Unreleased M4 MacBook Pro Offered For Sale on Russian Site (9to5mac.com) 23
9to5Mac: Following apparent photos of an M4 MacBook Pro box and a subsequent unboxing video, the unreleased model has now been offered for sale on a Russian classified ads site -- at a highly inflated price, naturally. Multiple units were advertised before the listings were removed, and it does now seem increasingly likely that the leaks are real.
Apple Pro tweeted a screengrab of one of the listing, which offered what appears to be the base model 14-inch M4 MacBook Pro, with the previously reported specs of 16GB unified memory, 512GB SSD, and three Thunderbolt 4 ports. We've also seen Geekbench results for a machine identified as "Mac 16,1" with performances in line with the reported specs. Rather than a one-off leak, it has been claimed that there are some 200 units out there. The ad on Avito was asking 720,000 rubles, which is around $7,400.
Apple Pro tweeted a screengrab of one of the listing, which offered what appears to be the base model 14-inch M4 MacBook Pro, with the previously reported specs of 16GB unified memory, 512GB SSD, and three Thunderbolt 4 ports. We've also seen Geekbench results for a machine identified as "Mac 16,1" with performances in line with the reported specs. Rather than a one-off leak, it has been claimed that there are some 200 units out there. The ad on Avito was asking 720,000 rubles, which is around $7,400.
less base ram then the m3 systems? (Score:2)
less base ram then the m3 systems?
Re: (Score:2)
Pretty sure the base m3 model has 8gb, and a quick check on Apple's site verifies it.
Duplicate Story (Score:2)
$7400 (Score:5, Funny)
So cheaper than Apple will charge for it?
Unified memory (Score:2)
Did Apple just re-invent shared CPU/GPU memory?
I remember many years ago specifying in the PC BIOS the amount of system RAM that was allocated to the onboard GPU. I guess Apple just decided that it's a selling point now...
Re: (Score:2)
A bit late to the party, aren't you? They've used unified memory in their M-series chips since the M1, and in their mobile products much longer than that.
Re: (Score:2)
Yep, I certainly am haha.
Re:Unified memory (Score:4, Informative)
Shared is different than unified.
With shared, you partition off a chunk of RAM and make it local to the GPU, and non-local to the CPU.
I.e., transferring data from one partition to the other is still a relatively slow process.
With unified, all RAM is local to the CPU and the GPU- there's no transferring of the data at all.
For example, say you've got a vertex buffer stored on disk in a format that works with your graphics library (Vulkan, Metal, DX12) on a system with unified memory.
You load that into a buffer, and simply hand the pointer off to the library, and instead of copying that to RAM that is local to the GPU, it just hands that pointer to the GPU and it uses it directly. No copy.
That being said, Intel GPUs have done this for a while now.
They still have an option to report a chunk of RAM as partitioned, but that's only for older graphics libraries expect discrete RAM (OpenGL, DX11).
AMD iGPUs still use partitioned RAM.
Re: Unified memory (Score:2)
Amd called it HSA and the ps4 cpu supports heterogeneous memory management. They also added it to a few desktop APUs in 2015
Itâ(TM)s part of the HSA standard
Re: (Score:2)
Zen and above do not, and likely never will.
The chiplet latency is too high for reasonable cache coherency.
Re: (Score:2)
Then what is the UMA Frame Buffer in all current Ryzen APU's?
As far as I'm aware, current AMD integrated GPU's support the optional UMA in Direct3D 12
Re: (Score:2)
Then what is the UMA Frame Buffer in all current Ryzen APU's?
Marketing co-option of a term.
The "UMA Frame Buffer" sets the partition size for the shared memory at runtime.
There is no UMA**
As far as I'm aware, current AMD integrated GPU's support the optional UMA in Direct3D 12
Nope.
They don't, can't, and won't.
It would be very, very bad for performance if they did.
The GPU is on the I/O die, and the CPU cores are across a very high latency link.
If you were to start a job with all of your CUs blasting away as hard as they could at a chunk of memory claimed in both locations, you would overwhelm the bus between the I/O die and the CCX with coherency tr
Re: (Score:2)
So you're saying cache flushing is more expensive than memory copying, and there is no shared memory between the CPU and GPU?
And somehow AMD has got this memory copying to be better performing than other integrated GPU's, like Intel, which has UMA
Re: (Score:2)
So you're saying cache flushing is more expensive than memory copying, and there is no shared memory between the CPU and GPU?
Very much so.
With an automatically coherent cache (where you're not manually executing the memory barriers) there must be a lot of coherency traffic since accesses can be random.
On architectures where the cache coherency is built into a shared cache (TLC/SLC/L4) or a very low-latency local fabric, the opposite is the case, since the coherency is effectively free (unless you've fucked up)
And somehow AMD has got this memory copying to be better performing than other integrated GPU's, like Intel, which has UMA
Of course. They have superior CUs, and ultimately- UMA is not by any means a guaranteed performance advantage.
See discr
Re: (Score:2)
You're looking for a heap with MEMORY_PROPERTY_DEVICE_LOCAL_BIT | MEMORY_PROPERTY_HOST_VISIBLE_BIT | MEMORY_PROPERTY_HOST_COHERENT_BIT
That means allocations there are local to both the CPU, the GPU, and cache coherent.
Why isn't Apple eating Nvidia's lunch? (Score:2)
It seems like the weak link in Nvidia gpu is the time it takes to move data to and from gpu memory makes mixed mode cpu/gpu processing prohibitive. Furthermore the unified systems tend to have vastly more memory which is often the bottleneck on AI models.
So why isn't any computer supporting unified memory just destroying Nvidia??
Re: (Score:2)
It seems like the weak link in Nvidia gpu is the time it takes to move data to and from gpu memory makes mixed mode cpu/gpu processing prohibitive.
Not at all. Given that literally most processing in... the entire world... happens in this way, what's your justification for saying it?
So why isn't any computer supporting unified memory just destroying Nvidia??
Because the premise of your assertion is false.
Stolen? (Score:4)
If this is true, this is not a "leak". This is outright theft from the company making these machines. Are these still being made in China, or did Apple get its India plant up and running?
Re: (Score:1)
Not necessarily: I can see making an early run of 200 machines with very restricted distribution. I can also see 1 of them being lost or stolen.
On the other hand, you could be right, this could be a single lot of 200 that was stolen.
Goodbye, $7400 (Score:2)
Good luck getting AppleCare (Score:2)
If your unreleased hardware has a problem, you're probably SOL.
FSCKing 512GB (Score:2)
One might think 512GB is plenty. My wife has about 130GB of music and another 130GB of photos. After system files that leaves her less than 10GB working storage. Move it to the network... nope! That corrupts the database of either app. I love her and all, but she doesn't need a $2,500 laptop to check her spam and browse the internet.
Hopefully once we get an alternative to the iPod we use for travel them music can finally be migrated off... but what a pain in the ass!
Re: (Score:2)
Worse, IMO, is that a simple microsd slot would solve that deficiency completely. You don't need an ultra fast SSD to store and play music and photos. It's beyond overkill. A HDD from the 90's would keep up just as well. All devices these days should have at least two tiers of storage - one for stuff that can benefit from the speed, and one for bulk storage.
FWIW, the Macbook Pro (at least one model I found specs for) has an SDXC slot that would work fine for that purpose.