The fastest way to find an answer on the internet is not to ask a question, but to make an incorrect statement. I don’t remember who said that, but time and time again, if I say something wrong on this blog, or any online forum, I’ll usually get a comment or an email in my inbox correcting me. It’s a fantastic way to learn.
That being said, I’ve made a few incorrect statements in my post about Marvell’s new Armada line that I’d like to address them now. Here are two emails I’ve received from Qualcomm and Marvell regarding their latest chips:
Hi Stefan,
Hope all is well.
Nice piece the other day about Marvell’s new Armada processor line, I just wanted to clarify one point:
While Marvell may now have a license to “tweak” ARM’s Cortex A8 architecture, Qualcomm also has an architectural license for the ARM v7 instruction set, which is what is used for the Cortex A8. Qualcomm did not use “ARM’s ready designed Cortex A8 processor” in the Snapdragon processor platform. Snapdragon’s processor was designed pretty much from scratch, so to speak.
If you’d like to compare it to the new Marvell family: First-generation Snapdragon chipsets run at over 1 gigahertz (GHz), while consuming only 350 milliwatts (mW) of power. Snapdragon also integrates a 600-MHz DSP, graphics, numerous 3G standards, etc. – all into a single chip. The second-generation QSD8650A Snapdragon (sampling later this year) runs at 1.3GHz and the dual-CPU version (QSD8672) runs at up to 1.5GHz. Both new chipsets are being fabricated with 45-nanometer technology. Snapdragon chipsets are already powering commercial devices – Toshiba’s TG01, Acer’s neoTouch and Liquid, and HTC’s HD2.
Is it possible to update the post to reflect that Snapdragon does not use an off-the-shelf A8?
I work for Qualcomm’s PR agency, and am happy to answer any questions you may have.
Thanks,
Alison
Hi Stefan,
Hope all is well. Really enjoyed reading your piece today with regards to the ARMADA announcement by Marvell. However, I’m a bit confused to what you meant by this statement, “Marvell has a license for the actual Cortex A8 architecture schematics, which they’ve taken the liberty of tweaking to make their Armada family more powerful. In the process however they managed to chop off support for ARM’s NEON SIMDfp instruction set”.
Marvell CPU cores are designed from the ground up, from the microarchitecture level. It is not a case where we have taken ARM’s design and modified it. Marvell has never seen any RTL from ARM and has made improvements to its own architecture for the past 13 years, taking the best of XScale and the best of Marvell designs.
Any chance you’d be open to a briefing with our executives for any follow-up questions you may have regarding the announcement?
Look forward to your interest and feedback.
Best,
Daniel
I learned something today and am simply passing on the knowledge.