Since last few days Apple iPhone 13 with a new processor A15 Bionic ChipWith 15 billion transistors, this is 27% more than the 2020 A14.
Most of these additional transistors can be used for new GPU (Graphics Processing Unit), the new AI Neural Engine and a few others.
The desktop version of the A15 is a new version with a faster clock speed Macbook Air, Macbook Pro, iMac When mac mini At the end of this year, we have double the cache (32MB). But what about it? mac pro workstation??
The more power you have, the more cash you get
Now, Apple posted an interesting job listing on September 10th. It advertised there CPU vacancies. Cache RTL Microarchitecture Based in the United States.
This is the fifth job posted by the Cupertino giant mentioning “CPU cache” and the third job mentioning multiprocessor systems. Now Apple is planning a product that uses more than one processor. Currently, the Mac Pro is the only candidate.
(Note that Apple can refer to multiple processor families in the same SoC, such as central processing units, graphics processing units, neural processing units, etc.).
The latest job description is “CPU Multi-Level Cache Subsystem Architecture and RTL Development for Multi-Processor Systems”, and how to cache (ultra-fast memory acting as the first port) for multiple cores and multiple physical processors. have to handle. Tells about. The amount of calls between the actual CPU cores and the rest of the system is very important.
one more job information (Linux embedded engineer), Apple said that a successful candidate was “one of the most visible teams verifying multidimensionally complex systems-on-chips in multiprocessor environments for future Apple products. This will be a department.” First, they’ll develop Linux The next generation Mac product environment that enables new cutting edge technologies (sic). ”
More Cache, Different Cache?
It turns out that the two high-performance cores of the Apple M1 each have 320KB of combined L1 cache and share 12MB of L2 cache. The remaining four energy-efficient cores each have 192KB of combined L2 cache, sharing 4MB of L2 cache. The M1 is basically inspired by the A14, so there is no L3 cache.
However, in order to boot the Xeon from a Mac Pro, Apple must have a different type of processor. It has a different cache architecture, has a faster clock speed, and can support much more memory than the currently available 16GB (located at: SoC with System-in-Package Configuration).
Due to the fact that it refers to multiprocessors rather than focusing on multicores, Apple believes that adding processors instead can improve performance while keeping the number of cores low.
The more cores you have, the greater the need for L3 cache. 64 cores each Threadripper Pro 3995WX For example, 64K L1, 512K L2, and 256MB L3 are shared (4MB each). By the way, this is slightly above Apple’s energy-efficient Core L1 + L2 cache quota. This may indicate that Apple is reluctant to add a more complex layer (i.e. L3 shared cache).
So you could end up with two hypothetical M2Xs that offer 64GB of RAM (32GB each) or four Mac Pros with 128GB of memory (and DDR5 in some cases). This includes three of the eight memory configurations, up to 1.5TB on current Xeon-based Mac Pros, but I still find it insufficient.
If you want to take this seriously, Apple needs to come up with an idea that will allow the M1’s successor to support more system memory.
no xserve server
But one thing that seems unlikely is that Apple will bring back the Xserve. Server A brand that provides rack servers to companies around the world. Nearly 13 years after the launch of the last Xserve, the market has changed beyond recognition. Dell, HP, Lenovo, etc. still exist, but the market dynamics have been replaced by hyperscalers such as Google, Facebook, Microsoft, Alibaba and Amazon.
These are companies that have a huge desire for computing power and aren’t afraid to set the agenda when they want it (hence why both AMD and Intel bought it). FPGA companies in recent years). I don’t want Apple to compete in that novel, low-margin environment.
However, it should come as no surprise that Apple follows all other hyperscalers and launches its own server chip dedicated to internal consumption. After all, with hundreds of millions of iCloud users and plans to become a service giant, it’s in Apple’s own interest to do what the customer does on the infrastructure side.
It reduces dependence on third parties by providing unique features to the owner and end users of the entire vertical stack that are not available anywhere else. supercharge Apple Privacy RelaySuper efficient video encoding technology, realistic vdiscussion meeting A device that works with low bandwidth. Dream of a pipe (line)? let’s watch.