GPU Guide

LLM token generation is memory-bandwidth bound — more GB/s means faster inference. Sort by bandwidth, filter by VRAM and budget, and click any card for full specs to find the right GPU for running models locally.

144 GPUs
#Card Generation VRAM Bandwidth MemoryCore / Boost
🥇NRTX 5090Blackwell32 GB
1792 GB/s
512-bit GDDR7 · 28 Gbps2017 / 2407 MHz
🥈NRTX PRO 6000 BlackwellPro96 GB
1792 GB/s
512-bit GDDR7 · 28 Gbps1590 / 2617 MHz
🥉NCMP 170HXno displayAmpere8 GB
1493 GB/s
HBM2e 4096-bit1140 / 1410 MHz
#4ARadeon VIIAMD · ROCmVega16 GB
1024 GB/s
HBM2 4096-bit1400 / 1750 MHz
#5NRTX 3090 TiAmpere24 GB
1008 GB/s
384-bit GDDR6X · 21 Gbps1560 / 1860 MHz
#6NRTX 4090Ada24 GB
1008 GB/s
384-bit GDDR6X · 21 Gbps2235 / 2520 MHz
#7NRTX 5080Blackwell16 GB
960 GB/s
256-bit GDDR7 · 30 Gbps2295 / 2617 MHz
#8ARX 7900 XTXAMD · ROCmRDNA324 GB
960 GB/s
384-bit GDDR6 · 20 Gbps2300 / 2499 MHz
#9NRTX 3090Ampere24 GB
936 GB/s
384-bit GDDR6X · 20 Gbps1395 / 1695 MHz
#10NRTX 3080 12GBAmpere12 GB
912 GB/s
384-bit GDDR6X · 19 Gbps1260 / 1710 MHz
#11NRTX 3080 TiAmpere12 GB
912 GB/s
384-bit GDDR6X · 19 Gbps1365 / 1665 MHz
#12NRTX 5070 TiBlackwell16 GB
896 GB/s
256-bit GDDR7 · 28 Gbps2295 / 2452 MHz
#13ARX 9070 XTAMD · ROCmRDNA416 GB
896 GB/s
256-bit GDDR6 · 28 Gbps2400 / 2970 MHz
#14NRTX A6000 AdaPro48 GB
864 GB/s
384-bit GDDR6 · 18 Gbps915 / 2520 MHz
#15ARadeon Pro W7900AMD · ROCmRDNA348 GB
864 GB/s
384-bit GDDR6 · 18 Gbps1900 / 2500 MHz
#16ARX 7900 XTAMD · ROCmRDNA320 GB
800 GB/s
320-bit GDDR6 · 20 Gbps2000 / 2394 MHz
#17NRTX A5000Pro24 GB
768 GB/s
384-bit GDDR6 · 16 Gbps930 / 1695 MHz
#18NRTX A6000Pro48 GB
768 GB/s
384-bit GDDR6 · 16 Gbps930 / 1800 MHz
#19NRTX 3080 10GBAmpere10 GB
760 GB/s
320-bit GDDR6X · 19 Gbps1440 / 1710 MHz
#20NRTX 4080Ada16 GB
736 GB/s
256-bit GDDR6X · 23 Gbps2205 / 2505 MHz
#21NRTX 4080 SuperAda16 GB
736 GB/s
256-bit GDDR6X · 23 Gbps2295 / 2550 MHz
#22NTITAN RTXTuring24 GB
672 GB/s
384-bit GDDR6 · 14 Gbps1350 / 1770 MHz
#23NRTX 4070 Ti SuperAda16 GB
672 GB/s
256-bit GDDR6X · 21 Gbps2340 / 2610 MHz
#24NRTX 5070Blackwell12 GB
672 GB/s
192-bit GDDR7 · 28 Gbps2164 / 2512 MHz
#25NQuadro RTX 6000Turing24 GB
672 GB/s
384-bit GDDR6 · 14 Gbps1440 / 1770 MHz
#26NQuadro RTX 8000Turing48 GB
672 GB/s
384-bit GDDR6 · 14 Gbps1395 / 1770 MHz
#27ARX 9070AMD · ROCmRDNA416 GB
672 GB/s
256-bit GDDR6 · 21 Gbps2070 / 2520 MHz
#28NRTX A4500Pro20 GB
640 GB/s
320-bit GDDR6 · 16 Gbps930 / 1650 MHz
#29NRTX 2080 TiTuring11 GB
616 GB/s
352-bit GDDR6 · 14 Gbps1350 / 1545 MHz
#30NRTX 3070 TiAmpere8 GB
608 GB/s
256-bit GDDR6X · 19 Gbps1575 / 1770 MHz
#31NRTX 5000 AdaPro16 GB
576 GB/s
256-bit GDDR6 · 18 Gbps1425 / 2550 MHz
#32ARX 7800 XTAMD · ROCmRDNA316 GB
576 GB/s
256-bit GDDR6 · 18 Gbps2124 / 2430 MHz
#33ARX 7900 GREAMD · ROCmRDNA316 GB
576 GB/s
256-bit GDDR6 · 18 Gbps1880 / 2245 MHz
#34ARadeon Pro W7800AMD · ROCmRDNA332 GB
576 GB/s
256-bit GDDR6 · 18 Gbps1800 / 2500 MHz
#35ARX 6950 XTAMD · ROCmRDNA216 GB
576 GB/s
256-bit GDDR6 · 18 Gbps2100 / 2310 MHz
#36iArc A770 8GBIntel · oneAPIArc8 GB
560 GB/s
256-bit GDDR6 · 18 Gbps2100 / 2400 MHz
#37iArc A770 16GBIntel · oneAPIArc16 GB
560 GB/s
256-bit GDDR6 · 18 Gbps2100 / 2400 MHz
#38NTITAN XpPascal12 GB
548 GB/s
384-bit GDDR5X · 11 Gbps1405 / 1582 MHz
#39ARX 6800AMD · ROCmRDNA216 GB
512 GB/s
256-bit GDDR6 · 16 Gbps1815 / 2105 MHz
#40ARX 6800 XTAMD · ROCmRDNA216 GB
512 GB/s
256-bit GDDR6 · 16 Gbps2015 / 2250 MHz
#41ARX 6900 XTAMD · ROCmRDNA216 GB
512 GB/s
256-bit GDDR6 · 16 Gbps2015 / 2250 MHz
#42ARadeon Pro W6800AMD · ROCmRDNA232 GB
512 GB/s
256-bit GDDR6 · 16 Gbps1700 / 2095 MHz
#43NRTX 4070Ada12 GB
504 GB/s
192-bit GDDR6X · 21 Gbps1920 / 2475 MHz
#44NRTX 4070 SuperAda12 GB
504 GB/s
192-bit GDDR6X · 21 Gbps1980 / 2475 MHz
#45NRTX 4070 TiAda12 GB
504 GB/s
192-bit GDDR6X · 21 Gbps2310 / 2610 MHz
#46NRTX 2080 SuperTuring8 GB
496 GB/s
256-bit GDDR6 · 16 Gbps1650 / 1815 MHz
#47NGTX 1080 TiPascal11 GB
484 GB/s
352-bit GDDR5X · 11 Gbps1480 / 1582 MHz
#48ARX Vega 64AMD · ROCmVega8 GB
484 GB/s
HBM2 2048-bit1274 / 1546 MHz
#49NTITAN X (Pascal)Pascal12 GB
480 GB/s
384-bit GDDR5X · 10 Gbps1417 / 1531 MHz
#50iArc B580 12GBIntel · oneAPIArc12 GB
456 GB/s
192-bit GDDR6 · 19 Gbps2170 / 2670 MHz
#51NRTX 2060 SuperTuring8 GB
448 GB/s
256-bit GDDR6 · 14 Gbps1470 / 1650 MHz
#52NRTX 2070Turing8 GB
448 GB/s
256-bit GDDR6 · 14 Gbps1410 / 1620 MHz
#53NRTX 2070 SuperTuring8 GB
448 GB/s
256-bit GDDR6 · 14 Gbps1605 / 1770 MHz
#54NRTX 2080Turing8 GB
448 GB/s
256-bit GDDR6 · 14 Gbps1515 / 1710 MHz
#55NRTX 3060 TiAmpere8 GB
448 GB/s
256-bit GDDR6 · 14 Gbps1410 / 1665 MHz
#56NRTX 3070Ampere8 GB
448 GB/s
256-bit GDDR6 · 14 Gbps1500 / 1725 MHz
#57NRTX 5060narrow busBlackwell8 GB
448 GB/s
128-bit GDDR7 · 28 Gbps2280 / 2497 MHz
#58NRTX 5060 Ti 8GBnarrow busBlackwell8 GB
448 GB/s
128-bit GDDR7 · 28 Gbps2377 / 2572 MHz
#59NRTX 5060 Ti 16GBnarrow busBlackwell16 GB
448 GB/s
128-bit GDDR7 · 28 Gbps2377 / 2572 MHz
#60NQuadro RTX 5000Turing16 GB
448 GB/s
256-bit GDDR6 · 14 Gbps1620 / 1815 MHz
#61NRTX A4000Pro16 GB
448 GB/s
256-bit GDDR6 · 14 Gbps735 / 1560 MHz
#62ARX 5700AMD · ROCmRDNA18 GB
448 GB/s
256-bit GDDR6 · 14 Gbps1465 / 1725 MHz
#63ARX 5700 XTAMD · ROCmRDNA18 GB
448 GB/s
256-bit GDDR6 · 14 Gbps1605 / 1905 MHz
#64ARadeon Pro W5700AMD · ROCmRDNA18 GB
448 GB/s
256-bit GDDR6 · 14 Gbps1688 / 1860 MHz
#65NP102-100 10GBno displayPascal10 GB
440 GB/s
320-bit GDDR5X · 11 Gbps1442 / 1582 MHz
#66NQuadro P6000Pascal24 GB
432 GB/s
384-bit GDDR5X · 9 Gbps1506 / 1645 MHz
#67NRTX 4000 AdaPro20 GB
432 GB/s
192-bit GDDR6 · 18 Gbps1500 / 2175 MHz
#68NRTX 4500 AdaPro24 GB
432 GB/s
192-bit GDDR6 · 18 Gbps1800 / 2580 MHz
#69ARX 7700 XTAMD · ROCmRDNA312 GB
432 GB/s
192-bit GDDR6 · 18 Gbps2171 / 2599 MHz
#70ARX 6750 XTAMD · ROCmRDNA212 GB
432 GB/s
192-bit GDDR6 · 18 Gbps2321 / 2600 MHz
#71NQuadro RTX 4000Turing8 GB
416 GB/s
256-bit GDDR6 · 13 Gbps1005 / 1545 MHz
#72ARX Vega 56AMD · ROCmVega8 GB
410 GB/s
HBM2 2048-bit1156 / 1471 MHz
#73ARX 6700 XTAMD · ROCmRDNA212 GB
384 GB/s
192-bit GDDR6 · 16 Gbps2321 / 2581 MHz
#74iArc A750Intel · oneAPIArc8 GB
384 GB/s
192-bit GDDR6 · 16 Gbps2050 / 2400 MHz
#75iArc B570 10GBIntel · oneAPIArc10 GB
380 GB/s
160-bit GDDR6 · 19 Gbps2500 / 2750 MHz
#76NRTX 3060Ampere12 GB
360 GB/s
192-bit GDDR6 · 15 Gbps1320 / 1777 MHz
#77NGTX 780 TiKepler3 GB
336 GB/s
384-bit GDDR5 · 7 Gbps875 / 928 MHz
#78NGTX 980 TiMaxwell6 GB
336 GB/s
384-bit GDDR5 · 7 Gbps1000 / 1076 MHz
#79NTITAN XMaxwell12 GB
336 GB/s
384-bit GDDR5 · 7 Gbps1000 / 1089 MHz
#80NGTX 1660 SuperTuring6 GB
336 GB/s
192-bit GDDR6 · 14 Gbps1530 / 1785 MHz
#81NRTX 2060Turing6 GB
336 GB/s
192-bit GDDR6 · 14 Gbps1365 / 1680 MHz
#82NRTX 2060 12GBTuring12 GB
336 GB/s
192-bit GDDR6 · 14 Gbps1470 / 1650 MHz
#83NRTX A3000 12GBPro12 GB
336 GB/s
192-bit GDDR6 · 14 Gbps1005 / 1560 MHz
#84ARX 5600 XTAMD · ROCmRDNA16 GB
336 GB/s
192-bit GDDR6 · 14 Gbps1375 / 1620 MHz
#85NGTX 1080Pascal8 GB
320 GB/s
256-bit GDDR5X · 10 Gbps1607 / 1733 MHz
#86NRTX 5050narrow busBlackwell8 GB
320 GB/s
128-bit GDDR6 · 20 Gbps2317 / 2572 MHz
#87NRTX 4000 SFF AdaPro20 GB
320 GB/s
160-bit GDDR6 · 16 Gbps1350 / 1800 MHz
#88ARX 6700AMD · ROCmRDNA210 GB
320 GB/s
160-bit GDDR6 · 16 Gbps2321 / 2424 MHz
#89NQuadro M6000Maxwell24 GB
317 GB/s
384-bit GDDR5 · 7 Gbps988 / 988 MHz
#90NGTX 780Kepler3 GB
288 GB/s
384-bit GDDR5 · 6 Gbps863 / 900 MHz
#91NQuadro P5000Pascal16 GB
288 GB/s
256-bit GDDR5X · 9 Gbps1607 / 1733 MHz
#92NGTX 1660 TiTuring6 GB
288 GB/s
192-bit GDDR6 · 12 Gbps1500 / 1770 MHz
#93NQuadro RTX 3000Turing6 GB
288 GB/s
192-bit GDDR6 · 12 Gbps1620 / 1620 MHz
#94NRTX 4060 Ti 8GBAda8 GB
288 GB/s
128-bit GDDR6 · 18 Gbps2310 / 2535 MHz
#95NRTX 4060 Ti 16GBAda16 GB
288 GB/s
128-bit GDDR6 · 18 Gbps2310 / 2535 MHz
#96NRTX A2000 12GBPro12 GB
288 GB/s
192-bit GDDR6 · 12 Gbps735 / 1200 MHz
#97ARX 9060 XT 8GBAMD · ROCmRDNA48 GB
288 GB/s
128-bit GDDR6 · 18 Gbps2200 / 2700 MHz
#98ARX 9060 XT 16GBAMD · ROCmRDNA416 GB
288 GB/s
128-bit GDDR6 · 18 Gbps2200 / 2700 MHz
#99ARX 7600AMD · ROCmRDNA38 GB
288 GB/s
128-bit GDDR6 · 18 Gbps2250 / 2655 MHz
#100ARX 7600 XT 16GBAMD · ROCmRDNA316 GB
288 GB/s
128-bit GDDR6 · 18 Gbps2310 / 2755 MHz
#101iArc A580Intel · oneAPIArc8 GB
288 GB/s
128-bit GDDR6 · 18 Gbps1700 / 2000 MHz
#102ARX 6650 XTAMD · ROCmRDNA28 GB
280 GB/s
128-bit GDDR6 · 18 Gbps2055 / 2635 MHz
#103NRTX 4060Ada8 GB
272 GB/s
128-bit GDDR6 · 17 Gbps1830 / 2460 MHz
#104NGTX 1070Pascal8 GB
256 GB/s
256-bit GDDR5 · 8 Gbps1506 / 1683 MHz
#105NGTX 1070 TiPascal8 GB
256 GB/s
256-bit GDDR5 · 8 Gbps1607 / 1683 MHz
#106ARX 6600 XTAMD · ROCmRDNA28 GB
256 GB/s
128-bit GDDR6 · 16 Gbps2359 / 2589 MHz
#107ARX 580 8GBAMD · ROCmPolaris8 GB
256 GB/s
256-bit GDDR5 · 8 Gbps1257 / 1340 MHz
#108ARX 590 8GBAMD · ROCmPolaris8 GB
256 GB/s
256-bit GDDR5 · 8 Gbps1469 / 1545 MHz
#109NQuadro P4000Pascal8 GB
243 GB/s
256-bit GDDR5 · 8 Gbps1202 / 1480 MHz
#110NRTX 3060 8GBnarrow busAmpere8 GB
240 GB/s
128-bit GDDR6 · 15 Gbps1320 / 1777 MHz
#111NGTX 770Kepler2 GB
224 GB/s
256-bit GDDR5 · 7 Gbps1046 / 1085 MHz
#112NGTX 970Maxwell4 GB
224 GB/s
256-bit GDDR5 · 7 Gbps1050 / 1178 MHz
#113NGTX 980Maxwell4 GB
224 GB/s
256-bit GDDR5 · 7 Gbps1126 / 1216 MHz
#114NRTX 3050 8GBAmpere8 GB
224 GB/s
128-bit GDDR6 · 14 Gbps1552 / 1777 MHz
#115NRTX 3050 16GBAmpere16 GB
224 GB/s
128-bit GDDR6 · 14 Gbps1042 / 1470 MHz
#116NRTX 2000 AdaPro16 GB
224 GB/s
128-bit GDDR6 · 14 Gbps405 / 2175 MHz
#117ARX 6600AMD · ROCmRDNA28 GB
224 GB/s
128-bit GDDR6 · 14 Gbps1626 / 2044 MHz
#118ARadeon Pro W6600AMD · ROCmRDNA28 GB
224 GB/s
128-bit GDDR6 · 14 Gbps1589 / 1920 MHz
#119ARX 5500 XT 4GBAMD · ROCmRDNA14 GB
224 GB/s
128-bit GDDR6 · 14 Gbps1607 / 1845 MHz
#120ARX 5500 XT 8GBAMD · ROCmRDNA18 GB
224 GB/s
128-bit GDDR6 · 14 Gbps1607 / 1845 MHz
#121ARX 570 8GBAMD · ROCmPolaris8 GB
224 GB/s
256-bit GDDR5 · 7 Gbps1168 / 1244 MHz
#122NQuadro M5000Maxwell8 GB
211 GB/s
256-bit GDDR5 · 7 Gbps1051 / 1051 MHz
#123NGTX 760Kepler2 GB
192 GB/s
256-bit GDDR5 · 6 Gbps980 / 1033 MHz
#124NQuadro M4000Maxwell8 GB
192 GB/s
256-bit GDDR5 · 6 Gbps773 / 773 MHz
#125NGTX 1060 3GBPascal3 GB
192 GB/s
192-bit GDDR5 · 8 Gbps1506 / 1709 MHz
#126NGTX 1060 6GBPascal6 GB
192 GB/s
192-bit GDDR5 · 8 Gbps1506 / 1709 MHz
#127NGTX 1650 SuperTuring4 GB
192 GB/s
128-bit GDDR6 · 12 Gbps1530 / 1725 MHz
#128NGTX 1660Turing6 GB
192 GB/s
192-bit GDDR5 · 8 Gbps1530 / 1785 MHz
#129NRTX A2000 6GBPro6 GB
192 GB/s
128-bit GDDR6 · 12 Gbps735 / 1200 MHz
#130ARX 7500 XTAMD · ROCmRDNA38 GB
192 GB/s
128-bit GDDR6 · 12 Gbps2254 / 2609 MHz
#131iArc A380Intel · oneAPIArc6 GB
186 GB/s
96-bit GDDR6 · 16 Gbps2000 / 2450 MHz
#132NRTX 3050 6GBnarrow busAmpere6 GB
168 GB/s
96-bit GDDR6 · 14 Gbps1552 / 1777 MHz
#133ARX 6500 XTAMD · ROCmnarrow busRDNA24 GB
144 GB/s
64-bit GDDR6 · 18 Gbps2610 / 2815 MHz
#134NGTX 1650Turing4 GB
128 GB/s
128-bit GDDR5 · 8 Gbps1485 / 1665 MHz
#135ARX 6400AMD · ROCmnarrow busRDNA24 GB
128 GB/s
64-bit GDDR6 · 16 Gbps1923 / 2321 MHz
#136NGTX 960Maxwell4 GB
112 GB/s
128-bit GDDR5 · 7 Gbps1127 / 1178 MHz
#137NGTX 1050Pascal2 GB
112 GB/s
128-bit GDDR5 · 7 Gbps1354 / 1455 MHz
#138NGTX 1050 TiPascal4 GB
112 GB/s
128-bit GDDR5 · 7 Gbps1290 / 1392 MHz
#139NGTX 950Maxwell2 GB
105 GB/s
128-bit GDDR5 · 7 Gbps1024 / 1188 MHz
#140NGTX 1630narrow busTuring4 GB
96 GB/s
64-bit GDDR6 · 12 Gbps1785 / 1800 MHz
#141iArc A310Intel · oneAPInarrow busArc4 GB
96 GB/s
64-bit GDDR6 · 12 Gbps1550 / 2000 MHz
#142NGTX 750 TiMaxwell2 GB
86 GB/s
128-bit GDDR5 · 5 Gbps1020 / 1085 MHz
#143NGTX 750Maxwell2 GB
80 GB/s
128-bit GDDR5 · 5 Gbps1020 / 1085 MHz
#144NGTX 1030narrow busPascal2 GB
48 GB/s
64-bit GDDR5 · 6 Gbps1227 / 1468 MHz