From WikiChip
Difference between revisions of "intel/microarchitectures/tremont"
< intel‎ | microarchitectures

(One-word typo fix. (Accross to across))
 
(34 intermediate revisions by 13 users not shown)
Line 5: Line 5:
 
|designer=Intel
 
|designer=Intel
 
|manufacturer=Intel
 
|manufacturer=Intel
|introduction=2018/2019
+
|introduction=2019
 +
|process=10 nm
 
|type=Superscalar
 
|type=Superscalar
 
|oooe=Yes
 
|oooe=Yes
Line 23: Line 24:
 
|extension 11=PCLMUL
 
|extension 11=PCLMUL
 
|extension 12=RDRND
 
|extension 12=RDRND
|extension 13=SHA
+
|extension 13=XSAVE
|core name=Gemini Lake
+
|extension 14=XSAVEOPT
 +
|extension 15=FSGSBASE
 +
|extension 16=PTWRITE
 +
|extension 17=RDPID
 +
|extension 18=SGX
 +
|extension 19=UMIP
 +
|extension 20=GFNI-SSE
 +
|extension 21=CLWB
 +
|extension 22=ENCLV
 +
|extension 23=SHA
 +
|core name=Elkhart Lake
 +
|core name 2=Jasper Lake
 +
|core name 3=Skyhawk Lake
 +
|core name 4=Lakefield
 +
|core name 5=Snow Ridge
 
|predecessor=Goldmont Plus
 
|predecessor=Goldmont Plus
 
|predecessor link=intel/microarchitectures/goldmont plus
 
|predecessor link=intel/microarchitectures/goldmont plus
 +
|successor=Gracemont
 +
|successor link=intel/microarchitectures/gracemont
 
}}
 
}}
'''Tremont''' is a successor to {{\\|Goldmont Plus}}, [[Intel]]'s future microarchitecture for Intel's ultra-low power line of microprocessors.
+
'''Tremont''' is [[Intel]]'s successor to {{\\|Goldmont Plus}}, a [[10 nm]] microarchitecture for ultra-low power devices and microservers.
  
 
== Codenames ==
 
== Codenames ==
{{empty section}}
+
{| class="wikitable"
 +
! Platform !! Core Name || PCH
 +
|-
 +
| || {{intel|Skyhawk Lake|l=core}} ||
 +
|-
 +
| Jacobsville || {{intel|Elkhart Lake|l=core}} || {{intel|Mule Creek Canyon|l=chipset}}
 +
|-
 +
| || {{intel|Jasper Lake|l=core}} ||
 +
|-
 +
| || {{intel|Lakefield|l=core}} ||
 +
|-
 +
| || {{intel|Snow Ridge |l=core}} ||
 +
|}
  
 
== Brands ==
 
== Brands ==
Line 37: Line 66:
  
 
== Release Dates ==
 
== Release Dates ==
{{empty section}}
+
Tremont was released in a number of products in late 2019.
  
 
== Technology ==
 
== Technology ==
{{empty section}}
+
Tremont uses Intel's [[10 nm process]].
 +
 
 +
== Compiler support ==
 +
{| class="wikitable"
 +
|-
 +
! Compiler !! Arch-Specific || Arch-Favorable
 +
|-
 +
| [[ICC]] || <code>-march=tremont</code> || <code>-mtune=tremont</code>
 +
|-
 +
| [[GCC]] || <code>-march=tremont</code> || <code>-mtune=tremont</code>
 +
|-
 +
| [[LLVM]] || <code>-march=tremont</code> || <code>-mtune=tremont</code>
 +
|-
 +
| [[Visual Studio]] || <code>/arch:?</code> || <code>/tune:?</code>
 +
|}
 +
 
 +
=== CPUID ===
 +
{| class="wikitable tc1 tc2 tc3 tc4"
 +
! Core !! Extended<br>Family !! Family !! Extended<br>Model !! Model
 +
|-
 +
| rowspan="2" | ? || 0 || 0x6 || 0x8 || 0x6
 +
|-
 +
| colspan="4" | Family 6 Model 134
 +
|}
  
 
== Architecture ==
 
== Architecture ==
 +
Tremont is designed with significant single-thread performance in mind while focusing on low-power small silicon area cores.
 
=== Key changes from {{\\|Goldmont Plus}} ===
 
=== Key changes from {{\\|Goldmont Plus}} ===
{{empty section}}
+
* Significant [[IPC]] uplift ([[Intel]] self-reported average 32% IPC across proxy benchmarks such as [[SPEC CPU2006]]/[[SPEC CPU2017]])
 +
* Front-end
 +
** Redesigned front-end
 +
*** New dual symmetric decode cluster
 +
**** Out-of-order decode
 +
**** 6-wide decode
 +
***** 3-way decode per cluster
 +
** Smarter [[prefetchers]]
 +
** Improved [[branch predictor]]
 +
*** Big-core level of performance
 +
* Back-end
 +
** larger ROB
 +
** wide issue (10-wide)
 +
* Execution Engine
 +
** 2x store data ports (up from 1)
 +
 
  
 
====New instructions ====
 
====New instructions ====
Termont introduced a number of {{x86|extensions|new instructions}}:
+
Tremont introduced a number of {{x86|extensions|new instructions}}:
  
 
* {{x86|CLWB|<code>CLWB</code>}} - Force cache line write-back without flush
 
* {{x86|CLWB|<code>CLWB</code>}} - Force cache line write-back without flush
Line 56: Line 124:
 
* User wait instructions: TPAUSE, UMONITOR, UMWAIT
 
* User wait instructions: TPAUSE, UMONITOR, UMWAIT
 
* Split Lock Detection - detection and cause an exception for split locks
 
* Split Lock Detection - detection and cause an exception for split locks
 +
 +
=== Block Diagram ===
 +
==== Individual Core ====
 +
:[[File:tremont block diagram.svg|850px]]

Latest revision as of 19:12, 10 December 2020

Edit Values
Tremont µarch
General Info
Arch TypeCPU
DesignerIntel
ManufacturerIntel
Introduction2019
Process10 nm
Pipeline
TypeSuperscalar
OoOEYes
SpeculativeYes
Reg RenamingYes
Instructions
ISAx86-64
ExtensionsMOVBE, MMX, SSE, SSE2, SSE3, SSSE3, SSE4.1, SSE4.2, POPCNT, AES, PCLMUL, RDRND, XSAVE, XSAVEOPT, FSGSBASE, PTWRITE, RDPID, SGX, UMIP, GFNI-SSE, CLWB, ENCLV, SHA
Cores
Core NamesElkhart Lake,
Jasper Lake,
Skyhawk Lake,
Lakefield,
Snow Ridge
Succession

Tremont is Intel's successor to Goldmont Plus, a 10 nm microarchitecture for ultra-low power devices and microservers.

Codenames[edit]

Platform Core Name PCH
Skyhawk Lake
Jacobsville Elkhart Lake Mule Creek Canyon
Jasper Lake
Lakefield
Snow Ridge

Brands[edit]

New text document.svg This section is empty; you can help add the missing info by editing this page.

Release Dates[edit]

Tremont was released in a number of products in late 2019.

Technology[edit]

Tremont uses Intel's 10 nm process.

Compiler support[edit]

Compiler Arch-Specific Arch-Favorable
ICC -march=tremont -mtune=tremont
GCC -march=tremont -mtune=tremont
LLVM -march=tremont -mtune=tremont
Visual Studio /arch:? /tune:?

CPUID[edit]

Core Extended
Family
Family Extended
Model
Model
 ? 0 0x6 0x8 0x6
Family 6 Model 134

Architecture[edit]

Tremont is designed with significant single-thread performance in mind while focusing on low-power small silicon area cores.

Key changes from Goldmont Plus[edit]

  • Significant IPC uplift (Intel self-reported average 32% IPC across proxy benchmarks such as SPEC CPU2006/SPEC CPU2017)
  • Front-end
    • Redesigned front-end
      • New dual symmetric decode cluster
        • Out-of-order decode
        • 6-wide decode
          • 3-way decode per cluster
    • Smarter prefetchers
    • Improved branch predictor
      • Big-core level of performance
  • Back-end
    • larger ROB
    • wide issue (10-wide)
  • Execution Engine
    • 2x store data ports (up from 1)


New instructions[edit]

Tremont introduced a number of new instructions:

  • CLWB - Force cache line write-back without flush
  • ENCLV - SGX oversubscription instructions
  • CLDEMOTE - Cache line demote instruction
  • SSE_GFNI - SSE-based Galois Field New Instructions
  • Direct store instructions: MOVDIRI, MOVDIR64B
  • User wait instructions: TPAUSE, UMONITOR, UMWAIT
  • Split Lock Detection - detection and cause an exception for split locks

Block Diagram[edit]

Individual Core[edit]

tremont block diagram.svg
codenameTremont +
designerIntel +
first launched2019 +
full page nameintel/microarchitectures/tremont +
instance ofmicroarchitecture +
instruction set architecturex86-64 +
manufacturerIntel +
microarchitecture typeCPU +
nameTremont +
process10 nm (0.01 μm, 1.0e-5 mm) +