Infrastructure Tools
GPU FLOPS Calculator
Add datacenter GPUs and systems to see total TFLOPS by precision, power draw (TDP), and VRAM instantly.
How it works
1 Add GPUs with the + button
2 Adjust quantities
3 View detailed FLOPS, power & VRAM results
GPUs & Systems
266| Name | Peak TFLOPS | Memory | TDP | |
|---|---|---|---|---|
HW Ascend 970 Huawei | 4,000 FP8 | 288 GB | — | |
M MTIA 450 Meta | 7,000 FP8 | 288 GB | 1400 W | |
HW Ascend 960 Huawei | 2,000 FP8 | 288 GB | — | |
QC AI250 Qualcomm | — | 768 GB | — | |
Instinct MI430X AMD · OAM | — | 432 GB | — | |
Instinct MI500 AMD | — | — | — | |
Vera Rubin NVIDIA · Superchip | 35,000 FP8 | 576 GB | — | |
Instinct MI455X AMD · OAM | 20,132.7 FP8 | 432 GB | — | |
Rubin NVIDIA · SXM | 17,500 FP8 | 288 GB | — | |
M MTIA 400 Meta | 6,000 FP8 | 288 GB | 1200 W | |
Maia 200 Microsoft | 5,050 FP8 | 216 GB | 750 W | |
Groq 3 LPU NVIDIA | 1,200 FP8 | — | — | |
M MTIA 300 Meta | 1,200 FP8 | 216 GB | 800 W | |
HW Ascend 950DT Huawei | 1,034 FP8 | 144 GB | — | |
HW Ascend 950PR Huawei | 1,000 FP8 | 128 GB | 900 W | |
HW Atlas 350 Huawei | 804 FP8 | 112 GB | 600 W | |
QC AI200 Qualcomm | — | 768 GB | — | |
Crescent Island Intel | — | 160 GB | — | |
Instinct MI440X AMD · OAM | — | — | — | |
Rubin CPX NVIDIA | — | 128 GB | — | |
TPU 8i Google | — | 288 GB | — | |
TPU 8t Google | — | 216 GB | — | |
Instinct MI355X AMD · OAM | 5,033.2 FP8 | 288 GB | 1400 W | |
GB300 NVIDIA · Grace Blackwell Ultra Desktop Superchip 252GB | 5,000 FP8 | 252 GB | — | |
Instinct MI350X AMD · OAM | 4,614 FP8 | 288 GB | 1000 W | |
TPU v7 Google · 192GB | 4,614 FP8 | 192 GB | — | |
B200 NVIDIA · SXM 180GB | 4,500 FP8 | 180 GB | 1000 W | |
B300 NVIDIA · SXM 288GB | 4,500 FP8 | 288 GB | 1400 W | |
Instinct MI325X AMD · OAM | 2,614.9 FP8 | 256 GB | 1000 W | |
AWS Trainium3 AWS | 2,517 FP8 | 144 GB | — | |
MT MTT S5000 Moore Threads | 1,000 FP8 | 80 GB | — | |
GeForce RTX 5090 NVIDIA · 32GB | 838.2 FP8 | 32 GB | 575 W | |
HW Ascend 910C Huawei | 800 FP16 | 128 GB | 600 W | |
GeForce RTX 5080 NVIDIA · 16GB | 450.2 FP8 | 16 GB | 360 W | |
H20 NVIDIA · 141GB HBM3e | 296 FP8 | 141 GB | 400 W | |
RTX PRO 6000 Blackwell NVIDIA · Workstation Edition | 126 FP32 | 96 GB | 600 W | |
M3 Ultra Apple | 32.8 FP32 | 512 GB | — | |
GB10 Grace Blackwell NVIDIA | 31 FP32 | 128 GB | 180 W | |
M5 Apple · Max | 16.6 FP32 | 128 GB | — | |
M5 Apple · Pro | 8.3 FP32 | 64 GB | — | |
M5 Apple | 4.2 FP32 | 32 GB | — | |
B30A NVIDIA | — | — | — | |
MX C600 MetaX | — | — | — | |
MX C700 MetaX | — | — | — | |
DGX Spark NVIDIA | — | 128 GB | 140 W | |
GB200 NVIDIA · Grace Blackwell Superchip 372GB | 10,000 FP8 | 372 GB | — | |
B100 NVIDIA · SXM 192GB | 3,500 FP8 | 192 GB | 700 W | |
GH200 NVIDIA · 144GB HBM3e | 1,979 FP8 | 144 GB | 1000 W | |
H200 NVIDIA · SXM 141GB | 1,979 FP8 | 141 GB | 700 W | |
Gaudi 3 Intel · 128GB | 1,678 FP8 | 128 GB | 900 W | |
AWS Trainium2 AWS | 1,299 FP8 | 96 GB | — | |
HW Ascend 910B Huawei | 400 FP16 | 64 GB | 400 W | |
BD Kunlun P800 Baidu · Kunlun III | 345 FP16 | — | — | |
H20 NVIDIA · 96GB | 296 FP8 | 96 GB | 400 W | |
L20 NVIDIA · 48GB | 239 FP8 | 48 GB | 275 W | |
GeForce RTX 4080 Super NVIDIA · 16GB | 208.9 FP8 | 16 GB | 320 W | |
MX N260 MetaX | 140 FP16 | 64 GB | 225 W | |
L2 NVIDIA · 24GB | 96.5 FP16 | 24 GB | — | |
H200 NVIDIA · NVL 141GB | 60 FP8 | 141 GB | 600 W | |
RTX 2000 Ada NVIDIA · 16GB | 48 FP8 | 16 GB | 70 W | |
M4 Max Apple | 18.4 FP32 | 128 GB | — | |
MT MTT S80 Moore Threads | 14.7 FP32 | 16 GB | — | |
M4 Pro Apple | 9.2 FP32 | 64 GB | — | |
M4 Apple | 4.3 FP32 | 32 GB | — | |
BR BR110 Biren | — | — | — | |
MT MTT S4000 Moore Threads | — | 48 GB | — | |
TPU v6e Google · 32GB | — | 32 GB | — | |
CS WSE-3 Cerebras | — | — | — | |
Instinct MI300X AMD · 192GB | 2,614.9 FP8 | 192 GB | 750 W | |
GH200 NVIDIA · 96GB HBM3 | 1,979 FP8 | 96 GB | 1000 W | |
Instinct MI300A AMD · 128GB | 1,961.2 FP8 | 128 GB | 760 W | |
Data Center GPU Max 1550 Intel · 128GB | 839 FP16 | 128 GB | 600 W | |
L40S NVIDIA · 48GB | 733 FP8 | 48 GB | 350 W | |
RTX 6000 Ada NVIDIA · 48GB | 728.5 FP8 | 48 GB | 300 W | |
RTX 5000 Ada NVIDIA · 32GB | 522.2 FP8 | 32 GB | 250 W | |
TPU v5p Google · 95GB | 459 FP8 | 95 GB | 450 W | |
Data Center GPU Max 1100 Intel · 48GB | 419.5 FP16 | 48 GB | 300 W | |
L40 NVIDIA · 48GB | 362 FP8 | 48 GB | 300 W | |
MX C500 MetaX | 280 FP16 | 64 GB | 350 W | |
L4 NVIDIA · 24GB | 242.5 FP8 | 24 GB | 72 W | |
GeForce RTX 4070 Ti NVIDIA · 12GB | 160.4 FP8 | 12 GB | 285 W | |
H100 NVIDIA · NVL 94GB | 60 FP8 | 94 GB | 400 W | |
M2 Ultra Apple | 27.2 FP32 | 192 GB | — | |
M3 Max Apple | 16.4 FP32 | 128 GB | — | |
MT MTT S3000 Moore Threads · 32GB | 15.5 FP32 | 32 GB | — | |
M2 Max Apple | 13.6 FP32 | 96 GB | — | |
HW Atlas 200I A2 Huawei · 20 TOPS | 10 FP16 | 12 GB | 25 W | |
M3 Pro Apple | 7.4 FP32 | 36 GB | — | |
M2 Pro Apple | 6.8 FP32 | 32 GB | — | |
HW Atlas 200I A2 Huawei · 8 TOPS | 4 FP16 | 4 GB | 21 W | |
M3 Apple | 3.5 FP32 | 24 GB | — | |
QC Cloud AI 100 Ultra Qualcomm | — | 128 GB | 150 W | |
CB MLU590 Cambricon | — | — | — | |
TPU v5e Google · 16GB | — | 16 GB | — | |
H100 NVIDIA · SXM5 80GB | 1,979 FP8 | 80 GB | 700 W | |
H100 NVIDIA · PCIe 80GB | 1,513 FP8 | 80 GB | 350 W | |
H800 NVIDIA · PCIe 80GB | 1,513 FP8 | 80 GB | 350 W | |
Gaudi 2 Intel · 96GB | 865 FP8 | 96 GB | 600 W | |
GeForce RTX 4090 NVIDIA · 24GB | 330.3 FP8 | 24 GB | 450 W | |
A800 NVIDIA · PCIe 80GB | 312 FP16 | 80 GB | 300 W | |
BR BR100 Biren | 256 FP32 | 64 GB | 550 W | |
GeForce RTX 4080 NVIDIA · 16GB | 195 FP8 | 16 GB | 320 W | |
Instinct MI210 AMD · PCIe | 181 FP16 | 64 GB | 300 W | |
HW Atlas 300I Duo Huawei · 96GB | 140 FP16 | 96 GB | 150 W | |
HW Atlas 300I Duo Huawei · 48GB | 140 FP16 | 48 GB | 150 W | |
Radeon RX 7900 XTX AMD · 24GB | 122.8 FP16 | 24 GB | 355 W | |
Radeon RX 7900 XT AMD · 20GB | 103.2 FP16 | 20 GB | 315 W | |
CB MLU370-X8 Cambricon | 96 FP16 | 48 GB | 250 W | |
A10G NVIDIA · 24GB | 70 FP16 | 24 GB | 300 W | |
HW Atlas 300V Pro Huawei | 70 FP16 | 48 GB | — | |
HW Atlas 300V Huawei | 50 FP16 | 24 GB | 72 W | |
GeForce RTX 3090 Ti NVIDIA · 24GB | 40 FP16 | 24 GB | 450 W | |
Arc A770 Intel · 16GB | 39.4 FP16 | 16 GB | 225 W | |
M1 Ultra Apple | 21.2 FP32 | 128 GB | — | |
Data Center GPU Flex Series Intel · 170 16GB | 16 FP32 | 16 GB | 150 W | |
Data Center GPU Flex Series Intel · 140 12GB | 8 FP32 | 12 GB | 75 W | |
RTX A2000 NVIDIA · 12GB | 8 FP16 | 12 GB | 70 W | |
M2 Apple | 3.6 FP32 | 24 GB | — | |
MI250X AMD · 128GB | 383 FP16 | 128 GB | 560 W | |
Instinct MI250 AMD · 128GB | 362.1 FP16 | 128 GB | 500 W | |
A100 NVIDIA · PCIe 80GB | 312 FP16 | 80 GB | 300 W | |
A30 NVIDIA · 24GB | 165 FP16 | 24 GB | 165 W | |
A40 NVIDIA · 48GB | 149.7 FP16 | 48 GB | 300 W | |
IX BI-V100 Iluvatar CoreX · Tiangai 100 | 147 FP16 | 32 GB | 250 W | |
EF Yunsui T20 Enflame | 134.4 FP16 | 32 GB | 300 W | |
EF Yunsui i20 Enflame | 128 FP16 | 16 GB | 150 W | |
A10 NVIDIA · 24GB | 125 FP16 | 24 GB | 150 W | |
HW Atlas 300I Pro Huawei | 70 FP16 | 24 GB | — | |
GeForce RTX 3080 Ti NVIDIA · 12GB | 34.1 FP16 | 12 GB | 350 W | |
RTX A5000 NVIDIA · 24GB | 27.8 FP16 | 24 GB | 230 W | |
RTX A4500 NVIDIA · 20GB | 23.7 FP16 | 20 GB | 200 W | |
RTX A4000 NVIDIA · 16GB | 19.2 FP16 | 16 GB | 140 W | |
M1 Max Apple | 10.4 FP32 | 64 GB | — | |
M1 Pro Apple | 5.3 FP32 | 32 GB | — | |
CB MLU290-M5 Cambricon | — | — | 350 W | |
A100 NVIDIA · SXM4 40GB | 312 FP16 | 40 GB | 400 W | |
A100 NVIDIA · PCIe 40GB | 312 FP16 | 40 GB | 250 W | |
A100 NVIDIA · SXM4 80GB | 312 FP16 | 80 GB | 400 W | |
Instinct MI100 AMD · 32GB | 184.6 FP16 | 32 GB | 300 W | |
RTX A6000 NVIDIA · 48GB | 154.8 FP16 | 48 GB | 300 W | |
V100S NVIDIA · PCIe 32GB | 130 FP16 | 32 GB | 250 W | |
GeForce RTX 3090 NVIDIA · 24GB | 71 FP16 | 24 GB | 350 W | |
GeForce RTX 3080 NVIDIA · 10GB | 59.5 FP16 | 10 GB | 320 W | |
GeForce RTX 3070 NVIDIA · 8GB | 40.6 FP16 | 8 GB | 220 W | |
M1 Apple | 2.6 FP32 | 16 GB | — | |
TPU v4 Google · 32GB | — | 32 GB | — | |
HW Atlas 650E Huawei 8 GPUs | — | 768 GB | 14.5 kW | |
HW Atlas 950 SuperPoD Huawei 1024 GPUs | — | 96 TB | 100 kW | |
DGX Rubin NVL8 NVIDIA 8 GPUs | — | 2.3 TB | 24 kW | |
Groq 3 LPX NVIDIA 256 GPUs | — | — | — | |
HGX Rubin NVL8 NVIDIA 8 GPUs | — | 2.3 TB | — | |
Helios AMD 72 GPUs | — | 30.4 TB | — | |
TPU 8i Pod Google 1024 GPUs | — | 288 TB | — | |
TPU 8t Superpod Google 9600 GPUs | — | 2,025 TB | — | |
Vera Rubin NVL72 NVIDIA 72 GPUs | — | 20.3 TB | — | |
HW Atlas 800I A3 Huawei 8 GPUs | — | 1 TB | 14.6 kW | |
HW Atlas 800T A3 Huawei 8 GPUs | — | 1 TB | — | |
HW Atlas 900 A3 SuperPoD Huawei 384 GPUs | — | 48 TB | — | |
HP Compute XD690 HPE 8 GPUs | — | 2.1 TB | 14.3 kW | |
DGX B300 NVIDIA 8 GPUs | — | 2.1 TB | 14 kW | |
DGX GB300 NVL72 NVIDIA 72 GPUs | — | 20.3 TB | — | |
DGX SuperPOD GB200 NVIDIA 2304 GPUs | — | — | 3840 kW | |
GB200 NVL32 NVIDIA 32 GPUs | — | — | 53.3 kW | |
GB200 NVL36 NVIDIA 36 GPUs | — | — | 60 kW | |
GB200 NVL576 NVIDIA 576 GPUs | — | — | 960 kW | |
GB200 NVL72 NVIDIA 72 GPUs | — | 13.1 TB | 120 kW | |
HGX B200 8-GPU NVIDIA 8 GPUs | — | 1.4 TB | 10.2 kW | |
HGX B300 8-GPU NVIDIA 8 GPUs | — | 2.1 TB | — | |
Instinct MI355X Platform AMD 8 GPUs | — | 2.2 TB | — | |
HP ProLiant Compute XD685 (HGX B200) HPE 8 GPUs | — | 1.4 TB | — | |
HP ProLiant Compute XD685 (HGX B300) HPE 8 GPUs | — | 2.1 TB | — | |
HP ProLiant Compute XD685 (HGX H200) HPE 8 GPUs | — | 1.1 TB | — | |
HP ProLiant Compute XD685 (Instinct MI355X) HPE 8 GPUs | — | 2.2 TB | — | |
TPU7x (Ironwood) Pod Google 9216 GPUs | — | 1,728 TB | — | |
AWS Trn3 UltraServer AWS 144 GPUs | — | 20.3 TB | — | |
CS CS-3 Cerebras 1 GPU | — | — | 27 kW | |
CS CS-3 Inference Cluster (16-node) Cerebras 16 GPUs | — | — | 475 kW | |
CS CS-3 Inference Cluster (32-node) Cerebras 32 GPUs | — | — | 945 kW | |
CS CS-3 Inference Cluster (64-node) Cerebras 64 GPUs | — | — | 1860 kW | |
CS CS-3 Inference Cluster (8-node) Cerebras 8 GPUs | — | — | 240 kW | |
CS CS-3 Inference Cluster (88-node) Cerebras 88 GPUs | — | — | 2540 kW | |
CS CS-3 Rack Cerebras 2 GPUs | — | — | 54 kW | |
HP Cray XD670 HPE 8 GPUs | — | 1.1 TB | — | |
DGX B200 NVIDIA 8 GPUs | — | 1.4 TB | 14.3 kW | |
DGX H200 NVIDIA 8 GPUs | — | 1.1 TB | 8.5 kW | |
AWS EC2 trn2.48xlarge AWS 16 GPUs | — | 1.5 TB | — | |
GI G593-SD1-LAX3 Gigabyte 8 GPUs | — | 1.1 TB | — | |
SU GPU SuperServer AS -8125GS-TNHR (HGX H100) Supermicro 8 GPUs | — | 640 GB | — | |
SU GPU SuperServer AS -8125GS-TNHR (HGX H200) Supermicro 8 GPUs | — | 1.1 TB | — | |
SU GPU SuperServer SYS-821GE-TNHR Supermicro 8 GPUs | — | 1.1 TB | — | |
Gaudi 3 HLB-325 Baseboard Intel 8 GPUs | — | 1 TB | 7.6 kW | |
HGX B100 8-GPU NVIDIA 8 GPUs | — | 1.5 TB | 5.6 kW | |
HGX H200 NVIDIA 8 GPUs | — | 1.1 TB | 8 kW | |
DE PowerEdge XE9680 Dell 8 GPUs | — | 1.1 TB | — | |
TPU v6e Pod Google 256 GPUs | — | 8 TB | — | |
AWS Trn2 UltraServer AWS 64 GPUs | — | 6 TB | — | |
SU X14 Gaudi 3 AI Training Platform Supermicro 8 GPUs | — | 1 TB | — | |
TPU v5e Pod Google 256 GPUs | — | 4 TB | — | |
TPU v5p Pod Google 8960 GPUs | — | 831.3 TB | — | |
DGX H100 NVIDIA 8 GPUs | — | 640 GB | 8.5 kW | |
HGX H100 4-GPU NVIDIA 4 GPUs | — | 320 GB | 4.2 kW | |
HGX H100 8-GPU NVIDIA 8 GPUs | — | 640 GB | 8 kW | |
IN NF5688M6 Inspur 8 GPUs | — | 640 GB | — | |
CB 玄思1000 (Xuansi 1000) Cambricon 4 GPUs | — | 128 GB | 2.3 kW | |
DGX A100 320GB NVIDIA 8 GPUs | — | 320 GB | 6 kW | |
DGX A100 640GB NVIDIA 8 GPUs | — | 640 GB | 6.5 kW | |
HGX A100 4-GPU NVIDIA 4 GPUs | — | 320 GB | 3.2 kW | |
HGX A100 8-GPU NVIDIA 8 GPUs | — | 640 GB | 6 kW | |
TPU v4 Pod Google 4096 GPUs | — | 128 TB | — | |
PE Altus XE2318GTv2 Penguin Solutions 4 GPUs | — | — | — | |
PE Altus XE4318GTS-DTC Penguin Solutions 8 GPUs | — | — | — | |
PE Altus XE5318GTO Penguin Solutions 8 GPUs | — | — | — | |
PE Altus XE5318GTSv2 Penguin Solutions 8 GPUs | — | — | — | |
IN Artemis II Rack Inventec 72 GPUs | — | — | 136 kW | |
IN Artemis III Rack Inventec 72 GPUs | — | — | 228 kW | |
IN Artemis Rack Inventec 72 GPUs | — | — | 125 kW | |
EF Cloud Blazer POD P2010 Enflame 32 GPUs | — | 1000 GB | — | |
EF Cloud Blazer POD P2110 Enflame 64 GPUs | — | 1000 GB | — | |
D- Corsair Inference Server d-Matrix 8 GPUs | — | 2 TB | — | |
D- Corsair SquadRack d-Matrix 64 GPUs | — | 16 TB | — | |
HP Cray Supercomputing EX254n HPE 8 GPUs | — | 768 GB | — | |
WI ES100G2 Wiwynn 2 GPUs | — | — | — | |
WI ES200G2 Wiwynn 2 GPUs | — | — | — | |
AS ESC N8-E11 ASUS 8 GPUs | — | 1.1 TB | — | |
AS ESC NB8-E11 ASUS 8 GPUs | — | 1.4 TB | — | |
IN GB2281A Ingrasys 8 GPUs | — | — | — | |
IN GB6181I Ingrasys 8 GPUs | — | — | — | |
TE Galaxy Blackhole Tenstorrent 32 GPUs | — | 1000 GB | 12 kW | |
IN IGS-HXR200 Ingrasys 8 GPUs | — | — | — | |
LE Lenovo NVIDIA GB300 NVL72 Lenovo 72 GPUs | — | — | 135 kW | |
MT MTT SGX5000 Moore Threads 8 GPUs | — | — | 11.4 kW | |
HP NVIDIA GB300 NVL72 by HPE HPE 72 GPUs | — | — | 132 kW | |
FU NXT RNGD Server Furiosa 8 GPUs | — | 384 GB | 3 kW | |
IN P5800G7 Inventec 8 GPUs | — | — | — | |
IN P9000AG7 (AC) Inventec 8 GPUs | — | — | — | |
IN P9000G6 (LC) Inventec 8 GPUs | — | — | — | |
IN P9000G7 (AC) Inventec 8 GPUs | — | — | — | |
FU PRIMERGY GX2480 M2s Fujitsu 8 GPUs | — | — | — | |
FU PRIMERGY GX2570 M8s (Air-Cooled) Fujitsu 8 GPUs | — | 1.4 TB | — | |
FU PRIMERGY GX2570 M8s (Liquid-Cooled) Fujitsu 8 GPUs | — | 1.4 TB | — | |
FU PRIMERGY GX2580 M8s Fujitsu 8 GPUs | — | — | — | |
QC QuantaGrid D74F-7U QCT 8 GPUs | — | 1.1 TB | — | |
QC QuantaGrid D75F-7U QCT 8 GPUs | — | 1.1 TB | — | |
QC QuantaGrid D75H-10U QCT 8 GPUs | — | — | — | |
QC QuantaGrid D75L-3U QCT 8 GPUs | — | 2.2 TB | — | |
QC QuantaGrid S74G-2U QCT 1 GPU | — | 96 GB | — | |
KU R480-X8 Kunlunxin 8 GPUs | — | 256 GB | — | |
RE RebelPOD (32-node) Rebellions 256 GPUs | — | 36 TB | 224 kW | |
RE RebelRack Rebellions 32 GPUs | — | 4.5 TB | 28 kW | |
RE RebelServer Rebellions 8 GPUs | — | 1.1 TB | 7 kW | |
PE Relion XE3314GTS Penguin Solutions 4 GPUs | — | — | — | |
PE Relion XE4418GTS-DTC Penguin Solutions 8 GPUs | — | — | — | |
PE Relion XE5318GTO Penguin Solutions 8 GPUs | — | — | — | |
PE Relion XE5318GTS Penguin Solutions 8 GPUs | — | — | — | |
PE Relion XE8318GTS Penguin Solutions 8 GPUs | — | — | — | |
PE Relion XE8418GTS Penguin Solutions 8 GPUs | — | — | — | |
SA SambaRack SN40L-16 SambaNova 16 GPUs | — | 1 TB | 10 kW | |
LE ThinkSystem SR680a V4 Lenovo 8 GPUs | — | — | — | |
LE ThinkSystem SR685a V3 Lenovo 8 GPUs | — | — | — | |
LE ThinkSystem SR780a V3 (HGX B200) Lenovo 8 GPUs | — | — | — | |
LE ThinkSystem SR780a V3 (HGX H100) Lenovo 8 GPUs | — | — | — | |
MX 曦云C500X 光链超节点 (Optical Supernode) MetaX 64 GPUs | — | — | — | |
MX 曦云C500服务器 (C500 Server) MetaX 8 GPUs | — | — | — | |
MX 曦云C550 3D Mesh 超节点 MetaX 64 GPUs | — | — | — | |
MX 曦云C550 Shanghai Cube 液冷整机柜 MetaX 128 GPUs | — | — | — | |
MX 曦云C550服务器 (C550 Server) MetaX 8 GPUs | — | — | — |
Sponsor This Page
Get your GPU cloud service in front of AI engineers using the FLOPS Calculator.
How to calculate TFLOPS
One TFLOPS (teraFLOPS) is one trillion floating-point operations per second. For a single GPU, peak throughput is:
TFLOPS = cores × boost clock (GHz) × 2 (FMA) ÷ 1000
Example: an NVIDIA H100 SXM5 has 16,896 FP32 cores running at 1.98 GHz, giving 16,896 × 1.98 × 2 ÷ 1000 ≈ 67 TFLOPS of FP32 performance. Tensor Core precisions (TF32, FP16, BF16, FP8) run far higher because the tensor units process whole matrix operations per cycle.
To size a cluster, multiply per-GPU TFLOPS by the number of GPUs, then apply a real-world efficiency factor (model FLOPS utilization, typically 40 to 60 percent for large-model training) to the vendor peak. This GPU FLOPS calculator does that math across mixed GPU and system configurations for every precision at once.
About the GPU FLOPS Calculator
This free GPU FLOPS calculator estimates total floating-point performance (TFLOPS by precision), GPU power consumption (TDP in kilowatts), and memory capacity (VRAM) for any combination of datacenter GPUs and pre-configured systems like NVIDIA DGX and GB200 NVL72. All figures use vendor-published peak TFLOPS adjusted by your chosen efficiency factor.