Geekbench AI

Sabroe SMC

Sabroe SMC

Grand Admiral Special
★ Themenstarter ★
Mitglied seit
14.05.2008
Beiträge
5.068
Renomée
716
Standort
Castrop-Rauxel
  • Docking@Home
  • BOINC Pentathlon 2011
  • BOINC Pentathlon 2012
  • BOINC Pentathlon 2013
  • BOINC Pentathlon 2014
  • BOINC Pentathlon 2015
  • BOINC Pentathlon 2016
  • BOINC Pentathlon 2017
  • BOINC Pentathlon 2018
  • BOINC Pentathlon 2019
  • BOINC Pentathlon 2021
  • BOINC Pentathlon 2022
  • BOINC Pentathlon 2023
  • BOINC Pentathlon 2024
  • BOINC Pentathlon 2025
  • BOINC Pentathlon 2026
Ein neuer Benchmark. Kann man hier: https://www.geekbench.com/ai/download/windows/ herunterladen. Mal sehen wie er sich so verhält.
Text ist von hier: https://www.smartdroid.de/geekbench-ai-bekannter-benchmark-testet-jetzt-auch-ki-faehigkeiten/

Das KI-Zeitalter erfordert neue Wege, um die Leistungsfähigkeit von Smartphones, Computern und Co. auf den Prüfstand zu stellen.

Primate Labs, die Entwickler des bekannten CPU-Benchmarks Geekbench, haben mit Geekbench AI einen neuen Benchmark vorgestellt, der die Leistung von Geräten bei Machine-Learning-Aufgaben misst. Der Cross-Plattform-Benchmark ist für Android, iOS, Linux, macOS und Windows verfügbar und testet die Performance von CPU, GPU und dedizierten KI-Chips (NPU für Neural Processing Unit, bei Pixel auch TPU für Tensor Processing Unit genannt) mit realistischen KI-Workloads.
Geekbench AI führt insgesamt zehn verschiedene KI-Workloads in drei Genauigkeitsstufen (Single Precision, Half Precision, Quantized) aus. Dabei kommen große Datensätze zum Einsatz, die praxisnahe Anwendungsfälle nachbilden, wodurch Entwickler und Anwender ein mehrdimensionales Bild der KI-Performance ihres Geräts erhalten sollen.
Eine Besonderheit ist, dass sich gezielt die Leistung von CPU, GPU oder NPU messen lässt. Auch die Wahl des KI-Frameworks ist möglich, etwa CoreML unter iOS/macOS oder QNN auf Snapdragon-basierten ARM-PCs. Entwickler können so die optimale Kombination aus Framework und Hardware für bestimmte Aufgaben ermitteln.
Die erzielten Ergebnisse werden über alle Plattformen hinweg vergleichbar gemacht, indem sie auf ein Baseline-System (Intel Core i7-10700) normiert werden. Ein Score von 1.500 Punkten bedeutet Gleichstand mit diesem Referenzsystem. Durch die Unterstützung vieler Betriebssysteme und Hardware-Konfigurationen lässt sich die KI-Leistung von Smartphones mit stromsparender NPU bis hin zu High-End-Workstations vergleichen.
Geekbench AI steht in einer kostenlosen Basisversion und als „Pro“-Variante für 99 US-Dollar zur Verfügung. Letztere bietet zusätzlich Automatisierungstools, einen Offline-Modus und eine kommerzielle Nutzungslizenz. Für Unternehmen gibt es spezielle Lizenzen mit dediziertem Support und Zugriff auf den Quellcode.

ASUS System Product Name​

Geekbench AI Score​

Single Precision Score 4491
Half Precision Score 1051
Quantized Score 5460
Geekbench AI 1.0.0 for Windows AVX2

Result Information​

Upload DateAugust 19 2024 02:12 PM
Views1

System Information​

AI Information
FrameworkONNX
BackendCPU
DeviceAMD Ryzen 9 7950X3D
System Information
Operating SystemMicrosoft Windows 11 Pro (64-bit)
ModelASUS System Product Name
MotherboardASUSTeK COMPUTER INC. TUF GAMING X670E-PLUS
Power PlanAusbalanciert
CPU Information
NameAMD Ryzen 9 7950X3D
Topology1 Processor, 16 Cores, 32 Threads
IdentifierAuthenticAMD Family 25 Model 97 Stepping 2
Base Frequency4.20 GHz
Cluster 116 Cores
Memory Information
Size64.00 GB

AI Performance​

WorkloadAccuracyScore
Image Classification (SP)100%1937
360.1 IPS
Image Classification (HP)100%175
32.5 IPS
Image Classification (Q)99%4468
833.3 IPS
Image Segmentation (SP)100%2484
40.3 IPS
Image Segmentation (HP)100%285
4.62 IPS
Image Segmentation (Q)98%3234
52.6 IPS
Pose Estimation (SP)100%15490
18.1 IPS
Pose Estimation (HP)100%10206
11.9 IPS
Pose Estimation (Q)95%45213
53.0 IPS
Object Detection (SP)100%2481
196.8 IPS
Object Detection (HP)100%246
19.5 IPS
Object Detection (Q)88%2285
183.2 IPS
Face Detection (SP)100%3846
45.7 IPS
Face Detection (HP)100%361
4.28 IPS
Face Detection (Q)97%5518
65.8 IPS
Depth Estimation (SP)100%10802
83.2 IPS
Depth Estimation (HP)99%1208
9.31 IPS
Depth Estimation (Q)76%12779
103.0 IPS
Style Transfer (SP)100%29293
37.7 IPS
Style Transfer (HP)100%24648
31.7 IPS
Style Transfer (Q)98%41237
53.2 IPS
Image Super-Resolution (SP)100%2921
107.9 IPS
Image Super-Resolution (HP)100%2512
92.7 IPS
Image Super-Resolution (Q)99%3988
147.7 IPS
Text Classification (SP)100%1611
2.15 KIPS
Text Classification (HP)100%681
909.6 IPS
Text Classification (Q)97%432
578.5 IPS
Machine Translation (SP)100%3155
54.3 IPS
Machine Translation (HP)100%716
12.3 IPS
Machine Translation (Q)60%3150
70.3 IPS
 
Zuletzt bearbeitet:
Ich gebe nicht besonders viel auf Geekbench für Vergleiche. Für eine Radeon VII bzw. Vega-GPU bin ich aber froh, dass das DirectML Backend zuverlässig funktioniert wenn es um KI-Anwendungen geht. Zum testen genügt das allemal, mit einer CPU ist das oft extrem mühsam.

Geekbench AI Score​

Single Precision Score 15275
Half Precision Score 16817
Quantized Score 10622
Geekbench AI 1.0.0 for Windows AVX2

System Information​

AI Information
FrameworkONNX
BackendDirectML
DeviceAMD Radeon VII

AI Performance​

WorkloadAccuracyScore
Image Classification (SP)100%6077
1.13 KIPS
Image Classification (HP)100%6661
1.24 KIPS
Image Classification (Q)100%4698
873.7 IPS
Image Segmentation (SP)100%8860
143.6 IPS
Image Segmentation (HP)100%8940
144.9 IPS
Image Segmentation (Q)98%7342
119.4 IPS
Pose Estimation (SP)100%100457
117.2 IPS
Pose Estimation (HP)100%143393
167.3 IPS
Pose Estimation (Q)97%83908
98.3 IPS
Object Detection (SP)100%6773
537.2 IPS
Object Detection (HP)100%6537
518.6 IPS
Object Detection (Q)89%5556
444.9 IPS
Face Detection (SP)100%19858
236.0 IPS
Face Detection (HP)100%20241
240.5 IPS
Face Detection (Q)97%14545
173.4 IPS
Depth Estimation (SP)100%36132
278.4 IPS
Depth Estimation (HP)99%39341
303.1 IPS
Depth Estimation (Q)72%25514
210.1 IPS
Style Transfer (SP)100%186254
239.4 IPS
Style Transfer (HP)100%259970
334.2 IPS
Style Transfer (Q)98%147933
190.8 IPS
Image Super-Resolution (SP)100%23299
860.3 IPS
Image Super-Resolution (HP)100%26922
994.1 IPS
Image Super-Resolution (Q)99%16374
606.3 IPS
Text Classification (SP)100%2047
2.73 KIPS
Text Classification (HP)100%2010
2.68 KIPS
Text Classification (Q)97%1099
1.47 KIPS
Machine Translation (SP)100%2960
51.0 IPS
Machine Translation (HP)100%2893
49.8 IPS
Machine Translation (Q)60%1150
25.9 IPS
Zum Vergrößern anklicken....
 
Zuletzt bearbeitet:

Hier mal meine Ergebnisse mit der GPU​

ASUS System Product Name​

Geekbench AI Score​

Single Precision Score 37292
Half Precision Score 51265
Quantized Score 27889

Geekbench AI 1.0.0 for Windows AVX2

Result Information​


Upload DateAugust 21 2024 09:17 AM
Views1

System Information​


FrameworkONNX
BackendDirectML
DeviceNVIDIA GeForce RTX 4090
AI Information
System Information
Operating SystemMicrosoft Windows 11 Pro (64-bit)
ModelASUS System Product Name
MotherboardASUSTeK COMPUTER INC. TUF GAMING X670E-PLUS
Power PlanAusbalanciert
CPU Information
NameAMD Ryzen 9 7950X3D
Topology1 Processor, 16 Cores, 32 Threads
IdentifierAuthenticAMD Family 25 Model 97 Stepping 2
Base Frequency4.20 GHz
Cluster 116 Cores
Size64.00 GB
Memory Information

AI Performance​


WorkloadAccuracyScore
Image Classification (SP)100%13402
2.49 KIPS
Image Classification (HP)99%17578
3.28 KIPS
Image Classification (Q)100%11246
2.09 KIPS
Image Segmentation (SP)100%23279
377.4 IPS
Image Segmentation (HP)100%29178
473.0 IPS
Image Segmentation (Q)98%21876
355.7 IPS
Pose Estimation (SP)100%305156
356.1 IPS
Pose Estimation (HP)100%539515
629.5 IPS
Pose Estimation (Q)97%266331
311.9 IPS
Object Detection (SP)100%18642
1.48 KIPS
Object Detection (HP)100%24268
1.92 KIPS
Object Detection (Q)89%15811
1.27 KIPS
Face Detection (SP)100%50714
602.6 IPS
Face Detection (HP)100%61899
735.5 IPS
Face Detection (Q)97%42281
504.1 IPS
Depth Estimation (SP)100%92295
711.1 IPS
Depth Estimation (HP)99%155415
1.20 KIPS
Depth Estimation (Q)75%66616
537.8 IPS
Style Transfer (SP)100%562846
723.5 IPS
Style Transfer (HP)100%1088107
1.40 KIPS
Style Transfer (Q)98%483614
623.6 IPS
Image Super-Resolution (SP)100%59358
2.19 KIPS
Image Super-Resolution (HP)100%78609
2.90 KIPS
Image Super-Resolution (Q)99%43695
1.62 KIPS
Text Classification (SP)100%3544
4.73 KIPS
Text Classification (HP)99%4376
5.84 KIPS
Text Classification (Q)97%2152
2.88 KIPS
Machine Translation (SP)100%5289
91.1 IPS
Machine Translation (HP)100%5186
89.3 IPS
Machine Translation (Q)60%2146
47.9 IPS
 
Hier mal von meinem Laptop ...

Geekbench AI Score​


7553
Single Precision Score
10464
Half Precision Score
5483
Quantized Score
Geekbench AI 1.0.0 for Windows AVX2

Result Information​


Upload DateAugust 22 2024 05:55 AM
Views1

System Information​


AI Information
FrameworkONNX
BackendDirectML
DeviceAMD Radeon 780M Graphics
System Information
Operating SystemMicrosoft Windows 11 Pro (64-bit)
ModelASUSTeK COMPUTER INC. ASUS TUF Gaming A17 FA707XI_FA707XI
MotherboardASUSTeK COMPUTER INC. FA707XI
Power PlanPerformance
CPU Information
NameAMD Ryzen 9 7940HS
Topology1 Processor, 8 Cores, 16 Threads
IdentifierAuthenticAMD Family 25 Model 116 Stepping 1
Base Frequency4.00 GHz
Cluster 18 Cores
Memory Information
Size32.00 GB

AI Performance​


WorkloadAccuracyScore
Image Classification (SP)100%2939
546.5 IPS
Image Classification (HP)100%4135
768.9 IPS
Image Classification (Q)100%2476
460.5 IPS
Image Segmentation (SP)100%3939
63.9 IPS
Image Segmentation (HP)100%4577
74.2 IPS
Image Segmentation (Q)98%3200
52.0 IPS
Pose Estimation (SP)100%36118
42.1 IPS
Pose Estimation (HP)100%60191
70.2 IPS
Pose Estimation (Q)96%27994
32.8 IPS
Object Detection (SP)100%2594
205.8 IPS
Object Detection (HP)100%3569
283.1 IPS
Object Detection (Q)89%2252
180.3 IPS
Face Detection (SP)100%5386
64.0 IPS
Face Detection (HP)100%9406
111.8 IPS
Face Detection (Q)97%4083
48.7 IPS
Depth Estimation (SP)100%16520
127.3 IPS
Depth Estimation (HP)99%23977
184.7 IPS
Depth Estimation (Q)73%12227
99.8 IPS
Style Transfer (SP)100%60861
78.2 IPS
Style Transfer (HP)100%98551
126.7 IPS
Style Transfer (Q)98%44710
57.7 IPS
Image Super-Resolution (SP)100%13249
489.2 IPS
Image Super-Resolution (HP)100%18740
692.0 IPS
Image Super-Resolution (Q)99%8866
328.3 IPS
Text Classification (SP)100%2497
3.33 KIPS
Text Classification (HP)100%2777
3.71 KIPS
Text Classification (Q)97%1767
2.37 KIPS
Machine Translation (SP)100%3110
53.6 IPS
Machine Translation (HP)100%3347
57.6 IPS
Machine Translation (Q)60%1407
31.4 IPS

​

​

​

Geekbench AI Score​

17299
Single Precision Score
30179
Half Precision Score
13479
Quantized Score
Geekbench AI 1.0.0 for Windows AVX2

Result Information​

Upload DateAugust 22 2024 06:04 AM
Views1

System Information​

AI Information
FrameworkONNX
BackendDirectML
DeviceNVIDIA GeForce RTX 4070 Laptop GPU
System Information
Operating SystemMicrosoft Windows 11 Pro (64-bit)
ModelASUSTeK COMPUTER INC. ASUS TUF Gaming A17 FA707XI_FA707XI
MotherboardASUSTeK COMPUTER INC. FA707XI
Power PlanPerformance
CPU Information
NameAMD Ryzen 9 7940HS
Topology1 Processor, 8 Cores, 16 Threads
IdentifierAuthenticAMD Family 25 Model 116 Stepping 1
Base Frequency4.00 GHz
Cluster 18 Cores
Memory Information
Size32.00 GB

AI Performance​

WorkloadAccuracyScore
Image Classification (SP)100%7373
1.37 KIPS
Image Classification (HP)99%12380
2.31 KIPS
Image Classification (Q)100%6542
1.22 KIPS
Image Segmentation (SP)100%12957
210.0 IPS
Image Segmentation (HP)100%17077
276.8 IPS
Image Segmentation (Q)98%11681
189.9 IPS
Pose Estimation (SP)100%82271
96.0 IPS
Pose Estimation (HP)100%256564
299.4 IPS
Pose Estimation (Q)97%73573
86.2 IPS
Object Detection (SP)100%8421
668.0 IPS
Object Detection (HP)100%15480
1.23 KIPS
Object Detection (Q)88%7700
617.5 IPS
Face Detection (SP)100%18980
225.5 IPS
Face Detection (HP)100%31452
373.7 IPS
Face Detection (Q)97%15649
186.6 IPS
Depth Estimation (SP)100%38167
294.1 IPS
Depth Estimation (HP)99%80202
617.9 IPS
Depth Estimation (Q)72%31389
259.4 IPS
Style Transfer (SP)100%134643
173.1 IPS
Style Transfer (HP)100%425831
547.4 IPS
Style Transfer (Q)98%141543
182.5 IPS
Image Super-Resolution (SP)100%34696
1.28 KIPS
Image Super-Resolution (HP)100%49607
1.83 KIPS
Image Super-Resolution (Q)99%23122
856.2 IPS
Text Classification (SP)100%2908
3.88 KIPS
Text Classification (HP)99%3614
4.82 KIPS
Text Classification (Q)97%1846
2.47 KIPS
Machine Translation (SP)100%3685
63.5 IPS
Machine Translation (HP)98%3876
67.0 IPS
Machine Translation (Q)60%1541
34.4 IPS

Geekbench AI Score​


2463
Single Precision Score
751
Half Precision Score
3901
Quantized Score
Geekbench AI 1.0.0 for Windows AVX2

Result Information​


Upload DateAugust 22 2024 06:10 AM
Views1

System Information​


AI Information
FrameworkONNX
BackendCPU
DeviceAMD Ryzen 9 7940HS
System Information
Operating SystemMicrosoft Windows 11 Pro (64-bit)
ModelASUSTeK COMPUTER INC. ASUS TUF Gaming A17 FA707XI_FA707XI
MotherboardASUSTeK COMPUTER INC. FA707XI
Power PlanPerformance
CPU Information
NameAMD Ryzen 9 7940HS
Topology1 Processor, 8 Cores, 16 Threads
IdentifierAuthenticAMD Family 25 Model 116 Stepping 1
Base Frequency4.00 GHz
Cluster 18 Cores
Memory Information
Size32.00 GB

AI Performance​


WorkloadAccuracyScore
Image Classification (SP)100%1447
269.1 IPS
Image Classification (HP)100%152
28.4 IPS
Image Classification (Q)99%4066
758.3 IPS
Image Segmentation (SP)100%897
14.5 IPS
Image Segmentation (HP)100%229
3.72 IPS
Image Segmentation (Q)98%2183
35.5 IPS
Pose Estimation (SP)100%4231
4.94 IPS
Pose Estimation (HP)100%3659
4.27 IPS
Pose Estimation (Q)95%18808
22.0 IPS
Object Detection (SP)100%1656
131.4 IPS
Object Detection (HP)100%209
16.6 IPS
Object Detection (Q)88%1716
137.6 IPS
Face Detection (SP)100%2954
35.1 IPS
Face Detection (HP)100%323
3.84 IPS
Face Detection (Q)97%4849
57.8 IPS
Depth Estimation (SP)100%5080
39.1 IPS
Depth Estimation (HP)99%978
7.53 IPS
Depth Estimation (Q)76%9525
76.7 IPS
Style Transfer (SP)100%10394
13.4 IPS
Style Transfer (HP)100%9349
12.0 IPS
Style Transfer (Q)98%23553
30.4 IPS
Image Super-Resolution (SP)100%1841
68.0 IPS
Image Super-Resolution (HP)100%1699
62.7 IPS
Image Super-Resolution (Q)99%3228
119.5 IPS
Text Classification (SP)100%1608
2.15 KIPS
Text Classification (HP)100%671
895.0 IPS
Text Classification (Q)97%411
550.0 IPS
Machine Translation (SP)100%1964
33.8 IPS
Machine Translation (HP)100%638
11.0 IPS
Machine Translation (Q)60%1977
44.1 IPS
[automerge]1724308662[/automerge]

Geekbench AI Score​


4217
Single Precision Score
4248
Half Precision Score
11901
Quantized Score
Geekbench AI 1.0.0 for Windows AVX2

Result Information​


Upload DateAugust 22 2024 06:36 AM
Views1

System Information​


FrameworkOpenVINO
BackendCPU
DeviceAMD Ryzen 9 7940HS w/ Radeon 780M Graphics
AI Information
System Information
Operating SystemMicrosoft Windows 11 Pro (64-bit)
ModelASUSTeK COMPUTER INC. ASUS TUF Gaming A17 FA707XI_FA707XI
MotherboardASUSTeK COMPUTER INC. FA707XI
Power PlanPerformance
CPU Information
NameAMD Ryzen 9 7940HS
Topology1 Processor, 8 Cores, 16 Threads
IdentifierAuthenticAMD Family 25 Model 116 Stepping 1
Base Frequency4.00 GHz
Cluster 18 Cores
Size32.00 GB
Memory Information

AI Performance​


WorkloadAccuracyScore
Image Classification (SP)100%3149
585.7 IPS
Image Classification (HP)100%3159
587.4 IPS
Image Classification (Q)100%8295
1.54 KIPS
Image Segmentation (SP)100%1915
31.0 IPS
Image Segmentation (HP)100%1919
31.1 IPS
Image Segmentation (Q)99%6940
112.5 IPS
Pose Estimation (SP)100%5365
6.26 IPS
Pose Estimation (HP)100%5596
6.53 IPS
Pose Estimation (Q)98%21923
25.7 IPS
Object Detection (SP)100%2400
190.3 IPS
Object Detection (HP)100%2507
198.9 IPS
Object Detection (Q)88%8353
669.6 IPS
Face Detection (SP)100%6717
79.8 IPS
Face Detection (HP)100%7252
86.2 IPS
Face Detection (Q)100%23087
274.3 IPS
Depth Estimation (SP)100%5311
40.9 IPS
Depth Estimation (HP)100%5372
41.4 IPS
Depth Estimation (Q)80%16645
131.9 IPS
Style Transfer (SP)100%14884
19.1 IPS
Style Transfer (HP)100%12930
16.6 IPS
Style Transfer (Q)98%61246
79.0 IPS
Image Super-Resolution (SP)100%3248
119.9 IPS
Image Super-Resolution (HP)100%3591
132.6 IPS
Image Super-Resolution (Q)99%7076
262.0 IPS
Text Classification (SP)100%3336
4.45 KIPS
Text Classification (HP)100%3054
4.08 KIPS
Text Classification (Q)92%5328
7.16 KIPS
Machine Translation (SP)100%3986
68.7 IPS
Machine Translation (HP)100%4079
70.3 IPS
Machine Translation (Q)85%6096
106.6 IPS
 
Zuletzt bearbeitet:

Ähnliche Themen

eratte
AMD Ryzen AI 9 HX3XX/HX4XX Notebook & Mini PC (ZEN5 12/10 Kerne, RDNA 3,5) Review und Diskussionsthread - Links in Post 1
6 7 8
Antworten
197
Aufrufe
12K
eratte
eratte
pipin
CRAY ANNOUNCES SHASTA SOFTWARE TO POWER THE EXASCALE ERA
Antworten
0
Aufrufe
661
pipin
pipin
pipin
HP Powers Breathtaking Gaming Experiences
Antworten
0
Aufrufe
1K
pipin
pipin
pipin
AMD Announces Worldwide Availability of AMD Ryzen™ PRO 3000 Series Processors Designed to Power the Modern Business PC
Antworten
0
Aufrufe
921
pipin
pipin
pipin
2nd Gen AMD EPYC™ Processors Set New Standard for the Modern Datacenter with Record-Breaking Performance and Significant TCO Savings
Antworten
0
Aufrufe
2K
pipin
pipin
Zurück
Oben Unten