PRADA SLUT posted:

Throttling aside, the performance of the M1 MBA and Pro are identical (assuming the same core configuration), correct? Like they have the exact same SoC?

hatty posted:

The only gaming I do on Mac are emulators and DOOM WADS and I haven’t seen any drops in performance from a 16” to a M1 Air. What kind of modern games do people play on macs?

mediaphage posted:

i mean it's going to be better than your current system in pretty much all ways save for where it's basically the same. if that's worth the upgrade to you now vs waiting for some potential future upgrade then you should go for it.

Boris Galerkin posted:

Can the M1 MBA replace my 2018 MBP? My MBP has 16 GB RAM and 500 GB SSD which is what I'd go for with the Air so no loss there. My MBP is the base model i5 2.3 GHz and from what I can tell the M1 blows it out of the water. Screen resolution and technology seems to be the same and I think the only thing I lose is the touchbar which I have literally never used in the 2 years I've owned this laptop. Going from 4 USB-C ports to 2 might be potentially an issue but I don't think I've ever needed more than 2 at once (but the option is nice of course).

I would use it for Office, PyCharm/VSCode/Xcode, and maybe get back into WoW. For the coding and gaming I'd probably plug it into a 24" or 27" 1080p/1200p monitor, nothing fancy.

Just wondering cause apparently Apple will give me a $750 trade-in for my MBP making the 500GB MBA $400. I looked on Swappa for other people selling my MBP model but they are listed by people with no pictures and no ratings and for $1100 and tbf I hate dealing with people trying to buy stuff from me. Oh and I have some scratches on the underside that I know for sure people are going to nitpick about and demand discounts on.

Kingnothing posted:

No so far. You need a display link adapter atm.

Splinter posted:

We're you already looking into upgrading your MBP? I think an M1 Mac would be able to replace your current laptop, and probably be better than your current laptop in most tasks, but your current laptop also seems like it still should be perfectly capable for what you do. Why not wait until next year when we're likely going to see a more "pro" M chip & GPU released, which for dev work and gaming might give you significantly more longevity than the M1 (and also that gives time for all the dev tools you may use to get properly ported to ARM, which might avoid some early adopter headaches).

TheMadMilkman posted:

Maxing out RAM and storage pushes the estimated delivery date back to late January for any of the M1 machines. That's absolutely wild to me.

TheMadMilkman posted:

Maxing out RAM and storage pushes the estimated delivery date back to late January for any of the M1 machines. That's absolutely wild to me.

Mister Facetious posted:

"Future Proofing".

I doubt the Techtubers actually have that much reach, but every one of them I've watched recommended 16gb+512 minimum, regardless of what performance they showed in their own videos, or the intended use case.

Mister Facetious posted:

"Future Proofing".

I doubt the Techtubers actually have that much reach, but every one of them I've watched recommended 16gb+512 minimum, regardless of what performance they showed in their own videos, or the intended use case.

Nitrousoxide posted:

The ram being physically on the die DOES reduce the need somewhat. Being physically close to the places that need the stored data means that it takes less time to retrieve data onboard there, less powerful hardware is needed to successfully transmit the data to the ram for storage due to electrical losses from transmission distance. Less transmission time means slightly less of a ram buffer is needed to store pending calculation data.

quote:

Also, if designed for it, the fact that the GPU and CPU can call on the same pool of ram for the relevant data rather than having separate copies stored locally does mean that the overall system cpu + gpu ram needed would be lessened.

quote:

That said, I'm not at all confident that this reduction altogether reduces the need in a meaningful way to the end-user. I'm sure there is a real reduction but it's probably not much.

MacOS just, in general, being more aggressive in throwing older applications and data into the swap than windows is probably does a lot more than any physical proximity of the ram does to reducing the utilization.

Hello Spaceman posted:

i wonder if this memory on package architecture boosts the effectiveness of the memory compression introduced in mavericks. or if that tech was designed with this in mind (or allowed this to be a hardware optimisation)

Boris Galerkin posted:

I get that the specs are better in every way but since we're all Apple users we should know specs aren't everything. I'm sure whatever Samsung or Google or Erikson phone has cameras that bow the iPhone cameras out of the water, and Google probably has a way better and fully featured system integrated smart assistant. But those are just the specs and the user experience, for I assume most of us, are unmatched when it comes to iPhone hardware.

So I was mainly looking for people who have had an Intel MBP and switched to a M1 MBA with similar use cases. When I bought my MBP I didn't do the air because the 11" or 12" screen was impossibly tiny and IIRC they maxed out at 8GB RAM and had like an ancient Atom or i3 processor which would not have worked well with my use case. Now I'm looking at the M1 MBP (8 core GPU one) and spec for spec it's better than my existing 2018 MBP but like I said above, specs aren't everything.

My use case is primarily Office applications, Zoom/Skype/whatever, PyCharm, VSCode, Xcode, and I'd like to get back into playing WoW. I have no idea if the fanless MBA can do all of that especially WoW. Alternatively the M1 500GB MBP is $650 after trade-in so only $150 more but that gives pause.

BobHoward posted:

No

- the DRAM isn't physically on the M1 die. It's standard LPDDR4x soldered to the M1 package

- M1 LPDDR4x runs at exactly the same bandwidth (4267 Mbps * 128 bits wide) that Intel supports in off-package mounted LPDDR4x in the 11th gen Core chips Apple would've put in the MBA and MBP13 if they hadn't switched to M1

- latency isn't any better because the time of flight of signals through the traces connecting DRAM to the processor is a negligible component of DRAM latency

- there may be slightly less power required to talk to the DRAM because the traces are short, but that doesn't translate into less memory use, just saves power

- "Less transmission time means slightly less of a ram buffer is needed to store pending calculation data." is a very conceptually confused sentence which is difficult to even respond to, but trust me it's very very wrong

(Mild apologies to mediaphage for my op overreaction but ideas like these are what's been spreading around the internet like wildfire and it's hard to stop them because most people just don't have the background in CS/EE to know any better)


While true, the gain is that Apple doesn't need to include an extra set of DRAM chips just for the GPU. But those were never helping with how much main RAM you needed.

Also, all the computers Apple put the M1 into replace Intel machines which never had a discrete GPU with its own RAM in the first place. Apple UMA is better than Intel UMA, but there won't be a profound difference in memory use between the two.

(This is the area I mentioned where there's a chance for real savings in applications well-optimized for Metal and the M1 GPU. However, for now, most non-Apple apps aren't that, most will never be that since they don't call Metal in the first place, and it's very unlikely to move the needle enough that you could get away with an 8GB RAM config when you'd otherwise need 16GB.)


If anything I'd say Windows is the more aggressive of the two when it comes to preemptively swapping stuff out. Windows swap management is weird and does dumb things because Windows.


Memory compression is completely orthogonal to all this. It's a part of the macOS kernel's VM subsystem which, when it's time to swap a page out, tries to keep it in RAM in compressed form instead. This helps reduce swapping out to disk, especially in circumstances where you only need a little more RAM than you physically have.

Compressed pages are still "swapped" in the sense that any process which attempts to read or write a compressed page will take the penalty of a page fault interrupt which invokes the VM system to bring the page back in a usable form. Decompressing it should be faster than reading it from disk, but it's still a form of paging AKA swapping.

Kingnothing posted:

Thank you for introducing sanity to the thread while it tried to convince a guy who held onto his previous machine for 7 years that he didn’t need 16GB of ram.

I get the “future proofing is stupid” mentality, but that extra couple 100$ could be the difference between like two or 3 OS generations in usability, never mind the video editing side.

Space Gopher posted:

There are plenty of systems that don't have a viable M1 replacement yet. And it's not like Apple is going to forcibly remote-disable x86 macs. The M1 is very good and it'll have an impact on the desirability of current systems, but they won't be worthless and people don't need to run out to replace working computers.

The long-term price curve for the last x86 macs will be interesting to see. I'm guessing it'll fall off quickly at first but stick on a higher-than-you'd-expect price for the people who have to nurse along some obscure but vital piece of software that doesn't work quite right under Rosetta and won't ever see a port or replacement. G5 PowerMac towers are worth more than Mac Pros that are several years newer, and far faster and more power-efficient, because people have to keep their legacy software going.

KOTEX GOD OF BLOOD posted:

I think the price curve for the machine I just sold (15" MBP 2015) is gonna be pretty awful though, since even the base Air trounces it in every way. That applies to the newer machines as well but I see your point about legacy software, I just think your notion applies more to the newer machines.

Splinter posted:

Yeah I could see there being extended demand for whatever ends up being the last top-ish spec Intel MBP/iMac/Mini/Pro, but prices for the older/slower Intel Macs probably won't be great.

Crunchy Black posted:

Amazon may have gotten themselves a custom SKU.

Kingnothing posted:

Thank you for introducing sanity to the thread while it tried to convince a guy who held onto his previous machine for 7 years that he didn’t need 16GB of ram.

I get the “future proofing is stupid” mentality, but that extra couple 100$ could be the difference between like two or 3 OS generations in usability, never mind the video editing side.

MarcusSA posted:

Possibly a dumb question but I haven’t seen it written anywhere yet.

How much ram is allocated to the GPU? Like with the Intel iris its 1.5 gigs or whatever.

Is it the same with the M1? Does it possibly scale if you have 16GB?

terre packet posted:

Can anyone recommend a long-form article that talks about ARM architecture and it's advantages & disadvantages vs. x86?

FCKGW posted:

Interesting, I know they sell a 128gb M1 Macbook Air to higher education for $799 so they probably do have special SKUs for certain customers

Crunchy Black posted:

FCKGW posted:

Interesting, I know they sell a 128gb M1 Macbook Air to higher education for $799 so they probably do have special SKUs for certain customers

Bob Morales posted:

I was waiting for the other Bob to post as soon as I read what that guy wrote, pretending he knew what he was talking about.

MarcusSA posted:

Possibly a dumb question but I haven’t seen it written anywhere yet.

How much ram is allocated to the GPU? Like with the Intel iris its 1.5 gigs or whatever.

BobHoward posted:

There is no fixed allocation, the GPU can access the full physical address space. According to one of Apple's WWDC presentations, on Apple Silicon every device like the Apple GPU gets its own IOMMU. I presume this means that any time the system wants to share some data with the GPU, it just creates some GPU IOMMU page table entries to map it into the GPU's view of memory.

I think I've read before that in principle recent Iris GPUs can also see everything, but legacy concerns keep them stuck on defining a window of physical addresses visible to the GPU. I've never cared enough to learn exactly what's going on, it sounds like classic weird legacy PC architecture bullshit. (Or maybe Intel's weird reluctance to put real IOMMUs into consumer hardware.)

Kingnothing posted:

It might be low cost retailer specific or Amazon specific cause it’s 128gb storage.

Apple doesn’t direct sell anything lower than 256gb@$899 (education) for the M1 airs.

Sad Panda posted:

Then an architecture question as you seem to be hot on that. While articles talk about the benefit of having unified memory because it reduces data duplication and the CPU and GPU can both process data in the same place, does this create a bottleneck in terms of access? Similar to my (possibly wrong) understanding of why there is separate data/instruction caches at L1?

BobHoward posted:

I mean I didn't write any of that to make that dude feel bad. I don't think any of that was original to him, literally all of it's retelling of ideas I've seen circulating in other forums and youtube vids and whatever. If it isn't your field of expertise it's pretty easy to believe.

BobHoward posted:

I mean I didn't write any of that to make that dude feel bad. I don't think any of that was original to him, literally all of it's retelling of ideas I've seen circulating in other forums and youtube vids and whatever. If it isn't your field of expertise it's pretty easy to believe.