Top AI Hardware & GPUs

GPUs and AI accelerators, with full specs, from the companies building them.

100+ parts · 24 manufacturers
// sort by
NVIDIA B200 NVIDIA · Datacenter · Blackwell
192GB HBM3e
8000 GB/s bandwidth
90 FP32 TFLOPS
1000W TDP
4nm process
Mar 2024 released

NVIDIA Blackwell flagship datacenter GPU for frontier model training

N/A
NVIDIA GB200 NVL72 NVIDIA · Datacenter · Blackwell
192GB HBM3e
8000 GB/s bandwidth
1200W TDP
4nm process
Mar 2024 released

Rack-scale Grace Blackwell superchip for trillion-parameter models

N/A
NVIDIA GB300 NVL72 NVIDIA · Datacenter · Blackwell Ultra
288GB HBM3e
8000 GB/s bandwidth
1400W TDP
4nm process
Jun 2025 released

Rack-scale Blackwell Ultra superchip with expanded memory

N/A
NVIDIA GeForce RTX 4090 NVIDIA · Consumer · Ada Lovelace
24GB GDDR6X
1008 GB/s bandwidth
83 FP32 TFLOPS
450W TDP
4nm process
Oct 2022 released

Flagship consumer GPU for gaming and local AI workloads

~$1599
NVIDIA GeForce RTX 5090 NVIDIA · Consumer · Blackwell
32GB GDDR7
1792 GB/s bandwidth
105 FP32 TFLOPS
575W TDP
4nm process
Jan 2025 released

NVIDIA flagship Blackwell consumer GPU

~$1999
NVIDIA H100 PCIe NVIDIA · Datacenter · Hopper
80GB HBM3
2000 GB/s bandwidth
51 FP32 TFLOPS
350W TDP
4nm process
Sep 2022 released

PCIe-form-factor Hopper datacenter GPU

~$25000
NVIDIA H100 SXM NVIDIA · Datacenter · Hopper
80GB HBM3
3350 GB/s bandwidth
67 FP32 TFLOPS
700W TDP
4nm process
Mar 2022 released

Hopper-architecture datacenter GPU for large-scale AI training

~$30000
NVIDIA H200 NVIDIA · Datacenter · Hopper
141GB HBM3e
4800 GB/s bandwidth
67 FP32 TFLOPS
700W TDP
4nm process
May 2024 released

High-memory Hopper refresh for large-model inference

~$32000
NVIDIA Rubin R100 NVIDIA · Datacenter · Rubin
288GB HBM4
22000 GB/s bandwidth
3nm process
Jun 2026 released

Successor to Blackwell built for agentic-scale AI

N/A
AMD Instinct MI250X AMD · Datacenter · CDNA 2
128GB HBM2e
3200 GB/s bandwidth
47.9 FP32 TFLOPS
560W TDP
6nm process
Nov 2021 released

CDNA 2 datacenter accelerator used in Frontier supercomputer

N/A
AMD Instinct MI300X AMD · Datacenter · CDNA 3
192GB HBM3
5300 GB/s bandwidth
163 FP32 TFLOPS
750W TDP
5nm process
Dec 2023 released

AMD flagship datacenter accelerator for large-model AI training

~$15000
AMD Instinct MI325X AMD · Datacenter · CDNA 3
256GB HBM3e
6000 GB/s bandwidth
163.4 FP32 TFLOPS
1000W TDP
5nm process
Oct 2024 released

High-memory CDNA 3 refresh for large-model inference

~$16000
AMD Instinct MI355X AMD · Datacenter · CDNA 4
288GB HBM3e
8000 GB/s bandwidth
1400W TDP
3nm process
Jun 2025 released

AMD's latest flagship AI training and inference accelerator

N/A
AMD Instinct MI450X AMD · Datacenter · CDNA 5
432GB HBM4
19600 GB/s bandwidth
2nm process
Sep 2026 released

CDNA 5 flagship of the MI400 series, built for the Helios rack

N/A
Google TPU 8i Google · Datacenter · TPU v8
Apr 2026 released

Google's eighth-generation TPU, purpose-built for inference

N/A
Google TPU 8t Google · Datacenter · TPU v8
Apr 2026 released

Google's eighth-generation TPU, split off to focus purely on training

N/A
Intel Data Center GPU Max 1550 Intel · Datacenter · Xe-HPC (Ponte Vecchio)
128GB HBM2e
3200 GB/s bandwidth
52 FP32 TFLOPS
600W TDP
5nm process
May 2023 released

Intel flagship datacenter GPU, used in the Aurora supercomputer

N/A
NVIDIA A100 40GB NVIDIA · Datacenter · Ampere
40GB HBM2
1555 GB/s bandwidth
19.5 FP32 TFLOPS
400W TDP
7nm process
May 2020 released

Original 40GB Ampere datacenter GPU

~$12500
NVIDIA A100 80GB NVIDIA · Datacenter · Ampere
80GB HBM2e
2039 GB/s bandwidth
19.5 FP32 TFLOPS
400W TDP
7nm process
May 2020 released

Ampere-generation datacenter GPU, still widely deployed

~$15000
NVIDIA B100 NVIDIA · Datacenter · Blackwell
192GB HBM3e
8000 GB/s bandwidth
700W TDP
4nm process
Mar 2024 released

Drop-in Blackwell successor to H100 for HGX systems

N/A
NVIDIA GeForce RTX 3090 Ti NVIDIA · Consumer · Ampere
24GB GDDR6X
1008 GB/s bandwidth
40 FP32 TFLOPS
450W TDP
8nm process
Mar 2022 released

Flagship Ampere consumer GPU

~$1999
AMD Instinct MI300A AMD · Datacenter · CDNA 3
128GB HBM3
5300 GB/s bandwidth
122.6 FP32 TFLOPS
550W TDP
5nm process
Dec 2023 released

APU combining CDNA 3 GPU and Zen 4 CPU cores

N/A
AMD Radeon RX 7900 XTX AMD · Consumer · RDNA 3
24GB GDDR6
960 GB/s bandwidth
61 FP32 TFLOPS
355W TDP
5nm process
Dec 2022 released

AMD flagship RDNA 3 consumer GPU

~$999
AMD Radeon RX 9070 XT AMD · Consumer · RDNA 4
16GB GDDR6
640 GB/s bandwidth
48 FP32 TFLOPS
304W TDP
4nm process
Mar 2025 released

AMD flagship RDNA 4 consumer GPU

~$599
Google TPU Ironwood (v7) Google · Datacenter · TPU v7
192GB HBM
7370 GB/s bandwidth
Apr 2025 released

Google's inference-optimized seventh-generation TPU

N/A
Google TPU v5p Google · Accelerator · TPU v5p
95GB HBM
2765 GB/s bandwidth
400W TDP
Dec 2023 released

Google custom AI accelerator for training large models on Cloud TPU pods

N/A
Google TPU v6e (Trillium) Google · Datacenter · TPU v6
32GB HBM
1640 GB/s bandwidth
May 2024 released

Fourfold performance-per-chip jump over TPU v5e

N/A
NVIDIA GeForce RTX 3090 NVIDIA · Consumer · Ampere
24GB GDDR6X
936 GB/s bandwidth
35.6 FP32 TFLOPS
350W TDP
8nm process
Sep 2020 released

Flagship Ampere consumer GPU

~$1499
NVIDIA GeForce RTX 4080 NVIDIA · Consumer · Ada Lovelace
16GB GDDR6X
717 GB/s bandwidth
48.7 FP32 TFLOPS
320W TDP
4nm process
Nov 2022 released

High-end Ada Lovelace consumer GPU

~$1199
NVIDIA GeForce RTX 4080 Super NVIDIA · Consumer · Ada Lovelace
16GB GDDR6X
736 GB/s bandwidth
52 FP32 TFLOPS
320W TDP
4nm process
Jan 2024 released

High-end Ada Lovelace consumer GPU

~$999
NVIDIA GeForce RTX 4080 SUPER NVIDIA · Consumer · Ada Lovelace
16GB GDDR6X
736 GB/s bandwidth
52.2 FP32 TFLOPS
320W TDP
4nm process
Jan 2024 released

Refreshed high-end Ada consumer GPU

~$999
NVIDIA GeForce RTX 5080 NVIDIA · Consumer · Blackwell
16GB GDDR7
960 GB/s bandwidth
56.3 FP32 TFLOPS
360W TDP
4nm process
Jan 2025 released

High-end Blackwell consumer GPU

~$999
NVIDIA H800 NVIDIA · Datacenter · Hopper
80GB HBM3
2000 GB/s bandwidth
67 FP32 TFLOPS
700W TDP
4nm process
Mar 2023 released

China-market Hopper GPU with reduced NVLink bandwidth

N/A
NVIDIA L40S NVIDIA · Datacenter · Ada Lovelace
48GB GDDR6 ECC
864 GB/s bandwidth
91.6 FP32 TFLOPS
350W TDP
4nm process
Aug 2023 released

Ada Lovelace datacenter GPU for inference and graphics workloads

~$8000
NVIDIA RTX 6000 Ada Generation NVIDIA · Workstation · Ada Lovelace
48GB GDDR6 ECC
960 GB/s bandwidth
91 FP32 TFLOPS
300W TDP
4nm process
Dec 2022 released

NVIDIA flagship Ada Lovelace workstation GPU

~$6800
NVIDIA RTX PRO 6000 Blackwell NVIDIA · Workstation · Blackwell
96GB GDDR7
1792 GB/s bandwidth
126 FP32 TFLOPS
600W TDP
4nm process
Apr 2025 released

Flagship Blackwell workstation GPU with 96GB VRAM

~$8565
Qualcomm AI200 Qualcomm · Datacenter · Hexagon NPU
768GB LPDDR
Jan 2026 released

Rack-scale inference accelerator built on the Hexagon NPU

N/A
AMD Radeon PRO W7900 AMD · Workstation · RDNA 3
48GB GDDR6 ECC
864 GB/s bandwidth
61 FP32 TFLOPS
295W TDP
5nm process
Apr 2023 released

AMD flagship RDNA 3 workstation GPU

~$3999
AWS Trainium2 Amazon · Datacenter · Trainium2
96GB HBM3
2900 GB/s bandwidth
Dec 2024 released

AWS's second-generation custom training chip

N/A
Cerebras WSE-3 Cerebras Systems · Datacenter · Wafer-Scale Engine 3
44GB SRAM
21000 GB/s bandwidth
23000W TDP
5nm process
Mar 2024 released

Largest chip ever built, with 4 trillion transistors

N/A
Google TPU v5e Google · Accelerator · TPU v5e
16GB HBM2
819 GB/s bandwidth
Aug 2023 released

Cost-efficient TPU generation for training and inference

N/A
Groq LPU Groq · Accelerator · GroqChip 1
0.23GB SRAM
80000 GB/s bandwidth
Jan 2023 released

Language Processing Unit purpose-built for low-latency LLM inference

N/A
Intel Gaudi 2 Intel · Datacenter · Gaudi 2
96GB HBM2e
2450 GB/s bandwidth
600W TDP
7nm process
May 2022 released

Second-generation Habana Labs training accelerator

N/A
Intel Gaudi 3 Intel · Accelerator · Gaudi 3
128GB HBM2e
3700 GB/s bandwidth
900W TDP
5nm process
Apr 2024 released

Intel third-generation AI training and inference accelerator

N/A
NVIDIA A800 NVIDIA · Datacenter · Ampere
80GB HBM2e
1935 GB/s bandwidth
19.5 FP32 TFLOPS
400W TDP
7nm process
Dec 2022 released

China-market Ampere GPU with reduced NVLink bandwidth

N/A
NVIDIA GeForce RTX 2080 Ti NVIDIA · Consumer · Turing
11GB GDDR6
616 GB/s bandwidth
13.4 FP32 TFLOPS
250W TDP
12nm process
Sep 2018 released

Flagship Turing consumer GPU

~$999
NVIDIA GeForce RTX 3080 NVIDIA · Consumer · Ampere
10GB GDDR6X
760 GB/s bandwidth
29.8 FP32 TFLOPS
320W TDP
8nm process
Sep 2020 released

High-end Ampere consumer GPU

~$699
NVIDIA GeForce RTX 3080 Ti NVIDIA · Consumer · Ampere
12GB GDDR6X
912 GB/s bandwidth
34.1 FP32 TFLOPS
350W TDP
8nm process
Jun 2021 released

High-end Ampere consumer GPU

~$1199
NVIDIA GeForce RTX 4070 SUPER NVIDIA · Consumer · Ada Lovelace
12GB GDDR6X
504 GB/s bandwidth
35.5 FP32 TFLOPS
220W TDP
4nm process
Jan 2024 released

Refreshed midrange Ada consumer GPU

~$599
NVIDIA GeForce RTX 4070 Ti NVIDIA · Consumer · Ada Lovelace
12GB GDDR6X
504 GB/s bandwidth
40.1 FP32 TFLOPS
285W TDP
4nm process
Jan 2023 released

Upper-midrange Ada consumer GPU

~$799
NVIDIA GeForce RTX 4070 Ti Super NVIDIA · Consumer · Ada Lovelace
16GB GDDR6X
672 GB/s bandwidth
44 FP32 TFLOPS
285W TDP
4nm process
Jan 2024 released

Upper-mid Ada Lovelace consumer GPU

~$799
NVIDIA GeForce RTX 5070 NVIDIA · Consumer · Blackwell
12GB GDDR7
672 GB/s bandwidth
30.9 FP32 TFLOPS
250W TDP
4nm process
Mar 2025 released

Midrange Blackwell consumer GPU

~$549
NVIDIA GeForce RTX 5070 Ti NVIDIA · Consumer · Blackwell
16GB GDDR7
896 GB/s bandwidth
44.2 FP32 TFLOPS
300W TDP
4nm process
Feb 2025 released

Upper-midrange Blackwell consumer GPU

~$749
NVIDIA GeForce RTX 5090 Laptop GPU NVIDIA · Consumer · Blackwell
24GB GDDR7
896 GB/s bandwidth
175W TDP
4nm process
Mar 2025 released

Top-end mobile Blackwell GPU

N/A
NVIDIA H20 NVIDIA · Datacenter · Hopper
96GB HBM3
4000 GB/s bandwidth
400W TDP
4nm process
Mar 2024 released

Export-compliant Hopper GPU for the China market

N/A
NVIDIA L4 NVIDIA · Datacenter · Ada Lovelace
24GB GDDR6
300 GB/s bandwidth
30.3 FP32 TFLOPS
72W TDP
4nm process
Mar 2023 released

Low-power Ada Lovelace inference GPU for cloud servers

~$2500
NVIDIA RTX 5000 Ada Generation NVIDIA · Workstation · Ada Lovelace
32GB GDDR6
576 GB/s bandwidth
65.3 FP32 TFLOPS
250W TDP
4nm process
Apr 2023 released

High-end Ada Lovelace workstation GPU

~$4000
NVIDIA RTX A6000 NVIDIA · Workstation · Ampere
48GB GDDR6 ECC
768 GB/s bandwidth
38 FP32 TFLOPS
300W TDP
8nm process
Dec 2020 released

Ampere-generation workstation GPU

~$4650
NVIDIA Tesla T4 NVIDIA · Datacenter · Turing
16GB GDDR6
320 GB/s bandwidth
8.1 FP32 TFLOPS
70W TDP
12nm process
Sep 2018 released

Ubiquitous low-power inference GPU

~$2500
NVIDIA V100 SXM2 NVIDIA · Datacenter · Volta
32GB HBM2
900 GB/s bandwidth
15.7 FP32 TFLOPS
300W TDP
12nm process
Dec 2017 released

First-generation Tensor Core datacenter GPU

~$10000
AMD Instinct MI210 AMD · Datacenter · CDNA 2
64GB HBM2e
1638 GB/s bandwidth
22.6 FP32 TFLOPS
300W TDP
6nm process
Mar 2022 released

Single-die PCIe CDNA 2 accelerator

N/A
AMD Instinct MI250 AMD · Datacenter · CDNA 2
128GB HBM2e
3277 GB/s bandwidth
45.3 FP32 TFLOPS
560W TDP
6nm process
Nov 2021 released

Dual-die CDNA 2 accelerator, cut-down MI250X

N/A
AMD Radeon Pro W7800 AMD · Workstation · RDNA 3
32GB GDDR6
576 GB/s bandwidth
45.2 FP32 TFLOPS
260W TDP
5nm process
Jun 2023 released

High-end RDNA 3 workstation GPU

~$2499
AMD Radeon RX 6800 XT AMD · Consumer · RDNA 2
16GB GDDR6
512 GB/s bandwidth
20.7 FP32 TFLOPS
300W TDP
7nm process
Nov 2020 released

High-end RDNA 2 consumer GPU

~$649
AMD Radeon RX 6900 XT AMD · Consumer · RDNA 2
16GB GDDR6
512 GB/s bandwidth
23 FP32 TFLOPS
300W TDP
7nm process
Dec 2020 released

Flagship RDNA 2 consumer GPU

~$999
AMD Radeon RX 6950 XT AMD · Consumer · RDNA 2
16GB GDDR6
576 GB/s bandwidth
23.7 FP32 TFLOPS
335W TDP
6nm process
May 2022 released

Refreshed flagship RDNA 2 consumer GPU

~$1099
AMD Radeon RX 7800 XT AMD · Consumer · RDNA 3
16GB GDDR6
624 GB/s bandwidth
37 FP32 TFLOPS
263W TDP
5nm process
Sep 2023 released

Upper-mid RDNA 3 consumer GPU

~$499
AMD Radeon RX 7900 XT AMD · Consumer · RDNA 3
20GB GDDR6
800 GB/s bandwidth
51.6 FP32 TFLOPS
300W TDP
5nm process
Dec 2022 released

Upper-high-end RDNA 3 consumer GPU

~$899
AMD Radeon RX 9070 AMD · Consumer · RDNA 4
16GB GDDR6
650 GB/s bandwidth
36.1 FP32 TFLOPS
220W TDP
4nm process
Mar 2025 released

High-end RDNA 4 consumer GPU

~$549
Apple M3 Ultra Apple · Consumer · Apple Silicon
512GB LPDDR5X (unified)
819 GB/s bandwidth
3nm process
Mar 2025 released

Dual-die Apple Silicon flagship with up to 512GB unified memory

N/A
Apple M4 Max Apple · Consumer · Apple Silicon
128GB LPDDR5X (unified)
546 GB/s bandwidth
3nm process
Oct 2024 released

Apple's highest-end SoC GPU, popular for local LLM inference

N/A
Cerebras WSE-2 Cerebras Systems · Datacenter · Wafer-Scale Engine 2
40GB SRAM
20000 GB/s bandwidth
15000W TDP
7nm process
Apr 2021 released

Second-generation wafer-scale AI chip

N/A
Google TPU v4 Google · Datacenter · TPU v4
32GB HBM2
1200 GB/s bandwidth
May 2021 released

Optically-reconfigurable TPU pod generation

N/A
Huawei Ascend 910B Huawei · Datacenter · Da Vinci
64GB HBM2e
819 GB/s bandwidth
400W TDP
7nm process
Aug 2023 released

China's leading domestic AI training accelerator

N/A
Intel Arc A770 Intel · Consumer · Alchemist
16GB GDDR6
560 GB/s bandwidth
19 FP32 TFLOPS
225W TDP
6nm process
Oct 2022 released

Intel flagship consumer GPU from its first Arc generation

~$349
NVIDIA A10 NVIDIA · Datacenter · Ampere
24GB GDDR6
600 GB/s bandwidth
31.2 FP32 TFLOPS
150W TDP
8nm process
Apr 2021 released

Ampere GPU for mixed graphics and AI inference

~$2500
NVIDIA A10G NVIDIA · Datacenter · Ampere
24GB GDDR6
600 GB/s bandwidth
35 FP32 TFLOPS
300W TDP
8nm process
Nov 2021 released

AWS-favored Ampere GPU for graphics and inference

N/A
NVIDIA A40 NVIDIA · Datacenter · Ampere
48GB GDDR6
696 GB/s bandwidth
37.4 FP32 TFLOPS
300W TDP
8nm process
Oct 2020 released

Ampere data center GPU for graphics and inference

~$5000
NVIDIA GeForce GTX 1080 Ti NVIDIA · Consumer · Pascal
11GB GDDR5X
484 GB/s bandwidth
11.3 FP32 TFLOPS
250W TDP
16nm process
Mar 2017 released

Flagship Pascal-era consumer GPU

~$699
NVIDIA GeForce RTX 2070 SUPER NVIDIA · Consumer · Turing
8GB GDDR6
448 GB/s bandwidth
9.1 FP32 TFLOPS
215W TDP
12nm process
Jul 2019 released

Refreshed midrange Turing consumer GPU

~$499
NVIDIA GeForce RTX 2080 SUPER NVIDIA · Consumer · Turing
8GB GDDR6
496 GB/s bandwidth
11.2 FP32 TFLOPS
250W TDP
12nm process
Jul 2019 released

Refreshed high-end Turing consumer GPU

~$699
NVIDIA GeForce RTX 3060 NVIDIA · Consumer · Ampere
12GB GDDR6
360 GB/s bandwidth
12.7 FP32 TFLOPS
170W TDP
8nm process
Feb 2021 released

Mainstream Ampere consumer GPU

~$329
NVIDIA GeForce RTX 3060 Ti NVIDIA · Consumer · Ampere
8GB GDDR6
448 GB/s bandwidth
16.2 FP32 TFLOPS
200W TDP
8nm process
Dec 2020 released

Popular midrange Ampere consumer GPU

~$399
NVIDIA GeForce RTX 3070 NVIDIA · Consumer · Ampere
8GB GDDR6
448 GB/s bandwidth
20.3 FP32 TFLOPS
220W TDP
8nm process
Oct 2020 released

Midrange Ampere consumer GPU

~$499
NVIDIA GeForce RTX 3070 Ti NVIDIA · Consumer · Ampere
8GB GDDR6X
608 GB/s bandwidth
21.7 FP32 TFLOPS
290W TDP
8nm process
Jun 2021 released

Midrange Ampere consumer GPU

~$599
NVIDIA GeForce RTX 3080 Laptop GPU NVIDIA · Consumer · Ampere
16GB GDDR6
448 GB/s bandwidth
150W TDP
8nm process
Jan 2021 released

High-end mobile Ampere GPU

N/A
NVIDIA GeForce RTX 4060 NVIDIA · Consumer · Ada Lovelace
8GB GDDR6
272 GB/s bandwidth
15.1 FP32 TFLOPS
115W TDP
4nm process
Jun 2023 released

Mainstream Ada Lovelace consumer GPU

~$299
NVIDIA GeForce RTX 4060 Ti NVIDIA · Consumer · Ada Lovelace
16GB GDDR6
288 GB/s bandwidth
22 FP32 TFLOPS
165W TDP
5nm process
May 2023 released

Mainstream Ada Lovelace consumer GPU

~$399
NVIDIA GeForce RTX 4070 NVIDIA · Consumer · Ada Lovelace
12GB GDDR6X
504 GB/s bandwidth
29.1 FP32 TFLOPS
200W TDP
4nm process
Apr 2023 released

Midrange Ada Lovelace consumer GPU

~$599
NVIDIA GeForce RTX 4070 Laptop GPU NVIDIA · Consumer · Ada Lovelace
8GB GDDR6
256 GB/s bandwidth
140W TDP
4nm process
Feb 2023 released

Midrange mobile Ada Lovelace GPU

N/A
NVIDIA GeForce RTX 4080 Laptop GPU NVIDIA · Consumer · Ada Lovelace
12GB GDDR6
432 GB/s bandwidth
175W TDP
4nm process
Feb 2023 released

High-end mobile Ada Lovelace GPU

N/A
NVIDIA GeForce RTX 4090 Laptop GPU NVIDIA · Mobile · Ada Lovelace
16GB GDDR6
576 GB/s bandwidth
39 FP32 TFLOPS
150W TDP
4nm process
Feb 2023 released

Top-tier Ada Lovelace mobile GPU for laptops

N/A
NVIDIA GeForce RTX 5060 NVIDIA · Consumer · Blackwell
8GB GDDR7
448 GB/s bandwidth
19.2 FP32 TFLOPS
145W TDP
4nm process
May 2025 released

Mainstream Blackwell consumer GPU

~$299
NVIDIA GeForce RTX 5060 Ti NVIDIA · Consumer · Blackwell
16GB GDDR7
448 GB/s bandwidth
23.7 FP32 TFLOPS
180W TDP
4nm process
Apr 2025 released

Entry-flagship Blackwell consumer GPU

~$429
NVIDIA GeForce RTX 5080 Laptop GPU NVIDIA · Consumer · Blackwell
16GB GDDR7
896 GB/s bandwidth
175W TDP
4nm process
Mar 2025 released

High-end mobile Blackwell GPU

N/A
NVIDIA Jetson AGX Orin NVIDIA · Edge · Ampere
64GB LPDDR5
205 GB/s bandwidth
5.3 FP32 TFLOPS
60W TDP
8nm process
Mar 2022 released

Flagship edge AI module for robotics and autonomous machines

~$1999
NVIDIA Quadro RTX 8000 NVIDIA · Workstation · Turing
48GB GDDR6
624 GB/s bandwidth
16.3 FP32 TFLOPS
295W TDP
12nm process
Aug 2018 released

Flagship Turing-era workstation GPU

~$5500
NVIDIA RTX 4000 Ada Generation NVIDIA · Workstation · Ada Lovelace
20GB GDDR6
360 GB/s bandwidth
26.7 FP32 TFLOPS
130W TDP
4nm process
Apr 2023 released

Midrange Ada Lovelace workstation GPU

~$1250
NVIDIA RTX 4500 Ada Generation NVIDIA · Workstation · Ada Lovelace
24GB GDDR6
432 GB/s bandwidth
39.6 FP32 TFLOPS
210W TDP
4nm process
Apr 2023 released

Upper-midrange Ada Lovelace workstation GPU

~$2250
NVIDIA RTX A5000 NVIDIA · Workstation · Ampere
24GB GDDR6
768 GB/s bandwidth
27.8 FP32 TFLOPS
230W TDP
8nm process
Apr 2021 released

High-end Ampere workstation GPU

~$2250
NVIDIA TITAN RTX NVIDIA · Consumer · Turing
24GB GDDR6
672 GB/s bandwidth
16.3 FP32 TFLOPS
280W TDP
12nm process
Dec 2018 released

Prosumer Turing card with 24GB VRAM

~$2499
NVIDIA V100 32GB PCIe NVIDIA · Datacenter · Volta
32GB HBM2
900 GB/s bandwidth
14 FP32 TFLOPS
250W TDP
12nm process
Mar 2018 released

PCIe-form-factor Volta datacenter GPU

~$8500
AMD Instinct MI100 AMD · Datacenter · CDNA
32GB HBM2
1229 GB/s bandwidth
23.1 FP32 TFLOPS
300W TDP
7nm process
Nov 2020 released

First-generation CDNA datacenter accelerator

N/A
AMD Radeon Pro W6800 AMD · Workstation · RDNA 2
32GB GDDR6
512 GB/s bandwidth
17.8 FP32 TFLOPS
250W TDP
7nm process
Apr 2021 released

High-end RDNA 2 workstation GPU

~$2249
AMD Radeon Pro W7700 AMD · Workstation · RDNA 3
16GB GDDR6
432 GB/s bandwidth
29.3 FP32 TFLOPS
190W TDP
5nm process
Oct 2023 released

Midrange RDNA 3 workstation GPU

~$999
AMD Radeon RX 5700 XT AMD · Consumer · RDNA
8GB GDDR6
448 GB/s bandwidth
9.75 FP32 TFLOPS
225W TDP
7nm process
Jul 2019 released

First-generation RDNA flagship consumer GPU

~$399
AMD Radeon RX 6700 XT AMD · Consumer · RDNA 2
12GB GDDR6
384 GB/s bandwidth
13.2 FP32 TFLOPS
230W TDP
7nm process
Mar 2021 released

Midrange RDNA 2 consumer GPU

~$479
AMD Radeon RX 6800 AMD · Consumer · RDNA 2
16GB GDDR6
512 GB/s bandwidth
16.2 FP32 TFLOPS
250W TDP
7nm process
Nov 2020 released

High-end RDNA 2 consumer GPU

~$579
AMD Radeon RX 7600 AMD · Consumer · RDNA 3
8GB GDDR6
288 GB/s bandwidth
21.5 FP32 TFLOPS
165W TDP
6nm process
May 2023 released

Mainstream RDNA 3 consumer GPU

~$269
AMD Radeon RX 7700 XT AMD · Consumer · RDNA 3
12GB GDDR6
432 GB/s bandwidth
35.2 FP32 TFLOPS
245W TDP
5nm process
Sep 2023 released

Midrange RDNA 3 consumer GPU

~$449
AMD Radeon RX 9060 XT AMD · Consumer · RDNA 4
16GB GDDR6
322 GB/s bandwidth
25.6 FP32 TFLOPS
180W TDP
4nm process
Jun 2025 released

Mainstream RDNA 4 consumer GPU

~$349
AMD Radeon VII AMD · Consumer · GCN 5.1 (Vega 20)
16GB HBM2
1024 GB/s bandwidth
13.8 FP32 TFLOPS
300W TDP
7nm process
Feb 2019 released

First consumer 7nm GPU, with 16GB HBM2

~$699
Apple M2 Ultra Apple · Consumer · Apple Silicon
192GB LPDDR5 (unified)
800 GB/s bandwidth
5nm process
Jun 2023 released

Dual-die Apple Silicon chip with up to 192GB unified memory

N/A
AWS Inferentia2 Amazon · Datacenter · Inferentia2
32GB HBM2e
820 GB/s bandwidth
Apr 2023 released

AWS's second-generation inference chip

N/A
AWS Trainium Amazon · Datacenter · Trainium
32GB HBM2e
820 GB/s bandwidth
Dec 2021 released

AWS's first purpose-built training chip

N/A
Groq GroqCard Groq · Datacenter · LPU v1
0.23GB SRAM
80000 GB/s bandwidth
300W TDP
14nm process
Nov 2020 released

Deterministic single-core language processing unit card

N/A
Intel Arc A750 Intel · Consumer · Xe-HPG (Alchemist)
8GB GDDR6
512 GB/s bandwidth
17.2 FP32 TFLOPS
225W TDP
6nm process
Oct 2022 released

Midrange Alchemist consumer GPU

~$249
Intel Arc B570 Intel · Consumer · Xe2 (Battlemage)
10GB GDDR6
380 GB/s bandwidth
11.5 FP32 TFLOPS
150W TDP
5nm process
Jan 2025 released

Entry Battlemage consumer GPU

~$219
Intel Arc B580 Intel · Consumer · Battlemage
12GB GDDR6
456 GB/s bandwidth
13.7 FP32 TFLOPS
190W TDP
5nm process
Dec 2024 released

Intel second-generation Battlemage consumer GPU

~$249
Intel Data Center GPU Max 1100 Intel · Datacenter · Xe-HPC
48GB HBM2e
1229 GB/s bandwidth
22.2 FP32 TFLOPS
300W TDP
5nm process
May 2023 released

PCIe variant of Intel's Ponte Vecchio datacenter GPU

N/A
Meta MTIA v2 Meta AI · Datacenter · MTIA v2
128GB LPDDR5
204 GB/s bandwidth
90W TDP
5nm process
Apr 2024 released

Meta's in-house recommendation and ranking inference chip

N/A
Microsoft Maia 100 Microsoft · Datacenter · Maia
64GB HBM2e
1800 GB/s bandwidth
500W TDP
5nm process
Nov 2023 released

Microsoft's first custom AI accelerator for Azure

N/A
NVIDIA A30 NVIDIA · Datacenter · Ampere
24GB HBM2
933 GB/s bandwidth
10.3 FP32 TFLOPS
165W TDP
7nm process
Apr 2021 released

Mainstream Ampere inference and analytics GPU

~$4500
NVIDIA GeForce GTX 1060 NVIDIA · Consumer · Pascal
6GB GDDR5
192 GB/s bandwidth
4.4 FP32 TFLOPS
120W TDP
16nm process
Jul 2016 released

One of the most widely used Pascal-era GPUs

~$249
NVIDIA GeForce GTX 1070 NVIDIA · Consumer · Pascal
8GB GDDR5
256 GB/s bandwidth
6.5 FP32 TFLOPS
150W TDP
16nm process
Jun 2016 released

Popular midrange Pascal-era GPU

~$379
NVIDIA GeForce GTX 1070 Ti NVIDIA · Consumer · Pascal
8GB GDDR5
256 GB/s bandwidth
8.2 FP32 TFLOPS
180W TDP
16nm process
Nov 2017 released

Upper-midrange Pascal-era GPU

~$449
NVIDIA GeForce GTX 1080 NVIDIA · Consumer · Pascal
8GB GDDR5X
320 GB/s bandwidth
8.9 FP32 TFLOPS
180W TDP
16nm process
May 2016 released

High-end Pascal-era consumer GPU

~$599
NVIDIA GeForce GTX 1660 Ti NVIDIA · Consumer · Turing
6GB GDDR6
288 GB/s bandwidth
5.4 FP32 TFLOPS
120W TDP
12nm process
Feb 2019 released

Budget Turing GPU without RT/Tensor cores

~$279
NVIDIA GeForce RTX 2060 NVIDIA · Consumer · Turing
6GB GDDR6
336 GB/s bandwidth
6.5 FP32 TFLOPS
160W TDP
12nm process
Jan 2019 released

Mainstream Turing consumer GPU

~$349
NVIDIA GeForce RTX 2060 SUPER NVIDIA · Consumer · Turing
8GB GDDR6
448 GB/s bandwidth
7.2 FP32 TFLOPS
175W TDP
12nm process
Jul 2019 released

Midrange Turing consumer GPU

~$399
NVIDIA GeForce RTX 2070 NVIDIA · Consumer · Turing
8GB GDDR6
448 GB/s bandwidth
7.5 FP32 TFLOPS
175W TDP
12nm process
Oct 2018 released

Midrange Turing consumer GPU

~$499
NVIDIA GeForce RTX 3050 NVIDIA · Consumer · Ampere
8GB GDDR6
224 GB/s bandwidth
9.1 FP32 TFLOPS
130W TDP
8nm process
Jan 2022 released

Entry-level Ampere consumer GPU

~$249
NVIDIA GeForce RTX 3070 Laptop GPU NVIDIA · Consumer · Ampere
8GB GDDR6
384 GB/s bandwidth
130W TDP
8nm process
Jan 2021 released

Midrange mobile Ampere GPU

N/A
NVIDIA GeForce RTX 4060 Laptop GPU NVIDIA · Consumer · Ada Lovelace
8GB GDDR6
256 GB/s bandwidth
115W TDP
4nm process
Feb 2023 released

Mainstream mobile Ada Lovelace GPU

N/A
NVIDIA Jetson Orin Nano NVIDIA · Edge · Ampere
8GB LPDDR5
102 GB/s bandwidth
0.65 FP32 TFLOPS
15W TDP
8nm process
Mar 2023 released

Compact edge AI module for makers and light robotics

~$499
NVIDIA Quadro RTX 6000 NVIDIA · Workstation · Turing
24GB GDDR6
624 GB/s bandwidth
16.3 FP32 TFLOPS
295W TDP
12nm process
Aug 2018 released

High-end Turing-era workstation GPU

~$4000
NVIDIA RTX 2000 Ada Generation NVIDIA · Workstation · Ada Lovelace
16GB GDDR6
224 GB/s bandwidth
12 FP32 TFLOPS
70W TDP
4nm process
Oct 2023 released

Compact entry Ada Lovelace workstation GPU

~$625
NVIDIA RTX A2000 NVIDIA · Workstation · Ampere
12GB GDDR6
288 GB/s bandwidth
8 FP32 TFLOPS
70W TDP
8nm process
Mar 2022 released

Compact entry Ampere workstation GPU

~$450
NVIDIA RTX A4000 NVIDIA · Workstation · Ampere
16GB GDDR6
448 GB/s bandwidth
19.2 FP32 TFLOPS
140W TDP
8nm process
Apr 2021 released

Midrange Ampere workstation GPU

~$1000
NVIDIA RTX A4500 NVIDIA · Workstation · Ampere
20GB GDDR6
640 GB/s bandwidth
23.7 FP32 TFLOPS
200W TDP
8nm process
Apr 2021 released

Upper-midrange Ampere workstation GPU

~$2250
NVIDIA Tesla P100 NVIDIA · Datacenter · Pascal
16GB HBM2
732 GB/s bandwidth
10.6 FP32 TFLOPS
300W TDP
16nm process
Jun 2016 released

First HBM2 datacenter GPU, pre-dates Tensor Cores

~$9000
NVIDIA TITAN V NVIDIA · Consumer · Volta
12GB HBM2
653 GB/s bandwidth
14.9 FP32 TFLOPS
250W TDP
12nm process
Dec 2017 released

Prosumer Volta card, first consumer GPU with Tensor Cores

~$2999
SambaNova SN40L SambaNova Systems · Datacenter · SN40L RDU
64GB HBM3
5nm process
Mar 2024 released

Reconfigurable dataflow chip for trillion-parameter inference

N/A
AMD Radeon Instinct MI60 AMD · Datacenter · GCN 5.1 (Vega 20)
32GB HBM2
1024 GB/s bandwidth
14.7 FP32 TFLOPS
300W TDP
7nm process
Nov 2018 released

Vega-based datacenter accelerator, AMD's first 7nm GPU

~$6500
AMD Radeon Pro W6600 AMD · Workstation · RDNA 2
8GB GDDR6
224 GB/s bandwidth
10.3 FP32 TFLOPS
100W TDP
7nm process
Oct 2021 released

Compact midrange RDNA 2 workstation GPU

~$649
AMD Radeon RX 5600 XT AMD · Consumer · RDNA
6GB GDDR6
336 GB/s bandwidth
7.19 FP32 TFLOPS
150W TDP
7nm process
Jan 2020 released

Midrange first-generation RDNA GPU

~$279
AMD Radeon RX 6600 AMD · Consumer · RDNA 2
8GB GDDR6
224 GB/s bandwidth
8.9 FP32 TFLOPS
132W TDP
7nm process
Oct 2021 released

Mainstream RDNA 2 consumer GPU

~$329
AMD Radeon RX 6600 XT AMD · Consumer · RDNA 2
8GB GDDR6
256 GB/s bandwidth
10.6 FP32 TFLOPS
160W TDP
7nm process
Aug 2021 released

Mainstream RDNA 2 consumer GPU

~$379
AMD Radeon RX Vega 64 AMD · Consumer · GCN 5 (Vega 10)
8GB HBM2
484 GB/s bandwidth
12.7 FP32 TFLOPS
295W TDP
14nm process
Aug 2017 released

Flagship Vega-generation consumer GPU with HBM2

~$499
AWS Inferentia Amazon · Datacenter · Inferentia
8GB DDR4
Dec 2019 released

AWS's first custom inference chip

N/A
Biren BR100 Biren Technology · Datacenter · BR100
64GB HBM2e
1638 GB/s bandwidth
550W TDP
7nm process
Aug 2022 released

Chinese chiplet-based general-purpose GPU for AI training

N/A
Etched Sohu Etched · Datacenter · Transformer ASIC
144GB HBM3e
4nm process
Jan 2025 released

ASIC hardwired to run only transformer models

N/A
Google TPU v3 Google · Datacenter · TPU v3
32GB HBM
900 GB/s bandwidth
May 2018 released

Liquid-cooled third-generation TPU

N/A
Graphcore Bow IPU Graphcore · Datacenter · Bow
0.9GB SRAM
65000 GB/s bandwidth
300W TDP
7nm process
Mar 2022 released

First wafer-on-wafer AI processor

N/A
Graphcore GC200 IPU Graphcore · Datacenter · Colossus Mk2
0.9GB SRAM
47500 GB/s bandwidth
300W TDP
7nm process
Jul 2020 released

Second-generation Intelligence Processing Unit

N/A
Huawei Ascend 910 Huawei · Datacenter · Da Vinci
32GB HBM2
1200 GB/s bandwidth
310W TDP
7nm process
Sep 2019 released

Huawei's first Ascend AI training chip

N/A
Intel Arc A580 Intel · Consumer · Xe-HPG (Alchemist)
8GB GDDR6
512 GB/s bandwidth
12.3 FP32 TFLOPS
175W TDP
6nm process
Oct 2023 released

Value Alchemist consumer GPU

~$179
Intel Data Center GPU Flex 170 Intel · Datacenter · Xe-HPG
16GB GDDR6
560 GB/s bandwidth
17 FP32 TFLOPS
150W TDP
6nm process
Aug 2022 released

Media transcode and inference GPU for cloud workloads

N/A
NVIDIA A16 NVIDIA · Datacenter · Ampere
64GB GDDR6
800 GB/s bandwidth
250W TDP
8nm process
Nov 2021 released

Quad-GPU Ampere card for virtual desktops

~$6500
NVIDIA GeForce GTX 1050 Ti NVIDIA · Consumer · Pascal
4GB GDDR5
112 GB/s bandwidth
2.1 FP32 TFLOPS
75W TDP
14nm process
Oct 2016 released

Budget Pascal-era GPU that needs no power connector

~$139
NVIDIA GeForce GTX 1650 NVIDIA · Consumer · Turing
4GB GDDR5
128 GB/s bandwidth
3 FP32 TFLOPS
75W TDP
12nm process
Apr 2019 released

Entry-level Turing GPU

~$149
NVIDIA GeForce GTX 1660 SUPER NVIDIA · Consumer · Turing
6GB GDDR6
336 GB/s bandwidth
5 FP32 TFLOPS
125W TDP
12nm process
Oct 2019 released

Budget Turing GPU refresh

~$229
NVIDIA GeForce GTX 980 Ti NVIDIA · Consumer · Maxwell
6GB GDDR5
336 GB/s bandwidth
6.1 FP32 TFLOPS
250W TDP
28nm process
Jun 2015 released

Flagship Maxwell-era consumer GPU

~$649
NVIDIA GeForce RTX 3060 Laptop GPU NVIDIA · Consumer · Ampere
6GB GDDR6
336 GB/s bandwidth
115W TDP
8nm process
Jan 2021 released

Mainstream mobile Ampere GPU

N/A
NVIDIA GeForce RTX 4050 Laptop GPU NVIDIA · Consumer · Ada Lovelace
6GB GDDR6
192 GB/s bandwidth
115W TDP
4nm process
Feb 2023 released

Entry mobile Ada Lovelace GPU

N/A
NVIDIA Quadro P4000 NVIDIA · Workstation · Pascal
8GB GDDR5
243 GB/s bandwidth
5.3 FP32 TFLOPS
105W TDP
16nm process
Feb 2017 released

Midrange Pascal-era workstation GPU

~$900
NVIDIA Quadro P6000 NVIDIA · Workstation · Pascal
24GB GDDR5X
432 GB/s bandwidth
12 FP32 TFLOPS
250W TDP
16nm process
Oct 2016 released

Flagship Pascal-era workstation GPU

~$5000
NVIDIA Tesla P40 NVIDIA · Datacenter · Pascal
24GB GDDR5
346 GB/s bandwidth
12 FP32 TFLOPS
250W TDP
16nm process
Sep 2016 released

Pascal-era inference GPU, still popular for cheap local LLM VRAM

~$5700
NVIDIA Titan Xp NVIDIA · Consumer · Pascal
12GB GDDR5X
547 GB/s bandwidth
12.1 FP32 TFLOPS
250W TDP
16nm process
Apr 2017 released

Prosumer flagship of the Pascal generation

~$1200
Qualcomm Cloud AI 100 Qualcomm · Datacenter · Cloud AI 100
32GB LPDDR4X
137 GB/s bandwidth
75W TDP
7nm process
Sep 2020 released

Power-efficient datacenter inference card

N/A
Qualcomm Cloud AI 100 Ultra Qualcomm · Datacenter · Cloud AI 100
128GB LPDDR4X
548 GB/s bandwidth
150W TDP
7nm process
Nov 2023 released

High-memory inference card for large generative models

N/A
Tenstorrent Blackhole Tenstorrent · Datacenter · Blackhole
32GB GDDR6
512 GB/s bandwidth
300W TDP
6nm process
Dec 2024 released

Tenstorrent's third-generation RISC-V AI accelerator

~$999
Tenstorrent Wormhole Tenstorrent · Datacenter · Wormhole
12GB GDDR6
336 GB/s bandwidth
160W TDP
12nm process
Jan 2023 released

RISC-V-based scalable AI accelerator card

~$1399
Alibaba Hanguang 800 Alibaba · Datacenter · Hanguang
12nm process
Sep 2019 released

Alibaba's first in-house AI inference chip

N/A
AMD Radeon Instinct MI25 AMD · Datacenter · GCN 5 (Vega 10)
16GB HBM2
484 GB/s bandwidth
12.3 FP32 TFLOPS
300W TDP
14nm process
Jun 2017 released

AMD's first Instinct-branded datacenter accelerator

N/A
AMD Radeon RX 6500 XT AMD · Consumer · RDNA 2
4GB GDDR6
144 GB/s bandwidth
5.8 FP32 TFLOPS
107W TDP
6nm process
Jan 2022 released

Entry-level RDNA 2 consumer GPU

~$199
Baidu Kunlun II Baidu · Datacenter · XPU-II
32GB GDDR6
512 GB/s bandwidth
120W TDP
7nm process
Aug 2021 released

Second-generation Baidu cloud AI chip

N/A
Biren BR104 Biren Technology · Datacenter · BR104
32GB HBM2e
819 GB/s bandwidth
300W TDP
7nm process
Aug 2022 released

Half-size sibling of the BR100 for inference workloads

N/A
Cambricon MLU370-X8 Cambricon Technologies · Datacenter · MLUv02
48GB HBM2
819 GB/s bandwidth
250W TDP
7nm process
Jun 2021 released

Dual-die AI training and inference card

N/A
Cambricon MLU590 Cambricon Technologies · Datacenter · MLUv03
64GB HBM2e
1229 GB/s bandwidth
310W TDP
7nm process
Jun 2023 released

Cambricon's flagship large-model training chip

N/A
d-Matrix Corsair d-Matrix · Datacenter · Digital In-Memory Compute
256GB Chiplet DIMC
6nm process
Nov 2024 released

In-memory-compute chiplet card for low-latency LLM inference

N/A
FuriosaAI RNGD FuriosaAI · Datacenter · Tensor Contraction Processor
48GB HBM3
1500 GB/s bandwidth
150W TDP
5nm process
Aug 2024 released

Power-efficient second-generation inference chip

N/A
Google Coral Edge TPU Google · Edge · Edge TPU
2W TDP
Mar 2019 released

Tiny USB and M.2 inference accelerator for embedded devices

~$25
Google TPU v2 Google · Datacenter · TPU v2
8GB HBM
600 GB/s bandwidth
May 2017 released

First TPU generation to support training, not just inference

N/A
Huawei Ascend 310 Huawei · Edge · Da Vinci
8GB LPDDR4X
8W TDP
12nm process
Oct 2018 released

Low-power edge inference chip

N/A
Intel Arc A380 Intel · Consumer · Xe-HPG (Alchemist)
6GB GDDR6
186 GB/s bandwidth
4.1 FP32 TFLOPS
75W TDP
6nm process
Jun 2022 released

Entry-level Alchemist consumer GPU

~$139
Intel Data Center GPU Flex 140 Intel · Datacenter · Xe-HPG
12GB GDDR6
280 GB/s bandwidth
8 FP32 TFLOPS
75W TDP
6nm process
Aug 2022 released

Low-power dual-GPU media and inference card

N/A
Intel Gaudi Intel · Datacenter · Gaudi 1
32GB HBM2
1000 GB/s bandwidth
350W TDP
16nm process
Jun 2019 released

Habana Labs' first-generation training accelerator

N/A
Moore Threads MTT S4000 Moore Threads · Datacenter · MUSA
48GB GDDR6
768 GB/s bandwidth
280W TDP
12nm process
Jan 2024 released

Chinese full-function GPU for AI training and inference

N/A
NVIDIA GeForce RTX 3050 Laptop GPU NVIDIA · Consumer · Ampere
4GB GDDR6
192 GB/s bandwidth
95W TDP
8nm process
Jan 2022 released

Entry mobile Ampere GPU

N/A
NVIDIA Tesla K80 NVIDIA · Datacenter · Kepler
24GB GDDR5
480 GB/s bandwidth
8.73 FP32 TFLOPS
300W TDP
28nm process
Nov 2014 released

Dual-GPU Kepler-era compute accelerator

~$5000
NVIDIA Tesla P4 NVIDIA · Datacenter · Pascal
8GB GDDR5
192 GB/s bandwidth
5.5 FP32 TFLOPS
75W TDP
16nm process
Sep 2016 released

Low-power Pascal-era inference GPU

~$2200
SambaNova SN30 SambaNova Systems · Datacenter · SN30 RDU
7nm process
Sep 2022 released

Second-generation reconfigurable dataflow chip

N/A
Samsung Mach-1 Samsung · Datacenter · Mach
4nm process
Jan 2025 released

Samsung's first in-house AI inference chip

N/A
Tenstorrent Grayskull Tenstorrent · Datacenter · Grayskull
8GB LPDDR4
118 GB/s bandwidth
200W TDP
12nm process
Oct 2021 released

Tenstorrent's first commercial AI inference card

~$799
FuriosaAI Warboy FuriosaAI · Datacenter · Tensor Contraction Processor
16GB LPDDR4X
180W TDP
14nm process
Nov 2021 released

FuriosaAI's first commercial inference chip

N/A
Google TPU v1 Google · Datacenter · TPU v1
Jan 2015 released

Google's first-generation inference-only TPU, never sold externally

N/A
Intel Arc A310 Intel · Consumer · Xe-HPG (Alchemist)
4GB GDDR6
112 GB/s bandwidth
3.1 FP32 TFLOPS
35W TDP
6nm process
Jan 2023 released

Entry HTPC and media transcode GPU

~$109
NVIDIA Tesla M60 NVIDIA · Datacenter · Maxwell
16GB GDDR5
320 GB/s bandwidth
9.6 FP32 TFLOPS
300W TDP
28nm process
Aug 2015 released

Dual-GPU Maxwell-era virtualization card

~$5000