Apple M4 Mac Mini With macOS vs. Intel / AMD With Ubuntu Linux Performance

[Craggles086]

Does the Apple M series answer the whole Risc vs Cisc debate that was the rage roughly 15 to 20 years ago?

[0xff]

Phoronix: Apple M4 Mac Mini With macOS vs. Intel / AMD With Ubuntu Linux Performance...

Why 7zip compiled without SIMD (SSE2) for test?

[mos87]

A lot of people are, myself included.

How exactly and how many is a lot?

[mdedetrich]

Do you realize what a shit solution that is? First, it requires the user to know which packages use a vulnerable library. Then, it requires them to know the compatibility matrix. Finally, it requires them to specifically "upgrade" each package that uses each vulnerable library.

It's a hack, at best. Not a real solution, at all.

I am not saying that its ideal, but I am saying that it is possible to do this. And at least for the usecase I described it is eloquent in how stupidly simple it is, you literally place a dll in the same folder as the executable.

Of course for proper "I want to upgrade the ssl library because there is some security vulnerability globally in the system" its definitely not ideal but for other cases it works quite well.

Also its not a hack, its a well documented feature in Windows and its entirely intended to be used that way.

[Times Two]

It would be more interesting if it made with TSMC's 3nm. Though it would be a competitor to Apple's M4 Pro and M4 Max chips, but without 3nm I can't see it outclassing it. Would make for an awesome gaming laptop though.

I wouldn't discount the possibility of Strix Halo being 3nm or at least some of the chiplets being 3nm.

[Dukenukemx]

The single most interesting x86 coming down the line is AMD's Strix Halo APU. Look's to be out in q1 2025. 256-bit memory bus and huge on-chip GPU.

It would be more interesting if it made with TSMC's 3nm. Though it would be a competitor to Apple's M4 Pro and M4 Max chips, but without 3nm I can't see it outclassing it. Would make for an awesome gaming laptop though.

[Dukenukemx]

That video doesn't prove a thing, it's some guy using a mocking voice to disparage Apple.

That video is referecing this video from Max Tech who did a memory test that shows that... 8GB for Mac isn't the same as 16GB for Mac, let alone Windows.

Yes Apple charges a premium and some may consider it price gouging but those same people do not have a problem with the System76 doing the same thing by charging $5400 for a $4000 video card or $9800 for $2000 worth of ram.

System76 is just one of many computers you can buy and install Linux onto. At the very least the ram is upgradable.

Linux users seem to like to pick and choose where to direct their faux outrage over corporate "greed".

If AMD does it, then good.

If S76 does it, then good.

System76 is not the equivalent to Apple.

Anyone else?

Bad.

We're not using open source because we trust corporations. Apple though does use open source often and their track record with it is horrible.

Also that test that was done of 8gb vs 16gb is flawed, the claim is that 8gb of ram on the Macs is the same as 16gb on a different platform but that guy tested two Macs.

He also made a video where he showed 8GB Mac vs 16GB Windows PC.

[Times Two]

While not cheap, the $2400 USD SKU MacBook Pro with the 14 core (10P/4E) CPU / 20 core GPU M4 Pro with 24GB RAM, 1 TB SSD, and 3 Thunderbolt 5 ports would be a beast of a laptop with great battery life. The warranty will be out on my M1 Pro work laptop in a few months, and as much as I dislike macOS, it's going to be hard not to refresh to an M4 Pro machine. I'll have to see what the x86 laptop options are at the time.

The single most interesting x86 coming down the line is AMD's Strix Halo APU. Look's to be out in q1 2025. 256-bit memory bus and huge on-chip GPU.

[coder]

Memory compression is not limited to Apple. It's basic functionality of an OS.

I've never configured ZRam, on my Linux boxes.

Heck, building Android requires at least 64GB of RAM, if you don't want to use SWAP.

I'm sure that depends on how many cores/threads you're using. For a long time, my heuristic for compiling a C++ tree was to have 1 GB of memory per hardware thread. After a while, I've increased this to 2 GB per thread, but even that might be slightly overkill for higher thread counts.

[gnattu]

The point of the post in general was simply that if the VM can be SO good then there could be an argument for running the same benchmarks in a Linux VM while no "native" alternative is currently available (could be a better comparison to the other machines than bringing macOS into the mix) - and for some users a VM might be a practical solution while not "messing" with the mac.

The perf cost of a VM depends on the workload and for some workload (IO intensive) the performance loss could be as huge as 40% which could be less comparable to the host performance.

But I do agree performance in a Linux VM is a real world use case and the benchmark in this environment is a useful.