GPUSPECINDEX

Every GPU With 16GB+ VRAM

A cross-vendor history of GPU variants with at least 16GB of VRAM, including launch years and memory technologies.

132variants
3vendors
13launch years

When 16GB stopped being exotic

Sixteen gigabytes of graphics memory began as a capacity associated with specialized workstations and compute hardware, where large scenes, scientific arrays, and professional visualization could justify both the chips and the wide memory subsystem. It later crossed into enthusiast gaming products, but not because capacity alone suddenly made a card fast. Higher-resolution textures, ray-tracing working sets, content creation, and longer product lifetimes all increased the value of keeping data close to the processor. The transition is interesting precisely because vendors reached the tier through different markets and at different times.

The table contains 132 variants with at least 16,384 MB, spanning 2014 through 2026. These are GPU variants. A model offered with two capacities contributes only the qualifying configuration, while separate catalog variants remain separate. Nvidia accounts for 73 variants, AMD for 51 variants, and Intel for 8 variants. The threshold is inclusive, so products above 16GB remain part of the historical tier rather than being split into an arbitrary premium page.

Capacity is not bandwidth

VRAM size answers how much can remain resident; memory type, bus width, clock, and cache design influence how quickly it can move. Two 16GB products can consequently serve very different workloads. Older professional boards may pair the capacity with architectures that are modest by current throughput standards, while a newer consumer design can use a narrower physical bus and compensate elsewhere. The cross-vendor table is useful because it keeps those dimensions beside the headline number instead of turning capacity into a performance ranking.

The history also explains why the tier can arrive unevenly. Memory-chip densities, package layouts, product segmentation, and software demand do not advance in lockstep. A vendor may reserve a capacity for professional products, introduce it on a flagship, then spread it downward when component economics change. This page does not label any amount “future proof,” and it does not infer benchmark results from specifications. It answers the narrower reproducible question: which cataloged variants reached or exceeded the tier, when did they launch, and what memory technologies carried them there?

Catalog variants

GPU variants in this editorial dataset
VendorVariantYearVRAMMemoryFP32 GFLOPSTDPConnector evidenceNodeTransistors M
amdFirePro S9150201416,384 MBGDDR55,070—not recorded by source28 nm—
nvidiaQuadro M6000201524,576 MBGDDR56,070250 Wnot recorded by source28 nm—
amdFirePro S9170201532,768 MBGDDR55,240—not recorded by source28 nm—
amdFirePro S7150 X2201616,384 MBGDDR57,540265 Wnot recorded by source28 nm—
nvidiaP100 GPU accelerator (mezzanine)201616,384 MBHBM2—300 Wnot recorded by source——
nvidiaP40 GPU accelerator201624,576 MBGDDR5—250 Wnot recorded by source——
nvidiaQuadro GP100201616,384 MBHBM210,336235 Wnot recorded by source16 nm—
nvidiaQuadro P5000201616,384 MBGDDR5X8,873180 Wnot recorded by source16 nm—
nvidiaQuadro P6000201624,576 MBGDDR5X10,882250 Wnot recorded by source16 nm—
nvidiaQuadro P5000 Mobile201716,384 MBGDDR56,197100 Wnot recorded by source16 nm—
nvidiaP6 GPU accelerator201716,384 MBGDDR56,16990 Wnot recorded by source——
amdRadeon Pro Vega 64 (Vega 10)201716,384 MBHBM2——not recorded by source14 nm—
amdRadeon Instinct MI25 (Vega 10)201716,384 MBHBM2—300 Wnot recorded by source——
amdRadeon Instinct MI6 (Polaris 10)201716,384 MBGDDR5—150 Wnot recorded by source——
amdRadeon Vega Frontier Edition (Liquid-cooled) (Vega 10)201716,384 MBHBM2—375 Wnot recorded by source14 nm—
amdRadeon Vega Frontier Edition (Vega 10)201716,384 MBHBM2—300 Wnot recorded by source14 nm—
amdRadeon Pro WX 9100 (Vega 10)201716,384 MBHBM2—230 Wnot recorded by source14 nm—
nvidiaQuadro P5200 Max-Q201816,384 MBGDDR57,578100 Wnot recorded by source16 nm—
nvidiaQuadro P5200 Mobile201816,384 MBGDDR59,006100 Wnot recorded by source16 nm—
nvidiaQuadro GV100201832,768 MBHBM214,800250 Wnot recorded by source12 nm—
nvidiaV100 GPU accelerator (PCIe FHHL card)201816,384 MBHBM213,210250 Wnot recorded by source——
nvidiaNvidia Titan V (CEO Edition)201832,768 MBHBM212,290250 Wnot recorded by source——
nvidiaQuadro RTX 5000201816,384 MBGDDR611,150—not recorded by source12 nm—
nvidiaQuadro RTX 6000201824,576 MBGDDR616,310—not recorded by source12 nm—
nvidiaQuadro RTX 8000201849,152 MBGDDR616,310—not recorded by source12 nm—
nvidiaT4 GPU accelerator (PCIe card)201816,384 MBGDDR68,10070 Wnot recorded by source——
amdRadeon Instinct MI60 (Vega 20)201832,768 MBHBM2—300 Wnot recorded by source7 nm—
nvidiaTitan RTX201824,576 MBGDDR612,442280 Wnot recorded by source—18,600
amdRadeon Pro Vega II (Vega 20)201932,768 MBHBM214,090475 Wnot recorded by source7 nm—
amdRadeon VII (Vega 20)201916,384 MBHBM2—300 Wnot recorded by source——
amdRadeon Pro Vega 64X (Vega 10)201916,384 MBHBM2——not recorded by source14 nm—
nvidiaQuadro RTX 5000 Max-Q201916,384 MBGDDR68,29480 Wnot recorded by source12 nm—
nvidiaQuadro RTX 5000 Mobile201916,384 MBGDDR69,492110 Wnot recorded by source12 nm—
nvidiaQuadro RTX 6000 Mobile201924,576 MBGDDR613,410200 Wnot recorded by source12 nm—
amdRadeon Pro W5700X (Navi 10)201916,384 MBGDDR6—250 Wnot recorded by source7 nm—
amdRadeon Pro 5700 XT (Navi 10)202016,384 MBGDDR6—130 Wnot recorded by source7 nm—
amdRadeon E9565 (Polaris10)202016,384 MBGDDR55,728135 Wnot recorded by source14 nm—
nvidiaGeForce RTX 3090202024,576 MBGDDR6X29,284350 Wnot recorded by source—28,300
nvidiaA40 GPU accelerator (PCIe card)202049,152 MBGDDR637,420300 Wnot recorded by source——
nvidiaRTX A6000202049,152 MBGDDR638,710300 Wnot recorded by source8 nm—
amdAMD Instinct MI100 (Arcturus)202032,768 MBHBM2—300 Wnot recorded by source7 nm—
amdRadeon RX 6800 (Navi 21)202016,384 MBGDDR6—250 Wnot recorded by source7 nm—
amdRadeon RX 6800 XT (Navi 21)202016,384 MBGDDR6—300 Wnot recorded by source7 nm—
amdRadeon RX 6900 XT (Navi 21)202016,384 MBGDDR6—300 Wnot recorded by source7 nm—
nvidiaA10 GPU accelerator (PCIe card)202124,576 MBGDDR631,240150 Wnot recorded by source——
nvidiaA30 GPU accelerator (PCIe card)202124,576 MBHBM210,320165 Wnot recorded by source——
nvidiaRTX A4000202116,384 MBGDDR619,170140 Wnot recorded by source8 nm—
nvidiaRTX A5000202124,576 MBGDDR627,770230 Wnot recorded by source8 nm—
nvidiaRTX A5000 Mobile202116,384 MBGDDR619,350150 Wnot recorded by source8 nm—
amdRadeon Pro W6800 (Navi 21)202132,768 MBGDDR6—250 Wnot recorded by source7 nm—
amdRadeon Pro W6800X (Navi 21)202132,768 MBGDDR6—300 Wnot recorded by source7 nm—
amdRadeon Pro W6900X (Navi 21)202132,768 MBGDDR6—300 Wnot recorded by source7 nm—
amdRadeon Pro V620 (Navi 21)202132,768 MBGDDR6—300 Wnot recorded by source7 nm—
nvidiaA2 GPU accelerator (PCIe card)202116,384 MBGDDR64,531—not recorded by source——
nvidiaRTX A4500202120,480 MBGDDR623,660200 Wnot recorded by source8 nm—
nvidiaGeForce RTX 3080 Ti 20 GB202220,480 MBGDDR6X27,341350 Wnot recorded by source—28,300
nvidiaGeForce RTX 3080 Ti Mobile/ Laptop202216,384 MBGDDR6——not recorded by source8 nm22,000
amdAMD Instinct MI210 (Aldebaran)202265,536 MBHBM2E—300 Wnot recorded by source6 nm—
nvidiaH100 GPU accelerator (PCIe card)202281,920 MBHBM2E51,200350 Wnot recorded by source——
nvidiaRTX A4500 Mobile202216,384 MBGDDR617,660140 Wnot recorded by source8 nm—
nvidiaRTX A5500202224,576 MBGDDR634,100230 Wnot recorded by source8 nm—
nvidiaRTX A5500 Mobile202216,384 MBGDDR622,270165 Wnot recorded by source8 nm—
nvidiaGeForce RTX 3090 Ti202224,576 MBGDDR6X33,546450 Wnot recorded by source—28,300
amdRadeon RX 6950 XT (Navi 21)202216,384 MBGDDR6—335 Wnot recorded by source7 nm—
nvidiaGeForce RTX 4090202224,576 MBGDDR6X73,236450 Wnot recorded by source—76,300
nvidiaL40 GPU accelerator202249,152 MBGDDR690,520300 Wnot recorded by source——
intelArc 7 A770 16GB202216,384 MBGDDR6—225 Wnot recorded by source6 nm21,700
nvidiaGeForce RTX 4080202216,384 MBGDDR6X42,900320 Wnot recorded by source—45,900
nvidiaRTX 6000 Ada Generation202249,152 MBGDDR691,060300 Wnot recorded by source4 nm—
amdRadeon RX 7900 XT (Navi 31)202220,480 MBGDDR6—315 Wnot recorded by source——
amdRadeon RX 7900 XTX (Navi 31)202224,576 MBGDDR6—355 Wnot recorded by source——
intelData Center GPU Max 1100202349,152 MBHBM2E—300 Wnot recorded by source—100,000
intelData Center GPU Max 15502023131,072 MBHBM2E—600 Wnot recorded by source—100,000
nvidiaGeForce RTX 4090 Mobile/ Laptop202316,384 MBGDDR6——not recorded by source4 nm45,900
nvidiaH800 GPU accelerator (SXM card)202381,920 MBHBM359,300700 Wnot recorded by source——
nvidiaL4 GPU accelerator202324,576 MBGDDR630,30072 Wnot recorded by source——
nvidiaRTX 4000 SFF Ada Generation202320,480 MBGDDR619,23070 Wnot recorded by source4 nm—
amdRadeon Pro W7800 (Navi 31)202332,768 MBGDDR645,250260 Wnot recorded by source——
amdRadeon Pro W7900 (Navi 31)202349,152 MBGDDR661,320295 Wnot recorded by source——
nvidiaGeForce RTX 4060 Ti 16 GB202316,384 MBGDDR620,106165 Wnot recorded by source—22,900
nvidiaRTX 4000 Ada Generation202320,480 MBGDDR626,730130 Wnot recorded by source4 nm—
nvidiaRTX 4500 Ada Generation202324,576 MBGDDR639,630210 Wnot recorded by source4 nm—
nvidiaRTX 5000 Ada Generation202332,768 MBGDDR665,280250 Wnot recorded by source4 nm—
amdRadeon RX 7800 XT (Navi 32)202316,384 MBGDDR6—263 Wnot recorded by source——
amdRadeon RX 7900M (Navi 31)202316,384 MBGDDR6—180 Wnot recorded by source——
amdRadeon Pro W7700 (Navi 32)202316,384 MBGDDR628,300190 Wnot recorded by source——
amdAMD Instinct MI300A (Antares)2023131,072 MBHBM3—550 Wnot recorded by source5 nm—
amdAMD Instinct MI300X (Aqua Vanjaram)2023196,608 MBHBM3—750 Wnot recorded by source5 nm—
nvidiaGeForce RTX 4090 D202324,576 MBGDDR6X66,540425 Wnot recorded by source—76,300
nvidiaRTX 5880 Ada Generation202449,152 MBGDDR669,270285 Wnot recorded by source4 nm—
amdRadeon RX 7600 XT (Navi 33)202416,384 MBGDDR6—190 Wnot recorded by source6 nm—
nvidiaGeForce RTX 4070 Ti Super202416,384 MBGDDR6X39,537285 Wnot recorded by source—45,900
nvidiaGeForce RTX 4080 Super202416,384 MBGDDR6X47,002320 Wnot recorded by source—45,900
nvidiaRTX 2000 Ada Generation202416,384 MBGDDR612,00070 Wnot recorded by source4 nm—
nvidiaGeForce RTX 4070 Ti Super202416,384 MBGDDR6X39,537285 Wnot recorded by source—76,300
amdRadeon Pro W7900 Dual Slot (Navi 31)202449,152 MBGDDR661,320295 Wnot recorded by source——
amdRadeon Pro V710 (Navi 32)202428,672 MBGDDR628,314158 Wnot recorded by source——
nvidiaB200 GPU accelerator2024184,320 MBHBM3E74,4501,000 Wnot recorded by source——
nvidiaH200 GPU accelerator (PCIe card)2024144,384 MBHBM3E60,320600 Wnot recorded by source——
amdRadeon Pro W7800 48GB (Navi 31)202449,152 MBGDDR645,250281 Wnot recorded by source——
nvidiaGeForce RTX 5080202516,384 MBGDDR749,352360 Wnot recorded by source—45,600
nvidiaGeForce RTX 5090202532,768 MBGDDR787,780575 Wnot recorded by source—92,200
nvidiaGeForce RTX 5090 D202532,768 MBGDDR787,780575 Wnot recorded by source—92,200
nvidiaGeForce RTX 5070 Ti202516,384 MBGDDR741,126300 Wnot recorded by source—45,600
amdRadeon RX 9070 (Navi 48)202516,384 MBGDDR6—220 Wnot recorded by source——
amdRadeon RX 9070 XT (Navi 48)202516,384 MBGDDR6—304 Wnot recorded by source——
nvidiaRTX PRO 4000 Blackwell202524,576 MBGDDR746,900140 Wnot recorded by source4 nm—
nvidiaRTX PRO 4500 Blackwell202532,768 MBGDDR754,940200 Wnot recorded by source4 nm—
nvidiaRTX PRO 6000 Blackwell Max-Q Workstation Edition202598,304 MBGDDR7110,100300 Wnot recorded by source4 nm—
nvidiaRTX PRO 6000 Blackwell Workstation Edition202598,304 MBGDDR7126,000600 Wnot recorded by source4 nm—
nvidiaGeForce RTX 5060 Ti 16 GB202516,384 MBGDDR722,183180 Wnot recorded by source—21,900
amdRadeon RX 9060 XT (Navi 44)202516,384 MBGDDR625,640160 Wnot recorded by source——
amdAMD Instinct MI350X2025294,912 MBHBM3e—1,000 Wnot recorded by source——
amdAMD Instinct MI355X2025294,912 MBHBM3e—1,400 Wnot recorded by source——
amdRadeon AI Pro R9700 (Navi 48)202532,768 MBGDDR6—300 Wnot recorded by source——
nvidiaRTX PRO 2000 Blackwell202516,384 MBGDDR717,03070 Wnot recorded by source4 nm—
nvidiaGeForce RTX 5090 D V2202524,576 MBGDDR787,780575 Wnot recorded by source—92,200
intelArc Pro B50202516,384 MBGDDR610,65070 Wnot recorded by source5 nm19,600
nvidiaB300 GPU accelerator2025294,912 MBHBM3E76,9901,400 Wnot recorded by source——
amdRadeon RX 7700 (Navi 32)202516,384 MBGDDR6—263 Wnot recorded by source——
amdRadeon AI Pro R9600D (Navi 48)202532,768 MBGDDR624,800150 Wnot recorded by source——
amdRadeon AI Pro R9700S (Navi 48)202532,768 MBGDDR6—300 Wnot recorded by source——
amdRadeon AI Pro R9600 (Navi 48)202632,768 MBGDDR624,800150 Wnot recorded by source——
amdAMD Instinct MI350P2026147,456 MBHBM3e—600 Wnot recorded by source——
amdRadeon RX 7900 GRE (Navi 31)—16,384 MBGDDR6—260 Wnot recorded by source——
intelArc 7 A770M—16,384 MBGDDR6—150 Wnot recorded by source6 nm21,700
intelArc Pro B60—24,576 MBGDDR612,800200 Wnot recorded by source5 nm19,600
intelArc Pro B70—32,768 MBGDDR622,940230 Wnot recorded by source5 nm—
intelData Center GPU Max 1350—98,304 MBHBM2E—450 Wnot recorded by source—100,000
nvidiaB100 GPU accelerator—196,608 MBHBM3E32,9501,000 Wnot recorded by source——
nvidiaRTX PRO 4000 Blackwell Mobile—16,384 MBGDDR723,04080 Wnot recorded by source4 nm—
nvidiaRTX PRO 5000 Blackwell Mobile—24,576 MBGDDR731,80095 Wnot recorded by source4 nm—
Source: Wikipedia, revision 1369488585 · catalog loaded 2026-08-16