The German enthusiast Roman “Der8auer” Hartung decided to check announcements of AMD concerning Ryzen Threadripper. Two of four crystals of Ryzen Threadripper processors are simply silicon preparations and do not comprise CPU kernels, a cache memory and other elements. For this Der8auer opened a cover of one more Threadripper processor…
He separated crystals from a substrate and studied their structure. The experiment quitted expensive: only CPU of model 1950X cost Roman approximately €1000. And it not to mention many hours of the free time. Experience consisted of several stages. On the first Der8auer manufactured the metal device for “scalping” of Threadripper — a prototype of the new Delid Die Mate version.
Further the processor it was successfully uncovered. Under it there were four Zeppelin crystals identical in appearance. They were required to be separated from a substrate, without having damaged the bottom (contact) face with “figure”. Roman achieved it, having lowered the processor crystals down on warmed-over copper bar.
The substrate went bubbles and fully darkened, but Zeppelin chips separated from textolite, having remained whole. Further by means of several grinding films 3M with grains to 3 microns Roman removed a silicon blanket from crystals from 40 microns, and as it was clarified, all of them turned on structural units from hundreds of 14-million nanometers of transistors.
These units are visible with the naked eye. What from this follows? First, we will mark insincerity of AMD concerning Ryzen Threadripper “stuffing”. These processors in many respects (minus Infinity Fabric, the additional PCI Express 3.0 lines and channels of memory) copy server EPYC. About any pigs instead of the disconnected crystals the speech does not go.
Secondly, it is quite possible that there is not a lot of suitable Zeppelin chips as marketing specialists of AMD want to show. The company already “miscalculated” earlier with the number of transistors APU Llano for the FM1 platform. At last, the third output which can be made following the results of Roman “Der8auer” Hartung’s experiment.
It consists in the long term appearances in the future of 24-and 32-core Ryzen Threadripper processors. It is possible that under favorable conditions for AMD would decide to release them even last summer.
However sales of EPYC in that case would suffer and it would be too precipitate to use “heavy artillery” in fight for an insignificant share of the market (tens of normal usually fall on each sold HEDT processor).