Intel 10nm blocked: new server jump ticket to 2020

The 10nm process has been slow to be put into mass production. The evolution of Intel's product lines has been greatly hindered. It can only delay the new platform and temporarily add various transitional products. The desktop is the same, the server is also the same. A PPT leaked by Central South University, Intel will launch the next generation server platform Cascade Lake-SP at the end of this year, but it is still 14nm process, up to 28 cores, the architecture is basically unchanged, just a refresh upgrade, and throughout In 2019, there was only this platform.

You know, AMD will come up with a new 7nm technology next year, and a new generation of EPYC Opteron servers with Zen2's new architecture.

Although it is impossible for AMD to cause fatal damage to Intel in a short period of time, with the strengthening of AMD products and market competitiveness, Intel's dominance in the server and data center market will be seriously threatened.

Former Intel CEO Ke Zaiqi has long admitted that it may be able to steal up to 20% of the server market share by AMD.

Intel's new 10nm platform will have to wait until 2020, but in addition to the officially publicly mentioned codenamed Ice Lake-SP, there is also a new Cooper Lake-SP.

Previously it was rumored that the Cooper Lake-SP is still based on the 14nm process, but from this roadmap, the two are obviously the same big platform, so it might just be a backup to prevent the 10nm process from happening again, Ice Lake -SP can't be launched smoothly.

In time, the Cooper Lake-SP is indeed earlier, will debut in 2019, and Ice Lake-SP will be in the second quarter of 2020.

Both support Barlow Pass DIMM technology, it is not clear, but it is likely to be a new generation of Optane DC Persistent Memory DIMM, Cascade Lake-SP will be supported for the first time.

In addition, both of them support eight-channel memory and will have major changes in the OmniPath bus.

As can be seen from the roadmap, there seems to be no new planning for the Xeon Phi product line. In the Ming Dynasty, it was Cascade Lake-AP. It is rumored to be a glue-packaged version of Cascade Lake-SP. The two cores are integrated. , thus providing up to 56 cores.

Considering that Xeon Phi products have begun to retire and there is no new generation, does this mean that the Xeon Phi coprocessor acceleration project ends here?