Llama.cpp

Llama.cpp is a port of Facebook's LLaMA model in C/C++ developed by Georgi Gerganov. Llama.cpp allows the inference of LLaMA and other supported models in C/C++. For CPU inference Llama.cpp supports AVX2/AVX-512, ARM NEON, and other modern ISAs along with features like OpenBLAS usage.

To run this test with the Phoronix Test Suite, the basic command is: phoronix-test-suite benchmark llama-cpp.

Project Site

github.com

Source Repository

github.com

Test Created

10 January 2024

Last Updated

2 June 2024

Test Maintainer

Michael Larabel 

Test Type

System

Average Install Time

43 Seconds

Average Run Time

16 Minutes, 59 Seconds

Test Dependencies

C/C++ Compiler Toolchain + BLAS (Basic Linear Algebra Sub-Routine)

Accolades

5k+ Downloads + Recently Updated Test Profile

Supported Platforms


Public Result Uploads *Reported Installs **Reported Test Completions **Test Profile Page ViewsOpenBenchmarking.orgEventsLlama.cpp Popularity Statisticspts/llama-cpp2024.012024.022024.032024.042024.052024.062K4K6K8K10K
* Uploading of benchmark result data to OpenBenchmarking.org is always optional (opt-in) via the Phoronix Test Suite for users wishing to share their results publicly.
** Data based on those opting to upload their test results to OpenBenchmarking.org and users enabling the opt-in anonymous statistics reporting while running benchmarks from an Internet-connected platform.
Data updated weekly as of 14 June 2024.

Revision History

pts/llama-cpp-1.1.0   [View Source]   Sun, 02 Jun 2024 10:36:25 GMT
Update against Llama.cpp upstream, switch to Llama 3 model.

pts/llama-cpp-1.0.0   [View Source]   Wed, 10 Jan 2024 18:02:20 GMT
Initial commit of llama.cpp CPU benchmark.

Suites Using This Test

Machine Learning

HPC - High Performance Computing

Large Language Models


Performance Metrics

Analyze Test Configuration:

Llama.cpp b1808

Model: llama-2-7b.Q4_0.gguf

OpenBenchmarking.org metrics for this test profile configuration based on 215 public results since 10 January 2024 with the latest data as of 23 May 2024.

Below is an overview of the generalized performance for components where there is sufficient statistically significant data based upon user-uploaded results. It is important to keep in mind particularly in the Linux/open-source space there can be vastly different OS configurations, with this overview intended to offer just general guidance as to the performance expectations.

Component
Percentile Rank
# Compatible Public Results
Tokens Per Second (Average)
93rd
8
29.7 +/- 0.2
91st
6
29.4 +/- 0.7
88th
3
27.5 +/- 0.8
87th
3
26.3 +/- 2.7
78th
4
18.1 +/- 0.3
Mid-Tier
75th
< 16.9
66th
8
16.5 +/- 0.1
66th
3
16.5 +/- 0.5
61st
9
16.4 +/- 0.2
59th
7
15.9 +/- 1.2
58th
6
15.7 +/- 1.3
55th
11
15.0 +/- 1.0
Median
50th
14.6
48th
11
14.4 +/- 0.8
34th
3
14.0 +/- 0.6
32nd
6
13.8 +/- 0.3
28th
4
13.5 +/- 0.1
Low-Tier
25th
< 12.9
22nd
5
12.2 +/- 0.4
20th
4
12.0 +/- 0.8
15th
4
10.7 +/- 0.7
12th
4
8.6 +/- 0.1
6th
7
4.0 +/- 0.2
OpenBenchmarking.orgDistribution Of Public Results - Model: llama-2-7b.Q4_0.gguf215 Results Range From 0 To 33 Tokens Per Second36912151821242730333620406080100

Based on OpenBenchmarking.org data, the selected test / test configuration (Llama.cpp b1808 - Model: llama-2-7b.Q4_0.gguf) has an average run-time of 2 minutes. By default this test profile is set to run at least 3 times but may increase if the standard deviation exceeds pre-defined defaults or other calculations deem additional runs necessary for greater statistical accuracy of the result.

OpenBenchmarking.orgMinutesTime Required To Complete BenchmarkModel: llama-2-7b.Q4_0.ggufRun-Time3691215Min: 1 / Avg: 1.88 / Max: 8

Based on public OpenBenchmarking.org results, the selected test / test configuration has an average standard deviation of 0.8%.

OpenBenchmarking.orgPercent, Fewer Is BetterAverage Deviation Between RunsModel: llama-2-7b.Q4_0.ggufDeviation3691215Min: 0 / Avg: 0.81 / Max: 8

Does It Scale Well With Increasing Cores?

Yes, based on the automated analysis of the collected public benchmark data, this test / test settings does generally scale well with increasing CPU core counts. Data based on publicly available results for this test / test settings, separated by vendor, result divided by the reference CPU clock speed, grouped by matching physical CPU core count, and normalized against the smallest core count tested from each vendor for each CPU having a sufficient number of test samples and statistically significant data.

OpenBenchmarking.orgRelative Core Scaling To BaseLlama.cpp CPU Core ScalingModel: llama-2-7b.Q4_0.gguf681232641.39792.79584.19375.59166.9895

Notable Instruction Set Usage

Notable instruction set extensions supported by this test, based on an automatic analysis by the Phoronix Test Suite / OpenBenchmarking.org analytics engine.

Instruction Set
Support
Instructions Detected
Used by default on supported hardware.
Found on Intel processors since Sandy Bridge (2011).
Found on AMD processors since Bulldozer (2011).

 
VZEROUPPER VEXTRACTF128 VPERMILPS VBROADCASTSS VPERM2F128 VINSERTF128 VBROADCASTSD VMASKMOVPS
Used by default on supported hardware.
Found on Intel processors since Haswell (2013).
Found on AMD processors since Excavator (2016).

 
VPBROADCASTQ VEXTRACTI128 VPBROADCASTD VPERMQ VGATHERQPS VPBROADCASTW VINSERTI128 VPBROADCASTB VPSLLVD VPERMD VPSRLVD VPERM2I128
Advanced Vector Extensions 512 (AVX512)
Used by default on supported hardware.
 
(ZMM REGISTER USE)
AVX Vector Neural Network Instructions (AVX-VNNI)
Used by default on supported hardware.
 
VPDPBUSD
FMA (FMA)
Used by default on supported hardware.
Found on Intel processors since Haswell (2013).
Found on AMD processors since Bulldozer (2011).

 
VFMADD231PS VFMADD132PS VFNMADD231PS VFMADD213PS VFNMADD231SS VFMADD132SS VFNMADD213SS VFMADD132SD VFMADD231SS VFMADD213SS
The test / benchmark does honor compiler flag changes.
Last automated analysis: 6 June 2024

This test profile binary relies on the shared libraries libopenblas.so.0, libm.so.6, libc.so.6, libgfortran.so.5, libquadmath.so.0.

Tested CPU Architectures

This benchmark has been successfully tested on the below mentioned architectures. The CPU architectures listed is where successful OpenBenchmarking.org result uploads occurred, namely for helping to determine if a given test is compatible with various alternative CPU architectures.

CPU Architecture
Kernel Identifier
Verified On
Intel / AMD x86 64-bit
x86_64
(Many Processors)
ARMv8 64-bit
aarch64
ARMv8 Neoverse-N1 128-Core

Recent Test Results

OpenBenchmarking.org Results Compare

1 System - 236 Benchmark Results

AMD Ryzen 9 7900 12-Core - ASRockRack 1U4LW-B650/2L2T B650D4U-2L2T/BCM - AMD Device 14d8

Ubuntu 24.04 - 6.8.0-31-generic - GNOME Shell 46.0

1 System - 3 Benchmark Results

AMD EPYC 7R13 48-Core - Supermicro H12SSL-I v1.02 - AMD Starship

EndeavourOS rolling - 6.8.2-zen2-1-zen - Xfce 4.18

1 System - 341 Benchmark Results

AMD Ryzen 9 7950X 16-Core - ASUS ProArt X670E-CREATOR WIFI - AMD Device 14d8

Pop 22.04 - 6.6.10-76060610-generic - GNOME Shell 42.5

2 Systems - 8 Benchmark Results

AMD EPYC 7R13 48-Core - Supermicro H12SSL-I v1.02 - AMD Starship

EndeavourOS rolling - 6.7.9-zen1-1-zen - X Server 1.21.1.11

1 System - 1 Benchmark Result

AMD EPYC 7R13 48-Core - Supermicro H12SSL-I v1.02 - AMD Starship

EndeavourOS rolling - 6.7.9-zen1-1-zen - X Server 1.21.1.11

3 Systems - 79 Benchmark Results

Intel Core i7-1185G7 - Dell XPS 13 9310 0DXP1F - Intel Tiger Lake-LP

Ubuntu 23.10 - 6.7.0-060700rc5-generic - GNOME Shell 45.1

2 Systems - 52 Benchmark Results

2 x INTEL XEON PLATINUM 8592+ - Quanta Cloud QuantaGrid D54Q-2U S6Q-MB-MPS - Intel Device 1bce

Ubuntu 23.10 - 6.6.0-060600-generic - GCC 13.2.0

2 Systems - 28 Benchmark Results

AMD EPYC 7R13 48-Core - Supermicro H12SSL-I v1.02 - AMD Starship

EndeavourOS rolling - 6.7.4-zen1-1-zen - Xfce 4.18

2 Systems - 23 Benchmark Results

AMD EPYC 7R13 48-Core - Supermicro H12SSL-I v1.02 - AMD Starship

EndeavourOS rolling - 6.7.4-zen1-1-zen - Xfce 4.18

2 Systems - 23 Benchmark Results

AMD EPYC 7R13 48-Core - Supermicro H12SSL-I v1.02 - AMD Starship

EndeavourOS rolling - 6.7.4-zen1-1-zen - Xfce 4.18

2 Systems - 18 Benchmark Results

AMD EPYC 7R13 48-Core - Supermicro H12SSL-I v1.02 - AMD Starship

EndeavourOS rolling - 6.7.4-zen1-1-zen - Xfce 4.18

2 Systems - 13 Benchmark Results

AMD EPYC 7R13 48-Core - Supermicro H12SSL-I v1.02 - AMD Starship

EndeavourOS rolling - 6.7.4-zen1-1-zen - Xfce 4.18

5 Systems - 587 Benchmark Results

Intel Core i5-14500 - ASUS PRIME Z790-P WIFI - Intel Raptor Lake-S PCH

Ubuntu 23.10 - 6.7.3-060703-generic - GNOME Shell 45.2

1 System - 12 Benchmark Results

AMD EPYC 7R13 48-Core - Supermicro H12SSL-I v1.02 - AMD Starship

EndeavourOS rolling - 6.7.4-zen1-1-zen - Xfce 4.18

1 System - 6 Benchmark Results

AMD EPYC 7R13 48-Core - Supermicro H12SSL-I v1.02 - AMD Starship

EndeavourOS rolling - 6.7.4-zen1-1-zen - Xfce 4.18

Most Popular Test Results

OpenBenchmarking.org Results Compare

6 Systems - 162 Benchmark Results

AMD Ryzen 7 5700G - ASUS TUF GAMING B550M-PLUS - AMD Renoir

Ubuntu 23.10 - 6.7.0-060700-generic - GNOME Shell 45.0

18 Systems - 154 Benchmark Results

AMD Ryzen 9 7950X 16-Core - ASRock B650 Pro RS - AMD Device 14d8

Ubuntu 23.10 - 6.7.0-060700-generic - GNOME Shell 45.0

5 Systems - 149 Benchmark Results

AMD EPYC 8534PN 32-Core - AMD Cinnabar - AMD Device 14a4

Ubuntu 23.10 - 6.6.9-060609-generic - GNOME Shell 45.0

5 Systems - 587 Benchmark Results

Intel Core i5-14500 - ASUS PRIME Z790-P WIFI - Intel Raptor Lake-S PCH

Ubuntu 23.10 - 6.7.3-060703-generic - GNOME Shell 45.2

16 Systems - 168 Benchmark Results

AMD Ryzen 5 5600G - ASUS TUF GAMING B550M-PLUS - AMD Renoir

Ubuntu 23.10 - 6.7.0-060700-generic - GNOME Shell 45.0

3 Systems - 83 Benchmark Results

AMD EPYC 8534P 64-Core - AMD Cinnabar - AMD Device 14a4

Ubuntu 23.10 - 6.5.0-5-generic - GNOME Shell

4 Systems - 100 Benchmark Results

AMD Ryzen Threadripper PRO 5965WX 24-Cores - ASUS Pro WS WRX80E-SAGE SE WIFI - AMD Starship

Ubuntu 23.10 - 6.5.0-13-generic - GNOME Shell 45.0

2 Systems - 52 Benchmark Results

2 x INTEL XEON PLATINUM 8592+ - Quanta Cloud QuantaGrid D54Q-2U S6Q-MB-MPS - Intel Device 1bce

Ubuntu 23.10 - 6.6.0-060600-generic - GCC 13.2.0

3 Systems - 69 Benchmark Results

Intel Xeon Silver 4216 - TYAN S7100AG2NR - Intel Sky Lake-E DMI3 Registers

Debian 12 - 6.1.0-11-amd64 - X Server

2 Systems - 262 Benchmark Results

AMD EPYC 7F32 8-Core - ASRockRack EPYCD8 - AMD Starship

Debian 12 - 6.1.0-11-amd64 - X Server

2 Systems - 31 Benchmark Results

Intel Core i9-10980XE - ASRock X299 Steel Legend - Intel Sky Lake-E DMI3 Registers

Ubuntu 22.04 - 6.2.0-39-generic - GNOME Shell 42.2

2 Systems - 28 Benchmark Results

AMD EPYC 7R13 48-Core - Supermicro H12SSL-I v1.02 - AMD Starship

EndeavourOS rolling - 6.7.4-zen1-1-zen - Xfce 4.18

2 Systems - 23 Benchmark Results

AMD EPYC 7R13 48-Core - Supermicro H12SSL-I v1.02 - AMD Starship

EndeavourOS rolling - 6.7.4-zen1-1-zen - Xfce 4.18

3 Systems - 27 Benchmark Results

AMD EPYC 7601 32-Core - TYAN B8026T70AE24HR - AMD 17h

Ubuntu 23.10 - 6.6.9-060609-generic - GNOME Shell 45.0

Find More Test Results