CHIP DESIGNER ARM has said that within 18 months its Mali GPU will be able to match the power of Sony's Playstation 3 or Microsoft's Xbox 360, but that Moore's Law was not the only way to achieve that.
ARM told The INQUIRER that it is already planning what its chips will be capable of in 2016 and in particular the graphics capabilities of its Mali GPU. Planning five years ahead is not all that surprising, as those who have seen roadmaps from AMD or Intel will agree, however ARM said that it needs to plan even further ahead, as the cadence of its designs appearing on devices is a further 18 to 24 months.
The message from ARM is achingly similar to that of AMD saying that it needed to take advantage of the CPU and the GPU in order to create the next generation user interfaces, gaming and tasks such as computational photography. ARM said that in order to make this happen it needs to "put huge amounts of computational power in the hands of developers".
In order to provide those "huge amounts of computational power", ARM said that it "can't throw hardware at it to solve the problem". When asked whether this means that ARM believes Moore's Law is not relevant any longer, ARM said that while it was still relevant there is also the need for other techniques in order to provide the computational power within the strict thermal design power requirements of embedded chips, such as aggressive power management.
ARM told The INQUIRER that "battery technology doesn't have that [Moore's Law] sort of growth" and that shrinking the process node does not necessarily solve the power problem. It said that a combination of technologies is needed, such as very aggressive power management and multiple GPUs that can power up as required.
Mali (and maybe Nvida with Kal-El upwards) are not trying to catch up with anyone with any in any logical sense due the very low power restriction. completely different markets with different goals/targets. Desktop GPU or a console could suck a power station and it wouldn't really matter to much.
"That we don't have a pocketable PS3 now (or in 18 months) from Sony themselves is likely more to do with them simply not working in that direction than it being an impressive feat"
If could be done now sure it would be. Look at at Sony NGP, they are trying and indeed getting closer
http://www.engadget.com/2011/01/27/the-sony-psp2/
The only thing holding them back further(And anyone else) is again lack of GPU power with use of batteries instead of a plug in brick PSU
"This means that it may well be possible to actually create said device now, but impossible to actually power it for a worthwhile duration"
Which is the point of the ARM statement (And impressive to boot). They are stating that they are going to be able to produce PS3 graphic power AND be usable with good battery life.
It really is about playing catch-up. That we don't have a pocketable PS3 now (or in 18 months) from Sony themselves is likely more to do with them simply not working in that direction than it being an impressive feat.
As is also mentioned in the article, power source improvements (batteries) aren't increasing at nearly the same rate. This means that it may well be possible to actually create said device now, but impossible to actually power it for a worthwhile duration. Odds are the same thing will still be basically true by the time ARM reaches that point, again making the entire thing largely irrelevant.
I stand by my "potentially impressive but hardly worth mentioning" views.
Its not about playing catch up, its about getting power consumption down v GPU Speed in a SoC in a tablet or mobile phone...
PlayStation 3 in your pocket within 18 months....sounds impressive to me..
So in 5 years time, a company will have caught up with what the big players were up to 3 years ago? Doesn't really seem like anyone should be worried about anything at all.
It's potentially impressive but it's hardly worth mentioning. In 5 years time, the power goals will have been moved significantly forward - both in performance and power consumption.
Any GPU on an ARM SoC that means we don't have to suffer the closed-source PowerVR shite gets a huge thumbs up in my book.
From what I understand, source code for the Mali drivers is available and open, meaning open source platforms such as MeeGo can potentially be built for a much wider range of ARM+Mali hardware without having to fail at the PowerVR hurdle.
Imagination Technologies - makers o PowerVR - and their refusal to open source their driver have seriously, SERIOUSLY, held back a number of open source projects when targeting common hardware, and not just ARM based hardware as a number of low cost Intel chipsets with embedded GPU also use PowerVR. Big mistake.
Fortunately with Mali it sounds like the end of Imagination Technologies. Good riddance.