Rockchip RK3288 vs RK3188 Performance Comparison

We were expecting some new processors at CES 2014, but this CES turned out to be mostly disappointing, and the vast majority of reporters present at the conference, do not seem interested in covering Chinese SoC vendors. Some information however surfaced on several foreign languages blogs and news sites about Rockchip RK3288. First, RK3288 is indeed based on ARM Cortex A12 as initially announced, and not some mysterious ARM Cortex A17 cores, but somebody at Rockchip likely made an embarrassing mistake, and Edward Weinert reported it was corrected at Rockchip booth with  pieces of paper as shown below.

So Rockchip RK3288 has four of these Cortex A12 core, a 4L2K H.265 / HEVC video decoder, and an high-end Mali T764 T624 GPU supporting OpenGL ES3.0 and OpenCL 1.1 that vastly outperforms Mali-400 MP4 found in RK3188, and is even significantly faster than Adreno 330 GPU found in Qualcomm Snapdragon 800 SoC in Kindle Fire HDX.

The first column is probably AMLogic M802 with Mali-450MP6. If you upgrade from RK3188 to RK3288, you’ll get 4 to nearly 6 times more 3D graphics performance based on GFXBench benchmarks. The company did not include the CPU performance, as Snapdragon 800 will clearly have an edge in this area.

The maximum frequency for RK3188 is 1.6GHz (despite being told it would go up to 1.8 GHz at launch), whereas RK3288 can be clocked up to 1.8 GHz, and moving from Cortex A9 to Cortex A12 improves performance from 2.5 DMIPS/Mhz to 3.3 DMIPS/MHz, meaning RK3288 is nearly 50% faster, in theory, compared to RK3188 when it comes to CPU performance.

We’ll have to get more benchmark results to get a complete picture of the performance, but we already know CPU performance should be pretty decent, and GPU performance be top of the class.

RK3288 is said to cost about $40 for now, with the price probably coming down to $35 in the second quarter. One of the first devices with RK3288 SoC will be allegedly be Yuandao M12 tablet.

Share this:

Support CNX Software! Donate via PayPal or cryptocurrencies, become a Patron on Patreon, or buy review samples

35 Replies to “Rockchip RK3288 vs RK3188 Performance Comparison”

  1. No info yet if the HVEC / H.265 decoding is really true hardware assisted video decoding on VPU, or if is simply software decoding on CPU or perhaps mainly software decoding on CPU with assistance from GPU shaders?

  2. @Harley
    “Real 4Kx2K H.265 Decoder”. So it would be highly misleading if it was not done by the VPU.
    4K2K HEVC decoding is simply impossible with the CPU only, and probably challenging (impossible?) with GPU compute.

  3. @cnxsoft
    I would think 4K HEVC decoding via OpenCL would be well within the capability of a Mali T760 series GPU. 4K HEVC encoding too, possibly. But using the VPU to do it instead is probably considerably more power efficient though.

  4. So where’s the Allwinner A80 that was supposed to be announced at CES 2014 ? I find that one much more interesting than this Rockchip and we still don’t know what GPU it will use.
    Whatever GPU the A80 will use I prefer raw CPU power to GPU power and the A80 will pretty much blow other Chinese SOCs out of the water when it comes to CPU power.

  5. @anonymous
    Yes, it appears you are correct. “Mali acceleration opens up the possibility of 1080p @ 60 fps and 4K x 2K @ 30 fps HEVC decode on mobile devices without dedicated HW”.

    Source: http://malideveloper.arm.com/downloads/Unleashing_Benefits_GPU_Computing_Mali-T600_ROBERTO_MIJAT_TECHSYMPOSIA.pdf

    @Marius
    No sources in English, so I’ve checked recent articles on Chinese sites, but they only talk about the A80 board, nothing about the GPU yet, nor pricing.

  6. I suppose there is a mistake on GPU too. That was first announced with a Mali-T624, and on the erroneous annouce with an inexistant Mali-T764. Now That’s a Mali-T76x ??? instead of a mali-T760 in the Mali-T7xx serie ???

    I suppose, as first announced, this summer, the GPU is instead a Mali-T624, does someone have the answer ?

    List of existing T6xx & T7xx Mali GPU on ARM website :
    http://www.arm.com/zh/products/multimedia/mali-high-end-graphics/index.php

  7. I do own 2 rock chip device rk3188 and one allwinner A31 and so far better graphics gpu ,brightness, sharpness, contrast.

  8. @Noloqoq
    Damn, I should have checked this too. Based on the 300Mtri/s, it can’t be Mali-T7xx either, so it must just be Mali-T624 instead, as they said the very first time. What did they smoke or drink when they prepared this?

  9. I spoke to the head technical guy at the Rockchip CES 2014 booth for a good 10 minutes on Thursday.
    While there was a bit of a language barrier at first, it was clear he expects Rockchip’s new quad core ARM SoC to outperform expectations..
    He was very excited about the RK3288 chip which he says will be available in small quantities around the first of February 2014.
    I asked him for his best estimate of the increase in “real world” Android tablet performance one could expect if one replaced an RK3188 chip directly with an RK3288.
    He seemed pretty confident in his claim that the RK3288 would outperform the RK3188 by 40%.
    He gave the following as reasons for the expected increase:
    1. The RK3288 uses the same 28nm manufacturing process, but unlike the RK3188, the RK3288’s cores are Cortex-A12 and they can actually run at 1.8Ghz.
    2. The RK3288 has a new dual channel DDR controller that he expects will greatly improve DDR memory access times.
    3. The new Quad-core MaliT76X will scream compared to the RK3188’s Mali 400.
    He also told me that the first Android tablet to hit the streets with the new chip will likely be an Asus.

  10. @cnxsoft
    Haha, I’ve seen this before with Chinese companies, they’re not very sure about what they’re making 😀 Kinda makes you wonder how they actually manage to get products out.
    I think the main thing here is that they don’t really put technical people in charge of this, it’s the marketing folks and stuff and most of the time they have no idea what a GPU is.

  11. @cnxsoft
    Yeah … fail but I’ll forgive them is they put a SATA port on it. And that fex thing of theirs is just great … you don’t need the binary kernel for that perticular gadget so it’s still better than Rockchip and the other unless they go the device tree way.
    Having a tablet and just putting in the Alwinner kernel and booting is more than you can expect from ARM hardware these days… kinda makes it feel like X86 🙂

  12. cnxsoft :
    @Noloqoq
    Damn, I should have checked this too. Based on the 300Mtri/s, it can’t be Mali-T7xx either, so it must just be Mali-T624 instead, as they said the very first time. What did they smoke or drink when they prepared this?

    Someone probably did it deliberately. Its common in China. IT would look “good” during chinese trade show – better numbers than competition. Here in English speaking world it just makes them look incompetent and retarded liars.

  13. @rasz_pl
    You are right. It is all too common within the industry in China to overstate or even flat out lie about specs on upcoming or even released products. There seems to be an attitude that “everyone else” is doing it, so we don’t want to look bad in comparison.

  14. Rafael duarte :
    Allwinner I like that idea to keep working closely with imagination technology

    I actually hope that Allwinner stops using PowerVR as they are infamous for using closed source drivers and slow with closed source driver updated on Linux.

    Well at least I would hope that Allwinner and others using IM GPUs would try to pressure them to open up their drivers more, preferably to open source, but at least more compatible licenses and better frequent linux support.

  15. Harley :

    Rafael duarte :
    Allwinner I like that idea to keep working closely with imagination technology

    I actually hope that Allwinner stops using PowerVR as they are infamous for using closed source drivers and slow with closed source driver updated on Linux.
    Well at least I would hope that Allwinner and others using IM GPUs would try to pressure them to open up their drivers more, preferably to open source, but at least more compatible licenses and better frequent linux support.

    Pretty hard to believe Allwinner would pressure someone else to release open source software when they’re not too good at it themselves.
    Even if they wanted to though I’m not sure they have the power ( pun intended ) to push PowerVR. The only ones that could do anything about it were TI which were close to making a deal by which Rob Clark would have made an open source PowerVR driver. TI then decided they don’t want to do multimedia SOCs and that was that.

  16. @rasz_pl
    Totaly agree, that’s like american companies like Intel that say their CPU only consume 10W (as mean value) and compare this to a whole system that consume 5W at TDP, or nvidia that say they created new technology for screen refresh rate, where all the other do the same, or Broadcomm that say they finnaly give full opensource driver for raspberry pi (and rasb pi that repeat, where most important things is kept closed ;). Go to your bank, you will probably have offer for « Really interesting » services too…

    Everyone in commerce lie (if at least commercial know the product) and no one is reliable if you don’t know the product you buy and hadn’t verified the fact.

  17. I’ve spent quite a bit of time with the RK3188 via my Minix Neo X7. I’ve ran this system on Android 4.2, Ubuntu 12.04-ARM, and PicUntu-4.5.

    Android has ran the best, likely due to the GPU drivers having hardware acceleration. I currently keep the Finless 1.7 firmware installed and it’s working very well overall.

    Ubuntu 12.04-ARM worked pretty well overall, though the lack of accelerated GPU drivers was a deal breaker, because I *can* get full 3D with Ubuntu on an x86 system, such as my Zotac ZboxHD-ID41-Plus. Outside of this one major issue, Ubuntu otherwise ran very well and had very few problems that I couldn’t simply work out within an hour or less. Wireless worked fine, though I don’t recall if I got eth0 up or not… it’s been several weeks and I didn’t document the procedure.

    PicUntu-4.5 was the easy to install, and could have been a winner… except for the fact that it’s worse off than the TronSmart rom I used for installing Ubuntu. The network drivers are very lacking and were unable to detect nor run wireless or ethernet. Also, PicUntu lacked full GPU acceleration, so like the Ubuntu image, it was a deal breaker.

    Since doing this project, I’ve done a bit of research and found systems that will work better for this type of project. Though, at this time –unless the GPU hardware acceleration is worked out– I think it would be a waste of money to proceed. I really love the idea of a pocket sized Linux embedded workstation. Unfortunately, while I’m one hell of a system designer, I have no background with writing drivers, or I would try to solve that GPU issue myself.

    Regardless, I look forward to products that include the RK32xx SoC. The RK3188 I’m currently using (Minix Neo X7) is pretty decent for *most* of the software I run on it. When I bought this system, I was looking for an alternative to the OUYA game console, and overall I’ve gotta give it to my Minix… it smokes the OUYA, if based on specs alone. But I’m really looking forward to a next gen SoC that I can use to replace my Minix as my main system. At that point, I can dedicate my Minix to ARM-Linux projects and take my time about working out the GPU acceleration, because all things considered, I’d really like to see Ubuntu running on this device and it be *completely* usable. The port-layout of the Minix Neo X7 is fantastic for a micro-PC running Linux and would give me exactly what I’m looking for in an embedded workstation. But, it has to be solid, because I do some pretty serious work with Linux (both professionally and as a hobby).

Leave a Reply

Your email address will not be published.

Advertisement
Advertisement