It is amazing to finish a chip design in just 2 months (started a chip in March, showed it in trade show in April and then shipped the samples in May). How much faster can they shorten the design cycle? I have no idea how innovation can take place if there is just little time for chip design to sit down to work, not even think. Personally, I don't think this can help the company to grow healthily and such strategy can't last long.
Rockchip outsources most of the design work to GUC which is a TSMC affliate. My sources told me it usually takes 3 iterations (sometimes 4) to get it right. So the March to May time frame could be the 3rd try, which could be a simple metal change, or from MPW shuttle to full-mask.
I think I am the guilty one here, not articulating this whole "design" process well in the story.
I will correct the story accordingly. Meanwhile, here's an e-mail I just received from Rockchip's Chen. He clarified this for us as follows:
"What I meant was, after we got the chip, we were able to get the chip in end products in less than 2 months."
Another very informative article - many thanks.
I do have an interesting question - since you are quoting Rockchip's CEO:
The RK31xx is a dual core Cortex A9 product designed for tablets and smartphones. Using Globalfoundries' high-k 28-nm process, “we can reduce cost, gain performance, but most importantly, we will reduce power consumption for tablets,” said Chen
Will the package interconnect for this application processor be Cu-wire wirebond or flip-chip bumps? I am curious how much wirebond for multi-core application processors even in 28nm.
Many thanks in advance if you can find and provide the answer.
Dear Junko, Great Post.
Enjoyed every word you've written. Also it's nice to see that Mr. Chen is open to questions.
Here is my question for Mr. Chen,
Developers are eager to support Amlogic chips, more then compared to Rockchip processors, as Amlogic have realeased Source code, which helps independent developers to develop further on platform. And thus create a favorable community support for Amlogic tablets .
Why Rockchip is not considering the same approach and release the Source code for their processors ?
And wish him Good Luck, eagerly waiting for the new RockChip processors.
Hi, axeteve. This just in from Rockchip's Feng Chen to respond to your question via e-mail:
"In the market, Rockchip’s software normally considered more stable than the competitors. As the Android
system upgrade real fast, in order to make sure the system stability we do not open the source code.
But considering the openness of the android community and Google’s upgrade cycle now getting longer,
we are considering open the SDK now for mature version."
Dear Junko - the article and the discussion is indeed refreshing and constructive - thank you.
As smartphones are heading toward $50 and tablets to $100 price, application processor vendors in China are exploring adequate low cost solutions. As evidenced by use of copper wires in 28nm for a dual core Cortex A-9 processor by Rockchip.
It would be useful to hear thoughts about which performance factor is driving Rockchip eventual plans for flip-chip interconnect and away from wire bond -- is it processor frequency, number of processor cores (from two toward four), or memory interface speed requirement, total number of wires (wire count), other?
Flip-chip reduces on-chip IR drop compared to wirebond but wirebond still meets noise margins in Rockchip's current products. Apparently as the saying goes - where is the will (to lower costs) there is a way (for wire bond) even at 28nm. So -- where are the limits to wirebond in mobile application processors? Many thanks in advance.
2-3 years ago, ordered straight out of Shenzhen a $100 7 inch tablet by Eken running Donut on some Rockchip model. As this was a dry run to test the form-factor (me like!) rather than a serious technology investment, I was neither surprised nor too disappointed when the bugger died in phases after a month, a common experience among other owners.
This doesn't necessarily reflect negatively on the RockChip part, but the company may be condemned, at best, to 'rule in hell', with no prospect of serving in the heaven of a line of better devices.
Dear Junko, Thankyou to you and Mr. Chen for answering my query. You guys are increasing your fanbase. :)
@Junko, I'll be waiting for for articles from you. Bookmark'd.
@Mr. Chen, I along with Android developer community is eagerly waiting for your positive decision towards creating a helpful growing developer community.
Best wishes for RockChip
Few things: There are high quality tablets using the RK3066. Currently writing from my one of such tablet and it has a 9.7" IPS Screen, Capacitive Touch (10 points touch) made of scratch-resistent glass, The back is made of metal, 8000mah battery, Dual Camera, HDMI, WiFi 802.11b/g/n with direct/USB Host and it is running Android 4.1 and it cost me just $180.
RK3066 is much faster than Mediatek's MT6577. For example it receive AnTuTu benchmark score of more than 8000 while MT6577 receive around 5400. (Screen resolution of the RK3066 tablet is 1024*768 while MT6577 phone 960*540 Both 1GB RAM)
the rhombus tech initiative, along with the EOMA-68 standard, has been designed to tackle the very problems that RockChip and other SoC vendors face. our strategy is very straightforward:
a) invite SoC vendors to release EVBs in a standardised modular form which can go straight into mass-production, needing only a very simple 2 to 4 layer PCB for the main I/O of any matching product.
b) standardise and therefore greatly simplify the software development. the advantage of having standard I/O boards (products into which the CPU Modules can fit) is that the software for those products will already have been written. porting a CPU Card to work in a range of existing hardware products is far, far simpler than forcing everyone to design complete products from scratch (including the software).
the cost savings and time savings should be evident, and this is absolutely critical and will only become more so as the prices are driven down further by 28nm and beyond, as well as the product lifecycles becoming shorter and shorter.
it's quite complex to explain initially but very straightforward once it's fully understood, and very exciting as well. would you be so kind as to mention to Mr Chen that we would love to work with him, especially to help introduce RockChip CPUs properly into the Free Software Community, which will result in considerable engineering cost saving for RockChip, apart from anything else? i am easy to find on the internet but here is my email address anyway: email@example.com
many thanks junko,
What are the engineering and design challenges in creating successful IoT devices? These devices are usually small, resource-constrained electronics designed to sense, collect, send, and/or interpret data. Some of the devices need to be smart enough to act upon data in real time, 24/7. Specifically the guests will discuss sensors, security, and lessons from IoT deployments.