Apple MacBook Pro 16-inch with M1 Max chip

[smatovic]

Anandtech's dive into M1 Pro and M1 Max:

https://www.anandtech.com/show/17019/ap ... erformance

M1 Max with 30W for CPU and 60W for GPU under load, with ~10 TFLOPs for GPU, sounds good for a mobile SoC, but that Apple has yet no ARM based chip and no own GPU solution for its Mac Pro workstation speaks for itself IMHO.

Further the OpenCL support on M1 seems crippled:

https://devtalk.blender.org/t/opencl-on ... m1/18446/8

and Apple starts supporting Blender for an own Metal GPGPU rendering backend:

https://www.blendernation.com/2021/10/1 ... -incoming/

doubt that a Lc0 Metal backend is on their wish-list.

--
Srdja

[h1a8]

ARM sucks for chess.

Buying an M1 Max for chess is quite possibly one of the most stupid decisions you can make.
Not even 5600X performance for quintuple the price....

The GPU is going to suck for Leela because not Nvidia.
CPU will suck for chess because ARM is slow at calculating bitboards and NEON is shit compared to AVX2.
And that's besides 8+2 cores for 2500+ being a horrible deal.

I can't seem to find a laptop with a such 5600x processor. Can you link me to one (or pm me)?
All I see is U, H, HS, and HX versions.

[Ras]

I like the Pulse 14 and Pulse 15

I have their Aura 15 with 4700U, 32GB RAM, 2TB SSD.

I can't seem to find a laptop with a such 5600x processor.

That's because it's a desktop CPU.

[wickedpotus]

I can't seem to find a laptop with a such 5600x processor. Can you link me to one (or pm me)?
All I see is U, H, HS, and HX versions.

My guess is x was supposed to be a generic sort of * (not a ref. to the desktop version) ... I don't think there is any HX versions of 5600 ? The 5600H is the fastest 5600 to get. But the 5800* and 5850 is not much higher priced so makes more sense to go with those wonderful 16-threads capable ones - or of course go with 5980HS (until 5980HX laptops starts popping up) which is sort of the highest end laptop CPU you can get currently.

[Milos]

I can't seem to find a laptop with a such 5600x processor. Can you link me to one (or pm me)?
All I see is U, H, HS, and HX versions.

My guess is x was supposed to be a generic sort of * (not a ref. to the desktop version) ... I don't think there is any HX versions of 5600 ? The 5600H is the fastest 5600 to get. But the 5800* and 5850 is not much higher priced so makes more sense to go with those wonderful 16-threads capable ones - or of course go with 5980HS (until 5980HX laptops starts popping up) which is sort of the highest end laptop CPU you can get currently.

Actually 11900H is marginally faster in most benchmarks (except multi-core CB R23 where is marginally slower) than 5980HS and is even better for chess.
I have it in the newest XPS15 and even though it's thermally limited due to XPS15 being an ultrabook, it kills the newest MBP in performance.
Moreover, there is also 11900HK that is particularly better in AVX512 ops (so NNUE should benefit a lot from it) and that is also better than 5980HX.
Unfortunately, there are really very few laptops with 11900HK one can buy atm.

[ChickenLogic]

I can't speak for the laptop/mobile versions of chips but in the desktop world the high end AMD chips use about 35W more than advertised ("TDP"). In my case it is not 105 Watts but around 140 Watts (5800X). While on the Intel world the top end CPUs like to overshoot by well over double of what is advertised - without any overclocking. I have friends who have reported this to me and it is discussed in forums as well. If the "mobile chips" of Intel behave this way as well I don't see them being useful for any mobile application.

[Raphexon]

I can't speak for the laptop/mobile versions of chips but in the desktop world the high end AMD chips use about 35W more than advertised ("TDP"). In my case it is not 105 Watts but around 140 Watts (5800X). While on the Intel world the top end CPUs like to overshoot by well over double of what is advertised - without any overclocking. I have friends who have reported this to me and it is discussed in forums as well. If the "mobile chips" of Intel behave this way as well I don't see them being useful for any mobile application.

They only consume more because the settings of the motherboard will let them do that.
Your BIOS should have options to configure TDP, and Intel will follow them down to a T. (PL1 value)

Laptop vendors tend to enforce TDP by default, so the TDP will be what it consumes.
Higher end laptops will either let you configure it or are unlocked to begin with. (In which case TDP no longer has meaning)

[wickedpotus]

They only consume more because the settings of the motherboard will let them do that.
Your BIOS should have options to configure TDP, and Intel will follow them down to a T. (PL1 value)

Laptop vendors tend to enforce TDP by default, so the TDP will be what it consumes.
Higher end laptops will either let you configure it or are unlocked to begin with. (In which case TDP no longer has meaning)

And of course most OS and modern laptop let's you switch to power saving modes that dial down power consumption even more if you wanna save power at the expense of maximum possible speed.

You can run a 5800 at 35% of TDP and it will still beat an M1 using its max TDP in stockfish speed..

[Ras]

I can't speak for the laptop/mobile versions of chips but in the desktop world the high end AMD chips use about 35W more than advertised ("TDP").

Actually, they don't, because AMD does advertise TDP, but does not claim that this is the physical power draw. Looking up how AMD defines TDP, it goes like this: If the CPU has temperature T1 which is high but still fine, and the case ambient has temperature T2, and the recommended thermal resistance of the cooling solution is R, then the amount of transferred thermal power is the advertised TDP (that's pretty much the definition of thermal resistance). Then they tweak these numbers so that they get a nice result like 65W, 95W, 105W, and so on. It indicates CPU classes, basically.

If your cooling solution were to fit these conditions, you would see the advertised TDP as electrical power draw because the CPU would be thermally throttling at sustained full load. Now, if you fit a better cooler, your TDP will increase because the thermal resistance goes down. At the same temperature difference, a lower thermal resistance allows more heat transfer. That continues until the chip hits its "package power limit" which in case of e.g. a 65W TDP CPU is 88W.

[ChickenLogic]

I can't speak for the laptop/mobile versions of chips but in the desktop world the high end AMD chips use about 35W more than advertised ("TDP").

Actually, they don't, because AMD does advertise TDP, but does not claim that this is the physical power draw. Looking up how AMD defines TDP, it goes like this: If the CPU has temperature T1 which is high but still fine, and the case ambient has temperature T2, and the recommended thermal resistance of the cooling solution is R, then the amount of transferred thermal power is the advertised TDP (that's pretty much the definition of thermal resistance). Then they tweak these numbers so that they get a nice result like 65W, 95W, 105W, and so on. It indicates CPU classes, basically.

If your cooling solution were to fit these conditions, you would see the advertised TDP as electrical power draw because the CPU would be thermally throttling at sustained full load. Now, if you fit a better cooler, your TDP will increase because the thermal resistance goes down. At the same temperature difference, a lower thermal resistance allows more heat transfer. That continues until the chip hits its "package power limit" which in case of e.g. a 65W TDP CPU is 88W.

That I simply didn't know. That makes sense to me then.

For Intel (I just looked it up) it says: "TDP stands for Thermal Design Power, in watts, and refers to the power consumption under the maximum theoretical load. Power consumption is less than TDP under lower loads. The TDP is the maximum power that one should be designing the system for. This ensures operation to published specs under the maximum theoretical workload."
So, maximum theoretical load should assume optimum cooling so it should never cross TDP and thus it should only ever cross it when the user tells the CPU to do so. There is no mobo out there (at least none that I know of) that overclock the CPU by default.