR-V, the code that runs on the GPU (or other offloading device, including the other cores of a CPU) in OpenCL and Vulkan is seeing a lot of community support. Projects are under way to allow developers to write GPU code in several interesting languages: Python, .NET (C#), Rust, Haskell, and many more. The slide lists nine that Khronos Group knows about, but those four are pretty interesting. Again, this is saying that you can write code in the aforementioned languages and have it run directly on a GPU. Curiously missing is HLSL, and the President of Khronos Group agreed that it would be a useful language. The ability to cross-compile HLSL into SPIR-V means that shader code written for DirectX 9, 10, 11, and 12 could be compiled for Vulkan. He expects that it won't take long for a project to start, and might already be happening somewhere outside his Google abilities. Regardless, those who are afraid to program in the C-like GLSL and HLSL shading languages might find C# and Python to be a bit more their speed, and they seem to be happening through SPIR-V.
As mentioned, we'll end on something completely different.
For several years, the OpenGL SC has been on hiatus. This group defines standards for graphics (and soon GPU compute) in “safety critical” applications. For the longest time, this meant aircraft. The dozens of planes (which I assume meant dozens of models of planes) that adopted this technology were fine with a fixed-function pipeline. It has been about ten years since OpenGL SC 1.0 launched, which was based on OpenGL ES 1.0. SC 2.0 is planned to launch in 2016, which will be based on the much more modern OpenGL ES 2 and ES 3 APIs that allow pixel and vertex shaders. The Khronos Group is asking for participation to direct SC 2.0, as well as a future graphics and compute API that is potentially based on Vulkan.
The devices that this platform intends to target are: aircraft (again), automobiles, drones, and robots. There are a lot of ways that GPUs can help these devices, but they need a good API to certify against. It needs to withstand more than an Ouya, because crashes could be much more literal.
Who Should Care? Thankfully, Many People
The Khronos Group has made three announcements today: Vulkan (their competitor to DirectX 12), OpenCL 2.1, and SPIR-V. Because there is actually significant overlap, we will discuss them in a single post rather than splitting them up. Each has a role in the overall goal to access and utilize graphics and compute devices.
Before we get into what everything is and does, let's give you a little tease to keep you reading. First, Khronos designs their technologies to be self-reliant. As such, while there will be some minimum hardware requirements, the OS pretty much just needs to have a driver model. Vulkan will not be limited to Windows 10 and similar operating systems. If a graphics vendor wants to go through the trouble, which is a gigantic if, Vulkan can be shimmed into Windows 8.x, Windows 7, possibly Windows Vista despite its quirks, and maybe even Windows XP. The words “and beyond” came up after Windows XP, but don't hold your breath for Windows ME or anything. Again, the further back in Windows versions you get, the larger the “if” becomes but at least the API will not have any “artificial limitations”.
Outside of Windows, the Khronos Group is the dominant API curator. Expect Vulkan on Linux, Mac, mobile operating systems, embedded operating systems, and probably a few toasters somewhere.
On that topic: there will not be a “Vulkan ES”. Vulkan is Vulkan, and it will run on desktop, mobile, VR, consoles that are open enough, and even cars and robotics. From a hardware side, the API requires a minimum of OpenGL ES 3.1 support. This is fairly high-end for mobile GPUs, but it is the first mobile spec to require compute shaders, which are an essential component of Vulkan. The presenter did not state a minimum hardware requirement for desktop GPUs, but he treated it like a non-issue. Graphics vendors will need to be the ones making the announcements in the end, though.
Subject: Graphics Cards, Shows and Expos | February 4, 2015 - 03:33 AM | Scott Michaud
Tagged: OpenGL Next, opengl, glnext, gdc 2015, GDC
The first next-gen, released graphics API was Mantle, which launched a little while after Battlefield 4, but the SDK is still invite-only. The DirectX 12 API quietly launched with the recent Windows 10 Technical Preview, but no drivers, SDK, or software (that we know about) are available to the public yet. The Khronos Group has announced their project, and that's about it currently.
According to Develop Magazine, the GDC event listing, and participants, the next OpenGL (currently called “glNext initiative”) will be unveiled at GDC 2015. The talk will be presented by Valve, but it will also include Epic Games, who was closely involved in DirectX 12 with Unreal Engine, Oxide Games and EA/DICE, who were early partners with AMD on Mantle, and Unity, who recently announced support for DirectX 12 when it launches with Windows 10. Basically, this GDC talk includes almost every software developer that came out in early support of either DirectX 12 or Mantle, plus Valve. Off the top of my head, I can only think of FutureMark as unlisted. On the other hand, while they will obviously have driver support from at least one graphics vendor, none are listed. Will we see NVIDIA? Intel? AMD? All of the above? We don't know.
When I last discussed the next OpenGL initiative, it was attempting to parse the naming survey to figure out bits of the technology itself. As it turns out, the talk claims to go deep into the API, with demos, examples, and “real-world applications running on glNext drivers and hardware”. If this information makes it out (and some talks remain private unfortunately although this one looks public) then we should know more about it than what we know about any competing API today. Personally, I am hoping that they spent a lot of effort on the GPGPU side of things, sort-of building graphics atop it rather than having them be two separate entities. This would be especially good if it could be sandboxed for web applications.
This could get interesting.
Subject: General Tech, Graphics Cards | January 16, 2015 - 10:37 PM | Scott Michaud
Tagged: Khronos, opengl, OpenGL ES, webgl, OpenGL Next
The Khornos Group probably wants some advice from graphics developers because they ultimately want to market to them, as the future platform's success depends on their applications. If you develop games or other software (web browsers?) then you can give your feedback. If not, then it's probably best to leave responses to its target demographic.
As for the questions themselves, first and foremost they ask if you are (or were) an active software developer. From there, they ask you to score your opinion on OpenGL, OpenGL ES, and WebGL. They then ask whether you value “Open” or “GL” in the title. They then ask you whether you feel like OpenGL, OpenGL ES, and WebGL are related APIs. They ask how you learn about the Khronos APIs. Finally, they directly ask you for name suggestions and any final commentary.
Now it is time to (metaphorically) read tea leaves. The survey seems written primarily to establish whether developers consider OpenGL, OpenGL ES, and WebGL as related libraries, and to gauge their overall interest in each. If you look at the way OpenGL ES has been developing, it has slowly brought mobile graphics into a subset of desktop GPU features. It is basically an on-ramp to full OpenGL.
We expect that, like Mantle and DirectX 12, the next OpenGL initiative will be designed around efficiently loading massively parallel processors, with a little bit of fixed-function hardware for common tasks, like rasterizing triangles into fragments. The name survey might be implying that the Next Generation OpenGL Initiative is intended to be a unified platform, for high-end, mobile, and even web. Again, modern graphics APIs are based on loading massively parallel processors as directly as possible.
If you are a graphics developer, the Khronos Group is asking for your feedback via their survey.
Subject: General Tech, Graphics Cards, Shows and Expos | August 15, 2014 - 08:33 PM | Scott Michaud
Tagged: siggraph 2014, Siggraph, OpenGL Next, opengl 4.5, opengl, nvidia, Mantle, Khronos, Intel, DirectX 12, amd
Let's be clear: there are two stories here. The first is the release of OpenGL 4.5 and the second is the announcement of the "Next Generation OpenGL Initiative". They both occur on the same press release, but they are two, different statements.
OpenGL 4.5 Released
OpenGL 4.5 expands the core specification with a few extensions. Compatible hardware, with OpenGL 4.5 drivers, will be guaranteed to support these. This includes features like direct_state_access, which allows accessing objects in a context without binding to it, and support of OpenGL ES3.1 features that are traditionally missing from OpenGL 4, which allows easier porting of OpenGL ES3.1 applications to OpenGL.
It also adds a few new extensions as an option:
ARB_pipeline_statistics_query lets a developer ask the GPU what it has been doing. This could be useful for "profiling" an application (list completed work to identify optimization points).
ARB_sparse_buffer allows developers to perform calculations on pieces of generic buffers, without loading it all into memory. This is similar to ARB_sparse_textures... except that those are for textures. Buffers are useful for things like vertex data (and so forth).
ARB_transform_feedback_overflow_query is apparently designed to let developers choose whether or not to draw objects based on whether the buffer is overflowed. I might be wrong, but it seems like this would be useful for deciding whether or not to draw objects generated by geometry shaders.
KHR_blend_equation_advanced allows new blending equations between objects. If you use Photoshop, this would be "multiply", "screen", "darken", "lighten", "difference", and so forth. On NVIDIA's side, this will be directly supported on Maxwell and Tegra K1 (and later). Fermi and Kepler will support the functionality, but the driver will perform the calculations with shaders. AMD has yet to comment, as far as I can tell.
Image from NVIDIA GTC Presentation
If you are a developer, NVIDIA has launched 340.65 (340.23.01 for Linux) beta drivers for developers. If you are not looking to create OpenGL 4.5 applications, do not get this driver. You really should not have any use for it, at all.
Next Generation OpenGL Initiative Announced
The Khronos Group has also announced "a call for participation" to outline a new specification for graphics and compute. They want it to allow developers explicit control over CPU and GPU tasks, be multithreaded, have minimal overhead, have a common shader language, and "rigorous conformance testing". This sounds a lot like the design goals of Mantle (and what we know of DirectX 12).
And really, from what I hear and understand, that is what OpenGL needs at this point. Graphics cards look nothing like they did a decade ago (or over two decades ago). They each have very similar interfaces and data structures, even if their fundamental architectures vary greatly. If we can draw a line in the sand, legacy APIs can be supported but not optimized heavily by the drivers. After a short time, available performance for legacy applications would be so high that it wouldn't matter, as long as they continue to run.
Add to it, next-generation drivers should be significantly easier to develop, considering the reduced error checking (and other responsibilities). As I said on Intel's DirectX 12 story, it is still unclear whether it will lead to enough performance increase to make most optimizations, such as those which increase workload or developer effort in exchange for queuing fewer GPU commands, unnecessary. We will need to wait for game developers to use it for a bit before we know.