Skip to comments.Qualcomm 1.2GHz dual core ARM demonstrated
Posted on 01/14/2011 2:00:32 PM PST by Ernest_at_the_Beach
Big GHz fight in 2011
Qualcomm has decided to show its dual Scorpion based MSM8x60 mobile CPU and it also told the world that this CPU can hit 1.2GHz, per core that is.
The chip has an Adreno 220 GPU that we know little about and of course its a 45nm chip since 28nm should be only ready in the latter part of 2011. The chip should be able to do high definition stereoscopic 3D video, and multi-party video conferencing as well as some nice gaming.
It is intended for tablets and phones and you can bet that you will find this core in one of the future HTC phones, simply as Qualcomm has quite a few shares in HTC, or so we heard.
With its 1.2GHz core clock MSM8x60 looks to be faster than Tegra 2, but Tegra 2 is available today and Qualcomm is still to ship its first tablet or phone based on this chip. Browsing with two cores should be significantly faster than with one core, which kind of makes sence, but again, lets see it shipping in real designs first.
Lets not forget Texas instruments who is about to ship its 1GHz OMAP 4430 chip and with hope to see 1.5GHz version later this year.
You can read more about the MSM8x60 and see Anand Lal Shimpi's fancy shoes here.
I’m kind of worried about the mobile GHz race we’re in right now. You was where it went with the MHz race in desktops.
What...bring it on...consumers benefit.
Mobile devices have two major concerns that desktops don’t (or at least not nearly to the same degree). One is energy consumption and the second is heat dissipation. Desktops aren’t being held, they are surrounded by multiple fans moving air in multiple directions, and they don’t have to be held. They are also plugged in, so battery life isn’t a concern.
Wake me up when ARM comes up with an FPU that isn’t outperformed by a 486DX.
That's what we thought last time, and then we ended up with grossly inefficient, power-sucking chips.
The thing is that mobile devices don’t need that much floating point. So more mips means faster interpreters..like dalvik!
ARM floating point depends on which implementation you’re getting. There is VFP, but it is pretty lame, but ARM also has NEON, providing low-end SIMD that can work as the FPU. But that’s not to say an ARM implementation even has to use ARM’s FPU solutions. Intel’s XScale is an ARM with an MMX unit.
And none of those implementations are great compared to other embedded RISC platforms or embedded x86 platforms.
Still I think better than 486, although probably far below the better embedded PowerPC.
And that is more or less what AMD (and others) were stubbornly insisting on the desktop until the K7. “What’s wrong with the K6? (or 6x86/MII or WinChip etc) Just look at the Winstone benchmarks!”
First - Android ain’t windows. So “Winstonee benchmarks” don’t buy you much. Another issue is that the K6 was done at maybe 350nanometers if memory serves. That is VERY different from 45nanometers. It just a completely different discussion.
And while Android isn't Windows - Windows 8 "Midori" will run on ARM. I have seen Windows running on ARM - it was painful. Microsoft has had Windows Vista, 7 and 8 running all running on ARM. They are bringing Windows on ARM to market because they don't want to be caught with their pants down like they were with Netbooks and ceede this space to Android and HP's WebOS devices.
It is clearly the intent of ARM and their licensees to push ARM cores into more mainstream performance intensive applications, but that will be very challenging in its present form. Hopefully Nvidia comes forward with something amazing. But the Cortex family isn't the basis of mainstream success.