@ mi7chy, I mean TSMC is huge too, or ASML's EUV and now their pelicles, but really, the ISA matters insomuch as it imposes costs and limitations as to the IPC, area and energy/power efficiency of your realized microarchitecture.
Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Qualcomm Nuvia vs Apple Mx
- Thread starter crazy dave
- Start date
-
- Tags
- amd apple silicon intel nuvia qualcomm
- Sort by reaction score
There could/would be fabric inside the "CPU" subsystem to get to/from elements inside of that locality. For cores to to shared L2 (or subsystem level L3) would be through a network. That could be independent of central backbone network of the whole SoC. It has to all be coherent (between caches et. al. ), but can have a much higher bandwidth inside the subsystem than between subsystems.
Seems doubtful that a small operation like Nuvia would be off inventing a new intersubsystem bus on generation 0.5 (or 1 ). Arm's Neoverse has an interconnect out to server level memory ( DDR+ECC). Take that off-the shelf and 'gut' the cores part. That would :
1. cut down time to market for a small company with limited budget.
2. faster exit of inevitable legal dust up with Apple. ( interconnect? we licenced it ... can't be stolen IP. PCI-controller.. ditto . security processor .. ditto. etc )
3. makes them a much more juicy acquihire target if CPU largely just plugs into other Arm standard IP that some other server implementer may have started on ( Amazon , Microsoft, etc. ) that may decided to pull inhouse. ( which is almost exactly what happened... only in slightly different market looking for a "plug in" CPU core technology. ). Somewhat standard silicon valley model these days. Create a collection of talent and a nominal product.. get acquired by bigger tech with deep pockets ( pay day for all the early investors ) and don't have to do the "grow the company over time" thing at all. Pretty good chance they were told to make themselves easily 'digestible' by at least one of the deep pocketed early investors.
An intra-subsystem fabric would mean couldn't mix and match with other Qualcomm cores though. They would need to compose a smaller core if going heterogenous CPU core.
It is just one (or two if "smaller core" is another subsystem ) adaptor to the internal bus that Qualcomm already had. There are more other subsystems in the SoC ( Modem , Hexagon/AI/ML , Grahpics, security ,etc.) doing an "adapter" to all of those to Nuvia's would be lots more work on stuff that was probably already in flight on design.
I completely understand what you mean and I think Qualcomm is in a godo position for their first chip. However if you think about it, it's really in the software side of things where the things will really show.
Apple's Rosetta 2 strategy worked wonderfully and it was almost seamless the question now is if Microsoft can do the same for Windows.
Regarding ARM on the Datacenter I can tell you that for general workloads yes, but specific workloads like SAP/HANA and some other ERP still require x86 due to binaries. Can't say more for now but cloud with ARM will become more cheap in the end and allow for amazing performance.
Totally agree but note that Windows on ARM has Dx12 (will be useful for DirectML on WSL 2 eventually, which is also on WoA), it has Photoshop, Lightroom, OpenGL, too as of 2020/2021, and Microsoft are now working to bring better support for 64-bit emulation with ported Aarch64 native binaries that have x64 extensions. Also, a lot of Rosetta was really about the inclusion of (likely not that expensive IMO) the ability to emulate the X64 memory model/total store ordering or what have you, not that Rosetta 2 itself was at all a bad translation layer - if anything all the more because of the tight integration.
Anyways, I agree Rosetta 2 and Apple's Arm transition was phemomnally well executed. It exceeded my expectations, even as someone who spent time arguing for the bullish position on the transition. Still loving the PC Gamer caucus cope on this to be honest.
Had to correct a poor fellow on fricking Zen IPC the other day RE: which I was assured was "no question" (rough translation) superior to anything from Arm. This is completely false lol. As of even Zen 3/3+, AMD are simply behind the ostensibly horrific Arm reference X2 cores by a good 15-20%, even with the paltry (certainly not as much L2/L3/SLC as Zen 3/3+ or Alder Lake include, or an A14/A15) cache allotments on the latter cores as seen from Qualcomm, MediaTek, lol.
Last edited:
Announcing ARM64EC: Building Native and Interoperable Apps for Windows 11 on ARM
Today, we’re announcing ARM64EC (“Emulation Compatible”), a new way to build apps for Windows 11 on ARM. With the latest Visual Studio tools in preview and the Windows 11 Insider SDK, you'll be able to take advantage of ARM64EC to incrementally
blogs.windows.com
RE porting.
Announcing ARM64EC: Building Native and Interoperable Apps for Windows 11 on ARM
Today, we’re announcing ARM64EC (“Emulation Compatible”), a new way to build apps for Windows 11 on ARM. With the latest Visual Studio tools in preview and the Windows 11 Insider SDK, you'll be able to take advantage of ARM64EC to incrementally
RE porting.
that's quite cool
Kuba Wojciechowski has leaked the characteristics of Qualcomm's Apple Mx rival SoC.
Not sure how that is going to complete. The iGPU is the exact same one as the 8cx gen 3. So any apps that lean on GPUs there will be a wide gap.
There are 8 memory channels. The M2/M1 have 8. Pro and up > 16 . In some sense not surprising that the iGPU has gone almost no where. There would be nothing to 'feed it' if it had more computation units. M2/M1 only go up to 4 P cores so somewhat questionable throwing another 4 Nuvia cores at just 8 channels is going to work well along side a iGPU also putting demands on the same bus network. ( benchmarks which largely get sucked into the 12MB system cache will work well. Apps with a heavy duty (broadly active) 8GB footprint less so. It is not balanced as well as the M1/M2 and in trying to get into a pissing match with the M1 Pro they left behind the memory bandwidth. )
Support for dGPUs ... so lower overall system perf/watt competitiveness. Although a better fit with the market dynamics in the Windows laptop space ( some folks have to have a Nvidia GPU to be 'happy'. )
Looking at the M2 Pro/Max I highly doubt Apple is going to loose a wink of sleep over anything listed there.
" ... For integrators not wanting to use NVMe for the boot drive, Qualcomm included a 2-lane UFS 4.0 controller with support of up to 1TB parts. ..."
Gotta wonder if going down the same rabbit hole the previous Qualcomm Windows SoC went where charge so much for the mandatory Cell Radio subsystem that system vendors pinch pennies on other parts of the system to reel the overall system price back into being competitive in the general Windows laptop space.
Even against AMD's new laptop line up rolling out this year (and maybe Intel Gen 14) there are some 'not so bright' spots for them. If the radios drive up the system price to high, it probably won't compete well in a broader market. Battery life will be a win and much of the rest much closer trade-offs.
P.S. It would be far, far , bar better for Qualcomm to copy Apple's marketing where mainly talk about the generation over generation improvement of their own stuff ( mainly compare M2 to M1 , M2 Pro to M1 Pro ) rather than trying to invite some Apple fight. Or ignoring their AMD/Intel competitors ( going to have to lean on battery life saving over "max awesome performance" on drag racing benchmarks. For folks willing to plug in a Windows Laptop .. they are not going to win. ). So 8cx gen 4 is big bump over gen 3 and mainly stick to that.
Last edited:
Windows 11 on ARM would be awesome as it introduces competition vs Intel/AMD duopoly.I think this interview on Anandtech was an interesting read but one thing in particular caught my eye:
It seems that when Nuvia-based SOCs finally release for PC the primary comparison Qualcomm will be drawing against will be Apple’s M-series SOCs rather than Intel/AMD chips.
That, to me, seems like a mistake. In the mobile space, Apple as the primary point of comparison makes sense. But while both Apple and Qualcomm will be making ARM-SOCs for the computer, the majority of customers that Qualcomm will be looking to acquire are going to be from the Windows-x86 side. For one thing, it’s a much bigger pool of people. I’d think they’d want to focus on giving those customers a reason to switch from x86 to their ARM cores.
Maybe they think their new SOCs will be so good that they won’t have to draw comparisons with AMD/Intel. But that’s not the answer he gave (despite Ian leaving the door open for that).
Anyway, these new chips are a long ways off but we’ll see how they perform and what Apple themselves have out by that point.
But a challenge here is process node die shrink. Apple appears to have 1st right of refusal for any future die shrink.
This is a competitive advantage in terms of performance per watt, raw performance, power draw and battery life.
My hope is that Windows 11 on ARM will help move Windows app towards a fat binary that leans more on ARM than x86 for the future.
So that x86 chips will only be relevant to legacy software.
To take market share from intel/AMD, these hypothetical ARM devices will have to do something better, and battery life isn't the answer people are looking for. Being plugged in most of the time eliminates that issue.
The something better will have to be performance or compatibility based, and that's not going to happen anytime soon.
Now if we ever come to a time where power is a problem, then all bets are off, and it may get that way in Europe, but not here in the U.S. unless there's a major war. (and I suspect I wont care, as I live in a ground 0 area)
That's where I think they'll sell.
Forget about trying is what I'd like to see them do. There's in absoultely no incentive for me to buy an ARM Windows PC right now.
But if they still try, actually come up with a reason for people to buy an ARM PC over an intel one, but keep the backwards compatibility ideal in mind...
It doesn’t have to be the answer but it sure is one.
And not needing to be plugged in most of the time in the first place will be good.
Battery life can’t be the only thing they sell these on but your underselling of the advantage they’ll have if they have impressive battery life is bizarre.
For the MBA, battery life is key, no question. For the M1/M2 Max and even Pro models, battery life isn't the only deciding factor - more so when its a desktop replacement and it is plugged in more often then not.
Plenty of people out there need to be mobile.
The unique value of Apple Silicon is that you don’t have to choose between performance and portability. You can work on the desk, train or your couch - doesn’t make a difference. This is something entirely impossible with x86 machines which take a big energy hit maintaining that performance.
Qualcomms marketing so far suggests that this is the niche they are going after as well. And personal mobile computing is the biggest market segment.
And I think it's just as bizarre that you think it's that important. It's just a total non issue for me. (and any of the laptops I buy for work)
Probably, but I don't know any of them.
You don't know many types of x86 laptops, but whatever (some have very long battery life as well, though they tend to be heavy!). Apple laptops don't run x86 Winodws and that's the biggest problem of all for me. Being able to run on battery for a long time doesn't mean much if it doesn't do what I need. I'd like an 14" M2 MBP myself, but it would just be for home for me, and actually more than I want to spend.
IBIWISI. But it's definitely not something I'm excited about seeing. (not even as much as my desire to get an M2 MBP) There's just no reason for me to get one, for work or home. I use an iPad or an android tablet at home, not a laptop -- they're even lighter and easier to carry and also have long battery lives. The android tablet I have even is OLED...
I tried one, it didn't work for me, portable or not. My work x86 laptop is lighter, cost pretty close to the same price once you put 16G of RAM in the MBA, has 32G of RAM, and has fans in it. It stays plugged in 99% of the time, and it's actually just as quiet.
That does not mean that those kind of people doesn't exist.
For example I never use a laptop as a plugin first device. For that kind of workflow I have some competent desktops.
Moreover, in my circle of friends no one uses a laptop for mainly stationary tasks.
This does not invalidate your experience but proves that different people have different needs.
In 2021 laptops outsold desktops 3 to 1 and the trend is upwards. If this continues the desktop will cease to exist as a relevant category fairly soon.
This is not about battery life, this is about performance on battery. No x86 laptop with powerful hardware offers full performance on battery (and if you change the power setting so it does, your battery life will be cut to almost zero).
I understand that, but that’s an entirely different topic.
What's your fairly soon? 10 years, twenty? 100?
I don't think they will never go away, but you might not see them anymore as a consumer eventually. If you need computer power, portable will never cut it. Most of the PC's we buy for work are still desktops, both because they are cheaper, and more powerful. The consumer market is different of course, and it will probably become portables, exclusively. That doesn't change business needs though.
That's a totally false statement. You just need a laptop with good cooling and a large battery, they are out there. Thinkpad T series are good examples.
Not at all when you try and say Arm will take over the Windows world...
Mainly because I've also been a Mac owner since the first intel model Mac Mini, and have spent thousands on many different Macs. That doesn't allow me to be interested in Macs and to post on this forum??
For work, yes, it's all Windows, but I do have a life outside of work occasionally too!
I would say 5-10 years. Projections say it will be 4 to 1 in 2025.
Modern x86 laptops with top CPUs throttle the performance by anywhere between 20-40% while running unplugged.
Dint believe me, check out some benchmarks: https://www.ultrabookreview.com/56783-amd-ryzen-7-6800h-vs-intel-core-i7-12700h/
I have never claimed anything like that. Frankly, I couldn’t care less about the Windows world. I have zero interest in that platform.
Last edited:
I love this thing you always do where you start with some claim about what "people" think the the moment there's pushback it's suddenly just about you and your use.
This doesn't seem like a claim that's just about you, but work.