What makes this fascinating is that Samsung doesn’t formally have a 12 nm FinFET course of. It has 14 nm, and the 11LPP, a 11 nm nodelet, which the corporate designed to compete with TSMC 12 nm. It would therefore be actually fascinating to listen to from NVIDIA on whether or not they’ve scaled out the “TU106” to 14LPP, or right down to 11LPP at Samsung. It’s fascinating to notice that the shrink in transistor sizes in these nodelets does not have an effect on die-sizes. We therefore see no die-size distinction between these Korea-marked chips, and people marked “Taiwan.” We’ve reached out to NVIDIA for remark.
Update July third: NVIDIA received again to us
The reply is actually easy and these markings should not new. Other Turing GPUs have had these markings prior to now. The chip is made at TSMC, however packaged in numerous areas. This one was carried out in Korea, therefore why his says “Korea”.On an unrelated notice: We already use each TSMC and Samsung, and qualify every of them for each course of node. We cannot remark in any additional element on future plans, however each stay terrific companions.