PowerVR G6630: go fast or go home

Here at Imagination we have a simple goal: to set a new benchmark in efficiency with each GPU core of the PowerVR graphics IP family we release. Efficiency essentially means that each core is optimized to deliver the highest performance possible with the lowest power consumption. The 2012 Embedded Technology Conference and Exhibition (ET) was the perfect place to launch the newest member of the PowerVR Series6 ‘Rogue’ family: the PowerVR G6630. ET is one of the world’s largest events for system designers and managers, covering embedded technology, electronics, software, mobile and embedded systems, and semiconductors and Imagination has maintained a strong presence at this event over the years.

Meet the PowerVR G6630: One core, six clusters, endless possibilities

The PowerVR G6630 is the fifth member of the Series6 ‘Rogue’ graphics IP family. It all started almost a year ago, when we introduced the first two members, the PowerVR G6200 and G6400, which were designed to deliver the best performance at the smallest area possible for two and four cluster architectures. With PowerVR G6230 and G6430 we went ‘all out’, adding incremental area to maximize performance whilst minimizing power consumption.

PowerVR GPU roadmap - Series5 Series5XT Series6PowerVR graphics IP roadmap: Series5/5XT SGX and Series6 ‘Rogue’

Now, PowerVR G6630 delivers jaw-dropping graphics and blisteringly fast GPU compute capabilities with six clusters packed inside one GPU core. Thanks to our hard-working engineering team, our unique Series6 ‘Rogue’ architecture allows Imagination to quickly scale the number of clusters inside a core. By offering a dynamic mixture of performance points at different power consumption levels, system designers can choose this six cluster configuration to address higher performance points within mobile processing. As we’ve recently seen, innovation in mobile computing is driven more and more by integrating the most appropriate GPU within the SoC because the embedded world is all about developing killer apps on the most efficient hardware – and often enough you will find that’s Imagination’s PowerVR graphics.

The PowerVR G6630 is the latest release of the Series6 family, demonstrating versatile scaling and efficiency. This core offers up to an amazing 60x compute and graphics throughput increase compared to our previous  generation graphics IP cores. Many of the platforms shipping from 2013 will be capable to take advantage of PowerVR G6630’s ability to deliver the latest features of the industry’s major APIs including OpenGL ES 3.0* and OpenCL 1.x from Khronos or DirectX10 from Microsoft.

PowerVR G6630 block diagram

PowerVR G6630 block diagram

All our ‘Rogue’ cores also include a robust power management mechanism. We’ve defined power islands which can be turned on or off dynamically to optimize power consumption. For example, the PowerVR G6630 enables each cluster pairs to be managed, allowing three modes of operations (X2, X4 and X6) which will help extend the ever-essential battery life of mobile and embedded computing systems.

Additional features include advanced autonomous processing and enhanced, dedicated scheduling via a C programmable microkernel based on Imagination’s Meta CPU technology and a dedicated 2D processor to offload common duties from the shader cluster core such as block image transfers allowing the core to perform basic graphics management tasks. All these improvements are designed to further reduce the overall system power consumption as well as deliver a highly latency tolerant graphics core that requires the lowest memory bandwidth in the industry.

PowerVR Series6 ‘Rogue’: The bigger picture

The PowerVR Series6 ‘Rogue’ architecture is based on a scalable number of compute clusters. This enables mobile computing platforms integrating PowerVR GPUs to be used for a growing range of applications from mobile to high performance embedded graphics including smartphones, tablets, ultra-portable laptops and PCs to automotive, STBs and smart TVs.

PowerVR Series6 GPUs utilize programmable computing elements known as clusters that are designed to offer the highest density of computing performance at the lowest possible power consumption. Clusters are arranged in multiple pairs to form an array of compute elements. Power and performance are further improved by providing optimized compute and graphics data paths, keeping processing on chip and by employing a wide range of lossless and lossy data compression techniques within the processing pipeline.

 

PowerVR Series6 family:  PowerVR G6630, PowerVR G6430 and PowerVR G6230

PowerVR Series6 graphics and compute performance overview

With PowerVR Series6 GPUs we’ve taken our patented tile-based deferred rendering (TBDR) technology  and created a design that reaches from a 100 GFLOPS up to more than one TFLOPS of computing performance with significantly greater efficiency relative to our previous cores. This essentially means our partners will deliver platforms with compute capabilities comparable to current generation mains-powered games console but battery powered. Graphics performance has also seen a significant increase compared to previous generations, as Rogue cores are designed to ensure superior fill rates and shader performance for the latest high-resolution smartphones, tablets and TVs as well as multi-screen support for the automotive industry and demanding console-quality gaming.

All PowerVR Series6 cores, including G6630, support the industry-leading PVRTC2™ and PVRTC™ texture compression standards, giving developers a high degree of flexibility and unprecedented image quality levels at 4bpp and 2bpp encoding modes as well as all major Khronos and Microsoft standards.

PowerVR graphics: one GPU solution for all standards

The PowerVR Series6 family is designed to meet all requirements for the latest graphics and compute APIs including OpenGL ES 3.0*/2.0/1.1, OpenGL 3.x/4.x, OpenCL 1.x and full WHQL-compliant DirectX10, with certain family members extending their capabilities to DirectX11.1. The unified shader array architecture offers linear compute scalability within a smaller area compared to competing GPUs, as much of the redundant multicore maintenance logic is no longer needed. One G6xxx core is thus able to offer more compute resources than a design that splits the GPU into independent cores.

Imagination’s GPU technology has seen an impressive growth over the past two decades. With roots in the desktop and console market, we now have one of the world’s largest engineering teams dedicated to graphics processor development. This industry-wide effort is complemented by our PowerVR Insider ecosystem of dedicated third party developers (now counting more than 35,000 members) and a powerful tool chain including our PowerVR Insider Graphics SDK available for free download.

The PowerVR GPU IP family maintains a leading position within the embedded graphics segment, with over 90 licenses by leading semiconductor companies which have shipped more than 600m devices with Imagination graphics inside. And the story continues: PowerVR Series6 has already secured an impressive number of lead licensees and partners, including industry-leading semiconductor and electronics companies like HiSilicon, ST-Ericsson, Texas Instruments, Renesas and MediaTek.

Excited about our new PowerVR Series6 line-up of graphics cores? Then don’t forget to follow us on Twitter (@ImaginationPR and @PowerVRInsider) for the latest announcements from Imagination, its partners and the mobile computing market.

 


* OpenGL ES 3.0 compliance will be subject to successfully passing OpenGL ES 3.0 conformance tests from Khronos (not yet released as of November 2012)

, , , , , , , , , , , , , , , , ,

  • Sean Lumly

    Wow… The 6630 sounds absolutely beastly, though efficiently arranged thanks largely to the multi-cluster configuration. What is the difference in die area compared to a traditional SGX similar-pipeline-count setup? I assume the GLES30 and CL1.1 functionality, as well as FB compression, PVRTC2, etc will probably add appreciably to the die size.

    I’m also very interested in learning more of how the new architecture improves latency tolerance over prior designs. Is this a compression thing or does the core somehow coordinate pending jobs with memory fetches?

    Thanks for the block diagram! With rogue, ImgTec is sure to maintain a healthy lead in per mm^2 performance compared to the competition. In fact, as an Android developer, I am very interested in targeting the ST-Ericsson NovaThor SoC as I understand that will be implementing one of these designs.

  • Ahmad Tariq

    so can we play pc games by this ?

  • alexvoica

    Hi Sean,

    Thanks for your comments, we’ve supported OpenCL 1.1 and PVRTC2 with Series5XT too, therefore the increase in area is only incremental. We’ve also launched the PowerVR G6200 and G6400, which are area-optimized cores for the best possible performance and worked closely with Khoronos to optimize our IP for OpenGL ES 3.0, making sure the new features are fully supported and not just tick-list implementations.

    For more details on the Series6 architecture and its new features, you can have a look at the micro-site dedicated to our graphics IP family

    http://www.imgtec.com/powervr/sgx_series6.asp

    Also, if you need help with your development work for Android, please don’t hesitate to use our dedicated PowerVR Insider forum, where our Dev Tech team can offer the support you need.

    http://www.imgtec.com/forum/default.asp

    We regularly attend events like GDC, gamescom and Droidcon, so look out for our presentations there too.

    Best regards,
    Alex.

  • alexvoica

    Our IP can be found in a number of devices, therefore it may be possible that you have already played a game on a laptop or netbook that has our graphics.

    With Series6, we’re targeting a wide range of applications and form factors, but above all, we’re offering high-performance graphics and compute that require very little power, making efficiency one of the key drivers for our architecture.
    Regards,
    Alex.

  • Jesse Lee

    I Wonder how many GFLOPS the PowerVR G6630 has?

  • alexvoica

    Hi Jesse,

    Although our Series6 cores can deliver impressive raw compute and graphics performance, we believe the focus in mobile computing should be on efficiency. We’ve designed and optimized these graphics cores to offer the right balance between performance, power and area, therefore making every FLOPS per clock count.

    For an idea of the theoretical peak performance of our GPU IP, have a look at the graph in the article below

    http://withimagination.imgtec.com/index.php/powervr/the-rise-of-gpu-compute

    Regards,
    Alex.

  • Jesse Lee

    I wonder if there are Linux/X11 drivers for PowerVR 6 series GPUs? Android is good but you can’t do very much with it.

  • Jesse Lee

    Wow the Series 6 makes the 5 and 5XT series look weak by comparison. I wonder where the G6630 is on that graph? The clocks on the 6 Series of GPUs are really high 600mhz that’s pretty high.

  • faisal

    i think g6630 theoretical peak performance 315gflops, am i right? what about texel?

  • alexvoica

    Hi,

    The theoretical peak performance varies with a lot of factors, including frequency.
    The real question is how close you can come to that theoretical peak performance in real-life usage scenarios. Our Series6 cores have been designed for intensive graphics and compute efficiency, therefore applications will be able to make full use of that raw performance.

    Regards,
    Alex.

  • alexvoica

    Hi Jesse,

    That is just an example, we’re working with our partners to understand what configuration suits them best.

    You have to remember though that the Series5/5XT cores have been able to behave quite well in terms of performance and power consumption, making even single core GPUs a very compelling solution at higher frequencies.

    Have a look at the TI’s OMAP 4 lineup

    http://en.wikipedia.org/wiki/OMAP#OMAP_4

    or

    Intel’s Atom Z2460 (Medfield)

    http://download.intel.com/newsroom/kits/ces/2012/pdfs/AtomprocessorZ2460.pdf

    and you’ll notice double the frequency of what we’ve quoted in our article for SGX.

    Regards,
    Alex.

  • alexvoica

    We’ve announced we would have certain DirectX 11 compatible cores.

    http://www.imgtec.com/News/Release/index.asp?NewsID=666

    “All members of the Series6 family support all features of the latest graphics APIs including OpenGL ES 3.0, OpenGL 3.x/4.x, OpenCL 1.x and DirectX10 with certain family members extending their capabilities to full WHQL-compliant DirectX11.1 functionality.”

  • Sean Lumly

    Thanks Alex!

    That information is very helpful, and I think I have a better idea of what’s going on, albeit from a very high level. Of course I have far more questions regarding the hardware, but I will refrain from asking them; if I don’t, they will never stop.. :) From a software perspective, I will likely take you upon your offer to join the insider forum.

    But I do have one more question that I am curious about: Does the 6×30 series support the ASTC format?

  • alexvoica

    Our PowerVR graphics architecture has and will continue to support all major texture compression standards alongside our PVRTC and PVRTC2 texture compression standard. ASTC support will be included in future graphics cores.

  • Sean Lumly

    Thanks again, Alex!

  • Ted

    It would be interesting to see how dual-core atom with G6630 on a windows 8 tablet will perform.

  • alexvoica

    Hi Ted,

    We’re looking forward to our partners starting to ship SoCs with Series6 as well. Until then, here are some previews of the CloverTrail platform with a PowerVR Series5 GPU

    http://www.fudzilla.com/home/item/28840-intel-shows-us-clover-trail-performance-on-a-tablet

    http://hothardware.com/Reviews/Intel-Clovertrail-Atom-Z2760-Windows-8-Tablet-Performance-Preview/

    and the latest article from Intel on developing and optimizing apps on Android for mobile GPUs

    http://software.intel.com/en-us/articles/intel-for-android-developers-learning-series-10understanding-gpus-on-android-for-the-intel

    Regards,
    Alex.

  • http://www.facebook.com/costelas.leontin Costelas Leontin

    Powervr series 6 GPUs have been announced in january 2012 and up until now( 3 days more till 2013) there hasn t been 1 mobile device to actually incorporate this type of GPU. When will devices with G6630 apear? In 2014 when the x60 boost wont be of any use? Its so odd to see a new generation that it is said to be the BEST in tones of news but there isnt 1 single device to support that.

  • alexvoica

    Hi,

    As you might know, there is always a gap between an IP product being announced and that certain IP product reaching consumers through devices. In the mobile world, this gap is usually about 12-18 months wide. This is valid not only for GPUs but for CPUs and other processors as well. Therefore, we expect to see PowerVR Series6 reaching a device near you in the upcoming year.

    In the meantime, you can follow us on Twitter for PowerVR-related reports and articles from various shows, conferences and events both Imagination and our partners attend world-wide, where we usually showcase some of the devices about to reach the market.

    Happy holidays,
    Alex.

  • Pingback: Next generation Xbox and PlayStation to come in March?

  • http://www.facebook.com/bonis.macho Bónis Machó

    I don’t think that directx support is relevant for linux users -.-

    (but this is an existing problem: if you won’t write drivers for linux – linux users will not choose your product plus they will cry in your forums: ‘I bought a netbook with intel atom, but ubuntu doesn’t give screen, what should i do?’ or something similar)

  • alexvoica

    Hi,

    We support all major operating systems, including Windows, Android and Linux. As you can see in the video below, we are running our OpenGL ES 3.0 demos from a Ubuntu machine so the Linux drivers are integrated on our FPGA configuration.

    http://www.youtube.com/watch?v=GwaUQPNsyUE

    Actually, we have submitted this exact platform to Khronos for OpenGL ES 3.0 compliance:

    “Conformance submission is on Imagination’s Rogue Test Chip on Linux 3.2.0.”

    http://www.imgtec.com/News/Release/index.asp?NewsID=723

    It is our partners’ choice to support specific operating systems on their platforms, which means some of these systems will target only one (and/or a limited number of) OS.

    Best regards,
    Alex Voica.

  • Pingback: Renesas introduces APE6 8-core processor for phones and tablets | vyagers

  • Pingback: Renesas introduces APE6 8-core processor for phones and tablets - Liliputing

  • Pingback: La guerra del hardware para 2013: Qualcomm, NVIDIA, Samsung y el resto | Andro4all

  • robert de niro

    Would be g6200 better than sgx 554mp4 in terms of pushing more polygons in games ? If not which variant will be better than sgx 554mp4 in that term and how much better approximately ?

  • alexvoica

    Hi,

    PowerVR Series6 GPU cores feature a number of improvements over Series5XT GPUs, therefore we will see performance gains in both fillrate, image quality, GFLOPS as well as lower power consumption and higher efficiency compared to our previous generation.

    When talking about fillrates in particular, the frequency at which each GPU is clocked at can heavily influence GPix/s results. For example, when theoretically comparing the PowerVR G6100 and SGX5xxMP2 GPU cores at similar frequencies, the first one provides a slightly better fillrate number than the latter.

    http://withimagination.imgtec.com/index.php/powervr/powervr-g6100-small-is-powerful

    Regards,
    Alex.

  • robert de niro

    tnx !

  • faisal

    Alex, like you said that each powervr rogue core containing up to 8 cluster, how to reach above 8 cluster? is multicore possible?

  • alexvoica

    Hi,

    PowerVR Series6 ‘Rogue’ cores deploy a pipeline cluster approach, with each GPU core scaling up to 8 clusters, and each cluster containing up
    to 16 pipelines.

    We’ve obviously got a roadmap that will continue to address multiple performance points. Unfortunately, I can’t discuss specific upcoming products right now so you’ll have to wait for the official announcement.

    Regards,
    Alex.

  • faisal

    Thanks for your reply Alex, i hope soon powervr rogue new configuration use tesselation co-processor and the graphic superior than ps4 or xbox infinity less than 3 years.

  • teamblackrogue

    Alex we’re small firm interested in launching a mobile android game console and we wanted to know from you if it is possible to get an SOC manufacturer to integrate this into an existing SOC, preferably quad core?

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

    Hi,

    There are currently several silicon providers shipping quad-core SoCs with PowerVR Series5XT-based graphics, including:
    - Allwinner A31, A31s
    - MediaTek MT6589, MT8125

    If you are interested in PowerVR Series6-based platforms, there are several upcoming customer announcements in the pipeline but we aren’t in a position to disclose their configurations yet.

    Best regards,
    Alex.

  • Pingback: Speculation time: To understand Apple’s A7, just look at its predecessors | LiMiT

  • Pingback: Speculation time: To understand Apple’s A7, just look at its predecessors |

  • sean chung

    iphone 5 gpu is faster than g6430 in triangle throughtput ……why

  • Jancox13

    Tegra K1 and Mali T760 still better

  • Jancox13

    PowerVR G6630 has 250 GFLOPS

    Tegra K1 has 365 GFLOPS
    Mali T760 has 326.4 GFLOPS

  • Anyone

    Where can I buy products NOW with any of the three?

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

    Where are you getting your numbers from and what frequency are you counting those GFLOPS at?

    PowerVR G6630 has a similar number of GFLOPS as the Tegra K1.

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

    Why?

  • Jancox13

    Those GPU not yet released or no devices use those GPU yet.

    maybe will released this year Q4…

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

    You can find out more about the GFLOPS count of PowerVR Rogue GPUs here

    http://blog.imgtec.com/powervr/graphics-cores-trying-compare-apples-apples

    PowerVR G6630 has 192 FP32 ALU cores with up to 384 FLOPs per cycle – 230.4 FP32 GFLOPS at 600MHz.

    Regards,
    Alex.

  • Jancox13

    I found “PowerVR GX6650: six Series6XT USCs – 192 FP32 ALU cores with up to 384 FLOPs per cycle – 384 FP16 ALU cores with up to 786 FLOPs per cycle. That means up to 460.8 FP16 GFLOPS and up to 230.4 FP32 GFLOPS at 600MHz” that was G6650 not 6630 :))

    That was fast… I dont even found any device available with SGX G6630 GPU then got another better PVR SGX series already.

    I hope that was right. Becoz I am true gamer very wanted a great GPU for my devices. don’t care which brand it come from. :))
    hope not for iPhone only. bcoz iPhone is too overprice here.

    thx for info.