Unconfigured Ad Widget

Collapse

Announcement

Collapse
No announcement yet.

The latest news about RK3288

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • Shomari
    replied
    The RK3288, although not an A17, is a very capable chip. I personally have no complaints at all at this stage of development. Finally a quad-core that has significant gains over the dual core performance of RK3066


    ...it's also a 40nm design like the RK3066, not 28nm like the RK3188. Still, the DVFS implementation and overall power consumption seems to be improved.

    Leave a comment:


  • Haba
    replied
    Originally posted by no_spam_for_me View Post
    But, Haba, your explanation doesn't really fits at all.
    It doesn't really fits at all because ARM talk about big.LITTLE at A17, so the explanation for the Allwinner A80 is ok http://www.allwinnertech.com/en/clq/processora/A80.html, because they talk about 8 cores (4 * A15 and 4 * A7) but your explanation doesn't fit to the RK3288, because they talk about 4 cores, so it must be [4 * A12] or if it is an A17 [2 * A1? and 2 * A7]???
    And, also in my eyes, the T764 isn't an indicator for the A17 because ARM doesn't know a T764 (I know that some tools tells it's a T764 but I don't know how and where they determine this information)???
    bigLITTLE is an option for A12-17 and is not required. You can have a four core A12-17 NOT bigLITTLE SOC such as the RK3288. The RK3288 has a four core A12 or 17 CPU and a Mali-T764 which is a Mali-T760 MP4 (four GPU cores). There is not a lot of difference between the 12 or 17, more importantly as stated testing is showing really good results. Perhaps only Rockchip can answer all the questions about testing and code checking showing A12 while all of their documentation clearly states it is a A17 chip.

    Leave a comment:


  • Shomari
    replied
    Originally posted by Haba View Post
    BigLITTLE is a combination of smaller high efficient cores like Cortex-A7 and high powered A12-17 cores. How many of each is up to the manufacturer up to a limit. Allwinner A80 has four A-7 and four A-15 cores for example . The A-15 are high power/power hungry and used only when needed, the A-7 are used most of the time to save battery and heat in mobile devices. TV boxes do not need this big little at all, I would rather have more gpu cores. The CPU's don't have to be bigLittle either instead they just use high or low powered cores like the RK3288 which uses four high powered cores. I still think the RK3288 has four A-17 because of the current documentation from Rockchip and elsewhere and the 700 series gpu. The A12 is always listed with the 600 series gpu. Software tools and code may not be detecting it properly.

    Do a dmesg and look at the read back of hw perfevents PMU driver. Every bit of documentation, including the architecture part identification code points to A12.

    Leave a comment:


  • no_spam_for_me
    replied
    But, Haba, your explanation doesn't really fits at all.
    It doesn't really fits at all because ARM talk about big.LITTLE at A17, so the explanation for the Allwinner A80 is ok http://www.allwinnertech.com/en/clq/processora/A80.html, because they talk about 8 cores (4 * A15 and 4 * A7) but your explanation doesn't fit to the RK3288, because they talk about 4 cores, so it must be [4 * A12] or if it is an A17 [2 * A1? and 2 * A7]???
    And, also in my eyes, the T764 isn't an indicator for the A17 because ARM doesn't know a T764 (I know that some tools tells it's a T764 but I don't know how and where they determine this information)???

    Leave a comment:


  • Haba
    replied
    Originally posted by no_spam_for_me View Post
    What I don't really understand is how do they count the cores at big.LITTLE?

    If the rk3288 is an A17 http://www.arm.com/products/processo...=TopNaviL27328 they count 4 cores (BTW: ARMv7-A), but if I look at the big.LITTLE http://www.arm.com/products/processo...processing.php they talk about "high-performance ARM CPU cores are combined with the most efficient ARM CPU cores", so in my eyes, if they count 2 hp-cores and 2 me-cores, the A12 (BTW: also ARMv7-A) must be better at TV Boxes...???
    BigLITTLE is a combination of smaller high efficient cores like Cortex-A7 and high powered A12-17 cores. How many of each is up to the manufacturer up to a limit. Allwinner A80 has four A-7 and four A-15 cores for example . The A-15 are high power/power hungry and used only when needed, the A-7 are used most of the time to save battery and heat in mobile devices. TV boxes do not need this big little at all, I would rather have more gpu cores. The CPU's don't have to be bigLittle either instead they just use high or low powered cores like the RK3288 which uses four high powered cores. I still think the RK3288 has four A-17 because of the current documentation from Rockchip and elsewhere and the 700 series gpu. The A12 is always listed with the 600 series gpu. Software tools and code may not be detecting it properly.

    Leave a comment:


  • Shomari
    replied
    Originally posted by no_spam_for_me View Post
    What I don't really understand is how do they count the cores at big.LITTLE?

    If the rk3288 is an A17 http://www.arm.com/products/processo...=TopNaviL27328 they count 4 cores (BTW: ARMv7-A), but if I look at the big.LITTLE http://www.arm.com/products/processo...processing.php they talk about "high-performance ARM CPU cores are combined with the most efficient ARM CPU cores", so in my eyes, if they count 2 hp-cores and 2 me-cores, the A12 (BTW: also ARMv7-A) must be better at TV Boxes...???
    I'm currently explaining the same thing to Bluetimes reps while trying to get them to update their product listings; from what I've seen, this generation of A12 Cortex chips (such as RK3288) are perfectly fine for tv boxes and arguably better suited than A17 would be, anyway.

    Leave a comment:


  • PolloLoco
    replied
    Originally posted by Javimetal View Post
    Do you have any proof? I'd like to ask Rockchip about this.


    About the A80, today I was checking a new board with this SoC, but it looks like need some time yet to be ready and stable.
    While it's slightly disappointing to hear that the Rk3288 is only an a12, its really the only game in town right now. I was interested in the a80 chip, but the only board taking orders (Optimus board) was selling for $345! That's insane.

    Leave a comment:


  • no_spam_for_me
    replied
    Originally posted by bluesmanuk View Post
    I've mailed Arm for clarification based upon the articles published so far and also asked the question in the Arm community forums.
    What I don't really understand is how do they count the cores at big.LITTLE?

    If the rk3288 is an A17 http://www.arm.com/products/processo...=TopNaviL27328 they count 4 cores (BTW: ARMv7-A), but if I look at the big.LITTLE http://www.arm.com/products/processo...processing.php they talk about "high-performance ARM CPU cores are combined with the most efficient ARM CPU cores", so in my eyes, if they count 2 hp-cores and 2 me-cores, the A12 (BTW: also ARMv7-A) must be better at TV Boxes...???

    Leave a comment:


  • Haba
    replied
    Originally posted by Javimetal View Post
    Well, today I have been with a member of ROCKCHIP, so I connected a RK3288 board to my laptop by OTG port, opened my windows cmd, adb shell.....

    cd sys/devices
    and there was the "ARMv7 Cortex-A12"

    He said must contact ARM to know more about this and give me a reply, we exchanged contacts and also have been talking about the 4K - 60Hz, he said this is already fixed, even they bought some TVs from Japan directly to test this.


    Well... more news soon about this as I have direct contact with Rockchip guys.
    I agree it would be good to find out more info. We do know the RK3288 is very advanced CPU with an advanced GPU that is at least a 40% more powerful per clock cycle and has additional features compared to CPU's such as the Cortex-A9 AML s802. The big-little design I see for tablets/phones when battery life/heat issues are a major concern, I don't see much interest in the tv box with this feature. I hope Rockchip will stand behind the RK3288 and work out the remaining issues so that everything is working properly and it's rock solid. The Rk3288 is not heating up nearly as bad as the S802 according to stress testing so Rockchip will be able to sell tons of these for not only tv boxes but tablets as well. TV boxes need proper ventilation holes at the bottom and top with a good thermal air flow with a large quality heat sink/ thermal paste is still the key to keeping the heat issue under control when users are gaming for hours. The lockup/reboot issues via overheating of the S802 boxes has been a disaster.

    Leave a comment:


  • bluesmanuk
    replied
    I've mailed Arm for clarification based upon the articles published so far and also asked the question in the Arm community forums.

    Leave a comment:


  • Javimetal
    replied
    Originally posted by rceccleston View Post

    What do you think to Rockchip misrepresenting / advertising the RK3288 as an A17, since its actually an A12, because it doesn't feature bigLITTLE.

    Considering Allwinner have recently released their A80, with ARMs bigLITTLE architecture, Rockchip are playing dirty marketing trick since their already a generation behind.
    Do you have any proof? I'd like to ask Rockchip about this.


    About the A80, today I was checking a new board with this SoC, but it looks like need some time yet to be ready and stable.

    Leave a comment:


  • Javimetal
    replied
    Originally posted by bluesmanuk View Post
    must contact ARM to know more about this

    That made me chuckle.

    How can they not know what their own product is?

    He said it is A17, just wants to ask ARM about why that "ARMv7 Cortex-A12" is there.

    Leave a comment:


  • bluesmanuk
    replied
    must contact ARM to know more about this

    That made me chuckle.

    How can they not know what their own product is?

    Leave a comment:


  • Javimetal
    replied
    Originally posted by bluesmanuk View Post
    Could it be that the latest iteration of 3288 is now compliant, therefore able to advertise as A17?
    Well, today I have been with a member of ROCKCHIP, so I connected a RK3288 board to my laptop by OTG port, opened my windows cmd, adb shell.....

    cd sys/devices
    and there was the "ARMv7 Cortex-A12"

    He said must contact ARM to know more about this and give me a reply, we exchanged contacts and also have been talking about the 4K - 60Hz, he said this is already fixed, even they bought some TVs from Japan directly to test this.


    Well... more news soon about this as I have direct contact with Rockchip guys.

    Leave a comment:


  • Shomari
    replied
    Originally posted by no_spam_for_me View Post
    Ok, I had hoped I could identify some tools, but it's only a marketing video...

    EDIT:
    ok, the second link is a little bit more helpful... I think they use the known directories (at kernel) to determine the values via USB... but I think no way to get this program...
    They're just essentially charting the core utilization per activity much like any other cpu monitoring app. Then they dressed it in their fancy branding. Other than that any cpu monitoring app run against a BIG.little chipset will output the same data.

    Leave a comment:

Working...
X