Live data from Hacker News

MacBook Pro M1X benchmarks have leaked

tomsguide.com

131–140 of 141 posts

Re: MacBook Pro M1X benchmarks have leaked

#131

Earlier quoted context omitted.

This does not address wether the GPU throttles in high CPU usage scenarios. That being said, RAM and storage is pretty negligible. LPDDR4/X uses around 0.2-0.5w, and around 0.1-0.05w for storage. There is also no reason to compare to Intel. The competition for Macs aren't older macs - newer macs are always going to be better than older macs. The competition is to what they could have been if they went for AMD, and to…

With regards to performance and power consumption, AMD is not far ahead of Intel. Yes, they have better performance when you factor in all the Intel "bugs" like branch predition, but it's only about 10%-20%, and TDP is more or less the same. Comparing it to a M1 will yield the same results. The lowest TDP i can could find was 65W for a Ryzen 3, which is still long way from M1 in both power consumption and performance…

Why are you comparing AMD desktop processors with laptop processors on power efficiency? That's incredibly contrived.

Compare the 5900HS with the M1, which is a 35W SoC, just like the M1, and much, much faster than any Rzyen 3.

Re: MacBook Pro M1X benchmarks have leaked

#132

i love my mac, can't wait for this upgrade people like to hate on the OS, but i find it super beautiful and easy to use i don't know how to describe it, but i feel good when i use my mac in contrast, everytime i have to boot windows, i feel like i'm oppressed by some corporate, i feel closed, the whole experience is boring, windows needs a serious REBOOT, not a reskin, a whole reboot, including deleting NTFS

>including deleting NTFS Imagine drinking so much kool-aid that you literally don't like a filesystem for no reason other than Apple good Microsoft bad.

NTFS is slow compared to EXT4 for my use cases, even slower than APFS

Re: MacBook Pro M1X benchmarks have leaked

#133

Doesn't say a lot. It's mostly speculation, the chip is listed here [1]. Notable info: Still at 5nm; double the GPU cores. Reports up to 3 displays, so it's safe to assume 2+1 external/internal. 8 High perf cores and 4 little perf cores. PCIE gen 4, no idea about number of lanes. [1] https://www.cpu-monkey.com/en/cpu-apple_m1x-1898

The other thing is RAM. Apple still has delusions of gimping it's buyers into 16/32GB for "Pro" level machines. I'd probably buy a 16" MBP with this chip if I could get it with socketed 64/128GB. But Apple's going to solder it to the board and charge a premium even if they did.

I believe the problem is that apple has stuck with LPDDR memory for performance (33% more memory bandwidth keeps integrated GPUs happy) and to hit power/efficiency targets (variable frequency).

Instead of a DIMM, LPDDR basically puts an entire DIMM on one chip soldered to the board (or package on the M1). LPDDR topologies tend to be limited in how many ranks of memory you can have, which limits the total capacity.

I think if Apple really wanted to mac a "HEDT SoC", they would switch to "quad channel" (256bit wide) memory, which along with the jump to LPDDR5 would let them sell 64GB/128GB machines. The only existing SoCs with wide memory controllers is the Tegra Xavier though.

Re: MacBook Pro M1X benchmarks have leaked

#134
post #17

Earlier quoted context omitted.

Yeah, I'm calling it now - no chance of 128GB in a Macbook Pro. 64GB even is doubtful.

It'd be fun if they got 32GB parts with double the memory bandwidth (which could make sense with the doubled Firestorm and GPU cores). Do we have perf counters that show how memory starved is M1 under pressure?

"Memory starving" a CPU is almost always a latency issue except with very specific workloads. A double wide memory controller would be very cool if they wanted to make a high performance integrated GPU though.

Re: MacBook Pro M1X benchmarks have leaked

#135

Earlier quoted context omitted.

>including deleting NTFS Imagine drinking so much kool-aid that you literally don't like a filesystem for no reason other than Apple good Microsoft bad.

NTFS is slow compared to EXT4 for my use cases, even slower than APFS

I haven't heard that, my PC runs NTFS on an M2 SSD and I have no complaints about the filesystem. The fastest SSDs get over 3600Mb/s using NTFS.

Is your use case something odd? I wouldn't expect to hit filesystem limits on a laptop anyway.

[1] https://ssd.userbenchmark.com/

Re: MacBook Pro M1X benchmarks have leaked

#136
post #79
post #44

Earlier quoted context omitted.

Class action lawsuits?

Because there’s a ceiling on RAM capacity? If that’s your requirement there are many people happy to sell you a device that will satisfy you. I started computing in the days when you could add your own instructions to your computer. I’m not suing anyone though that capability is no longer available

When the user guide prominently includes a wiring diagram for adding the memory necessary to run the product:

http://users.telenet.be/kim1-6502/6502/2ksa/2ksa.html

Cover:Schematics for a 5V, 3A regulated power supply and a 1K x 8 read/write memory block. The power supply and three such memory blocks can be added to the basic KIM-l microcomputer to provide the 4K RAM required by this assembler. Parts are available from Jameco Electronics.

Re: MacBook Pro M1X benchmarks have leaked

#137

Earlier quoted context omitted.

NTFS is slow compared to EXT4 for my use cases, even slower than APFS

I haven't heard that, my PC runs NTFS on an M2 SSD and I have no complaints about the filesystem. The fastest SSDs get over 3600Mb/s using NTFS. Is your use case something odd? I wouldn't expect to hit filesystem limits on a laptop anyway. [1] https://ssd.userbenchmark.com/

non realistic benchmark is not my usecase

compiling projects is, and it is slow with NTFS + slow windows process creation

Re: MacBook Pro M1X benchmarks have leaked

#138
post #80
post #29

Earlier quoted context omitted.

Also Apple didn't have to do ~80% of what Intel has to do - you know, the actual building of chip fabs as well as chip design. "merely" LOL.

Intel doesn’t have to build its own fabs. AMD sold its off and is doing fine. Fabless + foundry is an excellent model that lets each side concentrate on its strengths, that works for most companies. Intel started in an era when you did have to build your own fab. More importantly, even as others were going fabless, Intel continued to use its superiority in manufacturing as an additional advantage over their competiti…

Apple doesn't have to build it's own CPU, but often vertical integration gives one a significant advantage.

TSMC is slave to Dutch ASML, whos is the only one on the planet who can make the EUV lithography technology TSMC uses to make the latest chips.

Re: MacBook Pro M1X benchmarks have leaked

#139

Earlier quoted context omitted.

Rules change. I've got one of the 16GB RAM Macbook Pros. They've got it acting like it's L2 cache instead of RAM as we know it, so the place where you need more than 128G becomes, well, the SSD, your mass storage. These things already don't act as if it's RAM as we're used to considering it. They could probably have shipped the first ones with 4G RAM and had 'em act relatively normal… with heavy stuff like 4K video e…

Do you have any evidence for this claim? The actual concrete performance numbers for the M1 and Zen3 don't agree with you at all: M1: https://images.anandtech.com/doci/16252/latency-m1_575px.png Zen3: https://images.anandtech.com/doci/16214/lat-5950.png If anything the M1's ram has _higher_ latency than the Zen3. (Can't reply to the sibling comment but Intel's numbers are even better: https://images.anandtech.com/doc…

Well exactly how are you defining latency?

Is it random accesses you want to test? Or TLB thrashing?

The R per RV prange benchmark accesses memory randomly, but in a TLB friendly pattern (TLB thrashing real world workloads aren't particularly common). The M1 gets 30ns, the 5960X is 55ns and still climbing at the max array size.

The 5950X does better on "full" random, although that's unclear if that's a faster TLB, or the testing with a different page size. With Linux for TLB thrashing workloads you can always use HUGE pages, not sure if the M1 hardware supports that though.

Re: MacBook Pro M1X benchmarks have leaked

#140

Earlier quoted context omitted.

You're bringing up a limitation of the current tech, that it tethers you to a wall, to speak against this new tech. Does a student who has a gap between classes always want to find an outlet somewhere and deal with their computer sounding like it's about to liftoff? Does the video editor offloading some of their workload to GPU always want to find an outlet? Apple is selling super quiet hypermobile performance here,…

A lot of those students are using Chromebooks, which will get nothing but not cheaper. Video editors almost always want large monitors and top performance. Show me someone who works as a professional video editor who does it on a laptop screen. The best argument you could make for the MacBook in this regard is for developers who might be working as nomads, and musicians who often lug their machines to their gigs. But…

>Show me someone who works as a professional video editor who does it on a laptop screen.

I know a videographer who worked for news wire services who did just this and did his job on a MacBook sometimes on site. Helped him deliver news content faster. I don't think video editing on a MBP directly is as rare as you might assume.

No doubt that Mac's haven't turnee into ideal gaming machines, but I've gamed on ultraportables with a lot less GPU than this, I just never played a game like Tomb Raider.

Post reply on HN