The new CEO doesn’t start until next Monday. We don’t know whether Marketing ran this by him. That said, until a company has new products available, they will do everything they can to sell what they have.Boy is Intel in trouble if this is an example of their CEO's intelligence.This form of FUD might have worked 10 years ago but with everybody and his brother testing the M1 Macs against higher end Intel chips and showing the testing on Youtube it isn't going to work. Typical Intel though - use antiquated methods to deal with modern problems.
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.
Intel Downplays Apple's M1 Chip With 'Carefully Crafted' Benchmarks
- Thread starter MacRumors
- Start date
- Sort by reaction score
I realize that, but my statement t holds true, you cannot buy it from MSFT nor is it bundled with any computer other than the Surface
Lenovo sells Windows on Arm devices. There may be others. WoA is currently OEM only, Snapdragon only, but not Microsoft only. The last thing Intel wants is other OEMs seeing the M1 and starting or expanding their WoA lineups.
dell 32 gb i7 11th 2TB SSD is nearly 300€ less compared to the 16 gb M1 Macbook Pro 2TB SSD
Well, these PowerPoints must have ordered by Intel under command of its former CEO. I am not sure if, in only two weeks, the new CEO even had the time to look at it before it went public.
So are you saying that tribalism is okay with tech? I feel all tribalism is bad, as all it does is divide people. It's backward thinking, and has no place in humans in 2021. I would even go as far as saying that any type of tribalism shows a basic lack of humanity.
MS does license it to OEMs - that’s their whole business model. Lenovo, HP and Acer have produced Windows-on-ARM laptops, although they have not exactly set the world on fire. WoA on Mac would probably entail Parallels or someone making an OEM deal with MS to bundle WoA with their hypervisor: that used to happen with PPC products like SoftWindows. So while it is far from certain, it doesn’t involve any major U-turns from Microsoft.
Anyway, although the preview version was promising, it was clearly a way off being ready for the big time, so nothing will be announced until the kinks are out (for one thing, it sounds like some of the windows tools are currently ARM32 binaries which won’t run on M1 - but that’s likely something that MS will need to fix long term as other ARM chips drop 32 bit).
ISTR plenty of people here called Apple out for the vagueness of their figures. I recall posting comments along the lines of “its an ad, folks!”
Point is, though, that was followed by a ton of independent reviews (including non-fanboys such as Linus tech tips) showing that the M1 machines thrashed Intel, and while Apple’s claims were vague, they weren’t unrealistic.
With these intel claims, you can pretty much stop reading at the bit where they test performance and power consumption on different processor models (when M1’s “superpower” is low-power performance).
The m1 just one of my macs and I love it! I sold my 16" because so many plugins weren't working on
Catalina! There are many plugins that aren't ready for the m1 yet but Logic is so powerful
and chock filled with its own plugins that it doesn't bother me. That m1 air is a great deal!
Right but the m1 air has no fans and so far doesn't get hot. But I get your point. Thats why I sold my
16" macbook. loud fans. Im in Maine so the heat feels good!
I'm pretty much done here because you and few others are too lazy to read back through the thread. Had you read back through the thread you would've seen that the post I replied to was referring to the Intel-based Macs, not the M1. I'm sorry for you that you respond without educating yourself the information posted throughout this thread. 🙄
What if I want to do a taxing task while it's on my lap?
Everybody is missing a crucial point between the two platforms. Intel's chips are CISC (complicated instruction set) vs Apple's RISC chipset (reduced instruction set).
Reduced instruction set processors offload some processes to the application enabling the processor to complete an instruction in a single cycle. Complicated Instruction set processors handle all tasks, requiring several cycles to complete an instruction.
It is this difference between the two architectures that keep the M1 cool, while Intel's processors get hot trying to keep up.
Until Intel adopts the RISC architecture its processors will continue to lag the M1 (and its successors).
Reduced instruction set processors offload some processes to the application enabling the processor to complete an instruction in a single cycle. Complicated Instruction set processors handle all tasks, requiring several cycles to complete an instruction.
It is this difference between the two architectures that keep the M1 cool, while Intel's processors get hot trying to keep up.
Until Intel adopts the RISC architecture its processors will continue to lag the M1 (and its successors).
This is very wrong. There is no such thing as “offloading some processes to the application.” The application has no way to “process” anything other than having the processor do the processing. And many CISC operations also execute in a single cycle.
The differences between RISC and CISC are:
1) CISC uses microcode, which means there is a state machine which decodes, for *some* instructions, single instructions into multiple, sequenced, microops, which must be issued in sequence. This requires additional pipeline stages in the instruction decoder, and increases the penalty when a branch prediction is wrong or there is any other exception that causes the pipeline to have to be flushed.
2) CISC processors have fewer registers than RISC processors, which can reduce performance depending on workload. On the other hand, when an exception or process switch occurs, the CISC penalty may be less than RISC because it there is less data to write back into RAM.
3) CISC processors support more complicated instruction formats, typically with multiple instruction lengths. Often this requires a state machine or other complicated logic to determine where a particular instruction ends; again, this complicates the instruction decoder and has associated speed penalties.
4) CISC processors allow complex addressing formats, and allow arbitrary instructions, such as ALU operations, to access memory, and not just registers. This is usually accommodated by microcode, but it nonetheless creates complications in the Load/Store unit and makes it more difficult to do things like write coalescing, etc., which affect performance.
5) Both types of processors typically have multi-cycle instructions (it’s almost impossible, for example, to have a single-cycle integer multiply or divide instruction), but RISC keeps these to a minimum, which allows for simplicity in the reservation stations and scheduling units.
6) The extra hardware necessary for CISC takes up die area and power, and increases the length of the critical paths on the chip, affecting power usage and performance. This is particularly the case where things like complicated branch predictors are necessary to make up for CISC disadvantages.
7) CISC instruction streams typically take less space to be represented in memory (10-20%, though it can vary widely depending on the nature of the instruction stream), so smaller instruction caches can be used. This is less of an issue now that the data working set typically far exceeds the instruction working set.
A dell 32 gb i7 11th 2TB SSD is US$2,399.00 while a M1 Macbook Pro 2TB SSD US$2,299.00 which proves my point as the M1 is cheaper here in the states. I even used Amazon for the Dell and Apple's website for the Mac to make things even better for the PC and the PC still was more expensive. Also thanks to its unified memory the 16 gb M1 is on par with a 32 gb Intel machine. Heck, the M1 mini was beating the pants off the low end Intel Mac Pros (base US$5,999.00 through Apple) and those have a 3.5GHz 8‑core Intel Xeon W CPU.
Last edited:
As an Amazon Associate, MacRumors earns a commission from qualifying purchases made through links in this post.
Well. Originally. Apple stretched the truth there a LOT towards the end of the PPC era.
For those interested, I like this video that Engadget put up describing the differences between Intel and Apple Silicon.
www.engadget.com
It’s a good description of how processors work that builds on the descriptions to provide a good concluding explanations.
Apple's M1 isn't witchcraft, it's good chip design
Apple used a super-wide processor design paired with fast memory and cache to create a speedy, efficient processor for its new Macbook Pro.
www.engadget.com
It’s a good description of how processors work that builds on the descriptions to provide a good concluding explanations.
Already created a discussion item here:For those interested, I like this video that Engadget put up describing the differences between Intel and Apple Silicon.
Apple's M1 isn't witchcraft, it's good chip design
Apple used a super-wide processor design paired with fast memory and cache to create a speedy, efficient processor for its new Macbook Pro.
It’s a good description of how processors work that builds on the descriptions to provide a good concluding explanations.
Good balanced video on Apple's M1 chip design
may be it is different in europe:
https://www.dell.com/de-de/shop/laptops/neu-xps-13/spd/xps-13-9310-laptop/cn93227
: 2299€
https://www.apple.com/de/shop/buy-m...-m1-chip-mit-8‑core-cpu-und-8‑core-gpu-512gb#
: 2599€
"16 gb M1 is on par with a 32 gb Intel machine" -- wishful thinking - probably the contrary is the case (read cisc vs risc)
As an Amazon Associate, MacRumors earns a commission from qualifying purchases made through links in this post.
For those interested, I like this video that Engadget put up describing the differences between Intel and Apple Silicon.
Apple's M1 isn't witchcraft, it's good chip design
Apple used a super-wide processor design paired with fast memory and cache to create a speedy, efficient processor for its new Macbook Pro.
It’s a good description of how processors work that builds on the descriptions to provide a good concluding explanations.
Just a word of caution: in addition to the 1T/1C debacle in the video mentioned in jdb8167's thread, it also fails to attribute Andrei @ Anandtech for his work in breaking down the A14/M1's characteristics, which is not good. There may be other mistakes as well. The Anandtech guys were ... less than impressed by the video when it was brought to their attention.
Not to get too deep into 1T vs 1C debate here, but this whole thing was started by Usman at WccFtech (who is a senior tech editor there, yikes) because he was pissed by Apple's marketing claims about single core performance and he went and got himself horribly confused by what that actually means. Sadly this then spread because many tech journalists don't seem to understand the basics of what they are actually writing about. Normally I try not to be so caustic, because we all make mistakes including yours truly, but these are supposed to be professionals and this is not super advanced technical jargon. It can't be, *I* know it.
Truthfully, testing the performance of a "physical" core doesn't have much if any relevance for desktop, or laptop computers. [Side note: It does for HPC/cluster chips, maybe workstation depending on use, as often they are rented/utilized on a per-virtual-core, sometime physical-core, basis so understanding per-[physical] core performance can matter in these instances. Since such chips are often fully utilized at all times, the correct way to do this is to fully saturate the chip, get the performance numbers and then literally just divide by the number of physical or logical cores.]
The terminology can be confusing (though again these are supposed to be professionals so they should know this). The reason some benchmarks use the words "single core" and "multicore" instead "single thread" and "multithread" is 1) historical, and 2) you can think of "core" meaning "logical core" (some people call them "virtual cores") not "physical core". An SMT8 chip like POWER8, a distant cousin to the PPC chips in the G-series Macs, has 8 threads/core which for 4 physical cores will present 32 virtual cores. Contrary to the original claim, the benchmarks are generally agnostic to how the physical cores are structured, thus you can test them on SMT0,2,8,bazillion, it doesn't really matter. The reality is if you want to test single, lightly threaded, and multithreaded apps, some of which will encode affinity for their threads between cores, some of which will deliberately disaggregate threads, most won't care, then you just do that. The performance in these different contexts are what's important (and for your workloads you personally may care about one or the other more). If you want to normalize performance across CPUs, the standard ways are energy consumed and/or price.
Okay I guess I got kind of deep into it.
Last edited:
Actually, yes. Okay, I can't comment on the exact ratio but a lot of this is due to the fact that macOS on the M1 (and iOS/iPadOS on the A-series chips too) will make *very* heavy use of the SSD as a swap and is *very* optimized for that. Obviously this has limits, but it is shocking how well it seems to work.
Maybe people didn't look that hard previously, but it doesn't seem to do it as much on Intel chips - though technically I can't think of a reason why it couldn't if the computer has a fast enough SSD (someone more knowledgeable than me can chime in here) ... so as far as I know this isn't really about the CPU itself and it's possible macOS on an Intel CPU will behave the same way in Big Sur? Dunno. Anyway, this can have negative repercussions for SSDs (important since these computers are meant to last and the SSDs are not replaceable!), so we'll just have to see SSD lifetimes on the new machines as they age.
CISC vs RISC doesn't matter for this. CISC code can be smaller by about 10-20%, but code size such is a trivial part of RAM usage these days that this doesn't really matter anymore.
Last edited: