Project Denver is the codename of a microarchitecture designed by Nvidia that implements the ARMv8-A 64/32-bit instruction sets using a combination of simple hardware decoder and software-based binary translation (dynamic recompilation) where “Denver’s binary translation layer runs in software, at a lower level than the operating system, and stores commonly accessed, already optimized code sequences in a 128 MB cache stored in main memory”.[2] Denver is a very wide in-order superscalar pipeline. Its design makes it suitable for integration with other SIPs cores (e.g. GPU, display controller, DSP, image processor, etc.) into one die constituting a system on a chip (SoC).
Project Denver is targeted at mobile computers, personal computers, servers, as well as supercomputers.[3] Respective cores have found integration in the Tegra SoC series from Nvidia. Initially Denver cores was designed for the 28nm process node (Tegra model T132 aka “Tegra K1”). Denver 2 was an improved design that built for the smaller, more efficient 16nm node. (Tegra model T186 aka “Tegra X2”).
In 2018, Nvidia released an improved design (codename: “Carmel”, based on ARMv8 (64-bit; variant: ARM-v8.2[4] with 10-way superscalar, functional safety, dual execution, parity & ECC) got integrated into the Tegra Xavier SoC offering a total of 8 cores (or 4 dual-core pairs).[5][failed verification] The Carmel CPU core supports full Advanced SIMD (ARM NEON), VFP (Vector Floating Point), and ARMv8.2-FP16.[6] First published testings of Carmel cores integrated in the Jetson AGX development kit by third party experts took place in September 2018 and indicated a noticeably increased performance as should expected for this real world physical manifestation compared to predecessors systems, despite all doubts the used quickness of such a test setup in general an in particular implies.[7] The Carmel design can be found in the Tegra model T194 (“Tegra Xavier”) that is designed with a 12 nm structure size.
Overview[edit]
Pipelined processor with 7-way superscalar execution pipeline
128 KiB instruction + 64 KiB data L1 cache per core (both 4-way), 2 MiB L2 cache (16-way shared)[8]
Denver also sets aside 128 MiB of main memory as an interpretation cache, which is inaccessible to the main operating system.
Running at up to 2.5 GHz[9]
ARM code is translated either by a hardware translator or through software emulation to an instruction set that is internal to Project Denver. ARM instructions can be reordered, removed if they do not contribute to the end result, or otherwise optimized if software emulation is used.[2]
Chips[edit]
A dual-core Denver CPU was paired with a Kepler-based GPU solution to form the Tegra K1; the dual-core 2.3 GHz Denver-based K1 was first used in the HTC Nexus 9 tablet, released November 3, 2014.[10][11] Note, however, that the quad-core Tegra K1, while using the same name, isn’t based on Denver.
The Nvidia Tegra X2 has two Denver2 (ARMv8 64bit) cores inside and another four A57 (ARMv8 64bit) cores using a coherent HMP (Heterogeneous Multi-Processor Architecture) approach.[12] This pairs the units with a Parker-GPU.
The Tegra Xavier is pairing an Nvidia Volta-GPU and several special purpose accelerators with 8 CPU cores with the Carmel design. In this design 4 Carmel ASIC macro blocks (with each having 2 cores) are matched to each other with one more crossbar and 4 MiB of L3 memory.
History[edit]
The existence of Project Denver was revealed at the 2011 Consumer Electronics Show.[13] In a March 4, 2011 Q&A article CEO Jen-Hsun Huang revealed that Project Denver is a five-year 64-bit ARMv8-A architecture CPU development on which hundreds of engineers had already worked for three and half years and which also has 32-bit ARM instruction set (ARMv7) backward compatibility.[14] Project Denver was started in Stexar Company (Colorado) as an x86-compatible processor using binary translation, similar to projects by Transmeta. Stexar was acquired by Nvidia in 2006.[15][16][17]
According to Tom’s Hardware, there are engineers from Intel, AMD, HP, Sun and Transmeta on the Denver team, and they have extensive experience designing superscalar CPUs with out-of-order execution, very long instruction words (VLIW) and simultaneous multithreading (SMT).[18]
According to Charlie Demerjian, the Project Denver CPU may internally translate the ARM instructions to an internal instruction set, using firmware in the CPU.[19] Also according to Demerjian, Project Denver was originally intended to support both ARM and x86 code using code morphing technology from Transmeta, but was changed to the ARMv8-A 64-bit instruction set because Nvidia could not obtain a license to Intel’s patents.[19]
The first consumer device shipping with Denver CPU cores, Google’s Nexus 9, was announced on October 15, 2014. The tablet is manufactured by HTC and features the dual-core Tegra K1 SoC. The Nexus 9 is also the first 64-bit Android device available to consumers.[20]
See also[edit]
Comparison of ARMv8-A cores
References[edit]
^ NVIDIA Jetson AGX Xavier Delivers 32 TeraOps for New Era of AI in Robotics by Dustin Franklin (Nvidia development team for Jetson), December 12, 2018
^ a b Wasson, Scott (August 11, 2014). “Nvidia claims Haswell-class performance for Denver CPU core”. The Tech Report. Retrieved August 14, 2014.
^ Dally, Bill (January 5, 2011). “”PROJECT DENVER” PROCESSOR TO USHER IN NEW ERA OF COMPUTING”. Official Nvidia blog.
^ NVIDIA Jetson AGX Xavier Delivers 32 TeraOps for New Era of AI in Robotics by Dustin Franklin (Nvidia development team for Jetson), December 12, 2018
^ NVIDIA Drive Xavier SOC Detailed by Hassan Mujtaba on Jan 8, 2018 via WccfTech
^ NVIDIA Jetson AGX Xavier Delivers 32 TeraOps for New Era of AI in Robotics by Dustin Franklin (Nvidia development team for Jetson), December 12, 2018
^ “A Quick Test of NVIDIA’s “Carmel” CPU Performance”.
^ Hachman, Mark (August 11, 2014). “Nvidia reveals PC-like performance for ‘Denver’ Tegra K1”. PC World. Retrieved September 19, 2014.
^ Anthony, Sebastian (January 6, 2014). “Tegra K1 64-bit Denver core analysis: Are Nvidia’s x86 efforts hidden within?”. ExtremeTech. Retrieved January 7, 2014.
^ “Nexus 9 storms through Geekbench, Tegra K1 outperforms Apple iPhone 6’s A8”.
^ Shimpi, Anand (January 5, 2014). “NVIDIA Announces Tegra K1 SoC with Optional Denver CPU Cores”. Anandtech. Retrieved January 6, 2014.
^ NVIDIA Unveils Tegra Parker SOC at Hot Chips – Built on 16nm TSMC Process, Features Pascal and Denver 2 Duo Architecture, August 22, 2016
^ http://www.nvidia.com/object/ces2011.html Nvidia’s press conference webcast
^ Takahashi, Dean (March 4, 2011). “Q&A: Nvidia chief explains his strategy for winning in mobile computing”.
^ Valich, Theo (December 12, 2011). “NVIDIA Project Denver “Lost in Rockies”, to Debut in 2014-15″.
^ Miller, Paul (October 19, 2006). “NVIDIA has x86 CPU in the works?”. Engadget. Retrieved October 19, 2013.
^ Valich, Theo (March 20, 2013). “New Tegra Roadmap Reveals Logan, Parker and Kayla CUDA Strategy”.
^ Parrish, Kevin (October 14, 2013). “64-bit Nvidia Tegra 6 “Parker” Chip May Arrive in 2014. Devices with a 64-bit Tegra 6 could launch before the end of 2014″. Tom’s Hardware & ExtremeTech. Retrieved October 19, 2013.
^ a b Demerjian, Charlie (August 5, 2011). “What is Project Denver based on?”. Semiaccurate.
^ Amadeo, Ron (October 15, 2014). “Google announces Nexus 6, Nexus 9, Nexus Player, and Android 5.0 Lollipop”.
External links[edit]
Valich, Theo (September 20, 2012). “NVIDIA Project Boulder Revealed: Tegra’s Competitor Hides in GPU Group”.
Linley Gwennap (August 18, 2014). “Nvidia’s First CPU Is a Winner. Denver Uses Dynamic Translation to Outperform Mobile Rivals”. MPR, Linley Group.
