i dont like any of them, bit if i had to choose, id choose #2 but id get more ram. thats just me though.
i have #1 for reasons other than roon and hqplayer and i really do not like it. id sell you the 4 month old M1 studio i have but im willing to bet you wouldnt like it.
Did not pull the trigger on a Mac Studio. However, I just read on audiophilestyle (HQplayer 4 and Mac-mini M1) where a poster compared the Mac Studio M2 Ultra with the M1 Ultra and said there was no performance improvement with the M2 over the M1. Interesting analysis by the poster.
This may be one of the absolute best values in hqplayer hardware, especially used. I picked mine up used for $300. It can’t max out hqplayer options, but the ability to run all PCM options quickly, and almost anything at dsd256, and all in a small (mostly silent) form factor is solid for the money.
It makes me very curious where you end up on your shopping here. If the latest apple silicon can handle the full set of hqplayer options, then please just take my money now
Erik, thanks for the comment. You are absolutely correct. The Mac Mini M1 performs the same for me. I can stream at DSD512 with some modulators/filters as well. It certainly helps that I only have Roon/HQP5 installed on the Mac Mini
I am going to wait for some additional feedback before deciding on a new purchase.
I don’t believe that to be the case. I’m considering exactly the same upgrade, and Jussi suggests that an M2 with 6 performance cores or better will run DSD512 fine. Just remember that MacOS USB is limited to DoP rather than DSD Native, so bear that in mind. As a general rule, the Apple Silicon has much higher throughput/core and memory I/O than x64 architecture machines, and vastly better power efficiency when doing so. Where it runs into problems with HQP is (according to Jussi) with the specific CISC instruction sets that x64 has for some of the DSP work that HQP does - in a RISC system you effectively have to assemble complex instructions on the fly, which seems to be where it takes the hit - in this case, the x64s are analogous to an FPGA with a dedicated instruction.
Yes, I was referencing those, just providing a little context as to why the Apple Silicon still falls short in some areas, despite outperforming x64 in others. If I stick with MacOS for my Roon/HQP server, my use case tops out at DSD256, so an M2 Pro will do the job nicely. Otherwise I’ll custom build an ubuntu/AMD machine
It all depends on how many calculations single core can do per second, in practice.
Apple silicon runs at a bit over 3 GHz clock speeds. Latest Intel CPUs run at 6 GHz clock speed. Apple silicon can process for example two 64-bit values per clock cycle per core. Intel silicon can process four 64-bit values per clock cycle per core. With AVX512, Intel silicon can process eight 64-bit values per clock cycle per core.
But Apple silicon consumes less electricity per calculation though, and thus produces less heat as well.
I’m very happy with the M1 Mac Mini. As @elementze pointed out earlier, it is a really great cost effective solution for DSD256 as well as some 512.
I really want to stick with MacOS. That being said, I will only upgrade when Apple has a chip that will support DSD512 with all modulator/filter combinations.
Comparing a close-coupled RISC architecture with a discrete CISC (actually somewhat hybrid) architecture is a LOT more complex than that, but is getting seriously OT for the purposes
Vector operations, NEON vs AVX are not so different on these architectures. With non-vector instructions, Intel has deeper pipelining due to CISC, but since it is pipelined architecture, it can consume equivalent number of instructions per clock.
In the old days (i486 and such), before CISC moved to pipelined microcode instructions, there was more difference. And those differences have never applied to vector units anyway.
Still using my Mini M1 (16GB), now with HQPlayer 5 to drive Holo Spring 2 KTE at DSD256. The filter-modulator combo below do a great job, maybe the best piano sound quality I’ve heard outside a good live venue. Steady performance at ~65% of the Mini M1’s capacity.
