It's 7nm with some improved process. Or 10nm with 2 improvements, or 14nm with 3 improvements, etc.
The process names are completely detached from reality in terms of actual transistor feature size. The only thing we can be reasonably certain of is that 5nm has some kind of improved density over 7nm.
> Apple claims the M1 to be the fastest CPU in the world. Given our data on the A14, beating all of Intel’s designs, and just falling short of AMD’s newest 5950X Zen3 – a higher clocked Firestorm above 3GHz, the 50% larger L2 cache, and an unleashed TDP, we can certainly believe Apple and the M1 to be able to achieve that claim.
Probably Tiger Lake-U. I definitely believe M1 is faster.
Apple has a history of pretending things like Nvidia or Ryzen don't exist when it suits them so I'm sure there will be gotcha benchmarks down the line.
Apple also compared against "best-selling PCs" several times, but the best-selling PCs are the cheapest junk so obviously Macs will be faster than those.
“World’s fastest CPU core in low-power silicon”: Testing conducted by Apple in October 2020 using preproduction 13-inch MacBook Pro systems with Apple M1 chip and 16GB of RAM measuring peak single thread performance of workloads taken from select industry standard benchmarks, commercial applications, and open source applications. Comparison made against the highest-performing CPUs for notebooks, commercially available at the time of testing."
So, "Comparison made against the highest-performing CPUs for notebooks, commercially available [one month ago]". I guess there could be wiggle room on interpreting "highest-performing", but this seems pretty good.
> with up to 2.8x faster processing performance than the previous generation [2]
> Testing conducted by Apple in October 2020 using preproduction 13-inch MacBook Pro systems with Apple M1 chip, as well as production 1.7GHz quad-core Intel Core i7-based 13-inch MacBook Pro systems, all configured with 16GB RAM and 2TB SSD. Open source project built with prerelease Xcode 12.2 with Apple Clang 12.0.0, Ninja 1.10.0.git, and CMake 3.16.5. Performance tests are conducted using specific computer systems and reflect the approximate performance of MacBook Pro.
That's.. kind of weak. How many other perf tests did they throw away before taking this one because it showed so well? I guess we'll see the real-world benchmarks when people get their hands on them.
Geekbench is not a _great_ benchmark, but it's common enough that we could use it to roughly compare.
EDIT: Apparently there are Geekbench results that are unofficial that suggest it's faster than current MBPs, but we'll have to see.
I genuinely would love some info or statistics on this. AFAIC remember, laptops wake from sleep almost instantly to the lock screen. Is it a longer wait if one wakes directly to their desktop?
I'm with you in not fully understanding the benefit. Maybe this is a technology that is hard to imagine, but is difficult to go back from (60hz, Retina displays).
I think there's a conceptual difference. Macbooks wake up after a brief pause of a second or two; phones act like they never went to sleep. There's this perception of locking your phone being a zero-cost thing, which isn't quite true of putting your laptop to sleep. I assume this is the gap they're talking about bridging.
An 8 core processor on a Macbook Air that is also energy efficient? That is truly impressive. I never thought I would consider using Macbook Airs after all the years of using Macbook Pros, but Apple surprises me once again.
At 9:15 of the keynote they claim that the "high-efficiency" cores are as powerful as the outgoing dual-core MacBook Air's cores. Seems pretty good to me.
Anyone know how you interact with these cores as a developer/user? Say if I'm running some C code with OpenMP parallelism, can I bind it to three of the fast cores?
Binding to specific cores is not exposed to userspace, but you can influence which kinds of cores it's likely to be run on by setting thread priorities and QoS classes.
With ARM, yes, and you can also selectively turn on and off cores. For example when travelling with my pinebook pro I turn the big cores off to drastically improve battery life. However it's up to Apple to expose this functionality, and we all know how much control apple wants you to have of the computers you "license" from them.
Huh, the form factors for the Macs seem identical. I expected them to be thinner or something. Fanless is nice for the Air. But I'm really waiting to see what they do on the high end with the top MacBook Pro and even Mac Pro. Will we see an Apple discrete GPU? I guess we'll have to wait a year or two for that.
wow, 5nm, memory, gpu, and cpu on the same soc. RX 560 has the same 2.6 tflops as this gpu on this chip. They say 4x more efficient at 10w and 2x more powerful than intel.
It was to be expected because they save a lot of money that they were paying to Intel. It was estimated that they could shave something like $100 per computer by switching to an in house chip.
As one has a fan and the other not, they are probably clocked differently and the cooled one might sustain full power infinitely. The pro also has a larger battery, better speakers and microphones and the touch bar.
It's already the case that the Macbook Pro and Air have the "same" CPU - 1068NG7 and 1060NG7 are physically the same die, but with different power limits.
The 13” MacBook Pro doesn’t have a discrete gpu. They’re all using the integrated gpu of the M1 chip. We have idea how well it’ll perform in real world benchmarks yet.
Apple mentions TensorFlow explicitly in the ongoing presentation due to the new 16-core "Neural Engine" embedded in the M1 chip. Now that's an angle I did not expect on this release. Sounds exciting!
Edit: just to clarify, the Neural Engine itself is not really "new":
> The A11 also includes dedicated neural network hardware that Apple calls a "Neural Engine". This neural network hardware can perform up to 600 billion operations per second and is used for Face ID, Animoji and other machine learning tasks.[9] The neural engine allows Apple to implement neural network and machine learning in a more energy-efficient manner than using either the main CPU or the GPU.[14][15] However, third party apps cannot use the Neural Engine, leading to similar neural network performance to older iPhones.
But they needed to, MacBooks are simply no option if you want to train models. I dont expect crazy performance but would be great if MacBooks would be an option again for prototyping / model development at least
In fairness, it's been possible to convert a TensorFlow model to a CoreML model for a while, and in April TensorFlow Lite added a CoreML delegate to run models on the Neural Engine.
So don't think of it as Apple walked right in with this so much as Apple has been shipping the neural engine for years and now they're finally making it available on macOS.
I can't see how something that tiny can compete in any meaningful way with a giant nVidia type card for training. I'd imagine it's more for running models that have been trained already, like all the stuff they mentioned with Final Cut.
Yeah I would imagine it's intended for similar use-cases as they use for iOS - for instance image/voice/video processing using ML models, and maybe for playing around with training, but it's not going to compete with a discreet GPU for heavy-duty training tasks
For an 18-hour battery life computer (Macbook Air) that now doesn't even have a fan, it's for a complete different market segment from where nvidia cards dwell.
Not all NN models are behometh BERTs, U-Nets or ResNets. Person detection, keyword spotting, anomaly detection... there are lots of smaller neural nets that can be accelerated by a wide range of hardware.
That depends entirely on the hardware of both the ML accelerator and the GPU in question, as well as model architecture, -data and -size.
Unfortunately Apple was very vague when they described the method that yielded the claimed "9x faster ML" performance.
They compared the results using an "Action Classification Model" (size? data types? dataset- and batch size?) between an 8-core i7 and their M1 SoC. It isn't clear whether they're referring to training or inference and if it took place on the CPU or the SoC's iGPU and no GPU was mentioned anywhere either.
So until an independent 3rd party review is available, your question cannot be answered. 9x with dedicated hardware over a thermally- and power constrained CPU is no surprise, though.
Even the notoriously weak previous generation Intel SoCs could deliver up to 7.73x improvement when using the iGPU [1] with certain models. As you can see in the source, some models don't even benefit from GPU acceleration (at least as far as Intel's previous gen SoCs are concerned).
In the end, Apple's hardware isn't magic (even if they will say otherwise;) and more power will translate into higher performance so their SoC will be inferior to high-power GPUs running compute shaders.
On the NVidia A100, the standard FP32 performance is 20 TFLOPs, but if you use the tensor cores and all the ML features available then it peaks out at 300+ TFLOPs. Not exactly your question, but a simple reference point.
Now the accelerator in the M1 is only 11 TFLOPs. So it’s definitely not trying to compete as an accelerator for training.
But in the development phase, when you are testing on a smaller corpus of data, to make sure your code works, the on-laptop dedicated chip could expedite the development process.
I agree with the parent poster that it's probably more about inference, not training.
If ML developers can assume that consumer machines (at least "proper consumer machines, like those made by Apple") will have support to do small-scale ML calculations efficiently, then that enables including various ML-based thingies in random consumer apps.
It can be surprisingly cost-effective to invest a few $k in a hefty machine(s) with some high-end GPU's to train with due to the exceedingly hefty price of cloud GPU compute. The money invested up-front in the machine(s) pays itself off in (approximately) a couple of months.
The "neural" chips in these machines are for accelerating inference. I.e. you already have a trained model, you quantise and shrink it, export it to ONNX or whatever Apple's CoreML requires, ship it to the client, and then it runs extra-fast, with relatively small power draw on the client machine due to the dedicated/specialised hardware.
Cloud GPU instances are very expensive. If you get consumer GPUs not only do you save money, you can sell them afterwards for 50% of the purchasing price.
Sure, inferencing doesn't need floating point instructions, so NVIDIA will stay the only real solution for desktop/laptop based model training for a long time.
It is just not meant to be a training device so comparing with data Center or developer GPUs is useless. Faster inference for end-users is what is mentioned by Apple and the only use case where this hardware makes sense.
Exactly. Currently I am training my models using Google Colab and then exporting the model to run on my MBP. Would be interesting if I could do it locally
Another interesting thing is that ( if this is for training ) this will become the only accelerated version of Tensorflow for macOS as:
- No CUDA drivers for latest macOS
- AMD ROCHm only supports Linux runtime
Is this just opinion? Maybe they are designed ALSO for training. I wonder if this things can replace nVidia graphic cards on training? The neural core has a LARGE area on the chip design similar to the GPU area.
I think this is for inference not learning, even though they use the term machine learning. They seem to just mean running models based on machine learning approaches.
My thinking was along the same line as yours, but the way apple framed it seems to suggest that the M1 accelerates model training and not just inference. Here's the actual quote "It also makes Mac Mini a great machine for developers, scientist and engineers. Utilizing deep learning technologies like tensorflow ... which are now accelerated by M1".
It should be pretty straightforward to test this though: installing tensorflow-gpu on a mac mini and seeing the result.
I suspect, TF's latest branch should also indicate which GPUs are supported.
Curious to hear more thoughts on this.
They do not have any hardware combination which can actually support even modest GPU intensive training sadly, so much touting running models instead of training.
Yes, need to see more strong evidence that the new MBP's can handle large amounts of ML Training using TF or CreateML so we don't have to get NVIDIA machines/laptops.
huge own-goal on Apple's part. It would be literally impossible for me to use this machine to do my job. I guess I'll have to wait until M2.
Edit: If it wasn't clear, this was not a joke. I develop a relatively heavyweight service on the JVM, and between my IDE, the code I run, and all the gradle build daemon stuff, I regularly use up more than 32GB. Often over 50GB. (Although some swap is tolerable, having the majority of my resident set being swapped at any given time means things get very slow.)
And very likely the intel integrated graphics in existing mac's (aka not Xe). The AMD integrated graphics are easily 3x faster in similar product lines in many benchmarks vs the intel's. Which means its probably roughly the same as the Amd product lines. Large parts of the presentation perf improvements are probably GPU related.
Course i'm viewing this with a healthy dose of skepticism, having been around in the PPC days when you would think from apple's marketing that the PPC based macs were massively faster than your average PC. In reality they were ok, but rarely even the fastest device across a wide swath of benchmarks, mostly sort of middling.
Yes, but they seemed pretty ambivalent about the mini for a long time. Probably easier from a SKU perspective though than coming out with new iMac versions. (Even if iMacs overall outsell Minis--which I'm guessing they do--individual models may not.)
I think they're targeting a more powerful SoC at the iMac, or at least they didn't want to announce new iMacs without being able to replace the whole range.
I mean because they're going to bring out another one really soon, that will, presumably, be drastically better. I'm guessing (hoping) that beefing up the graphics is why the 16 is coming later.
After seeing how big a downgrade these machines are over their intel counterparts (16 GB max RAM, max 2x thunderbolt, max 1 external display on the laptops, no 10 GbE on the mini), I'd absolutely buy an Intel machine now to tide me over until Apple can catch up in a generation or two
I'm also disappointed by 16g when my current laptop has 64g. However, Apple mobile devices have a history of being way underpowered on RAM specs and outperforming in real usage.
That's their typical approach. They released iPad 1 with 256Mb only, then iPad 2 was released in short time while iPad 1 has become literally unusable after next software update. That's the lesson. I am certain they will upgrade CPU, webcam, RAM, and connectivity in the next version very soon.
All the Macs looked quite impressive! It was well worth waiting for these (at least for me). But I was disappointed that the maximum RAM is 16GB. I would’ve preferred a 32GB option for better future proofing (especially with web applications needing more and more memory).
Edit: Considering the fact that the RAM won’t be upgradeable (it’s part of the SoC), this limitation is a big bummer. What may be worse is that all these machines will start with an 8GB RAM configuration option at the low end, which isn’t going to age well at all in 2020.
Same! Mid-2012 rMBP. Turns out I could upgrade the AirPort card with a used one from ebay for $20, as well as the SSD (although that was maxed out at 1TB with mSATA). Eight years later, it's still a good computer.
> Considering the fact that the RAM won’t be upgradeable (it’s part of the SoC)
While I agree that RAM won't be upgradeable (as it hasn't been in all new models the past few years), are you sure that the RAM is part of the SoC? I believe what they labelled with "DRAM" in the M1 schmatics is very likely the L3 cache instead.
Adding RAM to the SoC would make little sense from a cost and yield perspective. I also believe that 16GB of DDR4 memory are much larger than the "DRAM" part of the SoC.
And here is the picture which confused by and tricked me into thinking the OP talked about the RAM integrated into the CPU (although upon closer inspection that picture also seems to cover the whole SoC package): https://www.apple.com/v/mac/m1/a/images/overview/chip_memory...
I did not see it mentioned but I hope an app comes with Big Sur which lets us see which apps will work with Rosetta2 through Finder or some sort of report
If it's anything like the transition from PowerPC architecture to Intel, you'll be able to see this information by selecting the app package in the Finder, and invoking the "Get Info" window (cmd-I) or palette (option-cmd-I).
The onboard graphics performance seems to be impressive (1050Ti - 1060 range). I wonder if Valve and Epic will start compiling games for ARM. This MacBook could be my first gaming laptop, who would have thought
I would guess that neither Valve or Epic will do that, it is up to game devs to do that. Mac is bad gaming platform, and dropping 32 bit x86 support from Mac OS didn't help
https://www.macgamerhq.com/opinion/32-bit-mac-games/
I think it is clear they dropped 32bit x86 support so it would be easier to develop Rosetta 2 (The x86 emulation layer). I don't really see why they would drop support otherwise.
> The onboard graphics performance seems to be impressive
I agree, although caveat emptor until independent benchmarks drop.
> I wonder if Valve and Epic will start compiling games for ARM
There are plenty of other problems besides lack of raw processing power that prevent the mac and macOS from being good gaming platforms. And of course, Epic and Apple are fighting it out over App Store policies, so they're unlikely to do each other any favors.
I think there's still a no fan vs fan distinction so they'll still be able to run the Pro more powerful, but it looks like it. I half suspect the Air was the one they wanted to release and they just didn't want there to be no Pro faster than an Air in the lineup.
Is there any indication at this point that they'll make that clear? It seems kind of strange to buy a computer without really knowing what you're buying.
I guess apple didn't officially say what intel processors they have in their computers, but at least you could look it up and know what you're getting.
They have also never shared this info for iPhones/iPads, so no reason to believe they will do it for Macs going forward. But we should be seeing third party benchmarks soon enough.
Yes, but they only replaced the lowest-end 13 inch pro at this point, and are still selling the higher-end Intels. The performance versions will follow next year.
I see no reason why it shouldn't. Most of homebrew is ruby script (that is architecture-agnostic), most of software installed through homebrew can be recompiled.
Also, mb air has no touch bar, but includes fingerprint scanner. I’m sure a lot hackernews will be pleased with that, despite it not being in a machine with the pro moniker.
1,381 comments
[ 2.9 ms ] story [ 376 ms ] threadhttps://www.extremetech.com/computing/315186-apple-books-tsm...
The process names are completely detached from reality in terms of actual transistor feature size. The only thing we can be reasonably certain of is that 5nm has some kind of improved density over 7nm.
> Apple claims the M1 to be the fastest CPU in the world. Given our data on the A14, beating all of Intel’s designs, and just falling short of AMD’s newest 5950X Zen3 – a higher clocked Firestorm above 3GHz, the 50% larger L2 cache, and an unleashed TDP, we can certainly believe Apple and the M1 to be able to achieve that claim.
https://www.anandtech.com/show/16226/apple-silicon-m1-a14-de...
EDIT: added [faster than the]
Apple has a history of pretending things like Nvidia or Ryzen don't exist when it suits them so I'm sure there will be gotcha benchmarks down the line.
Apple also compared against "best-selling PCs" several times, but the best-selling PCs are the cheapest junk so obviously Macs will be faster than those.
“World’s fastest CPU core in low-power silicon”: Testing conducted by Apple in October 2020 using preproduction 13-inch MacBook Pro systems with Apple M1 chip and 16GB of RAM measuring peak single thread performance of workloads taken from select industry standard benchmarks, commercial applications, and open source applications. Comparison made against the highest-performing CPUs for notebooks, commercially available at the time of testing."
So, "Comparison made against the highest-performing CPUs for notebooks, commercially available [one month ago]". I guess there could be wiggle room on interpreting "highest-performing", but this seems pretty good.
> with up to 2.8x faster processing performance than the previous generation [2]
> Testing conducted by Apple in October 2020 using preproduction 13-inch MacBook Pro systems with Apple M1 chip, as well as production 1.7GHz quad-core Intel Core i7-based 13-inch MacBook Pro systems, all configured with 16GB RAM and 2TB SSD. Open source project built with prerelease Xcode 12.2 with Apple Clang 12.0.0, Ninja 1.10.0.git, and CMake 3.16.5. Performance tests are conducted using specific computer systems and reflect the approximate performance of MacBook Pro.
Geekbench is not a _great_ benchmark, but it's common enough that we could use it to roughly compare.
EDIT: Apparently there are Geekbench results that are unofficial that suggest it's faster than current MBPs, but we'll have to see.
I'm with you in not fully understanding the benefit. Maybe this is a technology that is hard to imagine, but is difficult to go back from (60hz, Retina displays).
With my Mac Pro, as soon as I wake it, I can see the LED on my monitors are 'alive', but it still takes that same 10s or so to be displaying data.
Also on macOS, if you have an always-on VPN, that can absolutely cause wake time challenges.
To offline a core:
echo 0 > /sys/devices/system/cpu/cpuN/online
To online a core:
echo 1 > /sys/devices/system/cpu/cpuN/online
Where cpuN is 0-4. Keep in mind there's always one core you cannot disable to process interrupts.
There isn't an option like taskset on Linux to pin or move tasks among different cores, or like anything that's exposed in Linux's sysfs.
No fan however, is impressive....
They're looking for marketshare gains.
Some people would have paid more for iOS apps on their Mac.
Given this is an architecture shift, I guess it seems to make sense to test it out with a midrange product.
We will see when we get a full teardown
If we have same CPU on MacbookAir and MacbookPro - why would I get more expensive "Pro"? Can someone explain how is Pro faster than Air with same CPU?
Also, the "Windows Guy" bit is a bit lame IMO. I have two MacBooks and one custom built PC. The PC is faster than both MacBooks combined.
Accusing people not not reading the article is against the rules here.
> CPU, GPU, memory, I/O
Screens, batteries, form-factors.
Active cooling.
Edit: just to clarify, the Neural Engine itself is not really "new":
> The A11 also includes dedicated neural network hardware that Apple calls a "Neural Engine". This neural network hardware can perform up to 600 billion operations per second and is used for Face ID, Animoji and other machine learning tasks.[9] The neural engine allows Apple to implement neural network and machine learning in a more energy-efficient manner than using either the main CPU or the GPU.[14][15] However, third party apps cannot use the Neural Engine, leading to similar neural network performance to older iPhones.
Source: https://en.wikipedia.org/wiki/Apple_A11#Neural_Engine
Besides, can the neural engine be used to speed up other tasks?
https://blog.tensorflow.org/2020/04/tensorflow-lite-core-ml-...
So don't think of it as Apple walked right in with this so much as Apple has been shipping the neural engine for years and now they're finally making it available on macOS.
Unfortunately Apple was very vague when they described the method that yielded the claimed "9x faster ML" performance.
They compared the results using an "Action Classification Model" (size? data types? dataset- and batch size?) between an 8-core i7 and their M1 SoC. It isn't clear whether they're referring to training or inference and if it took place on the CPU or the SoC's iGPU and no GPU was mentioned anywhere either.
So until an independent 3rd party review is available, your question cannot be answered. 9x with dedicated hardware over a thermally- and power constrained CPU is no surprise, though.
Even the notoriously weak previous generation Intel SoCs could deliver up to 7.73x improvement when using the iGPU [1] with certain models. As you can see in the source, some models don't even benefit from GPU acceleration (at least as far as Intel's previous gen SoCs are concerned).
In the end, Apple's hardware isn't magic (even if they will say otherwise;) and more power will translate into higher performance so their SoC will be inferior to high-power GPUs running compute shaders.
[1] https://software.intel.com/content/www/us/en/develop/article...
Now the accelerator in the M1 is only 11 TFLOPs. So it’s definitely not trying to compete as an accelerator for training.
But in the development phase, when you are testing on a smaller corpus of data, to make sure your code works, the on-laptop dedicated chip could expedite the development process.
If ML developers can assume that consumer machines (at least "proper consumer machines, like those made by Apple") will have support to do small-scale ML calculations efficiently, then that enables including various ML-based thingies in random consumer apps.
It can be surprisingly cost-effective to invest a few $k in a hefty machine(s) with some high-end GPU's to train with due to the exceedingly hefty price of cloud GPU compute. The money invested up-front in the machine(s) pays itself off in (approximately) a couple of months.
The "neural" chips in these machines are for accelerating inference. I.e. you already have a trained model, you quantise and shrink it, export it to ONNX or whatever Apple's CoreML requires, ship it to the client, and then it runs extra-fast, with relatively small power draw on the client machine due to the dedicated/specialised hardware.
(If they are or can be, I'm interested)
Exactly. Currently I am training my models using Google Colab and then exporting the model to run on my MBP. Would be interesting if I could do it locally
Another interesting thing is that ( if this is for training ) this will become the only accelerated version of Tensorflow for macOS as: - No CUDA drivers for latest macOS - AMD ROCHm only supports Linux runtime
Tensorflow includes stuff for inference.
Edit: If it wasn't clear, this was not a joke. I develop a relatively heavyweight service on the JVM, and between my IDE, the code I run, and all the gradle build daemon stuff, I regularly use up more than 32GB. Often over 50GB. (Although some swap is tolerable, having the majority of my resident set being swapped at any given time means things get very slow.)
Are they really saying this vs dGPU?
Course i'm viewing this with a healthy dose of skepticism, having been around in the PPC days when you would think from apple's marketing that the PPC based macs were massively faster than your average PC. In reality they were ok, but rarely even the fastest device across a wide swath of benchmarks, mostly sort of middling.
Shame no 16 inch pro though. Surely they need to update that quick because who is going to want a 16 inch intel Mac now?
Signs point to another event in January, I'd expect it there with a heftier SoC.
Edit: Considering the fact that the RAM won’t be upgradeable (it’s part of the SoC), this limitation is a big bummer. What may be worse is that all these machines will start with an 8GB RAM configuration option at the low end, which isn’t going to age well at all in 2020.
https://www.apple.com/macbook-pro-13/specs/
“Memory 8GB unified memory Configurable to: 16GB”
While I agree that RAM won't be upgradeable (as it hasn't been in all new models the past few years), are you sure that the RAM is part of the SoC? I believe what they labelled with "DRAM" in the M1 schmatics is very likely the L3 cache instead.
Adding RAM to the SoC would make little sense from a cost and yield perspective. I also believe that 16GB of DDR4 memory are much larger than the "DRAM" part of the SoC.
Here is a picture of the SoC with the 16GB of DDR4 RAM: https://www.apple.com/v/mac/m1/a/images/overview/chip__fffqz...
And here is the picture which confused by and tricked me into thinking the OP talked about the RAM integrated into the CPU (although upon closer inspection that picture also seems to cover the whole SoC package): https://www.apple.com/v/mac/m1/a/images/overview/chip_memory...
My work machines on the other hand were all specced with 32GB.
I agree, although caveat emptor until independent benchmarks drop.
> I wonder if Valve and Epic will start compiling games for ARM
There are plenty of other problems besides lack of raw processing power that prevent the mac and macOS from being good gaming platforms. And of course, Epic and Apple are fighting it out over App Store policies, so they're unlikely to do each other any favors.
http://tim.id.au/laptops/apple/misc/pc_compatibility_card.pd...
Previous models had a massive delta in CPU performance based on using low power (Air, no fan) or medium power (Pro, with dual fans) Intel chips
I guess apple didn't officially say what intel processors they have in their computers, but at least you could look it up and know what you're getting.