Originally posted by pWe00Iri3e7Z9lHOX2Qx
View Post
Announcement
Collapse
No announcement yet.
Apple Announces The M2 Ultra SoC - 24 Core CPU, Up To 192GB Unified Memory
Collapse
X
-
Last edited by mitchellrenouf; 05 June 2023, 07:15 PM.
- Likes 3
-
Originally posted by sophisticles View PostI agree with Hector, if you spend the big bucks to buy a system with one of these and are stupid enough to want to run Linux on it, then don't harass the Asahi "developers" with questions as to when Vaporware Linux will finally be finished because it never will be.
Imagine someone spending 10+ grand on a high end system featuring one of these processors and wanting to void the warranty by wiping the excellent Max OS and installing a half-baked kludge of an OS in its place.
No thanks, you can keep your Linux, i will stick with Max OS if I ever have the dough to buy one of these.
- Likes 3
Comment
-
Originally posted by mitchellrenouf View PostThey are worth zero to me because they are too slow for what I do. I'm not going to be waiting extra hours for the same task I would rather buy a mac to run Linux but I can't unless they can overclock. Taking serious means features for people who actually know things about computers like being able to throw as much power as you want. Why would someone use a 95W TDP apu instead of having 400W just for the CPU power without having to build a cluster. You can probably double the performance of apples CPU with overclocking it doesn't matter if you max out their best mac it is still slower without overclocking. A i9 is 6Ghz stable overclocking right now if you buy one and throw as much power as you can at it they aren't even 4Ghz without doing that. You can even go near 7Ghz if you spend a lot on cooling and if you just don't care about money basically get 3 i9 processors for your dollar. You need a liquid helium cooling system to fully use any CPU.
these are 2 complete different cpu architectures and cpu ISAs ...
you can only detect who is faster by writing a benchmark.
"GHZ" is nonsense. theoretically you can build a computer chip who is faster at 1MHZ than your 6ghz chip.
it all dependents on how you implement stuff. for example the clock speed of a FPGA is much slower the mega65 is based on a FPGA and only has 40mhz but people tell that modern "6ghz" cpus have problem to emulate this chip in ja virtual machine...
if you see modern AI accelerators based on FPGA AMD Xilinx they are much faster than your 6ghz intel cpu at very low clock speeds.
thats possible because most of these AI operations are very simple operations at only 8bit or even only 4bit and a FPGA can create these 4bit and 8bit structures in massive quantity. and all these operations are all operated in paralell means there is no need for high frequencies at all.
​Phantom circuit Sequence Reducer Dyslexia
- Likes 15
Comment
-
-
Originally posted by qarium View Post
computer performance is not measured in GHZ...
these are 2 complete different cpu architectures and cpu ISAs ...
you can only detect who is faster by writing a benchmark.
"GHZ" is nonsense. theoretically you can build a computer chip who is faster at 1MHZ than your 6ghz chip.
it all dependents on how you implement stuff. for example the clock speed of a FPGA is much slower the mega65 is based on a FPGA and only has 40mhz but people tell that modern "6ghz" cpus have problem to emulate this chip in ja virtual machine...
if you see modern AI accelerators based on FPGA AMD Xilinx they are much faster than your 6ghz intel cpu at very low clock speeds.
thats possible because most of these AI operations are very simple operations at only 8bit or even only 4bit and a FPGA can create these 4bit and 8bit structures in massive quantity. and all these operations are all operated in paralell means there is no need for high frequencies at all.
​Last edited by mitchellrenouf; 05 June 2023, 08:27 PM.
- Likes 1
Comment
-
Originally posted by mitchellrenouf View PostIt is and isn't at the same time. Obviously the same chip running at a higher frequency is faster than not. I am also talking about a generalized CPU. Java is also really stupid to use for something that needs preformance if that's what you mean by ja.
if you compare apple M2 with intel cpus GHZ do not matter at all.
"I am also talking about a generalized CPU."
thats the point dude most intel cpus are generalized CPUs but the apple M1/M2 is not because the biggest parts of apple m2 is not a generalized CPU it is instead ASIC hardware
all the benchmarks people watch on phoronix.com is most of the time always only about the generlized cpu parts
and the ASICs are never tested but these ASIC parts are the relevant parts for apple M2 design.
for example Apple M3 will have AV1 decode and encode ASIC support.
also apple M3 will be in 3nm TSMC this will give intel a very hard time believe me.
Originally posted by mitchellrenouf View PostI would have to actually examine the software you are talking about to know why it is so slow.
is written in assembler and need to be clock precise means the virtual machine do not only need to calculate it but also need to sync it to the precise clock state... and this is a problem for modern 6ghz cpus to sync every state to make sure everthing is in place at a clock precise timing...
and for this it does not matter of your 6ghz cpu can execute 16 bit or 32bit or 64bit its emulate 8bit but the tact precise timing of all the components is the problem.
this means it is not only the cpu... all other components like sound chip and so one need to be clock precise just remember even the floppy disk of a commodore 64 has its own processor the same as the main processor and this also is everything written in assembler and need to be clock precise.
remember this is not a performance problem its a synconisation prolem because in this old systems like comodore 64 everything was developed to be clock precise with all other components liks flopy disk and so one.Phantom circuit Sequence Reducer Dyslexia
- Likes 2
Comment
Comment