I recently read an article by an Intel product manager on the need for “ECC” (error correction code) memory in CAD workstations. From the article: “Corrupted data can impact every aspect of your business, and worse yet you may not even realize your data has become corrupted. Error-correcting code (ECC) memory detects and corrects the more common kinds of internal data corruption.”
For some reason this triggered my memory of the sudden-acceleration Toyota Prius incident from 2010. The popular press latched on to the idea that cosmic rays were screwing with the electronics in the Prius. While theoretically possible, the probabilities of this were astronomically low. It did however, make for a great story and the FUD (fear uncertainty doubt) caused Prius prices to temporarily plummet and sales come to a crawl.
Back to ECC memory and CAD systems. Is there really a need for ECC memory in CAD or is it just FUD marketing to upsell hardware and make products sound more valuable than they really are? I decided to do a little research.
Who needs ECC memory and what is its role in professional & CAD workstation computing?
Naturally occurring cosmic rays can and do cause problems for computers down here on planet Earth. Certain types of subatomic particles (primarily neutrons) can pierce through buildings and computer components and physically alter the electrical state of electronic components. When one of these particles interacts with a block of system memory, GPU memory or other binary electronics inside your computer, it can cause a single bit to spontaneously flip to the opposite state. This can lead to an instantaneous error and the potential for incorrect application output and sometimes, even a total system crash. However, the theoretical chances of a single bit error caused by a cosmic ray strike on your PC or workstation’s memory is fairly rare — only about once every 9 years per 8GB of RAM, according to recent data.
ECC technology — used as both system RAM, and in devices such as high-end GPUs — can reliably detect and correct these errors, reducing the odds of memory corruption due to “single bit errors” down to about once every 45 years for 8GB of RAM. Of course, just like everything else in life there are always tradeoffs. ECC memory is typically up to 10% slower and significantly more expensive than standard non-ECC memory.
Because the odds of a cosmic ray strike increase in direct proportion to the physical amount of memory (and related components) inside a computer, this is a real concern for large scale, clustered supercomputing and other environments where computing tasks often include high-precision calculation sets that can take days or even weeks to complete. In the case of supercomputer clusters, which often contain hundreds or even thousands of connected computer nodes and terabytes of memory, the odds of cosmic ray strikes on the system are much more likely — and much more costly. Restarting a week-long calculation on a supercomputer can cost a facility many tens of thousands of dollars in lost time, electricity and manpower —not to mention lost productivity.
But for even very beefy PC CAD workstation configurations with loads of RAM on board, you are probably not at imminent risk from problems caused by cosmic ray strikes and the resulting single bit errors. Over the course of your work, you are much more likely to endure system crashes or application hangs dues to failing components, power fluctuations and software bugs than due to cosmic ray strikes. Additionally, many applications in the desktop design and engineering space can actually endure a single bit error without negatively impacting the computing process or product. For example, if the color or brightness of a single pixel on a display monitor is changed due to this type of memory corruption on the system’s GPU, nobody will ever see or notice it. There are many such examples of this type of error not really impacting ones everyday work.
This said, many leading technology manufacturers are enabling their high-end products with ECC memory for compute-heavy (especially clustered supercomputing) applications where the benefits of using error correcting memory outweigh any comparative speed/cost drawbacks. AMD for example, has engineered their new AMD FirePro W9000 and FirePro S9000 ultra-high-end GPU cards to include ECC memory which can selectively be enabled by the end user and used for many advanced computing purposes where rock-solid stability and protection from space rays is crucial.
Author: Tony DeYoung
The most compelling reason to install multiple GPUs is to drive multiple high-resolution displays. The secret’s out that “multi-mon” is the single best way to improve your productivity. Anyone who’s gone to two displays (or three — or more!) will tell you they could never go back to one. And more graphics cards can display more pixels across more monitors.
Which Graphics Card Works for You?
That said, you don’t necessarily need to populate two cards to run two monitors, so pay attention to the cards you’re selecting. NVIDIA’s Quadro with nView and Mosaic technology can support two displays across most of the product line. A single high-end AMD FirePro V7900, with its Eyefinity technology, can handle four on its own, thank you very much. As such, if your performance demands have you buying midrange or high-end cards, you might get all the screen real estate you want with one card. But if you’re much hungrier for pixels and screens than you are for polygons per second, you might consider two less-expensive, dual-monitor cards.
On top of multi-monitor support, you can use that extra slot to turn your workstation into a supercomputer. An exaggeration? Not to some. General-purpose computing on GPUs (GPGPU) technology is still evolving, but many of the applications that show the most promise are the ones of most interest to engineers and other CAD users: applications such as computational fluid
dynamics (CFD) and finite-element analysis (FEA). Simulation software developers such as ANSYS and Abaqus are porting code to harness GPUs to deliver big speed-ups — in many cases tenfold or even 100- fold increases — over CPU-only computation.
High-end graphics cards usually require more power than the 75 watts supplied by the typical x16 PCI Express interface. Workstation OEMs accommodate their extra needs via auxiliary power cables drawn from the supply. Some high-end and virtually all ultra high-end graphics cards are dual-slot thickness. They insert into one PCI Express x16 connector, but their thickness means an
adjacent x16 slot may be blocked and rendered useless.
Make the Right Choice
When purchasing a workstation online, the OEM’s product configurator should let you know if the chosen card or cards will mate to the chosen system, with respect to power supplies and connectors, the number of available PCI Express x16 slots, and whether a dual-slot card has sufficient clearance. For example, when outfitting graphics on a smaller chassis that can’t accommodate two dual-slot cards, chances are the OEM will only offer the option of two entry-level or two mid-range cards, both of which are single slot width.
For that matter, if you’re perusing the latest flavor of entry level workstation, full-length cards may not have clearance lengthwise. Again, the online configurator should ensure compatibility, so you shouldn’t have to worry about these issues.
A GPU manages how your computer graphics process and display and, thanks to parallel processing, is typically more efficient than a CPU. The GPUs that are best optimized for professional graphics-intensive applications, such as CAD, design visualization and analysis, are found in workstation caliber AMD FirePro and NVIDIA Quadro graphics cards.
Five Categories of GPUs
Such professional-caliber GPUs come in a variety of flavors for desktop as well as mobile form factors. In the more mature desktop arena, they tend to fall into five categories of add-in cards.
The first category is 2D GPUs. Professional 2D cards can manage some 3D processing, but are not optimized for regular or intensive 3D applications. They generally aren’t well suited for CAD use.
For professional-level CAD work, you’ll want a Quadro or FirePro 3D add-in card. Each of these product lines includes approximately half a dozen models that fall into the remaining four product categories, as defined here by Jon Peddie Research:
- entry-level: $350 or less
- mid-range: $350–$950
- high-end: $950–$1,500
- ultra high-end: $1,500 or more
There are always exceptions, but most buyers will want to match the performance and capabilities of the GPU with the rest of the system — that is, an entry-caliber card for an entry caliber workstation. Achieving good balance, where each component hits a performance level that is supported by the rest of the system, is the best way to maximize ROI for your workstation purchase and optimize your productivity.
Fortunately, most workstation OEMs today do this work for you, offering a subset of cards from AMD and NVIDIA that best match the capabilities of the model you’ve chosen.
Optimizing GPU Performance
Most graphics cards — and all performance-oriented models — slide into PCI Express x16 slots in the workstation. Graphics cards can be installed in open slots at the factory when ordering your new system, or anytime later if you buy a card off the shelf. A mid-life upgrade of your system with a latest-generation GPU can provide a cost-effective kick, for example if rendering becomes a bottleneck.
And unlike the machine that’s at your desk today, your new workstation (unless it’s a small–form factor model) will likely come equipped with at least two PCI Express x16 slots, able to accommodate two cards. Why would you want two (or more)? One reason is that multi-GPU technologies from NVIDIA (SLI) and AMD (CrossFire) allow the pairing of two cards (rendering alternate frames) to boost performance.
Where do you begin your quest for the right workstation? This particular hardware search should start with your software.
Let’s be real: Nobody relies on just one application over the course of a day. We’re all bouncing between disparate tasks and windows. But for the majority of CAD professionals, there is one application — or maybe a couple — that consumes the bulk of your hours at the desk. What’s the app that dominates your day? Got it? Now hit the web site of the software developer and find the minimum and recommended system requirements for your killer app. AutoCAD users can find this information at http://usa.autodesk.com/autocad/system-requirements.
Minimum is the Starting Point Only
In most cases, an application’s minimum requirements set an extremely low standard, as the software vendors begrudgingly must address the least common denominator of the installed base. We don’t recommend you follow these guidelines, but it’s worth making a note of the minimum graphics, system memory and CPU requirements. On the other hand, it’s highly likely that any new workstation on the market today will meet or exceed these numbers.
More interesting is the list of recommended or certified hardware. For SolidWorks, Dassault Systèmes (as of this writing) specifies a minimum of 1 GB RAM, but suggests 6 GB. Well, if you go with 1 GB, you’ll be sorry — even 6 GB isn’t necessarily the best choice, depending on your budget, and especially given the incredible amount of gigabytes/dollar that can be had today.
Similarly, Autodesk isn’t going to stop you from running a PC gamer graphics card, but the company will tell you which cards are optimized for performance and built for reliability when it comes to supporting AutoCAD or Autodesk Inventor.
Increasingly, the only CAD-certified graphics cards are professional-brand NVIDIA Quadro and AMD FirePro. That’s because software developers have consistently seen the fewest bugs and problems with cards that, like the system overall, have been exhaustively tested and tuned for professional workstation applications. In fact, the major CAD software developers will help you address issues related to running a Quadro or FirePro card, but they dedicate no support cycles to fixing bugs on consumer-class hardware.
It’s about time. After a hiatus from its role as a viable alternative to Intel for workstation-class CPUs, AMD is back. Instead of its traditional server/workstation focused Opteron line, this time the company is — quite wisely — choosing to target the market with a combination CPU/GPU part, what AMD refers to as an Accelerated Processing Unit, or APU.
New to the market are two, professional-caliber versions of its recent “Trinity” part, workstation-branded as the FirePro A300 and A4320. And while having two such parts represents a drop in the workstation bucket, as compared to Intel’s position, any new competition should only help CAD professionals find better products — or at least better deals on those products — in the future.
A New Strategy
While AMD has never given up plying its professional-brand FirePro GPUs in workstations, the same can’t be said for professional-brand CPUs. After a promising start and a firm foothold in the market, AMD’s CPUs are today, for all intents and purposes, absent in workstation platforms.
The company’s Opteron processor began making significant inroads into workstation platforms back in the mid-2000s. With Intel’s offerings at that time looking comparatively poor, Opteron steadily picked up workstation OEMs, but the end of 2003 having all major suppliers in tow with the exception of Dell. That increased OEM presence translated directly to increased market share, up to 4% of the overall market in mid-2006, and to a more 10% of dual-socket workstations shipped.
Then came the steady, inexorable decline, which by the end of 2011 left Opteron without any major OEM on board and virtually no market share. Truth be told, it wasn’t like AMD was ignoring the workstation CPU market out of ignorance or incompetence. Rather, it was a case of triage. The company knows full well it doesn’t have the wherewithal of its chief rival, Intel, and accordingly it’s always had to be careful about which markets it targets and which it doesn’t.
And that begs a question: why now does AMD think it should invest its time and money marketing CPUs for workstations, when it didn’t before? It’s not like Intel’s CPUs are struggling like they were back in 2005. Heck, more than ever, AMD is looking for arenas to sell CPUs that don’t directly compete with Intel. No, AMD’s renewed interest in workstation CPUs has more to do with its competitive positioning in GPUs than CPUs.
Ever since the CPU duo began building and marketing combination, all-in-one CPU+GPU parts (first with Intel’s Westmere in 2010, followed by AMD’s first Fusion parts), a unique opportunity fell into the AMD lap. As we’ve been pointing out for some time now, AMD now finds itself in the rare position where it can make a compelling, competitive case over both its chief rivals, Intel and Nvidia. Intel’s reputation for performance graphics has been poor, and despite the company’s largely successful attempt to boost its graphics profile (with 2011′s Sandy Bridge and 2012′s Ivy Bridge), AMD still owns the undeniable edge over Intel in graphics. Nvidia, meanwhile, which could argue graphics supremacy, doesn’t have x86 technology, making it impossible to compete in the new CPU+GPU segment.
Pitching an ISV-certified, professional-caliber version of Trinity to workstation OEMs can be convincing, especially given which end of the market that part could play. The dominant and still fastest growing segment of the workstation markets is the Entry class, particularly the low end of that class … precisely where the cost-effectiveness of the integrated part can appeal. The capabilities of parts like Trinity and GPU-integrated Ivy Bridge aren’t record-breaking, but they’re too good for workstation-shipping CAD professionals too ignore … especially those on tight budgets.
And given Intel virtually owns the market, OEMs like Dell and HP ought to welcome an enthusiastic re-entry by AMD. After all, no business wants to be beholden to one supplier, even if it’s a supplier of essentially infinite volume, like Intel.
What Does It Mean for CAD Professionals?
So after doing some comparison shopping, will you end up with a workstation with neither Intel nor Nvidia inside? Maybe, maybe not. But either way, you’ll be much more likely to get the machine you want at a lower price, regardless of whose brand is on it. Because while Intel’s been doing an impressive job as of late delivering the type of hardware professionals demand, any competition is welcome competition. And that not only benefits OEMs like HP and Dell, it should only help when it comes to keeping down IT costs for CAD.
AMD launched the AMD FirePro A300 Series Accelerated Processing Unit (APU) for entry-level and mainstream desktop workstations. Featuring AMD Eyefinity multi-display technology, the AMD FirePro A300 Series APUs are designed for CAD and media and entertainment (M&E) workflows.
AMD FirePro A300 Series APUs combine CPU and GPU functionality on a single chip to blend workstation performance and application-certified compatibility required to help keep design professionals productive in their work.
“Design professionals demand workstation-class tools that enable productivity and flexibility in their workflow, and the AMD FirePro A300 Series APUs enable workstation integrators and OEMs an exciting new computing platform on which to design and build powerful, entry-level desktop workstation configurations,” said Matt Skynner, corporate vice president and general manager of AMD Graphics.
According to the company, the AMD FirePro A300 Series APUs are the first single-chip processors capable of delivering the workstation-class visual computing performance required for advanced professional design workflows. The introduction of AMD FirePro A300 Series APUs is designed to allow OEMs and workstation integrators (WSIs) greater flexibility, enabling new workstation designs that help save space, are energy efficient, and have low heat and noise levels without compromising true workstation-class performance and reliability.
AMD FirePro A300 Series APUs were developed for the entry-level and mainstream workstation segments, providing a blend of CPU and GPU performance and industry-leading features to keep design professionals efficient:
- Support for AMD Eyefinity Technology for enhanced efficiency and immersive, multi-monitor productivity;
- AMD Turbo Core technology, where CPU and GPU performance are dynamically scaled depending on workload demands, effectively providing a more responsive experience;
- Support for horizontal display resolutions up to 10,240 x 1600 pixels, enabling large desktop spaces across multiple high-resolution display devices for advanced multitasking;
- Support for Discrete Compute Offload (DCO), allowing additional compute capability by using discrete AMD FirePro GPUs in parallel with APU graphics for extended GPGPU performance;
- 30-bit color support to enable image and color fidelity for advanced workflows such as color correction and image processing when using displays capable of 10-bit-per-channel operation;
- Dedicated UVD (universal video decoder/VCE, or video CODEC engine) media encoding hardware for faster “fixed function” GPU processing of H.264/MPEG4 files and other motion media formats when using compatible software, to free up CPU resources for other tasks.
Pricing and Availability
The AMD FirePro A300 Series APUs will be available in systems from a number of workstation integrators starting in August 2012.
|AMD FirePro A300 Series APUs|
|APU Model||TDP||CPU Cores||CPU Clock (Max/Base)||AMD Stream Processors||GPU Clock||Unlocked|
|AMD FirePro A300||65W||4||4 GHz / 3.4 GHz||384||760 MHz||No|
|AMD FirePro A320||100W||4||4.2 GHz / 3.8 GHz||384||800 MHz||Yes|
Author: CADspeed editors
If you’ve upgraded to the latest CAD software applications and your workstation is feel a little, well, overworked, AMD just might have the answer you’ve been looking in its latest line of workstation graphics launched this week.
The AMD FirePro W9000 GPU features incredible increased memory bandwidth and greater multi-display support performance than the competing solution. Following closely are the AMD FirePro W8000, W7000 and W5000 workstation graphics cards, all built on the AMD Graphics Core Next Architecture, and designed to balance compute and 3D workloads efficiently for computer-aided design and engineering, and for media and entertainment (M&E) professionals.
The AMD FirePro W9000, W8000, W7000 and W5000 GPUs are optimized and certified for leading software applications enabling users to unleash their creativity by ensuring ultra-high geometry performance. The latest AMD FirePro workstation graphics cards can enable smooth handling of complex models and feature dynamic power management that enables great performance and efficient power usage.
Using AMD Eyefinity technology, the AMD FirePro W9000, W8000, and W7000 GPUs can drive up to six, 30” independent displays via Multi-Stream Transport (MST) hubs for maximum workspace utilization at ultra-high 4096×2160 resolutions.
“As professionals work with larger data sets that demand advanced visualization and complex models, they need a graphics solution that is fast, powerful, and reliable,” said Matt Skynner, corporate vice president and general manager of AMD Graphics. “Certified for today’s software applications, the new AMD FirePro workstation graphics cards bring a range of features and capabilities for professionals working in digital signage, broadcast graphics, CAD/CAE and M&E, delivering the ideal balance of power, performance and reliability at the right price point.”
With the latest AMD FirePro workstation graphics offerings, graphics professionals can create more complex models and interact with them in real time, helping improve workflows and boost productivity. Through GCN and GeometryBoost, the state-of-the-art AMD FirePro W9000 workstation GPU delivers 1.95 billon triangles per second, which is 1.5 times as great as the competitor’s most powerful workstation graphics card, and up to 83 percent greater memory bandwidth than the competing solution, for outstanding application responsiveness.
The AMD FirePro W8000 workstation GPU, features Error Correcting Code (ECC) memory support and offers category leading dual-precision compute performance, up to 2.2 times as fast as the competing solution. This helps professionals experience greater accuracy in calculations performed for structural and molecular analysis and computational fluid dynamics without impacting application performance. Not to be outdone, the AMD FirePro W7000 workstation GPU is up to five times as fast as the competing solution in single-precision compute performance, while the AMD FirePro W5000 workstation GPU is the most powerful mid-range workstation graphics card ever created, delivering significantly better resolution, memory and display output performance than the competing card.
The AMD FirePro W9000, W8000, W7000 and W5000 workstation graphics cards are optimized and certified for leading workstation applications. Additionally, these new AMD FirePro cards support PCI Express 3.0 and AMD PowerTune and AMD ZeroCore Power technologies for dynamic power management.
AMD FirePro W9000
High-performance CAD engineers, media designers, digital signage professionals
4 TFLOPs single precision
1 TFLOP double precision
6GB of high speed GDDR5 memory
AMD FirePro W8000
High-performance CAD engineers, media designers, digital signage professionals
3.23 TFLOPs single precision
806 GFLOPs double precision
4GB of high speed GDDR5 memory
AMD FirePro W7000
Mid-range solution for CAD engineers, media designers, digital signage
2.4 TFLOPs single precision
152 GFLOPs double precision
4GB of high speed GDDR5 memory
AMD FirePro W5000
Mid-range solution for CAD engineers, media designers, digital signage
1.27 TFLOPs single precision
80 GFLOPs double precision
2GB of high speed GDDR5 memory
Last week, I talked about why Intel’s latest generations of graphics-enabled CPUs might make CAD professionals think twice about paying extra dollars for a discrete graphics card on their next workstations.
As I mentioned previously, the low-cost Entry 3D segment has seen steady gains over the years, for a logical reason … as average street prices fall and capabilities climb, the Entry class satisfies more and more of the workstation community. But then right around the start of 2011 — precisely when Sandy Bridge comes out of the chute in workstations like HP’s Z210 —Entry 3D shipments start to flatten and then decline (albeit modestly).
Why are Entry 3D sales more indicative than other segments of a possible erosion from integrated Sandy Bridge graphics? Well, if recent buyers were to opt for Sandy Bridge graphics, the discrete card they’d most likely be opting against would be an entry-class product. Those shopping for a mid-range or better card aren’t going to be enticed by CPU-integrated graphics. Such buyers have both the need for performance and the dollars to pay for it. So if Intel’s new push into professional-brand integrated graphics were to have an impact, we would logically see the effects first in Entry 3D. And that appears precisely to be the case, albeit at a far-from-dramatic rate.
Don’t expect the impact of CPU-integrated graphics to be either dramatic or fast-paced. For the near term, while Intel’s “good enough” graphics performance can satisfy a big chunk of the mainstream, it will be an appropriate choice for only the most budget-conscious professionals. Still, the trend line, as it was in mainstream graphics, is pointing just one way: up. Sandy Bridge’s successor, Ivy Bridge, has just recently begun shipping in the market, and it again provides a substantial bump in performance and features over its predecessor.
Give it time, and integrated solutions will eventually hold significant share among CAD pros … not to the extent it does in mainstream PC markets, but significant share nonetheless.
Intel had been promising that its latest generations of graphics-enabled CPUs would make CAD professionals think twice about paying extra dollars for a discrete graphics card on their next workstations. And it appears those promises are holding true … not in dramatic fashion, but valid nonetheless.
The thought of CPU-integrated graphics is a new proposition for buyers of professional-caliber looking to speed their CAD workflows. Prior to Intel’s Westmere generation, released in early 2010, virtually ever workstation shipped with a professional-brand graphics add-in card installed. The vast majority have been Nvidia Quadro models, with a minority share of units bearing AMD’s FirePro brand.
Westmere’s CPU+GPU combination first raised the question — could integrated graphics perform well enough for CAD duties to allow buyers to save some cash on the add-in card? The answer in 2010 was generally “no.” Performance was not up to snuff, even for entry-class CAD use, and as a result, most workstation OEMs still required the presence of a Quadro or FirePro card in any machine leaving the factory. That choice made sense, as the last thing HP or Dell would want for their professional customers is a poor graphics experience that might turn them off workstations altogether.
But then came 2011 and the launch of the Sandy Bridge generation of die-integrated graphics. With Sandy Bridge, Intel more than anything else focused performance improvements in graphics. And for the first time, the company began actively marketing its graphics for professional use (the “P” prefix in the P3000 signifying “professional” grade). The combination of Intel’s posture and Sandy Bridge’s substantially improved graphics were enough to get OEMs like HP to (for the first time) allow buyers to choose integrated graphics and pass on the graphics add-in card.
Now, Sandy Bridge’s graphics can’t compete head-to-head with Quadro or FirePro … it’s not intended to. What it is intended to do is provide competent graphics for CAD professionals who don’t have the highest demand for performance and whose budgets are especially tight. How did Intel do on its goals? Well, a look in the past few quarters at the add-in card attach rates for low-end systems and the distribution of the add-in cards sold should give a clue.
Anecdotally, OEMs are reporting that, while attach rates remain quite high, they have dropped with Sandy Bridge. And those reports seem to be validated by shipment numbers seen for professional graphics add-in card segments, specifically the low-cost Entry 3D segment. That segment sees steady gains over the years, for a logical reason … as average street prices fall and capabilities climb, the Entry class satisfies more and more of the workstation community. But then right around the start of 2011 — precisely when Sandy Bridge comes out of the chute in workstations like HP’s Z210 —Entry 3D shipments start to flatten and then decline (albeit modestly).
Next week, I’ll continue this discussion by explaining why Entry 3D sales more indicative than other segments of a possible erosion from integrated Sandy Bridge graphics.
Order Independent Transparency (OIT) in computer graphics programming terminology denotes any technique that can correctly render overlapping semi-transparent objects without having to sort them before they are being rendered. Rendering semi-transparent objects has always been a problem because the blending operation is order dependent: when a semi-transparent fragment is rendered, the underlying color (i.e. the background) is crucial for the final color to be correct.
OIT is a new option that can simply be enabled in Creo Parametric 2.0. With OIT, Creo Parametric 2.0 allows for pixel accurate rendering of overlapping semi-transparent objects without having to sort them before they are being rendered, providing up to 10 times performance of blended rendering in PRO/Engineer Wildfire 5.0 compared to when rendering transparency in Creo Parametric 2.0. This translates into less time waiting for your model to render and increased productivity over the long run.
This technique is easy to implement and add to an existing rendering pipeline: everything can be rendered as usual, semi-transparent or not. The technique here is fully implemented on the AMD FirePro professional graphics board, which totally frees the CPU from multiple render passes or face sorting. OIT only works with FirePro cards.
OIT assembles a pixel-accurate representation of the model and its surrounding geometry while maintaining user interactivity and visual quality. This creates a more practical transparent 3D viewpoint to continuously work within, helping improve the sense of design intuition and aid in better decision-making throughout the product development stages. It is also very accurate since the actual sorting that happens on the GPU is done per fragment.
The technique has a very low impact on the existing rendering pipeline and is therefore very easy to integrate in an existing rendering engine. As far as performance goes, the results speak for themselves: it achieves up to 10x faster frame rate compared to face sorting and regular blending.
How It Works
The technique is based on the usage of an A-buffer, a simple list of fragments per pixel, in its simplest form as a linked list of fragments per pixel. First, all primitives are rasterized to the A-Buffer, writing some color value and some depth value (Red-Green-Blue-Alpha-Depth), one index buffer (RAT) is used to keep the number of fragments in this pixel. Finally, a full screen shader pass will sort that A-Buffer according to the depth value and do the blending for each fragment according to their sorted indices.
Viewport performance with OIT enabled has been measured to increase up to ten times versus OIT disabled with transparency visual quality dramatically improved with pixel-accurate transparency rendering, solving visual artifact problems and z-ordering issues seen without OIT enabled. AMD developed the OIT implementation for PTC and the Creo Parametric 2.0 community, showing the company’s commitment to the market as an innovator – not just a product company.