'M2 Max' Geekbench Scores Leak Online, Revealing Rumored Specs and Performance

[sw1tcher]

The Geekbench results, first spotted on Twitter, are for a Mac configuration of with the M2 Max chip, a 12-core CPU, and 96GB of memory. The Mac listed has an identifier "Mac14,6," which could be upcoming MacBook Pros or the next-generation Mac Studio. Apple offers a maximum of 64GB of memory on the current 14-inch and 16-inch MacBook Pros, while the Mac Studio can be configured with up to 128GB of memory with the M1 Ultra.

According to the test, the M2 Max chip scored 1,853 in single-core and 13,855 in multi-core. For comparison, the M1 Max chip in the Mac Studio scored 1,755 in single-core and 12,333 in multi-core. If the M2 Max chip results are accurate, the performance increase will be relatively minor for the upcoming chip.

Hmm.... 5.6% improvement in single-core score and 12.3% in multi-core score. No sense in waiting for this if you want/need a new MBP now. Take advantage of the great deals going on right now.

The new max of 96GB of memory on a MBP sounds great, but that is going to be one darn expensive upgrade. It's already $400 to go from 32GB to 64GB on the MBP. That means it'll be another $400 to go from 64GB to 96GB :crazy:

 

[VaruLV]

I dont care that much for the lack of phenomenal increase in performance for M2 based Macbook Pro models, I'd rather see latest tech in SDXC card reader and HDMI port as well and FaceID as well.

 

[Wizec]

So the M2 Max is slower than M2?

M2 gets 1889 in single core, this article says the Max only achieves 1853?

 

[GMShadow]

Happily looking to trade in my MBP M1 Max for a newer version with 96GB of RAM, rather than this 64GB.

Here's hoping the base RAM increases, or the upgrade pricing returns to 'reality'.

96 GB of RAM on the M2 Max looks excellent... but that score is extremely underwhelming given that Raptor Lake is available with double the multicore score. Perhaps this score is not final and was run when the machine was under load. Yes RPL consumes a ton of power under heavy all-core loads, but under light load, it's relatively reasonable at anywhere between 15-30 Watts, and no more than 80W during things such as Windows update etc.

Is Raptor Lake doing that in a laptop?

 

[bradman83]

The performance improvements don't seem substantial, but my guess is that they focused on improving the intwerworking on the Ultra chips and allow it to scale better and utilise all the cores available. This was one of the drawback's of of M1, but should be their focus forward.

From a hardware perspective there isn't much preventing the M1 Ultra from utilizing all of its cores, hence why the CPU multicore benchmarks scale accordingly to the number of cores. The issue is that not all software workflows can take advantage of all of those cores simultaneously. To be clear there is very much optimization that can be done by developers to optimize their workflows across the board. macOS has had excellent parallelization APIs (Grand Central Dispatch) since the Snow Leopard era I believe, and Microsoft introduced updated APIs in Windows 11. Since multicore is the future hopefully that means more developers will be optimizing their software to scale accordingly.

 

[Zaydax333]

Looks like this is one of the lower end benchmarks for single core perf. The average for the single core on the M2 13 inch on geekbench seems to be 1899.

 

[james2538]

The 96 GB RAM is the most exciting part of the leak for me. I wonder if that means we'll now be starting off with 24 GB in the base models (96GB/2= 48GB/2 = 24GB). That would be a huge jump!

It's not surprising CPU performance isn't a dramatic increase though since it was rumored the additional cores would be efficiency ones (so 8 Performance + 4 Efficiency). I'm excited to see if this means we've improved yields enough that the 14" base model no longer gets binned chips. Perhaps will get some additional hours of battery life as well.

Can't forget we'll definitely be getting WiFi 6E and Bluetooth 5.3 with this too. Maybe they'll also throw in a faster SD Card Reader. All in all should be a modest upgrade.

If they really aren't 3nm chips then I'd expect we'd see them in January instead of March (which is traditionally around where Mac OS vX.2 versions are released)

 

[tipoo]

This was always rumored to be a GPU focused upgrade, so I'll hold judgement, CPU not moving much isn't too unexpected. I hope it adds hardware RT support, look at Optix results on Blender, using the RTX hardware, it extends the lead Nvidia already had with CUDA (and these results are only 2000 series, 4000/3000 are even faster on RT)

 

[AgeOfSpiracles]

This validates my decision to just pull the trigger on a 14" MBP M1 Pro. I had put it off for so long expecting for the M2 version to be announced, and when it became obvious that wasn't happening this year I just bit the bullet. The fact that it is such an incremental upgrade gives me peace.

 

[leman]

If they really aren't 3nm chips then I'd expect we'd see them in January instead of March (which is traditionally around where Mac OS vX.2 versions are released)

Frankly if the scores are already leaking online a release is likely to be imminent.

 

[leman]

This was always rumored to be a GPU focused upgrade, so I'll hold judgement, CPU not moving much isn't too unexpected. I hope it adds hardware RT support, look at Optix results on Blender, using the RTX hardware, it extends the lead Nvidia already had with CUDA (and these results are only 2000 series, 4000/3000 are even faster on RT)

I doubt we will see any new GPU capabilities, only performance improvements. RT and other stuff will likely arrive with 3nm next year.

 

[deconstruct60]

If this is true, it's extremely disappointing. Means no 3nm chips or new generation cores, same IPC as the M1.

Could be N3 , but in such a way that that dies are cheaper (smaller which gets higher yields and hopefully offsets the more expensive fab process ) and lower power consuming ( not bigger and faster). Winning the maximum prize on the Geekbench leader board doesn't have to be the primary objective. These scores are not anywhere close to being 'poor' / 'bad'. They are just not 'maximum bragging rights winning' scores. [ i.e., don't wring every lost drop of performance out of N3 this iteration ... just made it work and move in tick/tock fashion. A new MBP 14"/16" that is faster and gets even longer battery life ... how could that possibly sell well? *cough* Probably could. Even if AMD/Intel wrangle some pref/watt improvements in their 2023 laptop offerings; apple would just keep about the same gap. Windows on Arm with Qualcomm/Nuvia... not a problem. ]

Only disappointing because the hype level has been set so high. 12 cores was often spun into two more P cores. Doesn't look that way. Looks much like got faster 8P cores and now 4E cores to bump up multiple core score.


This could be N4 (or older) also and the battery life isn't going to move much.

Most likely no GPU raytracing either...

Geekbench doesn't measure that at all. ( the raytracing component to the score is a CPU specific raytrace. And more so , probably does have precision encoded raytrace to apple only library to snag any redirect to GPU specific hardware. ). Wasn't anything revolutionary likely coming there this iteration either outside the hype train. Decent chance raytraying shows up on the AR/VR headset first.

 

[akis-k]

I expected a lot more to be honest... I'm using 3d programs (c4d) for architectural visualization, and i hoped for a further improvement in speed. But it seems that Intel and Amd are taking the lead during this time of period...

 

[woolypants]

…and never will be?

Do you mean that x86_64 might catch up with Apple on performance per watt?

That's considered by experts to be impossible. Intel have tried several times but failed (e.g. Atom). Even the new efficiency cores aren't really very efficient, especially compared to Apple Silicon.

This decade belongs to ARM. x86 belongs in the past. It's a dead man walking.

 

[leman]

Could be N3 , but in such a way that that dies are cheaper (smaller which gets higher yields and hopefully offsets the more expensive fab process ) and lower power consuming ( not bigger and faster).

Do you believe that? What's more likely, that Apple uses N3 to make a slower CPU or that they use a horizontally scaler A16/M2 on the already relatively mature N5P? Not to mention that these scores are 100% in line with M2 performance (just add a P-cluster).

Only disappointing because the hype level has been set so high.

To me it's disappointing because I want three things. First, a split between consumer and prosumer desktop hardware. Second, new generation of P-cores (so far we had three generations of Apple Silicon on what is essentially the same P-core frontend and backend, with just few incremental tweaks and new instructions). Third, competition with x86 in the desktop segment. These M2 Max scores are enough to secure a lead among laptops in first half of 2023, but Apple is getting outgunned in the desktop space.

Geekbench doesn't measure that at all. ( the raytracing component to the score is a CPU specific raytrace. And more so , probably does have precision encoded raytrace to apple only library to snag any redirect to GPU specific hardware. ). Wasn't anything revolutionary likely coming there this iteration either outside the hype train. Decent chance raytraying shows up on the AR/VR headset first.

What I mean is that a M2 Max that uses A15/M2 CPU clusters is unlikely to feature any new GPU tech.

 

[Adreitz]

Some "incredibly sad guys scouting" somewhere are likely the source of the bulk of rumors we feast on each week. Have those "pitiful" guys give up sad pursuits and things would likely get remarkably boring around here real fast. If you enjoy rumors, embrace the sadness.

What's sad is how terrible the Geekbench results browser is. The fact that you can't sort search results by date or score or easily see stats such as average, median, 90th percentile, etc. for a given chip or system is mind-boggling considering how Geekbench has become a standard cross-platform benchmark. You can't even search for data in specific fields, such as all results with a certain score and Apple as the manufacturer. I've even seen cases in the past where a leaked score was posted on MR but I could not find it by searching the database.

 

[tipoo]

Do you mean that x86_64 might catch up with Apple on performance per watt?

That's considered by experts to be impossible. Intel have tried several times but failed (e.g. Atom). Even the new efficiency cores aren't really very efficient, especially compared to Apple Silicon.

This decade belongs to ARM. x86 belongs in the past. It's a dead man walking.

Name the experts considering this impossible please. x86 does have a penalty on using power on decode, but ISA is probably <3% of a CPU die these days, with a process node advantage I'm sure AMD or Intel could for periods peek out above AS. Just since the M1 launch, it went from completely dominating, to AMD matching their IGP and multicore, and Intel matching their single core and multi, albeit Intel burning by far the most power because they're on a node disadvantage.

This is also with Apple still commanding the lions share of the 5nm run.

There's a lot that could happen with nodes and different architecture choices that may see several leapfrogs and changes in leader over the next decade. Let's see what these 2025 Intel nodes are about too, they've been messing up on nodes for many years now but they're the big recipient of CHIPS act funding, and those new nodes might be good enough to even get some former competitors using them.

ARM has some advantages, but Apple's wins have been on the custom architecture and paying out for the best process nodes. RISC vs CISC is such an old outdated debate, thousands of other architecture choices and process node advantages outweigh it greatly.

 

[deconstruct60]

They decided not to release a Mac Pro and wait for M2 is an indication of that they didn't like the results of the Ultra.

The M1 Ultra has eight x1 PCI-e v4 lanes. How do you construct a reasonable Mac Pro with substantive I/O out of that? You can't without substantive hackery (e.g., taking the 'extra' TB controller and provisioning a x4 PCI-e v3 slot backhaul out of that with an internal discrete TB controllers.). There was not enough I/O in the basline M1 Max die to construct a reasonable Mac Pro. [ There was a rumor of a one-slot-wonder Mac Pro. That probably wouldn't have worked as it would be too close to a Studio in I/O limitations. ]

Apple needs something incrementally factored a different way to have something creditable for a Mac Pro. Something like two x16 PCI-e v4 complexes to provision slots out of (or better).

Solely just count CPU or GPU cores as criteria for a Mac Pro is a fail. ( Apple has a Mac Pro 2013 'successor' in the Mac Studio. Apple has about zero need for another studio in the line up. )

 

[_Spinn_]

Finally more information leaks! Excited to see the new devices.

 

[Adreitz]

So the M2 Max is slower than M2?

M2 gets 1889 in single core, this article says the Max only achieves 1853?

Another, faster benchmark result has already been uploaded. https://browser.geekbench.com/v5/cpu/18973011 1889 SC and 14586 MC.

 

[Adreitz]

Do you mean that x86_64 might catch up with Apple on performance per watt?

That's considered by experts to be impossible. Intel have tried several times but failed (e.g. Atom). Even the new efficiency cores aren't really very efficient, especially compared to Apple Silicon.

This decade belongs to ARM. x86 belongs in the past. It's a dead man walking.

You said "never". Never is a very long time.

 

[D-a-a-n]

Good I can now stick with my M1 Max MacBook Pro, Apple updates have been very poor lately

Omg, you have to stick with your 1 year old laptop now??? Oh the horror! You must be going through hell right now.

 

[k27]

There are some wealthy Apple fans on here who just want the best of the best, at all times. It's very comparable to car enthusiasts.

My two cents responding to all comments here so far:

1. Performance: The key thing people like MKBHD said about the M1 Max MBP was that it was a quantum leap in useful performance like video encode performance. For Marcus, it meant not having to take a Mac Pro on trips in a flight case. Now he could take the M1 Max MBP. And in-between times, his needs and indeed the needs of all of us haven't changed. It's not like YouTube now demands 8K videos, for example. 1080p is still the norm, and 4K for enthusiasts. There no need for another quantum leap in Apple Silicon performance. It's got space to gently evolve—hence 10% increases.

Other manufacturers also have hardware encoders.

The hardware encoder in the M1 is very fast, but the quality is not that great. If you want great quality with small file sizes, use software encoding.

The very good open source software encoders (x264, x265) seem to run slower on AppleSilicon than on x86_64:

https://forums.macrumors.com/threads/m1-handbrake-performance-very-disappointing.2345720/?post=31122609#post-31122609

1. Ray tracing GPU: Not a priority. Again, it's only really for enthusiast gamers (and then only really a subset of those!) , and Apple's shown no desire to appeal to that market.

I don't know if that's universally true.

Ray-Traced Shadows — Blender Manual

docs.blender.org

1. AV1 hardware code/decode: Very young technology and I don't believe Apple is a member of the Alliance for Open Media, which has created it. This is probably because it's not in Apple's particular interest to support it (and Apple's passion for open source is now dead anyway). I think we would expect to see it appear in Video Toolkit first, probably next year, as software support in the next macOS. And then in the silicon of the M3 in the Air/Pro. Then it might migrate to the M3 Pro/Max hardware video engines. With hardware MP4/HEVC and ProRes support, Apple's already touching the bases it wants to in terms of hardware encode/decode.

"AOMedia Video 1 (AV1) is an open, royalty-free video coding format initially designed for video transmissions over the Internet. It was developed as a successor to VP9 by the Alliance for Open Media (AOMedia),[2] a consortium founded in 2015 that includes semiconductor firms, video on demand providers, video content producers, software development companies and web browser vendors."

AV1 - Wikipedia

en.wikipedia.org

"The governing members of the Alliance for Open Media are Amazon, Apple, ARM, Cisco, Facebook, Google, Huawei, Intel, Microsoft, Mozilla, Netflix, Nvidia, Samsung Electronics and Tencent."

Alliance for Open Media - Wikipedia

en.wikipedia.org

Hardware decoding of AV1 is present in many chips (encoding in some). Unfortunately not in Apple (or it is not yet enabled). See e.g. the table in the Wikipedia article.

1. But AMD/Intel is better! Yes, they caught up—but you'll only enjoy that better performance for literally an hour on a laptop. Where Apple wins is performance per watt, and AMD/Intel are nowhere near close to that, and never will be.

That's probably an exaggeration. But I am not up to date. Very good laptops with AMD should also be quite good. However, they probably cannot keep up with Apple Silicon in this respect.

 

[torontotim]

If this is true, it's extremely disappointing. Means no 3nm chips or new generation cores, same IPC as the M1. Most likely no GPU raytracing either...

Do you really expect massive leaps in architecture and performance every 12-18 months? On what planet does that happen?

 

[leman]

Do you really expect massive leaps in architecture and performance every 12-18 months? On what planet does that happen?

A14 was released in September 2020, roughly 25 months ago. I was hoping that Apple would have the next generation of P-cores ready to ship by now. But both A15 and A16 only offer small evolutionary updates without any IPC improvements, which is the first time in Apple's history. Maybe they are moving to a 36-48 month schedule (understandable given the scale of the effort), which would be fine, except it makes them much more vulnerable to the competition who is advancing quickly.