Intel presented its long-awaited desktop counterpart to AMD this week. The latter had a difficult childhood and is now showing "interesting" results.
As is well known, Intel's chip development has not been going very well in the last few years. However, this has hardly had any influence on sales figures and profits, because chip manufacturers are selling everything that can be produced in any way or that can still be found in warehouse corners, not least because of the Corona crisis.
And Intel can produce a lot with its fabs. However, there are still almost no 10nm fabs with corresponding output, but the majority of production is still based on the already rather old 14nm process. From a positive point of view, this can be described as extremely mature.
Of course, chip development has not come to a standstill and Rocket Lake-S should have come off the production lines with 10nm long ago, but the capacity for it is simply not there yet (??). Intel already suspected this 24 months ago, which is why they have now ventured a so-called backporting with Rocket Lake S. The 10nm design will be produced in a few months. The 10nm design is produced in 14 nm, which brings a lot of challenges with it. Among other things, the dies will be significantly larger and the power consumption of the new chip is likely to set new negative records.
But apart from the power consumption, the chip is now probably not completely embarrassing. Rather, it is an interesting playground for overclockers with powerful mainboards. It achieves a boost clock of over 5 GHz and can thus at least give AMD a run for its money in some benchmarks.
The equipment, however, seems split. On the one hand, PCI-E 4.0 is finally supported, on the other hand, this connection is available as 16x and 4x ports. Since the GPU occupies the 16x port, only 4x lanes remain available for fast SSDs. In direct comparison, AMD offers an 8x connection that is twice as fast.
Special features such as AVX-512, VNNI or DLBoost currently only give Intel a measurable advantage in benchmarks - in real applications, broad support is still lacking. As far as the integrated hardware codecs are concerned, Intel highlights 12-bit HEVC and 10-bit AV1 as news.
However, Intel has created the biggest problem for itself with its own schedule: Alder Lake, a successor in the desktop sector, is to follow this autumn, which will then definitely be manufactured in 10nm and, with DDR5 and probably also PCIe 5.0, should really be on a technological par with the upcoming AMD processors. If Intel does not overturn this schedule at short notice.
Rocket Lake-S should officially be in the shops from 30 March. On that day, comparable benchmarks should also be available through the usual, suspicious trade media. Only then will a clearer picture emerge as to whether an investment in Intel's last 14nm generation is still fundamentally sensible.