The Radeon RX Vega series is a series of graphics processors developed by AMD. These GPUs use the Graphics Core Next (GCN) 5th generation architecture, codenamed Vega, and are manufactured on 14 nm FinFET technology. They are developed by Samsung Electronics and licensed to GlobalFoundries. The series consists of desktop graphics cards and APUs aimed at desktops, mobile devices, and embedded applications.

The lineup was released on 14 August 2017. It included the RX Vega 56 and the RX Vega 64, priced at $399 and $499 respectively. These were followed by two mobile APUs, the Ryzen 2500U and Ryzen 2700U, in October 2017. February 2018 saw the release of two desktop APUs, the Ryzen 3 2200G and the Ryzen 5 2400G, and the Ryzen Embedded V1000 line of APUs. In September 2018 AMD announced several Vega APUs in their Athlon line of products. Later in January 2019, the Radeon VII was announced based on the 7nm FinFET node manufactured by TSMC.

History

The Vega microarchitecture was AMD's high-end graphics cards line, and is the successor to the R9 300 series enthusiast Fury products. Partial specifications of the architecture and Vega 10 GPU were announced with the Radeon Instinct MI25 in December 2016. AMD later released the details of the Vega architecture.

Announcement

Vega was originally announced at AMD's CES 2017 presentation on 5 January 2017, alongside the Zen line of CPUs.

New features

Vega targets increased instructions per clock, higher clock speeds, and support for HBM2.

AMD's Vega has new memory hierarchy with high-bandwidth cache and its controller.[citation needed]

Support for HBM2 featuring double the bandwidth-per-pin over previous generation HBM. HBM2 allows for higher capacities with less than half the footprint of GDDR5 memory. Vega architecture is optimized for streaming very large datasets and can work with a variety of memory types with up to 512TB of virtual address space.[citation needed]

Primitive shader for improved geometry processing. Replaces vertex and geometry shaders in geometry processing pipelines with a more programmable single stage. The primitive shader stage is more efficient, introduces intelligent load balancing technologies and higher throughput.

NCU: The Vega GPU introduces the Next-Gen Compute Unit. Versatile architecture featuring flexible compute units that can natively process 8-bit, 16-bit, 32-bit or 64-bit operations in each clock cycle. And run at higher frequencies. Vega brings support for Rapid Packed Math, processing two half-precision (16-bit) in the same time as a single 32-bit floating-point operation. Up to 128 32-bit, 256 16-bit or 512 8-bit ops per clock are possible with the Vega architecture.

Draw Stream Binning Rasterizer designed for higher performance and power efficiency. It allows for "fetch once, shade once" of pixels through the use of a smart on-chip bin cache and early culling of pixels invisible in a final scene.[citation needed]

Vega bumps Direct3D feature level support from 12_0 to 12_1.[citation needed]

Vega's rasteriser brings hardware-acceleration support for Rasterizer Ordered Views and Conservative Rasterisation Tier 3.

Products

RX Vega branded discrete graphics

Model (Code name)ReleaseDate &PriceArchitecture &fabTransistors &die sizeCoreFillrateProcessingpower (GFLOPS)MemoryTBPBus interface
ConfigClock (MHz)Texture (GT/s)Pixel (GP/s)HalfSingleDoubleSize (GB)Bandwidth (GB/s)Bustype &widthClock (MT/s)
Radeon RX Vega 56 (Vega 10)Aug 28, 2017 $399 USDGCN 5 GloFo 14LPP12.5×109 486 mm23584:224:64 56 CU1156 1471258.9 329.573.98 94.1416,572 21,0888,286 10,544517.9 659.08409.6HBM2 2048-bit1600210WPCIe 3.0 ×16
Radeon RX Vega 64 (Vega 10)Aug 14, 2017 $499 USD4096:256:64 64 CU1247 1546319.2 395.879.81 98.9420,431 25,33010,215 12,665638.5 791.6483.81890295W
Radeon RX Vega 64 Liquid (Vega 10)Aug 14, 2017 $699 USD1406 1677359.9 429.389.98 107.323,036 27,47611,518 13,738719.9 858.6345W
  • v
  • t
  • e

Radeon VII branded discrete graphics

Model (Code name)Release Date & PriceArchitecture & fabTransistors & die sizeCoreFillrateProcessing power (GFLOPS)MemoryTBPBus interface
ConfigClock (MHz)Texture (GT/s)Pixel (GP/s)HalfSingleDoubleSize (GB)Bandwidth (GB/s)Bus type & widthClock (MT/s)
Radeon VII (Vega 20)Feb 7, 2019 $699 USDGCN 5 TSMC CLN7FF13.23×109 331 mm23840:240:64 60 CU1400 1800336.0 420.089.60 112.021,504 27,64810,752 13,8242,688 3,459161024HBM2 4096-bit2000300 WPCIe 3.0 ×16
  • v
  • t
  • e

Workstation GPUs

Model (Code name)Release date & priceArchitecture & fabTransistors & die sizeCoreFillrateProcessingpower (GFLOPS)MemoryTBPBus interfaceGraphic output ports
ConfigClock (MHz)Texture (GT/s)Pixel (GP/s)HalfSingleDoubleSize (GB)Bandwidth (GB/s)Bus type & widthClock (MT/s)
Radeon Vega Frontier Edition (Air-cooled) (Vega 10)Jun 27, 2017 $999 USDGCN 5 GloFo 14 nm12.5×109 494 mm24096:256:64 64 CU1382 1600353.8 409.688.4 102.422,643 26,21411,321 13,107707.6 819.216484HBM2 2048-bit1890300 WPCIe 3.0 ×163× DP 1.4a 1× HDMI 2.0b
Radeon Vega Frontier Edition (Liquid-cooled) (Vega 10)Jun 27, 2017 $1,499 USD375 W
  • v
  • t
  • e
Model (Code name)Release date & priceArchitecture & fabTransistors & die sizeCoreFillrateProcessing power (GFLOPS)MemoryTBPBus interfaceGraphic output ports
ConfigClock (MHz)Texture (GT/s)Pixel (GP/s)HalfSingleDoubleSize (GB)Bandwidth (GB/s)Bus type & widthClock (MT/s)
Radeon Pro Vega II (Vega 20)2019 $2,800 USDGCN 5 TSMC 7 nm13.23×109 331 mm24096:256:64 64 CU1720440.3110.128,18014,090880321024HBM2 4096-bit2000475 WPCIe 3.0 ×164× Thunderbolt 3 (USB Type-C) 1× HDMI 2.0b
Radeon Pro Vega II Duo (Vega 20)2019 $5,600 USD2× 4096:256:64 64 CU17202× 440.32× 110.12× 28,1802× 14,0902× 8802× 322× 1024HBM2 2× 4096-bit2000
2×4096:256:64 64 CU
  • v
  • t
  • e

Mobile workstation GPUs

Model (Code name)Release dateArchitecture &fabTransistors & die sizeCoreFillrateProcessing power (GFLOPS)MemoryTDPBus interface
ConfigClock (MHz)Texture (GT/s)Pixel (GP/s)HalfSingleDoubleSize (GB)Bandwidth (GB/s)Bus type & widthClock (MT/s)
Radeon Pro Vega 16 (Vega 12)Nov 14, 2018GCN 5 GloFo 14 nm?1024:64:32 16 CU815 119052.16 76.1626.08 38.083,338 4,8741,669 2,437104.3 152.34307.2HBM2 1024-bit240050 WPCIe 3.0 ×16
Radeon Pro Vega 20 (Vega 12)1280:80:32 20 CU815 128365.20 102.626.08 41.064,173 6,5692,086 3,285130.4 205.3189.4148050 W
Radeon Pro Vega 48 (Vega 10)Mar 19, 201912.5×109 495 mm23072:192:64 48 CU1200230.476.8014,7467,373460.88402.4HBM2 2048-bit1572?
Radeon Pro Vega 56 (Vega 10)Aug 17, 20173584:224:64 56 CU1138 1250254.9 280.072.83 80.0016,314 17,9208,157 8,960509.8 560.0120 W
Radeon Pro Vega 64 (Vega 10)Jun 17, 20174096:256:64 64 CU1250 1350320.0 345.680.00 86.4020,480 22,11810,240 11,059640.0 691.216?
Radeon Pro Vega 64X (Vega 10)Mar 19, 20194096:256:64 64 CU1250 1468320.0 375.880.00 93.9520,480 24,05110,240 12,026640.0 751.6512.02000
  • v
  • t
  • e

Desktop APUs

Raven Ridge (2018)

Common features of Zen based Raven Ridge desktop APUs:

  • Socket: AM4.
  • All the CPUs support DDR4-2666 (DDR4-2933 Ryzen) in dual-channel mode.
  • L1 cache: 96 KB (32 KB data + 64 KB instruction) per core.
  • L2 cache: 512 KB per core.
  • All the CPUs support 16 PCIe 3.0 lanes.
  • Includes integrated GCN 5th generation GPU.
  • Fabrication process: GlobalFoundries 14LP.
ModelCPUGPUTDPRelease dateRelease price
Cores (threads)Clock rate (GHz)L3 cache (total)ModelConfigClock (MHz)Processing power (GFLOPS)
BaseBoost
2 (4)3.2—N/a4 MBVega 3192:12:4 3 CU100038435 WSep 6, 2018US $55
OEM
3.4Dec 21, 2018US $65
3.5US $75
3.41100424.4Jul 7, 2019OEM
Sep 30, 2019
3.5Jul 7, 2019
Nov 19, 2019US $49
3.4Jul 21, 2020OEM
Ryzen 3 Pro 2100GE3.210003842019
4 (4)3.6Vega 8512:32:16 8 CU11001126Apr 19, 2018
May 10, 2018
3.53.765 WFeb 12, 2018US $99
May 10, 2018OEM
4 (8)3.23.8RX Vega 11704:44:16 11 CU1250176035 WApr 19, 2018
Vega 11May 10, 2018
3.63.9RX Vega 1165 WFeb 12, 2018US $169
Vega 11May 10, 2018OEM
  • v
  • t
  • e

Picasso (2019)

Common features of Zen+ based desktop APUs:

  • Socket: AM4.
  • All the CPUs support DDR4-2933 in dual-channel mode, while Athlon Pro 300GE and Athlon Silver Pro 3125GE support only DDR4-2666.
  • L1 cache: 96 KB (32 KB data + 64 KB instruction) per core.
  • L2 cache: 512 KB per core.
  • All the CPUs support 16 PCIe 3.0 lanes.
  • Includes integrated GCN 5th generation GPU.
  • Fabrication process: GlobalFoundries 12LP.
ModelCPUGPUTDPRelease dateRelease price
Cores (threads)Clock rate (GHz)L3 cache (total)ModelConfigClock (MHz)Processing power (GFLOPS)
BaseBoost
2 (4)3.4—N/a4 MBVega 3192:12:4 3 CU1100424.435 WSep 30, 2019OEM
Radeon GraphicsJul 21, 2020
4 (4)3.33.8
3.53.965 W
3.33.8Vega 8512:32:16 8 CU12001228.835 WJul 7, 2019
Sep 30, 2019
3.64.01250128065 WJul 7, 2019US $99
Sep 30, 2019OEM
3.33.9Radeon Graphics640:40:16 10 CU1200153635 WJul 21, 2020
4 (8)3.64.013001830.465 W
3.3Vega 11704:44:16 11 CU35 WJul 7, 2019
Sep 30, 2019
3.74.2RX Vega 1114001971.265 WJul 7, 2019US $149
Vega 11Sep 30, 2019OEM
  • v
  • t
  • e

Renoir (2020)

Common features of Ryzen 4000 desktop APUs:

  • Socket: AM4.
  • All the CPUs support DDR4-3200 in dual-channel mode.
  • L1 cache: 64 KB (32 KB data + 32 KB instruction) per core.
  • L2 cache: 512 KB per core.
  • All the CPUs support 24 PCIe 3.0 lanes. 4 of the lanes are reserved as link to the chipset.
  • Includes integrated GCN 5th generation GPU.
  • Fabrication process: TSMC 7FF.
Branding and modelCPUGPUTDPRelease dateRelease price
Cores (threads)Clock rate (GHz)L3 cache (total)Core ConfigModelClock (GHz)ConfigProcessing power (GFLOPS)
BaseBoost
Ryzen 78 (16)3.64.48 MB2 × 4Radeon Graphics2.1512:32:16 8 CU2150.465 WJul 21, 2020OEM
3.14.32.0204835 W
Ryzen 56 (12)3.74.22 × 31.9448:28:14 7 CU1702.465 WJul 21, 2020 (OEM) Apr 4, 2022 (retail)US $154
3.335 WJul 21, 2020OEM
Ryzen 34 (8)3.84.04 MB1 × 41.7384:24:12 6 CU1305.665 W
3.535 W
  • v
  • t
  • e

Cezanne (2021)

Common features of Ryzen 5000 desktop APUs:

  • Socket: AM4.
  • All the CPUs support DDR4-3200 in dual-channel mode.
  • L1 cache: 64 KB (32 KB data + 32 KB instruction) per core.
  • L2 cache: 512 KB per core.
  • All the CPUs support 24 PCIe 3.0 lanes. 4 of the lanes are reserved as link to the chipset.
  • Includes integrated GCN 5th generation GPU.
  • Fabrication process: TSMC 7FF.
Branding and modelCPUGPUThermal solutionTDPRelease dateMSRP
Cores (threads)Clock rate (GHz)L3 cache (total)Core configClock (MHz)ConfigProcessing power (GFLOPS)
BaseBoost
Ryzen 78 (16)3.84.616 MB1 × 82000512:32:8 8 CU2048—N/a65 W
Wraith StealthApr 13, 2021 (OEM), Aug 5, 2021 (retail)US $359
3.2—N/a35 W
Wraith StealthApr 13, 2021OEM
Ryzen 56 (12)3.61 × 61900448:28:8 7 CU1702.465 WJan 31, 2024US $140
3.94.4—N/a
Wraith StealthApr 13, 2021 (OEM), Aug 5, 2021 (retail)US $259
3.4—N/a35 W
Wraith StealthApr 13, 2021OEM
3.665 WJan 31, 2024US $125
Ryzen 34 (8)4.04.28 MB1 × 41700384:24:8 6 CU1305.6—N/a
OEMApr 13, 2021OEM
3.6—N/a35 W
OEMApr 13, 2021OEM
  • v
  • t
  • e

Mobile APUs

Raven Ridge (2017)

ModelRelease dateFabCPUGPUSocketPCIe lanesMemory supportTDP
Cores (threads)Clock rate (GHz)CacheModelConfigClock (MHz)Processing power (GFLOPS)
BaseBoostL1L2L3
2019GloFo 14LP2 (4)2.33.264 KB inst. 32 KB data per core512 KB per core4 MBRadeon Vega 3192:12:4 3 CU1000384FP512 (8+4)DDR4-2400 dual-channel12–25 W
Jan 6, 20192.43.3
Jan 8, 20182.53.41100422.4
Jan 6, 20192.63.51200460.8
Jan 8, 20184 (4)2.03.4Radeon Vega 6384:24:8 6 CU1100844.8
May 15, 2018
Oct 26, 20174 (8)3.6Radeon Vega 8512:32:16 8 CU1126.4
May 15, 2018
Sep 10, 20183.2DDR4-3200 dual-channel35–54 W
Oct 26, 20172.23.8Radeon RX Vega 10640:40:16 10 CU13001664DDR4-2400 dual-channel12–25 W
May 15, 2018Radeon Vega 10
Sep 10, 20183.3Radeon RX Vega 11704:44:16 11 CU1830.4DDR4-3200 dual-channel35–54 W
  • v
  • t
  • e

Picasso (2019)

Common features of Ryzen 3000 notebook APUs:

  • Socket: FP5.
  • All the CPUs support DDR4-2400 in dual-channel mode.
  • L1 cache: 96 KB (32 KB data + 64 KB instruction) per core.
  • L2 cache: 512 KB per core.
  • All the CPUs support 16 PCIe 3.0 lanes.
  • Includes integrated GCN 5th generation GPU.
  • Fabrication process: GlobalFoundries 12LP (14LP+).
Branding and ModelCPUGPUTDPRelease date
Cores (threads)Clock rate (GHz)L3 cache (total)Core configModelClock (GHz)ConfigProcessing power (GFLOPS)
BaseBoost
Ryzen 73780U4 (8)2.34.04 MB1 × 4RX Vega 111.4704:44:16 11 CU1971.215 WOct 2019
3750HRX Vega 10640:40:16 10 CU1792.035 WJan 6, 2019
3700C15 WSep 22, 2020
3700UJan 6, 2019
Ryzen 53580U2.13.7Vega 91.3576:36:16 9 CU1497.6Oct 2019
3550HVega 81.2512:32:8 8 CU1228.835 WJan 6, 2019
3500C15 WSep 22, 2020
3500UJan 6, 2019
3450U3.5Jun 2020
Ryzen 33350U4 (4)Vega 6384:24:8 6 CU921.6Jan 6, 2019
3300U
  • v
  • t
  • e

Dalí (2020)

ModelRelease dateFabCPUGPUSocketPCIe lanesMemory supportTDPPart number
Cores (threads)Clock rate (GHz)CacheModelConfigClock (GHz)Processing power (GFLOPS)
BaseBoostL1L2L3
Jan 6, 202014 nm2 (2)1.22.664 KB inst. 32 KB data per core512 KB per core4 MBRadeon Graphics (Vega)192:12:4 3 CU1.0384FP512 (8+4)DDR4-2400 dual-channel6 WYM3020C7T2OFG
Q1 20212.33.2128:8:4 2 CU1.1281.615 WYM3045C4T2OFG
Jan 6, 2020YM3050C4T2OFG
Sep 22, 2020YM305CC4T2OFG
Jan 6, 20202 (4)1.42.8192:12:4 3 CU1.03846 WYM3050C7T2OFG
Q1 20212.43.315 WYM3145C4T2OFG
Jan 6, 2020YM3150C4T2OFG
Sep 22, 2020YM315CC4T2OFG
Jan 6, 20202.63.51.2460.8YM3250C4T2OFG
Sep 22, 2020YM325CC4T2OFG
  • v
  • t
  • e

Renoir (2020)

Common features of Ryzen 4000 notebook APUs:

  • Socket: FP6.
  • All the CPUs support DDR4-3200 or LPDDR4-4266 in dual-channel mode.
  • L1 cache: 64 KB (32 KB data + 32 KB instruction) per core.
  • L2 cache: 512 KB per core.
  • All the CPUs support 16 PCIe 3.0 lanes.
  • Includes integrated GCN 5th generation GPU.
  • Fabrication process: TSMC 7FF.
Branding and ModelCPUGPUTDPRelease date
Cores (threads)Clock rate (GHz)L3 cache (total)Core configModelClock (GHz)ConfigProcessing power (GFLOPS)
BaseBoost
Ryzen 98 (16)3.34.48 MB2 × 4Radeon Graphics1.75512:32:8 8 CU179235–54WMar 16, 2020
3.04.335W
Ryzen 72.94.21.6448:28:8 7 CU1433.635–54W
35W
2.04.41.95512:32:8 8 CU1996.810–25WApr 13, 2021
1.84.21.751792Mar 16, 2020
8 (8)2.04.11.6448:28:8 7 CU1433.6
Ryzen 56 (12)3.04.02 × 31.5384:24:8 6 CU115235–54W
4600HS35W
2.1448:28:8 7 CU134410–25WApr 13, 2021
384:24:8 6 CU1152Mar 16, 2020
6 (6)2.3
Ryzen 34 (4)2.73.74MB1 × 41.4320:20:8 5 CU896
  • v
  • t
  • e

Lucienne (2021)

Common features of Ryzen 5000 notebook APUs:

  • Socket: FP6.
  • All the CPUs support DDR4-3200 or LPDDR4-4266 in dual-channel mode.
  • L1 cache: 64 KB (32 KB data + 32 KB instruction) per core.
  • L2 cache: 512 KB per core.
  • All the CPUs support 16 PCIe 3.0 lanes.
  • Includes integrated GCN 5th generation GPU.
  • Fabrication process: TSMC 7FF.
Branding and ModelCPUGPUTDPRelease date
Cores (threads)Clock rate (GHz)L3 cache (total)Core configModelClock (GHz)ConfigProcessing power (GFLOPS)
BaseBoost
Ryzen 78 (16)1.84.38MB2 × 4Radeon Graphics1.9512:32:8 8 CU1945.610–25WJan 12, 2021
Ryzen 56 (12)2.14.02 × 31.8448:28:8 7 CU1612.8
Ryzen 34 (8)2.63.84MB1 × 41.5384:24:8 6 CU1152
  • v
  • t
  • e

Cezanne (2021)

Common features of Ryzen 5000 notebook APUs:

  • Socket: FP6.
  • All the CPUs support DDR4-3200 or LPDDR4-4266 in dual-channel mode.
  • L1 cache: 64 KB (32 KB data + 32 KB instruction) per core.
  • L2 cache: 512 KB per core.
  • All the CPUs support 16 PCIe 3.0 lanes.
  • Includes integrated GCN 5th generation GPU.
  • Fabrication process: TSMC 7FF.
Branding and modelCPUGPUTDPRelease date
Cores (threads)Clock rate (GHz)L3 cache (total)Core configModelClock (GHz)ConfigProcessing power (GFLOPS)
BaseBoost (Single Core)Boost (All Core)
Ryzen 98 (16)3.34.84.416 MB1 × 8Radeon Graphics2.1512:32:8 8 CUs2150.435–54 WJan 12, 2021
3.04.035 W
3.34.64.235–54 W
3.04.035 W
Ryzen 73.24.42.0204835–54 W
2.835 W
2.04.515 WJan 4, 2022
1.94.43.410–25 WJan 12, 2021
Ryzen 56 (12)3.34.21 × 61.8448:28:8 7 CUs1612.835–54 W
3.035 W
2.34.33.615 WJan 4, 2022
4.210–25 WJan 12, 2021
4.08 MB1.6384:24:8 6 CUs1228.8
4 (8)3.34.21 × 41.81382.435–54 WJun 23, 2023
Ryzen 32.34.33.81.61228.815 WJan 4, 2022
2.74.110–25 WJan 12, 2021
2 (4)3.0—N/a—N/a1 × 21.2192:12:8 3 CUs460.815 WMay 5, 2022
  • v
  • t
  • e

Embedded APUs

ModelRelease dateFabCPUGPUMemory supportTDPJunction temp. range (°C)
Cores (threads)Clock rate (GHz)CacheModelConfigClock (GHz)Processing power (GFLOPS)
BaseBoostL1L2L3
February 2018GloFo 14LP2 (4)2.33.264KB inst. 32KB data per core512 KB per core4 MBVega 3192:12:16 3 CU1.0384DDR4-2400 dual-channel12–25W0–105
December 20184 (8)2.03.6Vega 8512:32:16 8 CU1.11126.4-40–105
2.2—N/a—N/a0–105
February 20182.03.6Vega 8512:32:16 8 CU1.11126.4
3.25DDR4-3200 dual-channel35–54W
December 20183.35—N/a
February 20183.8Vega 11704:44:16 11 CU1.31830.4
  • v
  • t
  • e
ModelRelease dateFabCPUGPUMemory supportTDP
Cores (threads)Clock rate (GHz)CacheModelConfigClock (GHz)Processing power (GFLOPS)
BaseBoostL1L2L3
February 25, 2020GloFo 14LP2 (2)1.22.664 KB inst. 32 KB data per core512 KB per core4 MBVega 3192:12:4 3 CU1.0384DDR4-2400 single-channel6W
2 (4)1.52.8DDR4-2400 dual-channel8-10W
April 16, 20192.43.312–25W
2.63.51.2460.8
  • v
  • t
  • e
ModelRelease dateFabCPUGPUSocketPCIe supportMemory supportTDP
Cores (threads)Clock rate (GHz)CacheArchi- tectureConfigClock (GHz)Processing power (GFLOPS)
BaseBoostL1L2L3
Nov 10, 2020TSMC 7FF6 (12)2.13.9532 KB inst. 32 KB data per core512KB per core8MBGCN 5384:24:8 6 CU1.51152FP620 (8+4+4+4) PCIe 3.0DDR4-3200dual-channel LPDDR4X-4266quad-channel10–25W
3.03.9535–54W
Jan 4, 20233.2448:28:8 7 CU1.61433.6
Nov 10, 20208 (16)1.74.1510–25W
2.94.2535–54W
  • v
  • t
  • e

See also

External links