David Witt(Chief for K8 generation secondary design) => 1997 homo-clustered architecture (Two monolithic clusters each with Int/FPU//LD/ST)
Jim Keller => 1999 hetero-clustered architecture (Three clusters, one with Int, one with FPU, one with LD/ST:;Proto-Bobcat // handled by the converted MIPS->x86 Alchemy team with Brad B.(Chief Arch(Bobcat))/McKinney(Chief Tech(Bobcat&Bulldozer))
Andy Glew => 2002-2004 K10 architecture (Independent from the above, but spinned out from LP-project::clusters are not yet cores)
Charles R. Moore => 2005-1H2007, Original Bulldozer architecture (Derived from Witt/Keller with four clusters: Two Int(CMT2), One LD/ST(SMT2), One FPU(SMT2)::Bulldozer)
Mike Butler => 2H2007-2013, K10 as "Bulldozer" (Derived from Glew's microarchitecture at AMD::clusters became cores)
Glew/Butler Architected design focused towards high performance => higher frequency
"Perhaps I should say here that my MCMT had a significant difference from
clustering in, say, the Alpha 21264,
Those clusters bypass to each other: there is a fast bypass within a
cluster, and a slightly slower (+1 cycle) bypass of results between
clusters. The clusters are execution units only, and share the data
cache. This bypassing makes it easy (or at least easier) to spread a
single thread across both clusters. My MCMT clusters, on the other
hand, do NOT bypass to each other. This motivates separate threads per
cluster, whether explicit or implicit."
Boss of Glew "K10 Architect" and Butler "K10 as Bulldozer Architect" wanted high frequency. As well not implementing SpMT or anything complex which could comprise frequency.
Moore Architected design focused towards low power and scalability => higher IPC(TLP->IPC[MT] and ILP->IPC[ST])
Boss of Moore's time as OG Bulldozer Architect wanted low power.
New naming scheme of ongoing AMD development;
HP1(launched) -> HP2(launched) -> HP3(dropped:2012->2013) // HP3 ST had access to 8 add/inc/dec/etc ALUs
LP1(hiatus:2014->2018) // Started while StoneyX was being done. FT4 socket was suppose to be Excavator and LP-NextGen(LP1).
ULP1(restart:LP1, resume on new target node:2019->2023]