What are the RAM specifications for the M4 MacBook?

MacTracker says that the memory is

• 7500 MT/s LPDDR5X – on 14" MBPs with plain M4 chips
• 8533 MT/s LPDDR5X – on 14" and 16" MBPs with M4 Pro and M4 Max chips

Note that you cannot expand RAM on any Apple Silicon MacBook Pro after purchase. There is no point in buying more RAM with these specifications because there is no way for you to install it.

If your interest is driven by curiosity, MacTracker lists memory bandwidth for the 14" and 16" M4-series MBPs as

• 120 GB/s (plain M4) (14" models only)
• 273 GB/s (M4 Pro)
• 410 GB/s (M4 Max with 14-core CPU)
• 546 GB/s (M4 Max with 16-core CPU)

If the M4 series is like the M1 series, then as you go up the processor line, there are more RAM chips surrounding the main chip, allowing the main chip to access several RAM chips at the same time.

In general, the architecture being used in these Macs is so different from conventional computers that just comparing RAM speeds does not begin to tell the whole story.

This wikipedia article has some background about some of that, including RAM speeds:

https://en.wikipedia.org/wiki/Apple_M4

Since 2012 Retina MacBook Pro models, RAM on MacBook and MacBook Pro and MacBook Air was fixed at the time of manufacture, and not replaceable later.

Apple-silicon Macs use a very fast variant of HBR memory stacked on the system-on-a-chip, and provide RAM access so fast, that separate Display RAM is no longer required. The marketing name for that is "unified RAM".

What exactly are you trying to find out?

As that article quickly points out, the standard Apple-Silicon M-series PRO and MAX interface to memory -- where that memory is located in the same chip-carrier -- is much wider than traditional systems. That in itself doubles the effective speed.

The RAM is different in the Pro or Max models compared to the “Non-Pro” models, although it is not rated by MhZ or GhZ. See the screenshots below.

For an explanation of MhZ vs MT/s, see: https://www.corsair.com/us/en/explorer/diy-builder/memory/mts-vs-mhz-explained/

Grant Bennet-Alder wrote:

In general, the architecture being used in these Macs is so different from conventional computers that just comparing RAM speeds does not begin to tell the whole story.

The Memory Bandwidth tells a much more accurate story – but even that is not complete.

When Anandtech tested a Mac with the M1 Max chip, which has a maximum bandwidth of 400 GB per second, they could not get the CPU cores alone to use much more than half of it. Heavy GPU loads did not consume all the available bandwidth, either. Using all that bandwidth might require a workload that would stress CPU cores, GPU cores, and media engines at the same time.

AnandTech (October 25, 2021) - Apple's M1 Pro, M1 Max SoCs Investigated: New Performance and Efficiency Heights

"Location, location, location" isn't a new idea – in real estate, or in computing. In the context of computers, the inability for signals to propagate any faster than the speed of light can matter.

A modern mechanical hard drive might have a seek time of about 9 milliseconds. We now consider these drives slow compared to SSDs, which may have seek times between 0.08 and 0.16 milliseconds.

The distance from New York City (East Coast) to Cupertino (West Coast) is around 2,563 miles. Light travels at roughly 186,000 miles per second. So it is a minimum of 13.78 milliseconds each way. If you were in NYC, and queried a server in Cupertino, it would be at least 27.56 milliseconds before you got an answer.

Recent Intel Mac Pro models had electrical (loading) issues on the DIMMs, and had to resort to Registered DIMMs, where an additional output Register (and its accompanying delays) are inserted into the circuits, slowing access times.

When the circuit traces are on an ordinary printed circuit card, wider busses to RAM memory go from merely difficult to unworkable, because they become big silver highways and have issues with slightly different propagation delays between signals, and they are harder to manufacture. If the total length of the traces gets close to half a foot, additional beefier driver circuits are needed. If you need to cross a card-edge connector, or support anybody's RAM chips, those drivers need to be even bigger. Apple's designs avoid all of that. Yup, not user upgradeable.

When it comes to connecting to memory, when you know your RAM devices are very close by (on the same carrier card) you can control for electrical issues like capacitance and propagation delays. You provide drivers on the CPU/GPU chip already tuned for the RAM you know will be connected, you can save many delays and inconsistencies.

Because they were able to get the initial M1 speeds up equal to Display RAM speeds, there is no need for separate, power-hungry Display RAM and no copying back and forth, and no special access to only Display RAM. Everything can be simpler, and that means faster.

here's a summary pull quote from the summary section of that anadtech article on the M1:

On the memory side, Apple has scaled its memory subsystem to never before seen dimensions, and this allows the M1 Pro & Max to achieve performance figures that simply weren’t even considered possible in a laptop chip. The chips here aren’t only able to outclass any competitor laptop design, but also competes against the best desktop systems out there, you’d have to bring out server-class hardware to get ahead of the M1 Max – it’s just generally absurd.

The M4 chips are similar, but evolutionary faster.