600mhz-1ghz?

[Kings]

ARM Cortex-A8

Acording to the ARM website the Cortex-A8 can "ability to scale in speed from 600MHz to greater than 1GHz".

Is the OMAP3530 locked at 600MHz? Will the ability for the Pandora to run at 1GHz allow for some Emulators thats seem unlikely to run at 600MHz? And would the extra MHz allow for better multitasking?

 

[Exophase]

Squidge can run his at 600MHz with standard voltage and 750MHz when overvolted. It has implied that users will be able to do the same, but it's completely unknown how well it'll clock (and at what voltages). I especially wouldn't expect 1GHz.

 

[Shiny]

It does not specify under which condition those numbers were taken. It could very well be in some special setup being cooled by liquid nitrogen to keep the chips from melting. I've seen this mentioned numerous times but never has it been clearly answered to my knowledge. That information may still be under NDA so they can't tell us yet.

 

[Holokauston]

well you could probably push the cpu to almost 1Ghz as ARM says but you might need to cool it and drain alot more power

 

[shmuck]

That's dealing with the clock speeds for the entire line of processors, it's not saying that every individual chip will be able to run at 1GHz. For example, the Pentium 4 scaled between 1.3GHz and 3.8GHz.

I believe that the Pandora's processor is one of the first implementations of this chip, so it's on the low end (~600MHz) 'cause faster chips don't exist yet. It's too bad, but maybe there'll be an Pandora update eventually ("Hope" or something) with a faster chip/extra features.

Of course, overclocking to some degree is always a possibility.

edit: The Pentium 4 analogy isn't perfect - that's really a brand name for several different chips, but you get the idea I think.

 

[Squidge]

Squidge can run his at 600MHz with standard voltage and 750MHz when overvolted.

Note that the voltage used never exceeeded ti's recommended maximum. If your prepared to run it out of spec, then you may achieve a greater overclock, but I obviously don't want to fry a $1500 dev board.

 

[Pickle]

There also info on : http://arm.com/products/CPUs/ARM_Cortex-A8.html

* Process 65nm (LP) 65nm (GP)
* Frequency* (MHz) 650+ 1100+
* Area with cache (mm²) <4 <4
* Area without cache (mm²)
* Power with cache† (mW/MHz) <0.59 <0.45

 

[Exophase]

Yeah, I think it's pretty clear now that OMAP3530 is using the LP process, someone please correct me if I'm wrong. The sad thing is that this version actually uses MORE power per MHz, but it has much better leakage. Personally I'd gladly give up good leakage for this much more power and better peak consumption rates.

 

[Kings]

Thanks for the clarification! 600MHz is enough and especially good if it runs for 10+ hours of heavy use.

Not to think too far ahead but is the Pandora PCB set up to only use the OMAP3530 LP? Might be interesting if a future revision could have a new OMAP3 without overhauling the PCB.

 

[icurafu]

It's obvious that TI would use LP for a mobile phone chipset, because avoiding leakage is the most important part.

 

[Exophase]

It's obvious that TI would use LP for a mobile phone chipset, because avoiding leakage is the most important part.

TI makes several different OMAP35xx, the highest power one is supposed to be geared for multimedia devices more than typical cell phones. Maybe it's not feasible for them to use different processes along the product line, I don't really know.

 

[Svartalf]

TI makes several different OMAP35xx, the highest power one is supposed to be geared for multimedia devices more than typical cell phones. Maybe it's not feasible for them to use different processes along the product line, I don't really know.

You don't shift gears on process in the middle of a production run... You run an entire run of something with the same process- and you set up multiple facilities with differing processes if you need different ones to make the same part class. TI's got fab facilities all over the place in the Dallas/Ft. Worth and Austin areas. These setups are pretty huge, really. Not sure if they're using the North Campus facility or the new one over on Alma Road in Richardson to make these...

 

[Exophase]

You don't shift gears on process in the middle of a production run... You run an entire run of something with the same process- and you set up multiple facilities with differing processes if you need different ones to make the same part class. TI's got fab facilities all over the place in the Dallas/Ft. Worth and Austin areas. These setups are pretty huge, really. Not sure if they're using the North Campus facility or the new one over on Alma Road in Richardson to make these...

What I meant was making the 3530 use the GP process and the lower ones use LP. They're obviously different chips, of course.

 

[ldesnogu]

What I meant was making the 3530 use the GP process and the lower ones use LP. They're obviously different chips, of course.

That's not that easy: processes are very complex beasts and each process requires a huge amount of engineering work; and I am ready to bet that OMAP34x and OMAP35x contain many hardmacros which make the process change more difficult.

 

[Exophase]

What I meant was making the 3530 use the GP process and the lower ones use LP. They're obviously different chips, of course.

That's not that easy: processes are very complex beasts and each process requires a huge amount of engineering work; and I am ready to bet that OMAP34x and OMAP35x contain many hardmacros which make the process change more difficult.

Yeah, I was thinking that too.. ahwell