GPU Vergleich


GPU CARD	RTX 6000 ADA	RTX A6000	RTX A5500	RTX A5000	RTX A4500	RTX A4000	RTX A2000
Streaming Multiprocessors	142	84	80	64	56	48	26
Tensor Cores	568	336	320	256	224	192	104
CUDA Cores	8,9	8.6	8,6	8,192	7,168	6,144	3,328
RayTracing Cores	142	84	80	64	56	48	26
Texture Mapping Unit	568	336	320	256	224	192	104
Render Output Unit	192	112	96	96	96	96	96
Base Clock (MHz)	915 MHz	1455 MHz	1080 MHz	1170 MHz	1050 MHz	735 MHz	562 MHz
Boost Clock (MHz)	2505 MHz	1860 MHz	1665 MHz	1695 MHz	1650 MHz	1560 MHz	1200 MHz
Memory Clock	2500 MHz 20 Gbps effective	2000 MHz	2000 MHz	2000 MHz	2000 MHz	1750 Mhz	1500 MHz
Total Video Memory	48 GB GDDR6	48 GB GDDR6	24 GB GDDR6	24 GB GDDR6	20 GB GDDR6	16 GB GDDR6	6 GB oder 12 GB GDDR6
Memory Interface	384-bit	384-bit	384-bit	384-bit	320-bit	256-bit	192-bit
Total Memory Bandwidth	960.0 GB/s	768.0 GB/s	768.0 GB/s	768.0 GB/s	640.0 GB/s	448.0 GB/s	288.0 GB/s
ECC Memory	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Pixel Rate	481.0 GPixel/s	208.3 GPixel/s	159.8 GPixel/s	162.7 GPixel/s	158.4 GPixel/s	149.8 GPixel/s	57.60 GPixel/s
Texture Rate	1,423 GTexel/s	625.0 GTexel/s	532.8 GTexel/s	433.9 GTexel/s	369.6 GTexel/s	299.5 GTexel/s	124.8 GTexel/s
FP16 / FP32 / FP64 in TFLOPS	91.06 / 91.06 / 1,423	40.00 / 40.00 / 1,250	34.10 / 34.10 / 1,066 GFLOPS	27.77 / 27.77 / 867.8	23.65 / 23.65 / 739.2	19.17 / 19.17 / 599.0	7.987 / 7.987 / 124.8
Fabrication Process	4 nm	8 nm	8 nm	8 nm	8 nm	8 nm	8 nm
Transistor Count	76,300 million	28,300 million	28,300 million	28,300 million	28,300 million	17,400 million	13,250 million
Connectors	4x DisplayPort 1.4a	4x DisplayPort	4x DisplayPort 1.4a	4x DisplayPort	4x DisplayPort	4x DisplayPort	4x DisplayPort
Form Factor	Dual Slot	Dual Slot	Dual Slot	Dual Slot	Dual Slot	Single Slot	Dual Slot
Power Connectors	16-pin EPS	8-pin EPS	1x 8-pin	8-pin EPS	1x 8-pin	1x 6-pin	None (Total board power: 70 W)
ZUM PRODUKT	6000 ADA	A6000	A5500	A5000	A4500	A4000	12 GB A2000 6 GB A2000


GPU CARD	L40	H100	A100	A16	A40	A30	A10	A2
Streaming Multiprocessors	142	114	108	10 x 4	84	56	72	10
Tensor Cores	568	456	432	40 x 4	336	56	288	40
CUDA Cores	8.9	9.0	6912	1280 x 4	10,752	3804	9216	2560
RayTracing Cores	142	-	108	10 x 4	84	64	72	10
Texture Mapping Unit	568	456	432	40 x 4	336	224	192	40
Render Output Unit	192	-	160	32 x 4	112	96	288	32
Base Clock (MHz)	735 MHz	1095 MHz	765 MHz	885 MHz	1305 MHz	930 MHz	885 MHz	1440 MHz
Boost Clock (MHz)	2490 MHz	1755 MHz	1410 MHz	1695 MHz	1740 MHz	1440 MHz	1695 MHz	1770 MHz
Memory Clock	2250 MHz 18 Gbps effective	1593 MHz 3.2 Gbps effective	1215 MHz	1812 MHz	1812 MHz	1215 MHz	1563 MHz	1563 MHz
Total Video Memory	48 GB HBM2e	80 GB HBM2e	80 GB HBM2e	16 GB GDDR6 x 4 (64 GB)	48 GB GDDR6	24 GB GDDR6	24 GB GDDR6	16 GB GDDR6
Memory Interface	384 bit	5120 bit	5120 bit	128-bit x 4	384-bit	384-bit	384 bit	128 bit
Total Memory Bandwidth	864.0 GB/s	2,039 GB/s	1,555 GB/s	231.9 GB/s x 4	695.8 GB/s	933.1 GB/s	600.2 GB/s	200.1 GB/s
ECC Memory	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Pixel Rate	478.1 GPixel/s	42.12 GPixel/s	225.6 GPixel/s	54.24 x 4 GPixel/s	194.9 GPixel/s	138.2 GPixel/s	162.7 GPixel/s	56.64 GPixel/s
Texture Rate	1,414 GTexel/s	800.3 GTexel/s	609.1 GTexel/s	67.80 x 4 GTexel/s	584.6 GTexel/s	322.6 GTexel/s	488.2 GTexel/s	70.80 GTexel/s
FP16 / FP32 / FP64 in TFLOPS	90.52 / 90.52 / 1,414 GFLOPS	204.9 / 51.22 / 25.61	77.97 / 19.49 / 9.746 (GFLOPS)	4.339 x 4 / 4.339 x 4 / 135.6 x 4 (GFLOPS)	37.42 / 37.42 / 1,169 (GFLOPS)	10.32 / 10.32 / 5.161	31.24 / 31.24 / 976.3	4.531 / 4.531 / 70.80
Fabrication Process	4 nm	4 nm	7 nm	8 nm	8 nm	7 nm	8 nm	8 nm
Transistor Count	76,300 million	80,000 million	54,200 million	-	28,300 million	54,200 million	28,300 million	-
Connectors	4x DisplayPort 1.4a	No outputs	No outputs	No outputs	3x DisplayPort	No outputs	No outputs	No outputs
Form Factor	Dual Slot	Dual Slot	Dual Slot	Dual Slot	Dual Slot	Dual Slot	Single Slot	Single Slot
Power Connectors	1x 16-pin	1x 16-pin	8-pin EPS	8-pin EPS	8-pin EPS	8-pin EPS	8-pin EPS	None (PCIe)
ZUM PRODUKT	L40	H100	A100	A16	A40	A30	A10	A2