What advantage will a 1.4nm chip have over a 4nm one? What new capabilities will this tech unlock on an edge device like my iPhone ?
Please don't mention lower power consumption.
Silicon is way outside my wheelhouse, so genuine question: why not mention power consumption? In the data center, is this not one of the most important metrics?
For instance, GK104 on 28nm was 3.5 billion transistors. AD104 today is 35 billion. Is Nvidia really paying 10x as much for an AD104 die as a GK104 die?
If your "cost per transistor" calculation includes amortization of the fixed costs of taping out a chip, over the expected production volume, then you can sometimes genuinely end up with newer process nodes being more expensive. Design for more advanced nodes keeps getting more expensive, and mask sets keep getting more expensive. Even more so if you're pricing out a mature process node compared to early in the production ramp up of a leading edge node.
There's significant demand for older process nodes and we constantly see new chips designed for older nodes, and those companies are usually saving money by doing so (it's rare for a new chip to require such high production volumes that it couldn't be made with the production capacity of leading-edge fabs).
Intel and AMD have both been selling for years chiplet-based processors that mix old and newer fab processes, using older cheaper nodes to make the parts of the processor that see little benefit from the latest and greatest nodes (eg. IO controllers) while using the newer nodes only for the performance-critical CPU cores. (Numerous small chiplets vs one large chip also helps with yields, but we don't see as many designs doing lots of chiplets on the same node.)
What google turns up when I google this is this statement by google [1], which attributes the low point to 28nm (as of 2023)... and I tend to agree with the person you are responding to that that doesn't pass the sniff test...
My laptop definitely dies significantly faster when I'm making it work instead of just mindlessly scrolling on it... since the display is on in both cases I don't see what that could be but chip powre consumption making a singificant difference.
My phone dies much faster when I am using it, but admittedly screen usage means I can't prove that's chip power consumption.
VR headsets get noticeably hot in use, and I'm all but certain that that is largely chip power usage.
Macbook airs are the same speed as macbook pros until they thermally throttle, because the chips use too much power.
I've not checked it, but AFAIK power consumption isn't really improved much if at all with dye shrinks. The main benefits are entirely around transistor density increases which allows for things like bigger caches.
It'll be beneficial to DRAM chips, allowing for higher density memory. And it'll be beneficial to GPGPUs, allowing for more GPU processors in a package.
> The main benefits are entirely around transistor density increases which allows for things like bigger caches
SRAM is probably the the worst example as it scales poorly with process shrinks. There are tricks still left in the bag to deal with this, like GAA, but caches and SRAM cells are not the headline here. It's power and general transistor density.
If the marketing naming is to be believed, in 1.4nm vs 4nm you'd be able to fit ~twice the transistors in your chip. That's twice the cores, twice the cache... That usually makes it faster.
A 1.4nm chip offers significant performance and capability improvements over a 4nm chip, primarily due to increased transistor density. This allows for more powerful and efficient on-device AI processing, enabling new features and capabilities on devices like an iPhone without relying on cloud-based services
I find it amusing how we’ve come from treating AI as a novelty to developing a sense of how it writes in the space of a few months. That parent comment doesn’t even have the famed em dashes, for instance. Still, we are able to recognize it as AI-generated just by looking at its syntax.
For me it is the lack of content, the blandness of the statement. You can tell it is just saying vague statements that could be true if you substituted 14nm and 8nm for 4nm and 1.4nm.
