Processors Raptor Lake Intel announced it as a series equipped with up to eight large and up to sixteen small cores, totaling a maximum of 24 cores. The editors of the Tom’sHardware website noticed that one of the wafers that Intel brought to the event was noticeably different (detail in the introduction + photo below this paragraph). The relatively large chips carried 34 large (P) cores.
The wafer was labeled “Raptor Lake-S, 34 core”. The editors of the Tom’sHardware website believe that this wafer made it to the event by mistake. Some websites speculate on the existence of a more powerful desktop version Raptor Lake, however this is most certainly not the case. Tom’sHardware offers a more rational explanation and that it is about Sapphire Rapids MCCwhich in the configuration for workstations could carry markings Raptor Lake.
Personally, I believe that the explanation can be even simpler, except for one discrepancy. It is definitely not a standard desktop processor, a backup in the form of a more powerful version Raptor Lake for socket LGA-1700. The chip carries an eight-channel memory controller, while the LGA-1700 socket does not support more than two channels.
The theory about configuration for workstations is also a little out of place. There were two series of leaks so far. The first is to bear the name Xeon-W Mainstream Workstation Platform (MWP) and build on a monolithic chip with 24 cores and a four-channel controller. The second will be the Xeon-W Expert Workstation Platform (EWP), coming out of tiles (chiplets) and carrying up to 56 cores and an eight-channel controller.
So a chip from a wafer does not fit into either the MWP specifications (which should have ten cores less and half the memory partition), or the EWP specifications (which should be chiplet). It is simply something in between.
So either the wafer shows a chip that really should have been something in between, but Intel for some reason canceled plans for its mass production, or it is, on the contrary, a newer design that has not yet appeared in Intel’s materials and which is supposed to strengthen the MWP series that would in 2023 with 24 cores in the workstation segment seemed rather weak (a situation could arise when the next generation of AMD desktop processors would significantly outperform Intel’s significantly more expensive product for workstations – which was already the case with the second generation Zenu 2 and resulted in killing the Intel HEDT product offering for several years).
The core from the exposed wafer is strikingly reminiscent of the Ice Lake-SP server configuration (28 cores) you see in the image above; it carries only large cores, they are not connected by a ring-bus, but by a network (mesh) and is equipped with more than a two-channel memory controller
However, the already mentioned option is also possible, that it is not a chip intended for sale. In that case, I would venture to disagree with Tom’sHardware that the wafer appeared at the event by mistake. When we read the label on the wafer case to the end, we come across the text: “INTEL COE/GCM Marketing RNB 650″. If the marketing department during the preparation of the Innovation event (=presentation Raptor Lake) requests a wafer with a 34-core configuration, prepares a case for it, and puts a sticker on it that says “Raptor Lake-S, 34c”, which is obviously nonsensical*, because , then it’s definitely intentional. Intel’s marketing department probably wanted to send a clear signal that it has something more powerful in store for the standard desktop, in short, that the 24-core configuration, which AMD competes with 16 cores with lower consumption, is not the best that Intel is able to offer.
*marking “S” belongs to the standard desktop series (LGA1700 / dual-channel) and is not used for the segment of workstations or servers
Source: Diit.cz by diit.cz.
*The article has been translated based on the content of Diit.cz by diit.cz. If there is any problem regarding the content, copyright, please leave a report below the article. We will try to process as quickly as possible to protect the rights of the author. Thank you very much!
*We just want readers to access information more quickly and easily with other multilingual content, instead of information only available in a certain language.
*We always respect the copyright of the content of the author and always include the original link of the source article.If the author disagrees, just leave the report below the article, the article will be edited or deleted at the request of the author. Thanks very much! Best regards!