The GeForce 30 series is a suite of graphics processing units (GPUs) designed and marketed by Nvidia, succeeding the GeForce 20 series. The GeForce 30 series is based on the Ampere architecture, which features Nvidia's second-generation ray tracing (RT) cores and third-generation Tensor Cores.[3] Through Nvidia RTX, hardware-enabled ray tracing is possible on GeForce 30 series cards.
  GeForce RTX 3090 Founders Edition | |
| Release date | September 17, 2020 |
|---|---|
| Manufactured by | Samsung |
| Designed by | Nvidia |
| Marketed by | Nvidia |
| Codename | GA10x |
| Architecture | Ampere |
| Models | GeForce RTX series |
| Transistors |
|
| Fabrication process | Samsung 8LPU |
| Cards | |
| Entry-level |
|
| Mid-range |
|
| High-end |
|
| Enthusiast |
|
| API support | |
| DirectX | Direct3D 12.0 Ultimate (feature level 12_2) Shader Model 6.8 |
| OpenCL | OpenCL 3.0[1][a] |
| OpenGL | OpenGL 4.6 |
| Vulkan | Vulkan 1.3[2] |
| History | |
| Predecessor | |
| Successor | GeForce 40 series |
| Support status | |
| Supported | |
The lineup, designed to compete with AMD's Radeon RX 6000 series of cards, consists of the entry-level and previously laptop-exclusive RTX 3050 and laptop-exclusive RTX 3050 Ti, mid-range RTX 3060, upper-midrange RTX 3060 Ti, high-end RTX 3070, RTX 3070 Ti, RTX 3080 10 GB, RTX 3080 12 GB and enthusiast RTX 3080 Ti, RTX 3090, and RTX 3090 Ti. This is the last generation from Nvidia to have official support for Windows 7 and 8.x as the latest drivers available for this generation require Windows 10.[4]
The GeForce 30 series began shipping on September 17, 2020.[5][6] The initial launch, consisting of the RTX 3070, RTX 3080, and RTX 3090, occurred during the 2020–2023 global chip shortage, resulting in widespread and notable shortages of the series as a whole that lasted from the series' launch until 2022.
The GeForce 30 series was succeeded by the GeForce 40 series, powered by the Ada Lovelace microarchitecture, which first launched in 2022.[7]
Release and availability issues
editLaunch day for the RTX 3080 was September 17, 2020.[8] The lack of pre-order functionality and high demand exacerbated by the COVID-19 pandemic, and boost in scalping, resulted in a large number of online retailers struggling with the sheer number of purchases.[8][9] Newegg had completely sold out as expected on Black Friday. Long lines formed outside physical stores with stock, such as Micro Center in the United States,[10] and Dospara in Japan.[11] Twitter users reported that they used bots to buy large numbers of cards to resell for higher prices.[12]
Nvidia released a statement the following day, apologizing for the difficulties with their online store, which went down on launch day due to high traffic.[13] On October 2, Nvidia announced that it would delay the release of RTX 3070 cards by two weeks to guarantee availability.[14] On October 5, Nvidia CEO Jensen Huang announced delays due to supply shortages, which were expected to go on until 2021.[15] On October 9, the company announced that all Founder's Edition graphic cards in the United States would temporarily be sold via Best Buy, while the official web store would be upgraded to improve the shopping experience.[16][17][18] In early December, Nvidia blamed continued component shortages on Samsung wafer shortages, resulting in chip shortages, among other factors.[19]
Shortages of RTX 30 series cards continued into 2021, and has been ongoing ever since its release.[20] In an effort to limit purchases by cryptominers, Nvidia announced in February that the RTX 3060 cards would be able to detect algorithms for mining of the Ethereum cryptocurrency and halve the hash rate.[21] Shortly after release, NVIDIA accidentally released a driver update which disabled the detection.[22] In March, TechRadar reported that the shortages could continue until the third quarter of the year, in part blaming a global GDDR6 memory shortage and the cards' supply being bought out by cryptominers.[23] In April, Hong Kong Customs and Excise seized 300 non-video CMP cards.[24]
Nvidia officially announced new RTX 3080, RTX 3070, RTX 3060 Ti Limited Hash Rate (LHR) SKUs on May 18, 2021, which limits the Ethash mining hash rate.[25]
Nvidia released the RTX 3080 Ti on June 3, and the RTX 3070 Ti one week later, on June 10. Both included the cryptocurrency mining hash rate limiter.[26]
There have been multiple attempts to meet the demand. EVGA has a system designed to put people into a queue; the goal is to prevent cards from selling out quickly and decrease wait time.[27]
Besides the Founders Edition, there are also a variety of aftermarket cards, referred to as "custom cards" by Nvidia. The graphics chip is taken from Nvidia and parts of the card are modified according to the company's specifications. Such modifications include: clock speeds, fans, heat sinks, connectors, and aesthetics. Such aftermarket cards tend to perform better than Nvidia's in-house cards, but such increased performance comes with tradeoffs in temperature and power consumption. Manufacturers include: Gigabyte, MSI, ZOTAC, Asus, EVGA, INNO3D.[28][29][30][31][32]
Nvidia officially released the GeForce RTX 3080 12GB graphics card on January 11, 2022 and the desktop GeForce RTX 3050 graphics card on January 27, 2022. The RTX 3050 had previously only had a laptop variant that was launched on May 11, 2021 along with the laptop variant of the RTX 3050 Ti that currently still has no desktop variant. Nvidia officially released the GeForce RTX 3090 Ti on March 29, 2022.
Details
editArchitectural improvements of the Ampere architecture include the following:
- CUDA Compute Capability 8.6[33]
- Samsung 8 nm 8N process (custom designed for NVIDIA)[34]
- Doubled FP32 performance per SM on Ampere GPUs
- Third-generation Tensor Cores with FP16, bfloat16, TensorFloat-32 (TF32) and sparsity acceleration
- Second-generation Ray Tracing Cores, plus concurrent ray tracing and shading and compute
- GDDR6X memory support (RTX 3060Ti GDDR6X Version, RTX 3070 Ti, RTX 3080, RTX 3080 12 GB, RTX 3080 Ti, RTX 3090, RTX 3090 Ti)
- PCI Express 4.0
- NVLink 3.0 (RTX 3090, RTX 3090 Ti)
- HDMI 2.1 supporting FRL6 (48 Gbit/s) transmission speed[35]
- PureVideo Feature Set K hardware video decoding with AV1 hardware decoding[36]
Desktop
edit- All the RTX 30 GPUs are made using the 8 nm Samsung node.[37]
- Only the RTX 3090 and RTX 3090 Ti support 2-way NVLink.
- RTX 3050 feature limited 8 lanes for the PCIe 4.0 bus interface. All other cards support the full ×16 bandwidth.
- Double-precision (FP64) performance of the Ampere chips are 1/64 of single-precision (FP32) performance.
| Model | Launch | Launch MSRP (USD) | Code name(s)[b] | Transistors (billion) | Die size (mm2) | Core config[c] | SM count[d] | L2 cache (MB) | Clock speeds[e] | Fillrate[f][g] | Memory | Processing power (TFLOPS)[h] | TDP (watts) | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Core (MHz) | Memory (GT/s)[i] | Pixel (Gpx/s) | Texture (Gtex/s) | Size (GB) | Band- width (GB/s) | Type | Bus width (bit) | Half (boost) | Single (boost) | Double (boost) | Tensor compute [sparse] | ||||||||||
| GeForce RTX 3050[40][41][42][43][44] | Feb 2, 2024 | $169 | GA107-325 | 8.7 | 200 | 2304 72:32:18:72 | 18 | 2 | 1042 (1470) | 14 | 33.34 (47.04) | 75.02 (105.8) | 6 | 168 | GDDR6 | 96 | 4.802 (6.774) | 4.802 (6.774) | 0.075 (0.105) | 30.1 (60.2) | 70 |
| Dec 16, 2022 | $249 | GA107-150 | 2560 80:32:20:80 | 20 | 1552 (1777) | 49.66 (56.86) | 124.2 (142.2) | 8 | 224 | 128 | 7.946 (9.098) | 7.946 (9.098) | 0.124 (0.142) | 36.4 [72.8] | 115 | ||||||
| Jul 18, 2022 | OEM[45] | GA106-150 | 13.25 | 276 | 2304 72:32:18:72 | 18 | 1515 (1755) | 48.48 (56.16) | 109.1 (126.4) | 6.981 (8.087) | 6.981 (8.087) | 0.109 (0.126) | 32.4[34] [64.8] | 130 | |||||||
| Jan 27, 2022 | $249 | 2560 80:32:20:80 | 20 | 1552 (1777) | 49.66 (56.86) | 124.2 (142.2) | 7.946 (9.098) | 7.946 (9.098) | 0.124 (0.142) | 36.4[46] [72.8] | |||||||||||
| GeForce RTX 3060[47][48][49] | Feb 25, 2021 | $329 | GA106-300 | 3584 112:48:28:112 | 28 | 3 | 1320 (1777) | 15.0 | 63.36 (85.29) | 147.8 (199) | 12 | 360 | 192 | 9.462 (12.74) | 9.462 (12.74) | 0.148 (0.199) | 51.2[34] [102.4] | 170 | |||
| May 2021 | GA106-302 | ||||||||||||||||||||
| Oct 2022 | 8 | 240 | 128 | ||||||||||||||||||
| Sep 2021 | GA104-150[50] | 17.4 | 392.5 | 12 | 360 | 192 | |||||||||||||||
| GeForce RTX 3060 Ti[47][51][52] | Dec 2, 2020 | $399 | GA104-200 | 4864 152:80:38:152 | 38 | 4 | 1410 (1665) | 14 | 112.8 (133.2) | 214.3 (253.1) | 8 | 448 | 256 | 13.72 (16.2) | 13.72 (16.2) | 0.214 (0.253) | 64.8[34] [129.6] | 200 | |||
| May 2021 | GA104-202 | ||||||||||||||||||||
| Oct 2022 | 9.5 | 608 | GDDR6X | ||||||||||||||||||
| Feb 23, 2022 | GA103-200[53] | 22 | 496 | 14 | 448 | GDDR6 | |||||||||||||||
| GeForce RTX 3070[40][54] | Oct 29, 2020 | $499 | GA104-300 GA104-302 | 17.4 | 392.5 | 5888 184:96:46:184 | 46 | 1500 (1725) | 144 (165.6) | 276 (317.4) | 17.66 (20.31) | 17.66 (20.31) | 0.276 (0.317) | 81.3[34] [162.6] | 220 | ||||||
| GeForce RTX 3070 Ti[40][55][56] | Jun 10, 2021 | $599 | GA104-400 | 6144 192:96:48:192 | 48 | 1580 (1770) | 9.5 | 151.6 (169.9) | 303.3 (339.8) | 608 | GDDR6X | 19.42 (21.7) | 19.42 (21.7) | 0.303 (0.34) | 87[57] [174] | 290 | |||||
| Oct 2022 | GA102-150[58] | 28.3 | 628.4 | 320 | |||||||||||||||||
| GeForce RTX 3080[40][59] | Sep 17, 2020 | $699 | GA102-200 GA102-202 | 8704 272:96:68:272 | 68 | 5 | 1440 (1710) | 138.2 (164.2) | 391.7 (465.1) | 10 | 760 | 320 | 25.07 (29.77) | 25.07 (29.77) | 0.392 (0.465) | 119[34] [238] | |||||
| GeForce RTX 3080 (12 GB)[40][60] | Jan 11, 2022 | $799 | GA102-220 | 8960 280:96:70:280 | 70 | 1260 (1710) | 120.9 (164.2) | 325.8 (478.8) | 12 | 912 | 384 | 22.58 (30.64) | 22.58 (30.64) | 0.353 (0.478) | 122[61] [244] | 350 | |||||
| GeForce RTX 3080 Ti[40][62] | Jun 3, 2021 | $1,199 | GA102-225 | 10240 320:112:80:320 | 80 | 6 | 1365 (1665) | 152.8 (186.5) | 436.8 (532.8) | 27.96 (34.1) | 27.96 (34.1) | 0.437 (0.533) | 136[57] [273] | ||||||||
| GeForce RTX 3090[40][63] | Sep 24, 2020 | $1,499 | GA102-250[64] GA102-300 | 10496 328:112:82:328 | 82 | 1395 (1695) | 9.75 | 156.2 (189.8) | 457.6 (556) | 24 | 936 | 29.28 (35.58) | 29.28 (35.58) | 0.458 (0.556) | 142[34] [284] | ||||||
| GeForce RTX 3090 Ti[40][65] | Mar 29, 2022 | $1,999 | GA102-350 | 10752 336:112:84:336 | 84 | 1560 (1860) | 10.5 | 174.7 (208.3) | 524.1 (625) | 1008 | 33.55 (39.99) | 33.55 (39.99) | 0.524 (0.625) | 160[66] [320] | 450 | ||||||
Laptop
edit- All mobile RTX 30 GPUs support fourth-generation Max-Q. The decision to enable Max-Q Technologies such as Dynamic Boost 2.0 and WhisperMode 2.0 on devices is handled by its manufacturer.[67][68][69]
- All models feature GDDR6 memory.
| Model | Launch | Code name(s) | Transistors (billion) | Die size (mm2) | Core config[c] | SM count[d] | L2 cache (MB) | Clock speeds[e][j] | Fillrate[f][g] | Memory | Processing power (TFLOPS) | TDP (watts) | ||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Core (MHz) | Memory (GT/s) | Pixel (Gpx/s) | Texture (Gtex/s) | Size (GB) | Band- width (GB/s)[j] | Bus width (bit) | Half (boost) | Single (boost) | Double (boost) | |||||||||
| GeForce RTX 3050 Laptop[70][71][72] | May 11, 2021 | GA107 (GN20-P0) | 8.7 | 200 | 2048 64:32:16:64 | 16 | 2 | 712–1530 (1057–1740) | 12 | 48.96 (55.68) | 97.92 (111.4) | 4 | 192 | 128 | 2.916 (4.329) | 2.916 (4.329) | 0.046 (0.067) | 35–80 |
| Jul 6, 2022 | GA107 (GN20-P0-R) | 2560 80:32:20:80 | 20 | 6 | 144 | 96 | ||||||||||||
| GeForce RTX 3050 Ti Laptop[70][73][74] | May 11, 2021 | GA107 (GN20-P1) | 735–1463 (1035–1695) | 46.82 (54.24) | 117 (135.6) | 4 | 192 | 128 | 3.763 (5.299) | 3.763 (5.299) | 0.059 (0.082) | |||||||
| GeForce RTX 3060 Laptop[70][75][76] | Feb 25, 2021 | GA106 (GN20-E3) | 12 | 276 | 3840 120:48:30:120 | 30 | 3 | 817–1387 (1283–1703) | 14 (12) | 66.58 (81.74) | 166.4 (204.4) | 6 | 336 (288) | 192 | 6.912 (10.94) | 6.912 (10.94) | 0.108 (0.171) | 60–115 |
| GeForce RTX 3070 Laptop[70][77][78] | Jan 12, 2021 | GA104-770 (GN20-E5) | 17.4 | 392 | 5120 160:80:40:160 | 40 | 4 | 780–1215 (1290–1620) | 97.2 (129.6) | 194.4 (259.2) | 8 | 448 (384) | 256 | 11.36 (15.97) | 11.36 (15.97) | 0.178 (0.249) | 80–125 | |
| GeForce RTX 3070 Ti Laptop[70][79][80] | Feb 1, 2022 | GA104 (GN20-E6) | 5888 184:96:46:184 | 46 | 510–1035 (1035–1485) | 99.36 (142.6) | 190.4 (273.2) | 12.93 (14.62) | 12.93 (14.62) | 0.202 (0.228) | ||||||||
| GeForce RTX 3080 Laptop[70][81][82] | Jan 12, 2021 | GA104-775 (GN20-E7) | 6144 192:96:48:192 | 48 | 780–1350 (1245–1710) | 129.6 (164.2) | 259.2 (328.3) | 8 16 | 13.64 (18.98) | 13.64 (18.98) | 0.213 (0.296) | 80–150 | ||||||
| GeForce RTX 3080 Ti Laptop[70][83][69] | Feb 1, 2022 | GA103 (GN20-E8) | 22 | 496 | 7424 232:96:58:232 | 58 | 585–1230 (1125–1590) | 16 (12) | 118.1 (152.6) | 285.4 (368.9) | 16 | 512 (384) | 14.43 (18.71) | 14.43 (18.71) | 0.225 (0.292) | |||
Reception
editRTX 3090 Ti
editAccording to TechRadar, "the Nvidia GeForce RTX 3090 Ti is an incredibly powerful graphics card". However, the problem would be that the RTX 3090 offers similar performance while "it costs significantly less."[84] In addition to its price, they criticized the RTX 3090 Ti's high power consumption.[84] The performance was praised, as the 3090 Ti "will likely be the go-to GPU for creative professionals that need brute force in their day-to-day work." In gaming, "the RTX 3090 Ti fares quite a bit better" compared to the RTX 3090, and even in 8K "will be able to hit a solid 60 fps in many games at high settings."[84]
RTX 3090
editPC Gamer described the Nvidia GeForce RTX 3090 as a "frankly enormous graphics card", which is "supremely powerful" and "more worthy of its Titan credentials than the GeForce branding", as for the average gamer "it doesn't deliver enough over the RTX 3080 to make sense, but for the pro-creator it's a workload-crushing card."[85]
RTX 3080
editGamesRadar praised the RTX 3080 for its capability to "run almost any game at 4K60fps with ray tracing" and found that it offers "4K gaming at an accessible price for the first time ever".[86] According to The Verge, "4K gaming arrived with the RTX 2080 Ti", but "the RTX 3080 does the same and more at a much more affordable price point."[87] Overall, the RTX 3080 would be "great for 1440p or 4K gaming"; however, they criticized the requirement of a 750W power supply as well as the 10GB of VRAM regarding future games.[87]
RTX 3070
editJim McGregor from Forbes found that "the RTX 3070 is the perfect solution for the more mainstream gamer", while "the competitive gamer or high-end creator will still likely opt for the GeForce RTX 3080 or RTX 3090."[88]
See also
editNotes
edit- ^ In OpenCL 3.0, OpenCL 1.2 functionality has become a mandatory baseline, while all OpenCL 2.x and OpenCL 3.0 features were made optional.
- ^ Models with GA1xx-x02 chips feature improved Lite Hash Rate (LHR) limiter for cryptomining.[38][39]
- ^ a b Shader Processors, Texture mapping units : Render output units : Ray tracing cores : Tensor Cores
- ^ a b The number of Streaming multi-processors on the GPU.
- ^ a b Core boost values (if available) are stated below the base value inside bracket.
- ^ a b Pixel fillrate is calculated as the number of render output units (ROPs) multiplied by the base (or boost) core clock speed.
- ^ a b Texture fillrate is calculated as the number of texture mapping units (TMUs) multiplied by the base (or boost) core clock speed.
- ^ Precision performance is calculated from the base (or boost) core clock speed based on a FMA operation. Calculation for accumulated Tensor (FP16) computation is: Tensor Cores * core clock * 256 / 1000.When used with sparsity feature, ratio is 2:1.
- ^ Cards with GDDR6X send two bits per transfer. The transfer rate should not be confused with the bit rate.
- ^ a b Max-Q design values are also accounted inside final clock rate value. For models with Max-Q technology the effective speeds of memory transfer and bandwidth is stated below inside brackets.
References
editExternal links
edit- Official website
- NVIDIA AMPERE GA102 GPU ARCHITECTURE whitepaper
Media related to Nvidia GeForce 30 series video cards at Wikimedia Commons
