>Be IBM>develop PPC G4>shits all over the Pentium 4 at a lower clock speed and smaller die size>get Steve fucking Jobs up on stage>he talks about how having a long pipeline is a problem and that the G4 is much more efficient>start to develop next-gen PPC 970/G5>Decide you're going to step on the same rake Intel did>extremely long pipeline, terrible branch predictor, cannot reach clock speed targets and hot as the sun>such a disaster it kills PPC client chips foreverWhat was the idea here exactly?
>>107534165iTODDLERS BTFO
>>107534165The smartest move Apple did was go to x64. The stupidest what to leave it.
>>107535448Articulate your claim, coalberg
>>107535469anyone with a functioning brain can see the reality of ARMshit that relies entirely on hardware accelerators to cheat at very specific applebench """"""""""tests"""""""
>>107535627>hardware acceleratorsusing hardware accelerators is not cheating, it's the entire reason you're able to write your shitposts, so is arm
>>107535627>hardware accelerator is le badlol, lmao even
>>107535769>>107535786street shit harder currynigger
>>107535874Sperg out all you want but it's not up to you to decide, or me. Today the current appers to be very much in the direction of arm and you will assimilate or stay on your sweltering 2030 shitbook forever ever
>>107535424dumb satania poster
>>107534165IBM developed PowerPC based on the current Power gen, Apple had very little say in it's development. IBM was aiming for the server market purely with 5, and this screwed Apple who couldn't use the chip even in it's PowerPC form for laptops, hence why they went for Intel. IBM just had different business goals and it did not involve Apple. Mind you, Apple's only real money generator at the time was the ipod, they were still recovering from being in the shitter for years. Not like they had the resources to pay IBM for custom chip designs, PowerPC was purely just off the shelf.
>>107535627>hardware acceleratorsMMX, SSE, AVX, AVX2, AVX512 etc etc etc
>>107535874>muh hardware accelerators bad>kvaazar uses x86_64 hardware acceleratorslmao dumb jeet
>>107534165Apple never used the Pentium 4 you retard, they used Core, which it self was P3 derivative.
>>107537492I didn't say they did, the point was Apple mocked the P4 for its flawed design, only for IBM to turn around and do the exact same thing
>>107535448>The stupidest what to leave it.If you actually used arm macs and mac prior you would know that arm switch was amazing
>>107537613If you actually used Linux you'd know that it's irrelevant what Apple uses to run this pile of shit operating system
>>107537732So you haven't used it. Got it.
>>107537744I have but let's pretend I haven't so that I can read more of your fanboy seething
>>107534165So>defeat the pentium 4 because it's a retarded long bipeline meme made to run at 15Ghz but physics get in the way>decides it's a good idea to copy the mistake pentium 4 did
>>107537390there's a large difference between using a baked in silicon solution to an arbitrary problem.. and doing more things in 1 clock cycle.regardless more perf is better, so both of you are retarded.
>>107537390>>107537430CPU instructions aren't hardware accelerators. Do you even know what AVX does?
>>107535769>>107535924Apple puts one SME core per chip, just so they can win in single-core benchmarks. Because of this, their multi-core SME performance is basically no better than single-core.Why would they do this if not to game benchmarks?>>107537390These are all per-core, unlike Apple's Special benchMark Execution units. Their performance grows linearly with the number of threads used, and because of this, their performance ceiling is MUCH higher.
>>107535448>x64. The stupidest what to leave it.This has to be in the top 10 dumbest /g/ takes I've ever read. Surely you're baiting.
>>107537775It not about fanboism. I used MacBooks from work updated every 2 years. Switch from intel to arm was amazing. It is fast, silent and powerful. I'm not sure why you don't like it.
>>107537935Apple ARM processors have four execution units capable of SIMD processing while consumer Intel processors only have three.
>>107538150Apple ARM processors have ONE execution unit capable of SME processing shared between ALL cores in a cluster.https://github.com/tzakharko/m4-sme-exploration/blob/main/reports/01-sme-overview.mdThe SIMD execution units you refer to are for NEON, and only support 128-bit vectors (lmao) compared to AVX2 which supports 256-bit. So Intel's three AVX units are equivalent to 6 of Applel's. AMD's chips support AVX-512, so one unit of those are equivalent to four of Apple's and Zen5 has two capable of doing FMAs, and four capable of integer math.
>>107538229NEON is faster than AVX>per cpu clusterM4 has 2 cpu clustersM4 Pro has 3 cpu clusters M4 Max has 5 cpu clusters
>>107538301>NEON is faster than AVXLmaoNEON on the M5 big core is 4x128b.AVX2 on Arrow Lake is 2x256b for floating point, 4x256b for integer/bitwise math, and AVX-512 on Zen5 is 2x512b for floating point and 4x512b for integer/bitwise.NEON is in no way comparable.>M4 Max has 5 cpu clusters9950X has 16 cores, all capable of AVX-512.
>>107538440all of this and still slower than arm
>>107538567>>107535627
>>107538603you can compile the kernel on macOS, they used an x86 emulator or a virtual machine
>>107538625>applel's processors so unusable they can't even RUN the benchmark without an emulator
>>107538649you can cherrypick any benchmark you wantHere is the M4 Max is faster than a dual socket 96 core EPYCThe M4 Pro is faster than a single socket 128 core EPYC
>>107538726That one isn't fair, since FFMPEG has a bunch of x86 asm that won't get compiled on the arm builds. It'd be fair if they were both cross compiling for the same architecture.
>>107535627i don't understand the point of claiming to be risc when you require 500 accelerators and extensions and addonsyou're just cisc with extra steps at that point imo
>>107538802>yeah both arm software and hardware are more efficient but it doesn't count
>>107538802asm doesn't need to be compiled. it just gets copied into binary while code that doesn't have arm asm needs to be compiled
>>107538865>benchmarking different things and extrapolating the results>>107538878>never heard of an assembler
>>107537613>and mac priorWhen every single mac is kneecapped with shitty cooling of course switching to a phone CPU half as good and a quarter as hot will seem to make it faster
>>107538825>i don't understand the point of claiming to be risc when you require 500 accelerators and extensions and addons>you're just cisc with extra steps at that point imo
>>107538958>assembleryou have asm code written by hand ready for assembly, on arm you have to compile all non asm code first
>>107539146And FFmpeg uses one of the slowest x86 assemblers in the world, nasm. Don't you think this might bias the results a bit?
>>107538973intel advertised their 50W cpus as 35W
>>107535924When's the last time you saw a functioning ARM computer older than 5 years? How about 10 years?x86 computers from 1980 are still in use to this day, running trains, airports, power plants, factories, even nuclear missile silos, and they're all completely serviceable with original parts from 1980, or after market parts made 4 decades later.The real world runs on x86, while your disposable currynigger trash shits the streets in 5 years maximum.
>>107539040That's the thing, x86 can keep adding instructions til the end of time. AArch64 is stuck with the instructions they can fit in 4 bytes.
>>107539325Want me to copy my previous post for you?
>>107534165>What was the idea here exactly?Sony paid them. Money talks. That's about it.
