Last month AMD held their 30 years of graphics celebration, during which they announced their next Radeon video card, the Radeon R9 285. Designed to be AMD’s new $249 midrange enthusiast card, the R9 285 would be launching on September 2nd. In…
Last month AMD held their 30 years of graphics celebration, during which they announced their next Radeon video card, the Radeon R9 285. Designed to be AMD’s new $249 midrange enthusiast card, the R9 285 would be launching on September 2nd. In…
In a surprising move this afternoon, NVIDIA has announced that they will be filing patent infringement complaints against both Qualcomm and Samsung. This complaint is centered around the alleged use of NVIDIA patented GPU technologies in both Qualcomm and Samsung’s SoC’s and the unwillingness of the respective companies to enter in to a licensing agreement. NVIDIA has filed complaints with both the US Federal Court and the International Trade Commission, and in the case of the latter is seeking an injunction against Samsung to prevent them from shipping several recent products, including the Galaxy S5, Galaxy Note 4, and Galaxy Edge.
Starting from the top, NVIDIA is of course no stranger to high profile lawsuits. However those suits are typically on the defensive side, such as the company suing Intel over the loss of rights to make chipsets for their products, or being sued by Rambus over the use of DDR signaling. This suit by comparison is unabashedly offensive – one NVIDIA calls the first time they have ever initiated a patent lawsuit – with NVIDIA going to the courts arguing that their GPU patents are being violated and seeking a resolution and compensation for those violations.
The use of patent infringement suits in the technology space is nothing new, and even in the GPU space they’re somewhat common. NVIDIA in particular is still the 800lb gorilla of the GPU world by patent portfolio size (and arguably by GPU R&D), so how they interact with other companies tends to depend on the size of those companies and what patents they have in turn. In the PC space NVIDIA and AMD have relatively strong cross-licensing agreements – AMD being the next largest GPU developer – followed by Intel who settled a suit out of court with NVIDIA 3 years ago over continued access to NVIDIA’s patents.
In the mobile SoC space however there are a much larger number of GPU manufacturers, and overall there is still a certain “wild west” aspect to patent licensing and infringement. On top of the larger number of GPU manufacturers there are even more companies involved once you discuss integration. In this case an Imagination PowerVR GPU may be licensed by an SoC integrator such as MediaTek, who in turn will sell the complete SoC to the device manufacturer such as HTC. In which case it’s not at all clear who is responsible for patent licensing, or indeed if parties are responsible at each and every step.
This leads us to today’s suit against Qualcomm and Samsung, as it’s based around both alleged patent infringement and arguments about who’s ultimately responsible for those infringements. As far as patents go NVIDIA is going to the courts with a number of patents, with some patents going as far back as the GeForce 2 era and even patents first devised by 3dfx (before NVIDIA acquired them). At first glance these appear to be technologies that are fundamental to modern GPU designs, in which case it is admittedly difficult to imagine other GPU designs not infringing on these patents.
One such example is called the ‘063 patent, which involves on-chip tiling and early visibility testing, and was first developed by a company purchased by 3dfx. This is a technique that all modern GPUs implement in some form, though these days the methods are much more advanced.
“The ʼ063 Patent was directed to this technology, which combined on-chip tiling with early visibility testing in the graphics pipeline. All of Samsung’s mobile products use GPUs that implement this patented invention.”
As is the case whenever anyone files a patent infringement suit, NVIDIA is telling the press and investors that they believe they have a strong case with ample evidence of infringement. The company ultimately believes that they would be victorious in court, though in many of these technology patent suits we see the involved parties settle out of court before any trial reaches its conclusion. Meanwhile if the case does go to trial, then NVIDIA has requested a full jury trial rather than a trial by judge.
Along with NVIDIA’s patent infringement claims, a big part of NVIDIA’s case will rest on allegations that Qualcomm and Samsung actively know about at least some of this infringement and have refused to settle the matter before now. NVIDIA’s claim notes that the company has been attempting to reach a license agreement with Samsung and Qualcomm since 2012, and that today’s suit is a result of their inability to come to an agreement over the last 2 years. Consequently because NVIDIA has been talking to these companies since 2012, they assert that this means that Qualcomm and Samsung have known about this infringement since NVIDIA presented their patents and proposed licensing agreement, meaning they have been willfully infringing on at least some of the involved patents over the last 2 years.
Complicating the matter is the question over who is responsible for patent violations. NVIDIA’s complaint alleges that Samsung believes this to be a supplier problem – that any patent violations are the responsibility of the company who designed the GPU, be it Qualcomm, Imagination, or ARM. As a result Samsung is unwilling to settle, and for that matter Qualcomm is no more willing to settle than Samsung is, though the complaint does not make it clear whether this is a disagreement over the patent claims or a belief that it’s the customer’s (Samsung’s) problem.
Ultimately this suit is focused around Samsung, as Samsung is both a device integrator and the manufacturer of the Exynos line of SoCs. Meanwhile because Samsung also uses Qualcomm’s Snapdragon SoCs in a number of products (including the North American editions of many phones), Qualcomm is being collared in to the suit on the basis that they are supplying some of the infringing GPUs.
This suit is not limited to just Qualcomm’s Adreno GPUs however, and also extends to ARM Mali and Imagination PowerVR GPUs as well, as Samsung has used both of those GPU families in their various Exynos designs. Overall NVIDIA believes that the Adreno 200, Adreno 300, and Adreno 400 families all violate NVIDIA’s patents, while the Mali-T628 and the PowerVR SGX (Series 5) are also explicitly named in claims.
As a result virtually all of Samsung’s modern products are affected by this suit. As part of their request for relief, NVIDIA is asking that the ITC prevent Samsung from importing products using the infringing SoCs, which would include all of Samsung’s latest products including the Note 4 and its Edge variant, the Galaxy S4 and S5, and the latest Tab tablets. As Qualcomm’s SoCs are among the accused products, this injunction would presumably extend to Snapdragon imports as well, which given Qualcomm’s heavy presence in high-end phones and tablets in the United States would make the final list of affected products far larger. However it should be noted that immediate injunctions are very rare, so it’s unlikely that an import ban would go into effect before the conclusion of the trial (if it even makes it that far).
Meanwhile in the Federal Court complaint, NVIDIA is asking for both unspecified damages (i.e. a monetary reward) and for further injunction to prevent Samsung and Qualcomm from infringing on NVIDIA’s patents. Though once again the real goal here seems to be about forming a licensing agreement rather than preventing Qualcomm and Samsung from shipping SoCs.
One of the accused infringing products: Samsung’s Galaxy S5
One of the big questions of course is why NVIDIA is going after these two companies in particular, especially since they are essentially claiming that ARM Mali, Imagination PowerVR, and Qualcomm Adreno GPU families all infringe on some of these patents. The answer in turn from NVIDIA is that Samsung is the largest phone supplier in the United States, and Qualcomm in turn is the largest SoC provider while also providing the SoCs for most of those Samsung phones, so it makes sense to start at the top. NVIDIA is not talking about further legal action at this time, but the outcome of this case could have an impact on whether NVIDIA targets the suppliers (e.g. ARM and Imagination) or goes solely after additional customers and their ilk. Ultimately who is responsible for this infringement – can and should Qualcomm indemnify their customers against infringement claims like these – may be just as important as the alleged infringement itself.
Which on that subject, as part of their claims NVIDIA also points out their existing license agreement with Intel. In this NVIDIA notes that their Intel agreement covers all SoCs and CPUs shipped by Intel, including those incorporating Imagination’s PowerVR GPUs. So in the case where the SoC integrator is responsible, their licensing agreements would seem to cover any infringement regardless of who designs the infringing GPU.
Wrapping things up, at this time Qualcomm and Samsung have yet to respond to this suit. However given the scale of the suit, the complexity of the underlying technology, and the vast number of products involved, barring a quick out of court settlement this is expected to be a case that could go on for years. And even longer if it goes to appeals and/or triggers counter-suits. Altogether the ITC should have an initial response to the filing within 35 days, and if no settlement is reached the expected court date is in 2-3 years. To that end today’s suit from NVIDIA will be the start of a much longer process for Qualcomm, Samsung, and the larger GPU industry as a whole.
If you go back far enough in the computer industry, there have been many successful video card companies. Back before the whole 3D craze kicked off, some of the fastest 2D video cards came courtesy of Matrox, and while they made some attempts at producing compelling 3D graphics cards, they were never able to grab the performance crown from NVIDIA or ATI. Their last real attempt at the 3D graphics market came in 2002 with the Parhelia-512, and as was the case with previous efforts it basically ended up falling short. Interestingly, the Parhelia-512 supported “surround gaming” long before AMD’s Eyefinity, and that may have opened the gates for what would become Matrox’s core focus over the next decade: multi-display video cards.
Since 2002, there haven’t been many reviews of Matrox cards because the focus shifted to industries that need not just two or three but potentially a dozen or more displays all running from a single system. Their last graphics card update was in 2009, and since then the top product has been the M9188, a single card capable of driving eight DisplayPort or DVI connections, with the possibility of using two cards to drive 16 displays. Who needs that many displays? Well, the financial and security markets are two easy examples, as they both have use cases where six or more displays is “reasonable”, and digital signage is another category where Matrox can provide useful technology. These are all professional markets, and the M9188 is priced accordingly ($1500+), but if you were looking to build a system with good graphics performance, Matrox basically hasn’t been relevant as their cards seem to focus almost exclusively on 2D performance these days.
That might be changing with future products given today’s announcement, as Matrox will be switching to AMD-designed GPUs for their next generation of multi-display products. These will continue to support Matrox’s PowerDesk desktop management software, but what’s not clear is whether Matrox will be doing much in the way of customized hardware. The announcement states that “key features of the selected AMD GPU include 28nm technology with 1.5 billion transistors; DirectX 11.2, OpenGL 4.4 and OpenCL 1.2 compatibility; shader model 5.0; PCI Express 3.0 and 128-bit memory interface.”
From that we can surmise that Matrox will be using a variant of the Cape Verde GCN core, which is one of the lower performance GCN parts from AMD. In fact, Matrox may actually be using AMD’s FirePro W600 cards, only with custom Matrox-developed software applications. This would also mean Matrox is looking at a maximum of six display outputs per graphics card (compared to eight on the M9188), but AMD already has the ability to run up to six GPUs in a system with the appropriate motherboard meaning up to 36 displays off a single system is theoretically possible.
The hardware is of course only part of the equation, and Matrox’s PowerDesk software is something that benefits many businesses and professionals. Matrox notes that “critical productivity-enhancing features available with Matrox PowerDesk software will continue to be supported on the next line of Matrox graphics cards designed with AMD GPUs.” These features include the ability to configure and manage multi-display setups, which can get tricky once you move past two or three displays. PowerDesk has tools to configure stretching, cloning, pivot, bezel management, and other items that are important for a professional multi-display configuration.
There are plenty of upsides to this announcement. For one, it allows Matrox to reallocate resources that are currently going into hardware development and instead focus on their core competency, which at this point is multi-display solutions. PowerDesk is well regarded in their target market, and this will allow Matrox to continue to improve the platform without trying to design their own hardware. AMD benefits as they’re able to partner with Matrox and potentially sell their GPUs at higher “professional” prices, and they may also increase their share of digital signage and other multi-display markets.
And of course the customers that purchase the cards benefit as they get to move to a modern platform with support for all the latest DirectX, OpenGL, and OpenCL libraries. Long-term, this also opens the doors for Matrox to offer substantially higher performance 3D solutions from AMD for customers that need such features. Overall, this announcement isn’t likely to affect most computer users, but it’s good to see Matrox still hanging around after several decades in the computer graphics industry, something many of their competition from the 90s didn’t manage to achieve.
It seems company sponsored live events are becoming more popular. Not long after AMD hosted their own gaming celebration, NVIDIA is upping the ante with a 24 hour live streamed event with several physical locations around the world, which they’re calling GAME24.
NVIDIA describes GAME24 as a celebration of gaming, and several gaming studios will also be involved in the event. However, NVIDIA has not yet announced the studio names. GAME24 will feature eSports competitions, record-breaking gaming stunt competitions, exclusive game demos, prizes, and developer interviews. And of course, along with a focus on gaming products and partnerships, there will be a significant amount of actual gaming taking place both remotely and in-person.
The event will take place starrting 6 PM PST on Thursday, September 18th, and it will last until the same time the next day. Streaming will be provided from start to finish via Twitch. To attend the event, register at http://www.game24.nvidia.com/. Attendance and registration are free. Registration ends at 5 PM PDT the day of the event but NVIDIA warns seats are limited, so it’s first come first served.
Confirmed locations include:
I have personally attended a few regional celebration events like this before, and I’ve always had a good time. Between the gaming, merchandise giveaways, and just the social aspects it should be a fun event.
Source: NVIDIA