The future of graphics is here: PowerVR Series6XT GPUs go Rogue at CES 2014

CES 2014 marks the expansion of the PowerVR Rogue architecture with a new family of GPU IP cores designed to enrich what is already the best roadmap for graphics the mobile and embedded industry has seen. The new PowerVR Series6XT Rogue architecture drives performance to new levels while reducing power consumption even further through advanced PowerGearing mechanisms.

During the event, we’re launching the first four members of our PowerVR Series6XT GPU IP family:

  • PowerVR GX6250, GX6450 and GX6650 are fully-featured implementations with two, four and six shading clusters respectively.
  • PowerVR GX6240 is an area-optimized two cluster implementation

The PowerVR Series6XT graphics IP family introduces even better levels of performance and the most comprehensive set of features designed to improve power management, reduce memory bandwidth requirements, offer superior image quality and accelerate time to market for our customers.

PowerVR Series6XT GPU Rogue (OpenGL ES 3.0, DirectX10, OpenCL 1.x)PowerVR Series6XT GPUs block diagram (OpenGL® ES* 3.0, DirectX 10, OpenCL™* 1.x)

Dramatic performance improvements

Let’s talk performance efficiency first. When we first introduced the PowerVR Rogue architecture, our goal was to deliver a new benchmark for performance in mobile. Every PowerVR Rogue GPU core has been fully optimised for maximum efficiency which means we’ve always focused on keeping power consumption to a minimum. Consequently, PowerVR Rogue-based platforms have been at the top of graphics benchmarks from Kishonti, Futuremark or Rightware – and will continue to do so thanks to the new performance improvements brought upon by PowerVR Series6XT.

In fact, we can now offer you a sneak peak at the next-generation GFXBench 3.0 rendering on our GPUs. The scene below offers a preview of how developers can use the latest OpenGL ES 3.0 features to achieve console quality effects such as deferred and volumetric lighting, fuzzy particle systems, dynamic video textures and depth of field on mobile devices.

Kishonti GFXBench 3.0 rendering on PowerVR - OpenGL ES 3.0 featuresA screenshot of Kishonti’s GFXBench 3.0 rendered by a PowerVR Rogue GPU

Thanks to numerous efficiency enhancements, our new PowerVR Series6XT GPU cores are up to 50% faster clock for clock, cluster for cluster compared to their Series6 counterparts; these performance measurements are based on frames from industry standard benchmarks.

And since PowerVR Series6 GPUs have already proven to be faster than any other mobile GPU on the market, PowerVR Series6XT is set to maintain and extend Imagination’s leadership when it comes to high performance, low power graphics cores. However, as performance keeps ramping up, it is important to remember that power and area are still the driving issues for mobile designs.

We’ve addressed both these issues by providing a comprehensive roadmap of GPU cores that have a set of carefully selected features which enable the best possible performance for each segment of the mobile market.

PowerVR Series6 Series6XT Series6XE GPU - PowerVR graphics roadmapThe current PowerVR Rogue graphics IP roadmap

Power is particularly relevant for mid- and high-end devices; independent reviewers have seen countless examples of platforms throttling under graphics-intensive tasks. All PowerVR GPUs are designed to deliver sustained high performance while minimising heat dissipation thanks to a low power feature set that has been designed for real-world use cases. This feature set includes lower-power GFLOPS and PVR3C compression technologies – you can find out why PVR3C is vital for mobile by reading this blog article, but we’ll expand on this in later blog articles.

Meanwhile, for silicon vendors that are targeting entry level platforms, area is becoming more and more relevant due to the costs associated with manufacturing larger SoCs. To address this, our partners can use our Design Optimization Kits (DOKs) to achieve an ideal balance of performance, power and area for their specific application. For example, the first DOK co-developed with Synopsys has reduced area up to 30% for PowerVR Series6 GPUs, while lowering dynamic power consumption significantly.

PowerVR Series6XT GPU - OpenGL ES 3.0 Soft Kitty demoSoft Kitty – our latest OpenGL ES 3.0 demo rendering on a PowerVR Rogue GPU

What’s new for PowerVR Series6XT GPUs

PowerVR Series6XT GPUs feature a number of performance improvements which have been achieved thanks to a combination of enhancements. We’ve added more low-power GFLOPS and made architectural tweaks that have boosted efficiency all-round; the list of enhancements includes:

  • An improved instruction set for even better application performance and increased GPU efficiency
  • A next-gen Hierarchical Scheduling Technology (HST) for further increased resource utilisation; this involves a combination of driver-based scheduling, optional multi-core scheduling (demand driven workload distribution across GPU cores), Microkernel firmware scheduling, coarse grain scheduling (programmable, per-cluster distribution) and cluster thread scheduling (data driven, massively multi-threaded and multi-tasking)
  • Sustained polygon throughput and pixel fillrate improvements for more ever more responsive and fluid graphics
  • GPU compute data path optimizations for higher parallel processing performance tuned with real world practical applications in mind

With the improvements listed above, Series6XT GPUs are able to offer amazing computing horsepower per mm2 and mW, making them the perfect choice for handling complex user-interfaces, ultra-realistic 3D gaming, image/video processing, augmented/virtual reality and many other demanding graphics and/or parallel compute applications.

And the list doesn’t stop here: we’ve added Deep Colour support for Ultra HD market segments. Silicon vendors can license our PowerVR Series6XT GPUs, PowerVR Series5 hardware video transcoders and the upcoming Series2 ‘Raptor’ camera/imaging IP and create a vision platform that maintains perfect colour fidelity from capture all the way to display.

PowerVR Series6XT GPU - Ultra HD DeepColourA comparison of image quality with and without Deep Colour

Moreover, the PowerGearing G6XT advanced power management technology provides the means for PowerVR Series6XT GPUs to offer more for less: more performance for lower power, when compared to previous generation cores. PowerGearing G6XT is a Series6XT-specific implementation of our advanced PowerGearing techniques and includes a combination of automatically enabling clock gating, enabling/disabling power to clusters, and intelligently using the Microkernel firmware to offer lower latency workload feedback into DVFS and power management decision process.

PVR3C triple compression for less bandwidth and lower power

PowerVR Series6, Series6XE and Series6XT GPU cores feature PVR3C, a technology encompassing three types of compression designed to lower bandwidth traffic and power consumption at the system level.

PowerVR Series6XT GPU - PVR3C compression technologies (PVRTC/ASTC, PVRIC, PVRGC)PVR3C includes a unique combination of lossy texture compression (PVRTC2/ASTC), lossless compression for dynamic resources and uncompressed texture resources (PVRIC2) and geometry compression for parameter memory (PVRGC)

Firstly, PowerVR Series6XT supports lossy texture compression. This is designed to reduce static texture bandwidth and can go down to as low as less than one bit per pixel (1bpp). PowerVR Series6XT GPUs offer developers the possibility to use any of the existing texture formats below, creating the widest possible choice of industry texture formats:

  • PVRTC/PVRTC2 (PowerVR Texture Compression), the industry’s most used low power texture compression format
  • ASTC (Adaptive Scalable Texture Compression), a new optional extension for both OpenGL* 4.x and OpenGL ES 3.0 released and promoted by the Khronos Group
  • ETC1 (Ericsson Texture Compression) and ETC2 (backwards compatible with ETC1 and mandatory in the OpenGL ES 3.0 graphics standard)
  • DXTn, DXTC or BCn for DirectX9_3/10 capable cores

As the second and third parts of PVR3C, PowerVR Series6XT also delivers lossless image and geometry compression:

  • PVRIC2 (2nd generation PowerVR Image Compression) aims at reducing framebuffer and render target bandwidth; PVRIC has a typical 2:1 compression rate with zero quality loss and integrates with display controllers.

PVRIC2 for PowerVR Series6XT GPU - Lossless image compression2nd generation lossless image compression (PVRIC2) for PowerVR Series6XT GPUs

  • PVRGC (PowerVR Geometry Compression) typically reduces parameter bandwidth by 30% with zero quality loss and comes in handy for applications with complex geometry (high polygon count)

Tools and ecosystem support

On top of the usual RTL deliverables, we’ve created flexible physical Design Optimization Kits to provide silicon vendors with reference design flows, reference layouts and documentation that enables them to tune for performance, power and area in an intuitive way that translates to a faster time to market with better resulting products.

Finally, PowerVR GPU cores continue to have the widest ecosystem support in the industry. We offer the best graphics SDK available for iOS, Android and Windows and provide training and support to developers big or small.

PowerVR Insider EcosystemThe PowerVR Insider ecosystem

We also promote the widespread adoption of GPU compute in mobile by working with open source platform vendors and contribute to the development of OpenCL, Renderscript/Filterscript and other compute APIs.

For more news and announcements related to our PowerVR Rogue GPUs, please come back to our blog and follow us on Twitter (@ImaginationPR, @PowerVRInsider, @GPUCompute). If you are attending CES or MWC 2014, stop by Imagination’s booth to have a look at the latest demos and applications running on our customers’ PowerVR-based platforms.

 

Editor’s Note

* PowerVR Series6XT Rogue cores are based on published Khronos specifications, and are expected to pass the Khronos Conformance Testing Process. Previous generation PowerVR cores have already achieved conformance. Current conformance status can be found at www.khronos.org/conformance.

OpenCL and the OpenCL logo are trademarks of Apple Inc. used by permission by Khronos.

OpenGL is a registered trademark and the OpenGL ES logo is a trademark of Silicon Graphics Inc. used by permission by Khronos.

, , , , , , , ,

  • Pingback: Low-cost mobile 3D graphics will get much better with latest Imagination PowerVR designs | VentureBeat | Gadgets | by Dean Takahashi

  • Pingback: Low-cost mobile 3D graphics will get much better with latest Imagination PowerVR designs | My Website

  • Jusephe

    Wow that was fast, really fast… I would like to ask, if this is “XT” series like series 5 and series 5XT, will it be possible to scale those series 6 IP cores into something like Power Vr G6650MP2 or not ?

  • Pingback: Low-cost mobile 3D graphics will get much better with latest Imagination Technology’s latest designs | BaciNews

  • MrPoletski

    Will you PLEASE make me a high end GPU for my PC already, I’ve only been waiting since the KYRO for one!

  • Ishant Sharma

    When can we expect to see multi core variants or greater than 6 clusters?
    What can we read into with lack of DX11_1 support?
    How competitive is Rogue v a Qualcomm adreno v tegra K1 v Mali 7xx – is there a benchmark that you can show us that truly outlines the merits of Rogue.
    How many lead partners have taken the design and is their broader interest than original series 6.

  • desi

    G6400 @ 300 Mhz == 120 GFLOPS (estimated from graphs from this blog)

    Tegra K1 @ 950 Mhz == 365 GLOPS

    so if you up the frequency of G6400 * 3 == 360 GFLOPS

    Don’t have the numbers for other gpus.

  • dr.no

    Is 2X increase on same 28 nm or 20 nm chip design.
    same frequency of 300 Mhz as other Rogues GPU #s given.
    Or this just theoretical numbers.

    Also, What is the estimate of H265 encodes being on hardware chip
    timeline wise.

  • dr.no

    Since Intel licenses GPU blocks from Imagination.
    Doubt full any other company can release such a product.
    besides Desktop GPU is dead end as
    all of concerted effort will go on GPU+CPU sharing memory.
    That is why OpenCL put all their effort.

    Even supercomputing Nvidia couldn’t compete without collaborating
    with IBM and even trying to license their GPU cores to smartphone and tablet OEMS.
    and Nvidia announced that any new GPU will come to mobile first.

  • B3nlok

    Id like to know about that as well. Right at the time Series 6 announcement there was all that talk about Multicore vs Multicluster and how the latter is a better and more efficient approach because avoids redundant elements and unnecessarily silicon costs. I wonder if this time around there will be a more flexible approach to the multicore solution.

  • Luca De Marco

    I guess Alex my question is like others, which says whether the series 6xt will be scalable multiprocessor like it was the series 5xt? But I guess this sort of details cannot be disclosed right now.
    Good stuff as usual!!
    Cheers
    L

  • kadajawi

    Interesting. What would interest me most is if anyone is using the D5500 already. i.e. is there any, say MediaTek processor that supports Hi10P, XAVC, HEVC, …

    Also is Raptor meant for cameras as the imaging processor that is connected to the sensor? Any chance of getting it into say a Pentax DSLR, or is it not suitable for that? I find that Fujitsu, the provider of the Pentax processors is seriously lagging behind.

  • http://rys.pixeltards.com/ Rys Sommefeldt

    Hi Jusephe, B3nlok,

    We’re not focussing on multi-core in that way with Series 6 and 6XT, in the sense that we scale all of the architectural elements to create MP configurations. Scaling the compute cluster, texture hardware and render back-ends is the right thing to do today, and that’s how 6 and 6XT scale across the core families.

    We always look at the best way to achieve greater performance for a given IP, so we always consider all available options as we develop future technology, but for today’s market that needs a certain set of features, a certain level of performance, and in a certain area and power budget, the current way of scaling suits our customers’ needs best.

    Cheers,

    Rys Sommefeldt
    Senior Engineer, PowerVR

  • http://withimagination.imgtec.com/index.php/author/alexvoica Alexandru Voica

    Hi Luca,

    Rys has already answered the same question above.

    Regards,
    Alex.

  • http://withimagination.imgtec.com/index.php/author/alexvoica Alexandru Voica

    Hi,

    I’ve stated in the blog article that PowerVR Series6XT GPU cores are up to 50% faster clock for clock, cluster for cluster compared to their Series6 counterparts and that these performance measurements are based on frames from industry standard benchmarks.

    We haven’t released any H.265 (HEVC)-capable encoders yet (only a PowerVR D5500 decoder) but we are working with ecosystem partners to accelerate parts of the encoding algorithm using OpenCL/Renderscript on PowerVR Rogue GPUs.

    Regards,
    Alex.

  • http://withimagination.imgtec.com/index.php/author/alexvoica Alexandru Voica

    Hi,

    There is no theoretical limit to the amount of cluster scaling we can do. Our roadmap is driven, among other factors, by our customers and we haven’t had any requests for a GPU with more than six clusters so far.

    DirectX 11_1 implies the addition of certain features that can impact area and power; again, our customers feel that OpenGL ES 3.0/DirectX 10 is the right combination of features and performance that they need right now. But the Rogue architecture can support DirectX11_1 (as stated when it was first introduced) and you will be seeing this API supported by a PowerVR GPU in the future.

    The article also references the upcoming GFXBench 3.0 benchmark from Kishonti (OpenGL ES 3.0 capable); but there are other relevant benchmarks out there, including 3DMark from Futuremark or Basemark from Rightware where PowerVR Rogue GPUs lead the pack for mobile devices. Moving forward, I think it will be interesting to see how PowerVR ranks up against its competitors when you take power efficiency into account; for example, running the same benchmark 10-20 times and observing how the framerate is impacted.

    Finally, our financial statement states:

    43 licenses including 18 for PowerVR GPU, 14 for MIPS CPU, eight for
    PowerVR VPU (including one for new Raptor camera vision IP) and three
    for Ensigma RPU

    As you can see, we have had an increase in the number of GPU licenses, particularly in markets like Asia.

    Best regards,
    Alex.

  • http://withimagination.imgtec.com/index.php/author/alexvoica Alexandru Voica

    Hi,

    AnandTech usually provides accurate estimates for theoretical peak FP32 performance.

    http://www.anandtech.com/show/7629/imagination-technologies-announces-powervr-series6xt-architecture-available-for-immediate-licensing

    Therefore, if we are comparing based on theoretical data, a G6630/GX6650 at 950 MHz would be equivalent to Tegra K1.

    Obviously, we have to remember that mobile devices have more realistic power envelopes (a smartphone SoC will go up to around 1-2W of peak power consumption while tablets can go up to 4-5W) so frequencies tend to be lower for these types of devices. We’ve also added a number of new and unique power-saving features like ASTC support, PVRIC2 and PVRGC, which enable our customers to achieve much better (and more importantly, sustained) performance without the overheating/throttling issues we’ve seen in competing solutions.

    Regards,
    Alex.

  • http://withimagination.imgtec.com/index.php/author/alexvoica Alexandru Voica

    Hi,

    As much as I yearn for the PowerVR KYRO days, our primary market right now is mobile and embedded (smartphones, tablets, portable game consoles, smart TVs, etc.).

    However, if the right partner comes along, there is nothing preventing our architecture to target the PC and console space or even scale up into power-efficient HPC.

    Best regards,
    Alex.

  • http://withimagination.imgtec.com/index.php/author/alexvoica Alexandru Voica

    Let’s not forget that we are founding members of both the Khronos Group (OpenGL, OpenGL ES, OpenCL, OpenVX, etc.) and the HSA Foundation (HSAIL); both organizations target a very wide market of devices and we are active contributors to all relevant API standards they promote.

    Regards,
    Alex.

  • http://withimagination.imgtec.com/index.php/author/alexvoica Alexandru Voica

    Hi,

    We’ve seen a lot of interest in our video and vision cores from silicon vendors targeting mobile, embedded and automotive markets. Both the PowerVR D5500 and the PowerVR Series2 Raptor ISP IP have been licensed by a number of lead partners.

    The PowerVR Series2 ‘Raptor’ imaging pipeline architecture is designed from the ground-up to be optimized for integration into next-generation System-on-Chips (SoCs) for a broad range of imaging and vision applications. The architecture features simultaneous multi-sensor
    support, 16-bit color depth capability, and 4K support.

    More details on PowerVR Series2 can be found here:

    http://withimagination.imgtec.com/index.php/news/introducing-powervr-raptor-isp-architecture

    Best regards,
    Alex.

  • Pingback: Imagination cranks its PowerVR graphics up to Series 6XT - Liliputing

  • Kingu Prima

    Could it be the GPU of the next iPhone6 ? No DX11 or H.265 support as the NVidia Tegra K1 ? The Tegra K1 platform includes support for H.265 video decode as well, but this isn’t accelerated fully in hardware, rather the decode is split across NVENC and CPU.

  • dr.no

    if G6630 and GX6650 are equivalent in theoretical speed.
    than where is 50% increase coming from.

  • dr.no

    doubtful.

    iphone 6 probably will use 6630
    or faster frequency if using 20 nm chip design
    which gives around 40% increase.

    or Apple will have their own version Imagination GPU called
    Apple GPU.

    It takes 1.5 to 2 years from the time Imagination announces
    and vendor ships in the chip design.

  • Sean Lumly

    Wow.. The Tegra K1 certainly is kicking up a fuss. I have little doubt that these new PowerVR GPUs compete favourably when considering perf/mm2 and perf/watt compared to the K1. Imagination has some impressive tech for increasing performance/memory efficiency generation-over-generation.

    The K1 looks to be a tablet SoC at the rated power draw (5W). I just hope that chip designers that use Imagination IP also create more power-hungry/performant chips that can take advantage of the larger tablet form factor and can compete with the K1.

    It will be interesting to see how 2014 unfolds. But in this industry, and the speed at which it moves, it is generally interesting!

  • http://withimagination.imgtec.com/index.php/author/alexvoica Alexandru Voica

    Hi,

    I think I need to make one clarification first. Tegra K1 is an application processor, complete with a CPU, a GPU, and other relevant blocks. If we are comparing GPU architectures, then we’re talking about PowerVR Rogue vs Nvidia Maxwell.

    H.256 is a video standard that has little relevance when talking about mobile graphics.

    Now, moving on, all PowerVR GPUs are based on a unique TBDR (Tile Based Deferred Rendering) architecture that makes it perfect for mobile. We’ve been refining this architecture for over two decades and the current implementation (Rogue) is the most advanced yet. The package of features presented above provide the perfect solution for providing leading performance while keeping power consumption to a minimum.

    All other competing architectures (Maxwell included) are either IMR (immediate Mode Rendering) or TBR (Tile Based Rendering)-based hybrids that struggle to meet the basic need of any successful graphics family: offer the best performance in a limited power envelope.

    You can claim features, compare at unrealistic frequencies and overload on API support all you want, but at the end of the day, if the device overheats or starts throttling under load, having DirectX 11 or DirectX 10/OpenGL 3.x or 4.x won’t really matter because your application will start to struggle, lag and eventually the device will be too hot to hold.

    Regards,
    Alex.

  • roninja

    To stick to a max of 6 clusters seems a touch conservative to me. Is there more to the Rogue roadmap for instance is the 1tflops performance simply a theoretical marketing gimmic?

    Is Series 6XT designed to sit alongside a future PowerVR RTU?

    What’s the motivation for Series 6XE and 4 different designs – wearables? Cheap handsets/tablets, lo cost TVs and automotive?

    Of the 18 or so new GPU licences in 2013 H1 are these all Series 6XE/XT wins or a mix of older designs even Series 5XT? When can we expect to see this new ip in product is it the usual 18-24 months cycle or sooner?

    Are there customers who see merit in combining MIPS warrior + Series 6XT and committed to this kind of integration?

  • roninja

    When could we expect first series 7 designs and will this gpu ip family include a ray tracing component or will that remain separate ip block?

  • http://withimagination.imgtec.com/index.php/author/alexvoica Alexandru Voica

    As the figure above shows, we’ve added low power ALUs and improved throughput efficiency on the front-end and existing FP32 units; all these combined provide the performance boost.

    Regards,
    Alex.

  • Pingback: Imagination Unveils OpenGL ES 3.0 Compliant Entry Level PowerVR Series6XE GPUs, and Series6XT Architecture Promising TFLOPS Performance

  • Luca De Marco

    Indeed I agree with Alex. Also I think at this stage there is no point to look at a PC market where it’s certainly in crisis. Dont get me wrong, I used to love PC games, but now days most of profit comes from mobile and embedded. It’s better focusing in efficient chip where they can deliver high performance, able to run beautiful games just in the size of your pockets.

  • http://withimagination.imgtec.com/index.php/author/alexvoica Alexandru Voica

    Hi,

    I can’t comment on any specifics regarding our future roadmap products at the moment but the ray tracing IP is a certainty (i.e. it will be introduced as we have promised).

    Best regards,
    Alex.

  • http://withimagination.imgtec.com/index.php/author/alexvoica Alexandru Voica

    The 1 TFLOPS performance promise is not a marketing gimmick, the Rogue architecture can scale to that level. It’s just that most of our customers target mobile and embedded devices that do not require that kind of performance (which you would typically use for desktop/HPC-class devices).

    We will disclose more on our GPU and RTU plans later this year, I can’t make any comments right now. All I can say is things are very exciting from the tech demos I’ve seen.

    Series6XE has been introduced because we’ve seen some of our customer base requesting GPU IP to target the low-end/mid-range mobile and embedded market. There is great opportunity for growth in this segment and we’d like to capitalize on it as much as possible. Offering a diverse, fully-featured and complete roadmap has always been one of our major strengths. The examples you’ve given are very appropriate.

    Most of these design wins are based on Rogue, but Series5XT is still a viable solution for some. Given that the Series6XT designs have already been delivered to customers, we might see them in actual devices sooner than the usual design cycle of 1 1/2 – 2 years (but that obviously can depend on how things evolve from the market’s point of view).

    We are providing support for several projects combining MIPS and PowerVR (and/or more Imagination IP).

    Best regards,
    Alex.

  • Pingback: CES 2014: Imagination zapowiada PowerVR Series6XT - przyszłość GPU w tabletach => Tablety.pl

  • Lucad

    Hi Alex, is It reasonable to think we are gonna to see the series 6xt in final product around end of 2014?
    Cheers
    L

  • ailuros

    Hi Alex,

    Do you mean you’ve added more low power ALUs to 6650 compared to 6630? Because if yes, you also just said that 6630/6650 at 950MHz provide comparable peak FP32 performance to the GK20A GPU in Tegra K1.

  • http://withimagination.imgtec.com/index.php/author/alexvoica Alexandru Voica

    We can’t comment on when OEM partners are planning to integrate processors with PowerVR Series6XT GPUs into a product because a chipset has multiple elements besides a GPU that are not under our control.

    Best regards,
    Alex.

  • http://withimagination.imgtec.com/index.php/author/alexvoica Alexandru Voica

    I was offering either of the two as an example of comparable peak performance; both G6630 or GX6650 are high-end GPUs (one is Series6, the other Series6XT) that can offer that type of raw horsepower, if need be.

    Furthermore, both cores (and Series6XT GPUs, in general) feature PowerGearing, a technology that enables us to (among other things) shut down clusters, which makes a dramatic difference for power consumption. So these cores are also much better at ramping up performance when they have to and then scale back to a power saving state when they’re sitting idle.

    This doesn’t contradict my statement that PowerVR Series6XT GPUs offer better overall performance that Series6 counterparts. I was simply saying that any of the G6630 or GX6650 is a viable candidate.

    Regards,
    Alex.

  • Pingback: Imagination Technologies presenta le nuove GPU PowerVR 6XT e 6XE

  • Pingback: Imagination Technologies presenta le nuove GPU PowerVR 6XT e 6XE | Frog In The Box

  • Pingback: Siti Internet Aziendali » Imagination Technologies presenta le nuove GPU PowerVR 6XT e 6XE

  • Virat Kohli

    What new Series 6 Rogue SoCs are in the pipeline for 2014 … Intel, LG, Mediatek and Apple would be obvious candidates?

  • http://withimagination.imgtec.com/index.php/author/alexvoica Alexandru Voica

    There will be several SoCs with Series6 GPUs released next year. Unfortunately, I can’t disclose which exactly since we are under NDA with our customers. I think you should see more disclosures around MWC (and shortly after) but obviously it all depends on when our licensees decide to introduce them publicly.

    Best regards,
    Alex.

  • Lucad

    Hi Alex, why there are few discussion around the web (Anand included) pointing out that Apple owns his GPU in the A7 chip? Does the A7 has the G6430 or a different GPU design?
    Cheers
    L

  • Roninja

    Let me speculate on Intel Merrifield, LG ODIN octa, hopefully a mobile SoC from Mediatek although they seem to lean on cheap and cheerful Mali.
    A real shame the efforts from STE and TI won’t see light of day and I await what Broadcom have in store via their Renesas mobile acquisition. Whether we see anything from Huawei or unannounced partners is something I eagerly await. Also keen to understand if what Rogue configurations these new SoCs go with.

  • verdantchile

    Take note of the description of the “Hierarchical Scheduling Technology” mentioned in the article. It includes the capability for demand-driven distribution of workload and tiles across multiple GPUs/cores, in the event such a configuration were present.

    The overhead of MP scaling makes it less than ideal. Its flexibility may never be needed within the product lifetime of Series6XT, but the future-proofing is, at least, in there.

  • Virat Kohli

    Just seen the Allwinner A80 announced at CES what Rogue part is it?

  • http://withimagination.imgtec.com/index.php/author/alexvoica Alexandru Voica

    Apple is a licensee and, as any licensee, can access the full range of Imagination IP. Beyond that, I cannot make any comments regarding which specific GPU variant and/or configuration they are using.

    I wouldn’t trust Internet speculation; instead, look for teardowns from third parties. Chipworks is a good place to start.

    Best regards,
    Alex.

  • http://withimagination.imgtec.com/index.php/author/alexvoica Alexandru Voica

    I cannot comment on their announcement at this point. I’ll have to come back to you on this one (either here or in a separate article).

    Regards,
    Alex.

  • Luca De Marco

    Hi Alex, let me be bit naughty and let me attach this link:

    http://www.tomshardware.com/news/lenovo-thinkvision-28-nvidia-tegra-k1-android,25733.html?&_suid=138968438967505111403206828982

    I know it is not your style to comment competitor product, which is fair enough and quite respectable.
    But it would be good to understand your feeling about this benchmark of pre product Tegra K1 inside a Lenevo tablet which will come out on July.
    I know these are merely speculation and early benchmark, but it is quite impressive seeing the gpu of K1 almost doubling the Apple A7 in the gfx 2.7 benchmark.
    We don’t know what it will be the final frequency in the final product, but would you expect similar performance from Imagination design and their licensees?
    Regards
    Luca

  • http://withimagination.imgtec.com/index.php/author/alexvoica Alexandru Voica

    Hi,

    As long as the conversation is civilized and related to Imagination and our technologies, I don’t mind answering questions related to our company’s products and performance.

    Before I do, I think it’s worth pointing out that Lenovo ThinkVision 28 is not a tablet. It’s a professional touchscreen monitor that acts as an AIO (All In One) PC – essentially a desktop computer with the monitor and processor in the same case.

    These devices have slightly different specifications; for example, they can incorporate active or passive cooling and thus can handle higher power consumption and dissipation. They can also be (and typically are) clocked higher than a traditional smartphone/tablet (a form factor usually below 13″, much thinner and battery-powered). I think it’s important we wait until we have an apples to apples comparison (smartphone vs. smartphone or tablet vs. tablet) before jumping to conclusions related to performance.

    We believe we currently have the technology that enables our licensees to offer the same or better levels of graphics and compute performance while (and this is very important) keeping the power consumption adequate for a mobile device. Our unique PowerGearing G6XT technology is essential here. It allows our highest-end GPU designs (GX6650 and beyond) to ramp up raw performance when required and then turn large parts of the design off when unused. This prevents throttling and ensures a reliable, solid user experience. If you add to this the improvements in efficiency we’ve made to our TBDR architecture (all competing solutions are either IMR or brute-force, tile-based designs) and other features like PVR3C, Series6XT offers the best PowerVR GPU IP available now.

    One final point: An important new feature for mobile graphics benchmarks (and something we’ve been advocating for quite some time now) is testing how these processors behave under a typical load (i.e. using the device continuously for 15-20 minutes or more). For example, one could see either how the peak performance is affected after running the benchmark in a loop for, say, 20-30 iterations. It would be very interesting to look at those results too.

    Best regards,
    Alex.

  • Luca De Marco

    Thanks Alex, your are always professional providing exhaustive answers.
    2014 looks pretty exciting!!
    Regards
    Luca

  • Pingback: PowerVR GX6650: redefining performance in mobile with 192 cores | Smartphone & Tablet Design

  • Pingback: Imagination details mucho macho PowerVR GX6650 mobile GPU | htaccess

  • Pingback: Imagination details mucho macho PowerVR GX6650 mobile GPU | facebookgroup

  • Pingback: Imagination details mucho macho PowerVR GX6650 mobile GPU | MyFamilyWay.com

  • Pingback: The Blow Magazine | Imagination details mucho macho PowerVR GX6650 mobile GPU