Power users are pretty good at finding uses for better hardware. For example, if you are into local LLMs, you could find immediate use for a 1000x faster GPU.
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.
M4+ Chip Generation - Speculation Megathread [MERGED]
- Thread starter GoetzPhil
- Start date
- Sort by reaction score
M3 ultra is way more than 35% faster in gpu for many loads. Just look at blender scores. 4000 -> 7000 or so. Btw, m5 shows up there now. 70% faster than m4. Wow. https://opendata.blender.org/benchm...PI&group_by=device_name&blender_version=4.5.0
Yep, any AI is going to benefit from whatever GPU and memory that can be thrown at it.
AI's pushing things in a somewhat different direction with its needs compared to others, but power users are always going to demand more performance no matter what, AI or not. Applications and needs evolve. Back in the day, it was "I need to edit HD video in real time." Then UHD, 4k, 8k. At 240 fps. Plus VR. And whatever else gets tacked on that's still to come. I don't think there's any danger of running out of needs for more powerful computers anytime soon. Then there are folks doing stuff like this: https://forums.macrumors.com/threads/new-m1-studio-crashes-with-heavy-load.2427579/
The need for more isn't going to be sated by M5 or M6 or M20.
That's quite a feat! Solid improvement. Can't wait to see what the bigger SoCs can offer.
Given the performance and feature set of my M4 Mac mini, I don't think I'll be interested in an upgrade until the base Mx chip gets Thunderbolt 5 / DisplayPort 2.1, and I don't think that will happen until M7 or so.
I wouldn’t hold my breath on TBv5 .
Thunderbolt qualification is primarily an Intel thing. Intel is not the 800 lbs gorilla in the market anymore. The market push to make TBv5 wildly pervasive is likely not going to be on Intel’s top 20 list for the immediate future.
Upcoming Panther Lake chip from intel have one Platform controller chiplets. This just has TBv4
Inside ‘Panther Lake’: What to Know About Intel’s Crucial First 18A Processors
All eyes are on the chip maker as it prepares a full-court press into a new process technology, spearheaded by chips for laptops, as well as for the data center. Here's the download on what's new in the latest 'Lake.'
The overall platform has TBv5, but they appear to be leaning on discrete controllers for that provisioning .
If Intel isn’t pushing TBv5 as entry level compute , how likely is it going to get traction in overall market? Not likely.
It is more likely going to be a ‘feature’ to segment upper half products from the affordable mainstream. I/O features are not shrinking as well. There are other base function units in the plain Mncompeting for limited die space area .
The underlying core of USB is likely just about as comatose . Ther gap between USB 1 and 2 is ~4 yrs , 2 -> 3. ~8 yrs , 3 -> 4 ~11 yrs . USB 5 likely isn’t coming any time soon.
Getting entry level just to top end DPv2.1 (was typo 1.2) is going to take more time. [ so that might work later. ]
In general, PC market Qualcomm has no Thunderbolt coverage at all . AMD is very slightly better than nothing. As Intel looses substantial share to them it will be very hard to claw that back. Those two will quite happy to follow the USB committee lull of USB4.
Last edited:
Thunderbolt 5 is USB 4. No need for a new USB 5. Those Thunderbolt 5 speeds are already part of the USB 4 v2 spec, which actually came out 3 years ago back in 2022.
Also, Panther Lake is essentially a soon-to-be-current 2025 chip, not a far-into-the-future chip. Furthermore, cheap third party Chinese discrete chipsets for USB 4 v2 should be out by 2027. For Macs, I was talking about Apple integrating Thunderbolt 5 in say M7 Macs or later, several years from now. Regardless of what Intel does, Apple already incorporates TB 5 into its mid-range and upper chips now, so it's a not a big stretch to think they'll also incorporate TB 5 into its base chips several years from now.
As for going beyond DisplayPort 1.2, it seems DP 1.4 is already very common with the vast majority of mid and upper range monitors supporting DP 1.4 (and some supporting DP 2.1), and Apple's base Macs already all start with at least DP 1.4. Basically, all I want... several years from now... is all base Macs to have DP 2.1. (I'm not counting the rumoured future Ax series MacBooks.)
In case you're wondering why it matters to me... It's for ultra high resolution monitor compatibility. On paper it really shouldn't matter, since it's very unlikely I will be moving beyond 6K monitors anytime soon, and 6K 10-bit 4:4:4 60 Hz is already supported on DP 1.4 (but not DP 1.2). However, to implement 6K 10-bit 4:4:4 60 Hz on DP 1.4 requires DSC. That's fine, but there are a few buggy implementations out there that screw up DSC compatibility, meaning that some of these new non-Apple 6K monitors drop down to 4:2:2 chroma at 6K 60 Hz on Macs. Having DP 2.1 removes the requirement for DSC, and removes that source of compatibility issues.
That said, if say by the time I really want to upgrade several years down the line, I'll still have the option of getting an Mx Pro, which already incorporates TB 5 / DP 2.1. Also, HDMI 2.1 is already on current Mac minis, and that also supports 6K 10-bit 4:4:4 60 Hz without DSC, albeit without the bonus of Thunderbolt 4 and USB 4 peripheral support.
Last edited:
I just want to point out that you can connect whatever you like to those 12 PCIe lanes.I wouldn’t hold my breath on TBv5 .
Thunderbolt qualification is primarily an Intel thing. Intel is not the 800 lbs gorilla in the market anymore. The market push to make TBv5 wildly pervasive is likely not going to be on Intel’s top 20 list for the immediate future.
Upcoming Panther Lake chip from intel have one Platform controller chiplets. This just has TBv4
Inside ‘Panther Lake’: What to Know About Intel’s Crucial First 18A Processors
All eyes are on the chip maker as it prepares a full-court press into a new process technology, spearheaded by chips for laptops, as well as for the data center. Here's the download on what's new in the latest 'Lake.'www.pcmag.com
The overall platform has TBv5, but they appear to be leaning on discrete controllers for that provisioning .
If Intel isn’t pushing TBv5 as entry level compute , how likely is it going to get traction in overall market? Not likely.
It is more likely going to be a ‘feature’ to segment upper half products from the affordable mainstream. I/O features are not shrinking as well. There are other base function units in the plain Mncompeting for limited die space area .
The underlying core of USB is likely just about as comatose . Ther gap between USB 1 and 2 is ~4 yrs , 2 -> 3. ~8 yrs , 3 -> 4 ~11 yrs . USB 5 likely isn’t coming any time soon.
Getting entry level just to top end DPv1.2 is going to take more time.
In general, PC market Qualcomm has no Thunderbolt coverage at all . AMD is very slightly better than nothing. As Intel looses substantial share to them it will be very hard to claw that back. Those two will quite happy to follow the USB committee lull of USB4.
Adding TB5 or USB4 v2 will require external chips but can be done on Panther Lake. We are not going to see laptops or desktop computers with 8 x USB 2.0 and 2 x USB 3.2 and 4 x Thunderbolt 4 ports 😂
Sometimes I get the impression that the Mx performs like the Mx-1 Pro, depending on the specific test. The same goes for the Mx-2 Max. What do you think?
Depends what you want to do with it. For example, Mx Max is way faster than Mx Pro for video exports.
What is left, and which is right? I assume M5 is right, so is the left M1,2,3 or 4?
If it is not on the base platform controller tile/chiplet is that going to show up on the lowest priced i3 type systems? No. As long as TBv5 is mainly a discrete controller option it is more likely going to just be kept on the 'upper half" systems as a product differentiation tool.
Very similar thing happened with USB 2 and USB 3. USB 3 was either absent or susbtantively limited after it arrived as a 'add on' solution. Southbridge/PCH chips didn't go 100% USB 3 provisioning until generations later. Affordably priced PC systems had USB 2.0 sockets for many YEARS after USB 3.0 came out.
Similar issue with Ethernet 1GbE that stayed on PCs for many years after 10GbE . Many network gear folks didn't want 10GbE to get cheaper.
Pragmatically in the plain Mn systems. that Apple makes there is no room for discrete Thunderbolt solutions. The logic board design presumes the TB is integrated.
You are missing the point that something 'next gen' is a major contributor to the current gen being pushed into the default platform at the lower levels. The issue here is why does the iPad Air need super duper I/O. It doesn't .
The more mature tech is always going to be cheaper to implement and when get to the lower end of the line up where costs matter more.
If there is no USB 4 v3 , v4 etc. spec there is little chance for there to b e a Thunderbolt follow on. Thunderbolt is a "less optional stuff than USB" type of spec at this point. Intel was a major ( and probably critical) factor and pushing USB 4 v2 through. If Intel isn't pushing new stuff, then who will? Apple trying to do it all by themselves. Apple doesn't have the leverage with the committee. Intel being a dominate almost monopoly player is important piece of the puzzle of progress.
The USB is spec is primarily built so to make it easier to stick to the minimal cost option as your standard for base offerings. ( e.g., minimal required speed for Type-C connectors is USB 2.0 . that allows implementors to spend about as little money as they want. ).
AMD and Qualcomm are far more likely to fall in behind pushing more Oculink provisioning than Thunderbolt-like alternative. It is cheaper.
I suspect Intel will survive as a significant play in 3-4 years, but until then they have got lots of internal problems that leaves little time, money , and resources to be fighting USB committee on progress they generally don't want to make. Apple has distractions in Vision/Glasses custom chips for those. Making their modem line work long term, and even getting Ultra chips on a regular basis that doesn't loose money.
cheaper discrete chipset is all the more likely that this won't make it to the base Platform chiplet. If Intel was the overwhelming dominate player perhaps that would trigger a platform tile inclusion, but they aren't any more. Intel has been in trouble for years. This three way split of the basic CPU package is more expensive than what AMD, Qualcomm, MediaTek/Nvidia are doing. There is still going to be cost pressures to keep the Platform chiplet cheaper in Nova Lake also. ( Intel killing off rentable units and having to throw more money at the CPU chiplet redesign only makes cost control worse. )
Apple kept FW800 segregated on the 'upper half' units right up until the end. The increasing horsepower that Apple is putting into the base level units has more folks moving down in the product line up ( because it is fast enough). Segregating on I/O will partially offset that. Apple isn't going to be in trouble if folks spend $400 more to get the higher end I/O.
Sorry that was a digit flip typo. It should have been 2.1. Apple doing a 2.1 pass through doesn't require much in terms of additional switch complexity work. It is also less asymmetric bandwidth pressure on the internal SoC data bus from Display Processors to the TB controllers.
The base ones getting DP 2.1 passthrough wouldn't necessarily mean they were getting TBv5.
so yeah DP v2.1 pass through would solve the problem. Non buggy DSC from Apple would help even more.
I think the 'tag wags the dog' pops up where next gen Pro XDR has to have TBv5 so therefore Apple has to rollout out TBv5 to iPads so that they can run $5,000 monitors at max refresh rates. ( versus folks who buy $5000 monitors just buy a $500 more expensive Mac. )
I would expect provisioning backhault to HDMI 2.2 out of Apple on the base level chip support before TBv5. By M7 era that would be more common on TV's of that era than trying to do upmarket >= 6K stuff.
? I didn't say anything about the iPad Air. I don't expect the iPad Air to get even Thunderbolt 3 anytime soon. Incorporating Thunderbolt x into the base SoC doesn't mean all devices like iDevices will that SoC get Thunderbolt x. The Macs like the Mac mini would get that Thunderbolt x upgrade though.
That's true, but I'd be surprised if Apple decoupled DP 2.1 support from Thunderbolt 5.
I expect all future Thunderbolt 4 Macs to support up to DisplayPort 1.4 only.
Last edited:
Not sure if this is interesting to anyone
Got my iPad Pro M5 (4P+6e)
Unlike the M4 model, the Geekbench 6 single thread score doesn't increase much with additional cooling. Too early to make any judgements, but it seems like M5 can hit near peak single-thread performance at normal temperatures unlike M4. Curious how it will hold up when it heats up.
M5 (4P+6e):
Idle/room temperature 4117 / 15845 https://browser.geekbench.com/v6/cpu/14596883
On ice 🥶 4179 (101.5%) / 16876 (106.5%) https://browser.geekbench.com/v6/cpu/14600126
M4 (4P+6e):
Idle/room temperature 3698 / 14739 https://browser.geekbench.com/v6/cpu/14600365
On ice 🥶 3975 (107.5%) / 15039 (102.0%) https://browser.geekbench.com/v6/cpu/14600606
Got my iPad Pro M5 (4P+6e)
Unlike the M4 model, the Geekbench 6 single thread score doesn't increase much with additional cooling. Too early to make any judgements, but it seems like M5 can hit near peak single-thread performance at normal temperatures unlike M4. Curious how it will hold up when it heats up.
M5 (4P+6e):
Idle/room temperature 4117 / 15845 https://browser.geekbench.com/v6/cpu/14596883
On ice 🥶 4179 (101.5%) / 16876 (106.5%) https://browser.geekbench.com/v6/cpu/14600126
M4 (4P+6e):
Idle/room temperature 3698 / 14739 https://browser.geekbench.com/v6/cpu/14600365
On ice 🥶 3975 (107.5%) / 15039 (102.0%) https://browser.geekbench.com/v6/cpu/14600606
when will we get increased core counts for mx pro/max
like maybe 20 cores for the max and 16 for the pro
like maybe 20 cores for the max and 16 for the pro
I doubt we’ll get much increase. I actually could see them pull an M3 Pro and change the M5 Pro to 8P+6E to keep it at 14 cores, with the Max going to 12P+6E for 18 cores. I just don’t see more P cores, but I could be wrong.
I mean M1 to M2 saw a 10 to 12 core increase and m3 got to 16 cores
m4 stayed the same but buffed the pro versions
i could see a higher core count for the max this time with the extra time and the chiplet thingy that tsmc has
also x2 elite has 18 cores so probably to compete with them
(even though the x2 elite is still cooked)