Mainstream HW vs. Custom Niche HW

[som99]

Nice little read, thanks. Im sitting on the 1.42GHz Mac mini so only 32MB vram but im quite happy with it and MorphOS, im thinking of using a decent PATA SSD in it also, I am trying to get my hands on a PowerMac but all ive found are Geforce ones so far (within driving distance).

But I would also like to see a migration but since the x86 market is BIG and full of hardware that needs drivers I would think focusing on ARM would be a great way to go, for me TI's OMAP boards would be a good pick of the bunch, cheep and easy to get your hands on and a nice userbase and since they use PowerVR for graphics the route would be fast and nice.

So my vote would go for ARM and OMAP boards IF AT ALL MOVING TO A NEW architecture, but that's just my pieces of eight.

[som99]

If an ISA change does occur I'd rather see it be to ARM as it is a more broadly licensed archetecture (and its not X86).
BUT ARM just isn't powerful enough (yet).

As do I, tho checking crossplatform benchmarks of CPU/Memory performance a OMAP 4430 @ 1000MHz (2 cores) preforms as a Mac mini 1,42GHz and the Exynos 4412 1600 MHz (4 cores) preforms as the PowerPC G5 (970FX) 2300 MHz (2 cores) and the ARM boards consumes ****loads less power and are small and the G5 power horses wight like 50lb (22KGs).

So they are getting there and fast, a generation or two I can see ARM boards out preforming the PowerPC G5 (970MP) 2500 MHz (4 cores).

Now there are more factors to consider then these tests but overall ARM are not as far behind the top PPC's as one might think in some parts.

Tho im not saying that MorphOS should go ARM, but if they should migrate I would think ARM would be the best bet, and they are not as weak as people might think they are starting to get horse power under the hood and they are still growing and will for a long time and it feels more viable in the long run, besides old Mac hardware there is not much to see in the future of PPC hardware at acceptable price ranges while SoC's will be cheep and easy to get.

Don't get me wrong, I love my Mac mini and MorphOS and I would support them if they stay at PPC, tho I think it's interesting to think about MorphOS and ARM.

Edit: Do not kill me, im just speaking my mind.

[som99]

I've seen benchmark of the Cortex A15 @ 1.7GHz that give it parity with the G4 (not the G5), so I'm not sure about the benchmarks you're quoting.

But don't get me wrong, I'm a big fan of the ARM ISA.
And once it moves to 64 bits things are really going to take off.

BTW - I haven't killed anybody lately (that I'll admit to). :lol:

I do not know how liable and conclusive the benchmark I found was but reading what is tested it seems fair enough for a rough est.
Tho im happy to get pointed in the right direction

The program I looked at was Geekbench 2 and the tests used is the following:

Code: [Select]

Integer Performace
Integer benchmarks measure integer performance by performing a variety of processor-intensive tasks that make heavy use of integer operations. None of the integer benchmarks perform any file I/O in order to isolate the work done to just the processor and the memory subsystem.

Blowfish encrypts and decrypts memory using the Blowfish algorithm.
bzip2 Compress and bzip2 Decompress compress and decompress a text file in memory using libbzip2.
Image Compress and Image Decompress compress and decompress an image in memory using libjpeg.
Lua executes a script written in the Lua Programming Language. The Lua script is a prime number sieve that finds all prime numbers below 200,000.
Floating Point Performance
Floating point benchmarks measure floating point performance by performing a variety of processor-intensive tasks that make heavy use of floating-point operations. None of the floating point benchmarks perform any file I/O in order to isolate the work done to just the processor and the memory subsystem.

Mandelbrot renders the Mandelbrot set.
Dot Product computes the dot product of two vectors.
LU Decomposition computes the LU decomposition of a 128x128 matrix.
Primality Test performs the first few iterations of the Lucas-Lehmer test on a particular Mersenne number to determine whether or not it is prime.
Sharpen Image and Blur Image apply a convolution filter to an image in memory. These filters are similar to the filters found in graphics editors like Adobe Photoshop.
Memory Performance
Memory benchmarks measure not only the performance of the underlying memory hardware, but also the performance of the functions provided by the operating system used to manipulate memory.

Read Sequential loads values from memory into registers.
Write Sequential stores values from registers into memory.
Stdlib Allocate allocates and deallocates blocks of memory of varying sizes using functions from the C Standard Library.
Stdlib Write writes a constant value to a block of memory using functions from the C Standard Library.
Stdlib Copy copies values from one block of memory to another using functions from the C Standard Library.
Stream Performance
Stream benchmarks measure both floating point performance and sustained memory bandwidth. Geekbench 2 uses benchmarks based on the STREAM benchmarks developed John D. McCalpin. None of the stream benchmarks perform any file I/O in order to isolate the work done to just the processor and the memory subsystem.

Stream Copy computes a(i) = b(i), where a and a b are arrays.
Stream Scale computes a(i) = q * b(i), where a and b are arrays, and q is a constant.
Stream Add computes a(i) = b(i) + c(i), where a, b, and c are arrays.
Stream Triad computes a(i) = b(i) + q * c(i), where a, b, and c are arrays, and q is a constant.
 But ill take your word for it and agree until I find better benchmarks

The move to 64 bit ARM is not to far away now, im happily waiting for a SoC to play with in the future

Also im glad I did not get killed today (as far as I know im not on anyones list yet)