Watch Out Raspberry Pi, Here Comes The Pine A64

[g00ey]

There is no value in a 64bit cpu until memory is over 4GB. The 64 bit program cannot be exercised usefully without perhaps 8gb available. A 1GB memory is only using 1/4 of the 32 bit program address space.
Please offer reasonable configurations!

That is mostly inaccurate. Get a 64-bit computer with 4 GB of memory and you can run essentially everything the same as a 64-bit computer with 8 GB of memory, just you have to watch you don't have too many programs clogging up the RAM. With 1 GB of RAM, you can still use 64-bit programs without issue, there is more to 64-bit processing than how much memory the system can see. So used for developing 64- bit applications or for basic tasks there isn't really any issue in terms of the specs.

In general, it isn't that clear-cut as to what the benefits are of a 64-bit CPU over a 32-bit CPU, apart from the memory addressing. The obvious things are that you get wider registers and you can use higher precision decimal numbers straight into the CPU. But there are improvements that are not that obvious too.

In the case of the introduction of x86_64 architecture, the 64-bit CPU not only got wider registers but a lot of them compared to it's predecessor. Along with that, the CPUs have come with extensions such as the SSE3, SSE4, AVX, AVX2, ... that eventually got adapted to the 64bit architecture, but still, one could argue that similar extensions would be conceivable for a 32 bit CPU.

What is the case for the x86 architecture is not necessarily the case for the ARM architecture as I believe that a plethora of added extensions are not desired for a RISC based CPU. It kind of goes against the RISC philosophy; if you add a lot of instructions, the CPU is no longer a RISC CPU.

As for the MIPS CPUs that are also popular in mobile devices these days, they have been 64bit since the '90s. Don't know about the Freescale CPUs (former PowerPC for those of you who remember the good old PowerMacs), I think they still are 32bit unless IBM or whoever maintains Freescale today has come up with an update... EDIT: Turns out that Freescale is 64bit and has been for quite a number of years. Probably the development of the PS3 boosted the funding for such an architecture. There are a lot of PowerPC based chips out there that are both 32bit and 64bit. ..

 

[willy481]

In general, it isn't that clear-cut as to what the benefits are of a 64-bit CPU over a 32-bit CPU, apart from the memory addressing. The obvious things are that you get wider registers and you can use higher precision decimal numbers straight into the CPU. But there are improvements that are not that obvious too.

In the case of the introduction of x86_64 architecture, the 64-bit CPU not only got wider registers but a lot of them compared to it's predecessor. Along with that, the CPUs have come with extensions such as the SSE3, SSE4, AVX, AVX2, ... that eventually got adapted to the 64bit architecture, but still, one could argue that similar extensions would be conceivable for a 32 bit CPU.

What is the case for the x86 architecture is not necessarily the case for the ARM architecture as I believe that a plethora of added extensions are not desired for a RISC based CPU. It kind of goes against the RISC philosophy; if you add a lot of instructions, the CPU is no longer a RISC CPU.
................
As for the MIPS CPUs that are also popular in mobile devices these days, they have been 64bit since the '90s. .....

Interesting.... so as per an earlier post correcting my missing that the A64 Cortex A53 uses the (read much more advanced) ARMv8 instruction set, than the older ARMv7 instruction set used in the Pi2, perhaps the more advanced instruction set, in itself facilitated by the move to a 64bit architecture, makes a significant contribution to the Pine A64 using less energy than the Pi2 for the same work ?

Another aspect is the info on the Pine kickstarter site mentions use of 'power management', presumably supported by hardware, and which although mentioned as supported in Linux manuals (degrees of sleep states etc), seems to get no mention in Pi documentation. So that may be another area where the Pine A64 achieves an advantage for IoT applications etc.

 

[g00ey]

In general, it isn't that clear-cut as to what the benefits are of a 64-bit CPU over a 32-bit CPU, apart from the memory addressing. The obvious things are that you get wider registers and you can use higher precision decimal numbers straight into the CPU. But there are improvements that are not that obvious too.

In the case of the introduction of x86_64 architecture, the 64-bit CPU not only got wider registers but a lot of them compared to it's predecessor. Along with that, the CPUs have come with extensions such as the SSE3, SSE4, AVX, AVX2, ... that eventually got adapted to the 64bit architecture, but still, one could argue that similar extensions would be conceivable for a 32 bit CPU.

What is the case for the x86 architecture is not necessarily the case for the ARM architecture as I believe that a plethora of added extensions are not desired for a RISC based CPU. It kind of goes against the RISC philosophy; if you add a lot of instructions, the CPU is no longer a RISC CPU.
................
As for the MIPS CPUs that are also popular in mobile devices these days, they have been 64bit since the '90s. .....

Interesting.... so as per an earlier post correcting my missing that the A64 Cortex A53 uses the (read much more advanced) ARMv8 instruction set, than the older ARMv7 instruction set used in the Pi2, perhaps the more advanced instruction set, in itself facilitated by the move to a 64bit architecture, makes a significant contribution to the Pine A64 using less energy than the Pi2 for the same work ?

Another aspect is the info on the Pine kickstarter site mentions use of 'power management', presumably supported by hardware, and which although mentioned as supported in Linux manuals (degrees of sleep states etc), seems to get no mention in Pi documentation. So that may be another area where the Pine A64 achieves an advantage for IoT applications etc.

I'd guess that ARMv7 and ARMv8 are different architectures that features different instruction sets. I'd want to imagine that the ARMv specs not only describes the instruction sets supported but also how these instructions are implemented on the wafer (silicon substrate), and here there may be improvements between the versions. Improvements that even an assembler programmer won't be directly aware of. Other improvements might involve things such as better cache predictions or more efficient memory controller, also these things are not directly visible to the programmer as these improvements don't affect the instruction set.

I'd also like to imagine that it is a "natural" process that newer versions of the ARM architecture continuously gets improvements regardless of the introduction of 64bit support. I mean, it could be moot whether all of the ARMv8 improvements over ARMv7 are due to the introduction of the 64bit architecture or just improvements that could as well be applied without the 64 bit parts. As far as I can read; ARMv7 features the A32 and T32 instruction sets whereas ARMv8 features or supports the A64, A32 and T32 instruction sets.

The bottom line is that it is not that clear as to what improvements one may expect from a 64bit architecture over a 32bit architecture.

 

[1stFalloutBoy]

Note 2GB version is available..