We understand as a manufacturing node all the technology used to manufacture processors of a certain size in their transistors. New nodes appear regularly that allow chips with higher capacities.
But for quite some time now, we have seen how every new node is not released to PC but to PostPC devices like smartphones, tablets etc. But all this has a very simple reason to understand.
New nodes debut on PostPC devices
In recent years, we have been able to observe how new manufacturing nodes of processors of all kinds have not come out with high performance products, be it CPUs, GPUs and even SoCs.
Whenever a new node appears we hope to see which high-end processor we are going to see, but that is not true right off the bat and we have to wait several years to finish seeing the use of these manufacturing nodes in high-end processors. .
This has the consequence that, for example, TSMC’s 7nm node was first released in a smartphone SoC, along with the recent 5nm node from the same foundry.
Manufacturing maturity and cost of nodes
Processor manufacturing nodes mature over time, which means they improve their performance. But what do we mean by performance in this case? Well, due to the fact that the number of errors per mm of manufactured chip is decreasing more and more. When the error rate is high, then it reaches the point where making a chip from a certain size becomes counterproductive due to the fact that the number of chips that would come out per wafer becomes so low that it is not profitable to manufacture.
Smartphone SoCs are typically less than 120mm2, they are small enough to launch a new manufacturing node. In addition, the high margins in th is market allow them to finance the high cost of deploying the new manufacturing node. Another reason is related to the clock speed and consumption, it is directly related to the voltage and speeds of use of PostPC devices. clock and therefore voltages lower than those of the office.
Manufacturing nodes start out with clock speeds well below their ideal, which is why data is often leaked by future PC processors who manufacture samples that run at much slower speeds than the final processor, especially s ‘they are. they use a new knot for “brand new”.
On PC, the first cores to free a node are those with poor performance
As the manufacturing nodes have passed, the deployment of each new node has become more and more expensive. If, for example, we had seen AMD, Intel or NVIDIA deploy one of their new architectures directly under one of the new nodes, not only the inventory would have been negatively affected, but also the cost.
A curious case is AMD’s decision with Zen 2, the first version of which was the Ryzen 3000, built by multiple chips, part of which uses a less advanced node while Zen 2 cores used the 7nm node, but measuring less. 100 mm. Over time, AMD released a monolithic version of Zen 2 in the form of the Ryzen 4000, but it took over a year to do so.
Another example is the Intel Lakefield, half of which is built at 10nm and it is a chip with very low power consumption and size, being the first chip built under Intel’s 10nm node.
New nodes will continue to debut at PostPC
Since each new manufacturing node gets more expensive and part of the cost of initial deployment includes the cost of developing the new node, by logic we will see not only how the trend continues, but also the devices. Premium PostPC. they will be more and more expensive and the period of maturity of the nodes will be much longer.
Although the marketing and predictions speak of a certain node for a certain date in the future, these are really very optimistic predictions and we can see that architectures designed for a specific manufacturing node are delayed or backported to a lesser node. advanced.
While future fabrication nodes can in many cases already craft, that doesn’t mean they’re ready to go in the latest high-end processor of whatever nature. that is why we will continue to see them for the first time on PostPC devices.