Developer Guide

  • 2022.1
  • 09/08/2022
  • Public
Contents

Intel Atom® x6000E Series Processor

The following table provides a brief description of each preset.
Preset Details
Description
Usage Notes
Dedicated cache for 1 real-time workload, no software SRAM, shared cache for best-effort workloads on CPU + GPU
Enables 1 real-time workload to run on the system by minimizing interference from GPU. For this preset, disable the Software SRAM Setting.
Dedicated cache for 1 real-time workload, 1 x L3 software SRAM (1 way), shared cache for best-effort workloads on CPU + GPU
Similar to Preset 1, but adds software SRAM in L3 cache. A good starting point for developers looking to experiment with Software SRAM.
Dedicated cache for 2 real-time workloads, 1 x L3 software SRAM (1 way), shared cache for best-effort workloads on CPU, dedicated cache for GPU
Similar to Preset 2, but adds cache partitioning to support 2 independent real-time workloads.
Dedicated cache for 2 real-time workloads, 1 x L3 software SRAM (2 ways), 1 x L2 software SRAM, shared cache for best-effort workloads on CPU + GPU
Similar to Preset 3, but adds L2 software SRAM and increases L3 software SRAM size. Useful for real-time workloads that have access latency requirements beyond what L3 cache can provide.
Dedicated cache for 1 real-time workload, 1 x L3 software SRAM (2 ways), 1 x L2 software SRAM, shared cache for best-effort workloads on CPU, dedicated but minimal cache for GPU
Similar to Preset 4, but only one real-time workload and with a smaller dedicated cache region for GPU.
If you followed the Get Started Guide for UEFI BIOS (Public Document) or the Get Started Guide for Slim Bootloader (Public Document), as part of the process Preset 4 was configured as the default. Otherwise, run Cache Configurator to see what is the current cache partition. See Setup Software SRAM Script for information about the default presets for all supported processors.

Preset Configuration Details

The following tables show the configuration of each preset (assuming 12-way L2, 16-way L3).
Key:
Label
Name
Description
S
Software SRAM
Cache ways dedicated for software SRAM buffer
C[<number>]
CPU[COS ID]
Cache ways assigned to CPU use
G
GPU
Cache ways assigned to GPU use
Preset 1
This preset configures L3 cache only.
This preset allows non-real-time (best-effort) workloads running on the CPU and GPU to share cache ways 15:4. Cache ways 3:0 are dedicated for use by 1 real-time application.
Cache Way
Cache Level 3
15
C[0,2,3], G
14
C[0,2,3], G
13
C[0,2,3], G
12
C[0,2,3], G
11
C[0,2,3], G
10
C[0,2,3], G
9
C[0,2,3], G
8
C[0,2,3], G
7
C[0,2,3], G
6
C[0,2,3], G
5
C[0,2,3], G
4
C[0,2,3], G
3
C[1]
2
C[1]
1
C[1]
0
C[1]
Preset 2
The only difference between Presets 1 and 2 is that cache way 15 is software SRAM.
Cache Way
Cache Level 3
15
S
14
C[0,2,3], G
13
C[0,2,3], G
12
C[0,2,3], G
11
C[0,2,3], G
10
C[0,2,3], G
9
C[0,2,3], G
8
C[0,2,3], G
7
C[0,2,3], G
6
C[0,2,3], G
5
C[0,2,3], G
4
C[0,2,3], G
3
C[1]
2
C[1]
1
C[1]
0
C[1]
Preset 3
The only difference between Presets 2 and 3 is that Preset 3 supports 2 real-time workloads.
Cache Way
Cache Level 3
15
S
14
C[0,3]
13
C[0,3]
12
C[0,3]
11
C[0,3]
10
C[0,3]
9
C[0,3]
8
C[0,3]
7
C[0,3]
6
C[0,3]
5
C[0,3]
4
C[0,3]
3
C[2]
2
C[2]
1
C[1]
0
C[1]
Preset 4
This preset configures L2 cache and L3 cache.
In L3 cache, cache ways 15:14 are software SRAM. Cache ways 13:4 are shared among best-effort workloads running on CPU and GPU. Cache ways 3:2 are reserved for the first real-time workload, and cache ways 1:0 are reserved for the second real-time workload.
In L2 cache, cache way 11 is software SRAM. The rest of the cache is for best-effort workloads.
Cache Way
Cache Level 3
15
S
14
S
13
C[0,3], G
12
C[0,3], G
11
C[0,3], G
10
C[0,3], G
9
C[0,3], G
8
C[0,3], G
7
C[0,3], G
6
C[0,3], G
5
C[0,3], G
4
C[0,3], G
3
C[2]
2
C[2]
1
C[1]
0
C[1]
Cache Way
Cache Level 2
11
S
10
C
9
C
8
C
7
C
6
C
5
C
4
C
3
C
2
C
1
C
0
C
Preset 5
This preset configures L2 cache and L3 cache.
In L3 cache, cache ways 15:14 are software SRAM. Cache ways 13:5 are shared among best-effort workloads running on the CPU. Cache ways 4:1 are reserved for the real-time workload. Cache way 0 is reserved for GPU.
In L2 cache, cache way 11 is software SRAM. The rest of the cache is for best-effort workloads.
Cache Way
Cache Level 3
15
S
14
S
13
C[0,2,3]
12
C[0,2,3]
11
C[0,2,3]
10
C[0,2,3]
9
C[0,2,3]
8
C[0,2,3]
7
C[0,2,3]
6
C[0,2,3]
5
C[0,2,3]
4
C[1]
3
C[1]
2
C[1]
1
C[1]
0
G
Cache Way
Cache Level 2
11
S
10
C
9
C
8
C
7
C
6
C
5
C
4
C
3
C
2
C
1
C
0
C

Product and Performance Information

1

Performance varies by use, configuration and other factors. Learn more at www.Intel.com/PerformanceIndex.