Jump to content

RDNA 3: Difference between revisions

From Wikipedia, the free encyclopedia
Browse history interactively
← Previous editNext edit →
Content deleted Content added
VisualWikitext
→Compute Units: clarify the nature of AI acceleration on RDNA 3, replace AMD's performance claims with an independent benchmark.
Line 85: Line 85:
=== Graphics Compute Die (GCD) ===
=== Graphics Compute Die (GCD) ===
==== Compute Units ====
==== Compute Units ====
RDNA 3's Compute Units (CUs) for graphics processing are organized in dual CU Work Group Processors (WGPs). Rather than including a very large number of WGPs in RDNA 3 GPUs, AMD instead focussed on improving per-WGP throughput.
RDNA 3's Compute Units (CUs) for graphics processing are organized in dual CU Work Group Processors (WGPs). Rather than including a very large number of WGPs in RDNA 3 GPUs, AMD instead focused on improving per-WGP throughput.
This is done with improved [[Superscalar processor|dual-issue]] shader ALUs with the ability to execute two instructions per cycle. It can contain up to 96 graphics Compute Units that can provide up to 61 TFLOPS of compute.<ref name="Gula">{{Cite web |last1=Gula |first1=Damien |date=November 3, 2022 |title=AMD's RDNA 3 GPUs are Way Cheaper Than the RTX 4090 |url=https://gizmodo.com/amd-rdna-3-gpu-rtx-4090-4080-rx-7900-xtx-xt-price-date-1849741000 |website=Gizmodo |language=en-US |access-date=April 8, 2023}}</ref> Each RDNA 3 Compute Unit has dedicated AI acceleration with Wave MMA ([[Matrix multiplication|matrix]] [[multiply-accumulate]]) instructions,<ref name="wmma-gpuopen">{{cite web |last=Vasishta |first=Aaryaman |date=January 10, 2023 |title=How to accelerate AI applications on RDNA 3 using WMMA |url=https://gpuopen.com/learn/wmma_on_rdna3/ |website=GPUOpen |language=en-US |access-date=August 14, 2023 |archive-url=https://web.archive.org/web/20230110194947/https://gpuopen.com/learn/wmma_on_rdna3/ |url-status=live |archive-date=January 10, 2023}}</ref> which can improve AI-based performance by 2.7x and also benefits ray tracing instructions, similar to [[Nvidia]]'s Tensor cores.<ref name="Gula"/>
This is done with improved [[Superscalar processor|dual-issue]] shader ALUs with the ability to execute two instructions per cycle. It can contain up to 96 graphics Compute Units that can provide up to 61 TFLOPS of compute.<ref name="Gula">{{Cite web |last1=Gula |first1=Damien |date=November 3, 2022 |title=AMD's RDNA 3 GPUs are Way Cheaper Than the RTX 4090 |url=https://gizmodo.com/amd-rdna-3-gpu-rtx-4090-4080-rx-7900-xtx-xt-price-date-1849741000 |website=Gizmodo |language=en-US |access-date=April 8, 2023}}</ref>

While RDNA 3 doesn't include dedicated execution units for AI acceleration, the efficiency of running inference tasks on [[Half-precision floating-point format|FP16]] execution resources is improved with Wave MMA ([[Matrix multiplication|matrix]] [[multiply-accumulate]]) instructions, leading to increased performance over RDNA 2.<ref name="tomshw-rdna3">{{cite web |last1=Walton |first1=Jarred |title=AMD RDNA 3 and Radeon RX 7000-series GPUs: Everything we know |url=https://www.tomshardware.com/features/amd-radeon-rx-7000-rdna-3-price-performance-benchmarks-release-date#section-amd-rdna-3-architecture-ai-accelerators |website=Tom's Hardware |access-date=20 July 2024 |date=15 June 2024}}</ref><ref name="tomshw-interview">{{cite web |last1=Walton |first1=Jarred |last2=Alcorn |first2=Paul |title=Into the GPU Chiplet Era: An Interview With AMD's Sam Naffziger |url=https://www.tomshardware.com/features/gpu-chiplet-era-interview-amd-sam-naffziger |website=Tom's Hardware |access-date=20 July 2024 |date=23 June 2022 |quote=We asked whether AMD would include some form of tensor core or matrix core in the architecture, similar to what both Nvidia and Intel are doing with their GPUs. He responded that the split between RDNA and CDNA means stuffing a bunch of specialized matrix cores into consumer graphics products really isn't necessary for the target market, plus the FP16 support that already exists in previous RDNA architectures should prove sufficient for inference-type workloads.}}</ref> WMMA supports FP16, BF16, INT8, and INT4 data types.<ref name="wmma-gpuopen">{{cite web |last=Vasishta |first=Aaryaman |date=January 10, 2023 |title=How to accelerate AI applications on RDNA 3 using WMMA |url=https://gpuopen.com/learn/wmma_on_rdna3/ |website=GPUOpen |language=en-US |access-date=August 14, 2023 |archive-url=https://web.archive.org/web/20230110194947/https://gpuopen.com/learn/wmma_on_rdna3/ |url-status=live |archive-date=January 10, 2023}}</ref> ''[[Tom's Hardware]]'' found that AMD's fastest RDNA 3 GPU, the RX 7900 XTX, was capable of generating 26 images per minute with [[Stable Diffusion]], compared to only 6.6 images per minute of the RX 6950 XT, the fastest RDNA 2 GPU.<ref name="tomshw-sd">{{cite web |last1=Walton |first1=Jarred |title=Stable Diffusion Benchmarks: 45 Nvidia, AMD, and Intel GPUs Compared |url=https://www.tomshardware.com/pc-components/gpus/stable-diffusion-benchmarks |website=Tom's Hardware |access-date=20 July 2024 |date=15 December 2023}}</ref>


==== Ray tracing ====
==== Ray tracing ====

Revision as of 10:12, 20 July 2024

RDNA 3
LaunchedDecember 13, 2022; 23 months ago (2022-12-13)
Designed byAMD
Manufactured by
Fabrication process
Codename(s)
  • Navi 31: Plum Bonito
  • Navi 32: Wheat Nas
  • Navi 33: Hotpink Bonefish
Product Series
Desktop
Professional/workstation
Specifications
Compute
  • Up to 122.8 TFLOPS (FP16)
  • Up to 61.42 TFLOPS (FP32)
  • Up to 1.919 TFLOPS (FP64)
Clock rate1500 MHz to 2500 MHz
Shader clock rate2269 MHz
L0 cache64 KB (per WGP):
  • 32 KB instructions cache
  • 16 KB K data cache
L1 cache256 KB (per array)
L2 cache6 MB
L3 cacheup to 96 MB (16 MB per MCD)
Memory supportGDDR6
Memory clock rateup to 20 Gbps
PCIe supportPCIe 4.0
Supported Graphics APIs
Direct3DDirect3D 12.0 Ultimate (feature level 12_2)
Shader ModelShader Model 6.7
OpenCLOpenCL 2.1
OpenGLOpenGL 4.6
VulkanVulkan 1.3
Media Engine
Encode codecs
Decode codecs
Color bit-depth
  • 8-bit
  • 10-bit
  • 12-bit
Encoder(s) supported
  • AMF
  • VCE
Display outputs
History
PredecessorRDNA 2
VariantCDNA 3 (datacenter)
SuccessorRDNA 4
Support status
Supported

RDNA 3 is a GPU microarchitecture designed by AMD, released with the Radeon RX 7000 series on December 13, 2022. Alongside powering the RX 7000 series, RDNA 3 is also featured in the SoCs designed by AMD for the Asus ROG Ally and Lenovo Legion Go consoles.

Background

On June 9, 2022, AMD held their Financial Analyst Day where they presented a client GPU roadmap which contained mention of RDNA 3 coming in 2022 and RDNA 4 coming in 2024.[1] AMD announced to investors their intention to achieve a performance-per-watt uplift of over 50% with RDNA 3 and that the upcoming architecture would be built using chiplet packaging on a 5 nm process.[2]

A sneak preview for RDNA 3 was included towards the end of AMD's Ryzen 7000 unveiling event on August 29, 2022. The preview included RDNA 3 running gameplay of Lies of P, AMD CEO Lisa Su confirming that a chiplet design would be used, and a partial look at AMD's reference design for an RDNA 3 GPU.[3]

Full details for the RDNA 3 architecture were unveiled on November 3, 2022 at an event in Las Vegas.[4]

Architecture

Chiplet packaging

For the first time ever in a consumer GPU, RDNA 3 utilizes modular chiplets rather than a single large monolithic die. AMD previously had great success with its use of chiplets in its Ryzen desktop and Epyc server processors.[5] The decision to move to a chiplet-based GPU microarchitecture was led by AMD Senior Vice President Sam Naffziger who had also lead the chiplet initiative with Ryzen and Epyc.[6] The development of RDNA 3's chiplet architecture began towards the end of 2017 with Naffziger leading the AMD graphics team in the effort.[7] The benefit of using chiplets is that dies can be fabricated on different process nodes depending on their functions and intended purpose. According to Naffziger, cache and SRAM do not scale as linearly as logic does on advanced nodes like N5 in terms of density and power consumption so they can instead be fabricated on the cheaper, more mature N6 node. The use of smaller dies rather than one large monolithic die is beneficial for maximizing wafer yields as more dies can be fitted onto a single wafer.[7] Alternatively, a large monolithic RDNA 3 die built on N5 would be more expensive to produce with lower yields.

RDNA 3 uses two types of chiplets: the Graphics Compute Die (GCD) and Memory Cache Dies (MCDs). On Ryzen and Epyc processors, AMD used its PCIe-based Infinity Fabric protocol with the package's dies connected via traces on an organic substrate. This approach is easily scalable in a cost-effective manner but has the drawbacks of increased latency, increased power consumption when moving data between dies at around 1.5 picojoules per bit, and it cannot achieve the connection density needed for high-bandwidth GPUs.[8] An organic package could not host the number of wires that would be needed to connect multiple dies in a GPU.[9]

RDNA 3's dies are instead connected using TSMC's Integrated Fan-Out Re-Distribution Layer (InFO-RDL) packaging technique which provides a silicon bridge for high bandwidth and high density die-to-die communication.[10] InFO allows dies to be connected without the use of a more costly silicon interposer such as the one used in AMD's Instinct MI200 and MI300 datacenter accelerators. Each Infinity Fanout link has 9.2 Gbps in bandwidth. Naffziger explains that "The bandwidth density that we achieve is almost 10x" with the Infinity Fanout rather than the wires used by Ryzen and Epyc processors. The chiplet interconnects in RDNA achieve cumulative bandwidth of 5.3 TB/s.[10]

Memory Cache Dies (MCDs)

With a respective 2.05 billion transistors, each Memory Cache Die (MCD) contains 16 MB of L3 cache. Theoretically, additional L3 cache could be added to the MCDs via AMD's 3D V-Cache die stacking technology as the MCDs contain unused TSV connection points.[11][12] Also present on each MCD are two physical 32-bit GDDR6 memory interfaces for a combined 64-bit interface per MCD.[13] The Radeon RX 7900 XTX has a 384-bit memory bus through the use of six MCDs while the RX 7900 XT has a 320-bit bus due to its five MCDs.

Graphics Compute Die (GCD)

Compute Units

RDNA 3's Compute Units (CUs) for graphics processing are organized in dual CU Work Group Processors (WGPs). Rather than including a very large number of WGPs in RDNA 3 GPUs, AMD instead focused on improving per-WGP throughput. This is done with improved dual-issue shader ALUs with the ability to execute two instructions per cycle. It can contain up to 96 graphics Compute Units that can provide up to 61 TFLOPS of compute.[14]

While RDNA 3 doesn't include dedicated execution units for AI acceleration, the efficiency of running inference tasks on FP16 execution resources is improved with Wave MMA (matrix multiply-accumulate) instructions, leading to increased performance over RDNA 2.[15][16] WMMA supports FP16, BF16, INT8, and INT4 data types.[17] Tom's Hardware found that AMD's fastest RDNA 3 GPU, the RX 7900 XTX, was capable of generating 26 images per minute with Stable Diffusion, compared to only 6.6 images per minute of the RX 6950 XT, the fastest RDNA 2 GPU.[18]

Ray tracing

RDNA 3 features second generation ray-tracing accelerators. Each Compute Unit contains one ray tracing accelerator. The overall number of ray tracing accelerators is increased due to the higher number of Compute Units, though the number of ray tracing accelerators per Compute Unit has not increased over RDNA 2.

Clock speeds

RDNA 3 was designed to support high clock speeds. On RDNA 3, clock speeds have been decoupled with the front end operating at a 2.5 GHz frequency while the shaders operate at 2.3 GHz. The shaders operating at a lower clock speed gives up to 25% power savings according to AMD and RDNA 3's shader clock speed is still 15% faster than RDNA 2.[19]

Cache and memory subsystem

RDNA 3 increased the capacity of L1 and L2 caches. The 16-way associative L1 cache shared across a shader array is doubled in RDNA 3 to 256 KB. The L2 cache increased from 4 MB on RDNA 2 to 6 MB on RDNA 3. The L3 Infinity Cache has been lowered in capacity from 128 MB to 96 MB and latency has increased as it is physically present on the MCDs rather than being closer to the WGPs within the GCD.[20] The Infinity Cache capacity was decreased due to RDNA 3 having wider a memory interface up to 384-bit whereas RDNA 2 used memory interfaces up to 256-bit. RDNA 3 having a wider 384-bit memory means that its cache hitrate does not have to be as high to still avoid bandwidth bottlenecks as there is higher memory bandwidth.[20] RDNA 3 GPUs use GDDR6 memory rather than faster GDDR6X due to the latter's increased power consumption.

Media engine

RDNA 3 is the first RDNA architecture to have a dedicated media engine. It is built into the GCD and is based on VCN 4.0 encoding and decoding core.[21] AMD's AMF AV1 encoder is comparable in quality to Nvidia's NVENC AV1 encoder but can handle a higher number of simultaneous encoding streams compared to the limit of 3 on the GeForce RTX 40 series.[22]

Supported encoding frame rates (FPS) per resolution and video coding format[23]
Resolution H.264 H.265 AV1
1080p60 360 360 360
1440p60 360 360 360
4K60 180 180 240
8K60 48 48 60

Display engine

RDNA 3 GPUs feature a new display engine called the "Radiance Display Engine". AMD touted its support for DisplayPort 2.1 UHBR 13.5, delivering up to 54Gbps bandwidth for high refresh rates at 4K and 8K resolutions.[24] The Radeon Pro W7900 and W7800 support the 80Gbps UHBR20 standard. DisplayPort 2.1 can support 4K at 480 Hz and 8K at 165 Hz with Display Stream Compression (DSC). The previous DisplayPort 1.4 standard with DSC was limited to 4K at 240 Hz and 8K at 60 Hz.

Power efficiency

AMD claims that RDNA 3 achieves a 54% increase in performance-per-watt which is in line with their previous claims of 50% performance-per-watt increases for both RDNA and RDNA 2.

Graphics Compute Die (GCD) Memory Cache Die
(MCD)
Navi 31[13] Navi 32 Navi 33
Launch Dec 2022 Sep 2023 Jan 2023 Dec 2022
Codename Plum Bonito Wheat Nas Hotpink Bonefish —
Compute units
(Stream processors)
[FP32 cores] 		96
(6144)
[12288] 		60
(3840)
[7680] 		32
(2048)
[4096]
Process TSMC N5 TSMC N6
Transistors 45.4 bn. 19.9 bn. 13.3 bn. 2.05 bn.
Transistor density 150.2 MTr/mm2 101.5 MTr/mm2 65.2 MTr/mm2 54.64 MTr/mm2
Die size 304.35 mm2 196 mm2 204 mm2 37.52 mm2
Max TDP 405 W 263 W 200 W —
Products Consumer Desktop
  • RX 7900 GRE
  • RX 7900 XT
  • RX 7900 XTX
  • RX 7700 XT
  • RX 7800 XT
  • RX 7600
  • RX 7600 XT
  • RX 7700 XT (3×)
  • RX 7800 XT (4×)
  • RX 7900 GRE (4×)
  • RX 7900 XT (5×)
  • RX 7900 XTX (6×)
Mobile
  • RX 7900M
 —
  • RX 7600S
  • RX 7600M
  • RX 7600M XT
  • RX 7700S
  • RX 7900M (4×)
Workstation Desktop
  • W7800
  • W7900
  • W7700
  • W7500
  • W7600
  • W7700 (4×)
  • W7800 (4×)
  • W7900 (6×)
Mobile — — — —

Products

Gaming

Desktop

Model
(Code name) 		Release Date
& Price 		Architecture
& fab 		Chiplets Transistors
& die size[i] 		Core Fillrate[ii][iii][iv] Processing power[ii][v]
(TFLOPS) 		Infinity Cache Memory TBP Bus
interface
Config[vi] Clock[ii]
(MHz) 		Texture
(GT/s) 		Pixel
(GP/s) 		Half Single Double Size Bandwidth
(GB/s) 		Size Bandwidth
(GB/s) 		Bus type
& width 		Clock
(MT/s)

Radeon RX 7600
(Navi 33)[25] 		May 25, 2023
$269 USD 		RDNA 3
TSMC N6 		Monolithic 13.3×109
204 mm2 		2048:128:64:32:64
32 CU 		1720
2655 		220.2
339.8 		110.1
169.9 		28.18
43.50 		14.09
21.75 		0.440
0.680 		32 MB 476.9 8 GB 288 GDDR6
128-bit 		18000 165 W PCIe 4.0
×8
Radeon RX 7600 XT
(Navi 33)[26][27] 		Jan 24, 2024
$329 USD 		1720
2755 		220.2
352.6 		110.1
176.3 		28.18
45.14 		14.09
22.57 		0.440
0.705 		16 GB 190 W
Radeon RX 7700 XT
(Navi 32)[28] 		Sep 6, 2023
$449 USD 		RDNA 3
TSMC N5 (GCD)
TSMC N6 (MCD) 		1 × GCD
3 × MCD 		28.1×109
346 mm2 		3456:216:96:54:108
54 CU 		1900
2544 		410.4
549.5 		182.4
244.2 		52.53
70.34 		26.27
35.17 		0.821
1.099 		48 MB 1995 12 GB 432 GDDR6
192-bit 		245 W PCIe 4.0
×16
Radeon RX 7800 XT
(Navi 32)[29] 		Sep 6, 2023
$499 USD 		1 × GCD
4 × MCD 		3840:240:96:60:120
60 CU 		1800
2430 		432
583.2 		172.8
233.2 		55.30
74.65 		27.64
37.32 		0.864
1.166 		64 MB 2708 16 GB 624 GDDR6
256-bit 		19500 263 W
Radeon RX 7900 GRE
(Navi 31)[30] 		Jul 27, 2023
China only,
Feb 27, 2024
$549 USD 		57.7×109
529 mm2 		5120:320:192:80:160
80 CU 		1270
2245 		406.4
718.4 		243.8
431.0 		52.02
91.96 		26.01
45.98 		0.813
1.437 		2250 576 18000 260 W
Radeon RX 7900 XT
(Navi 31)[31] 		Dec 13, 2022
$899 USD 		1 × GCD
5 × MCD 		5376:336:192:84:168
84 CU 		1500
2400 		504.0
806.4 		288.0
460.8 		64.51
103.2 		32.26
51.61 		1.008
1.613 		80 MB 2900 20 GB 800 GDDR6
320-bit 		20000 315 W
Radeon RX 7900 XTX
(Navi 31)[32] 		Dec 13, 2022
$999 USD 		1 × GCD
6 × MCD 		6144:384:192:96:192
96 CU 		1900
2500 		729.6
960.0 		364.8
480.0 		93.39
122.9 		46.69
61.44 		1.459
1.920 		96 MB 3500 24 GB 960 GDDR6
384-bit 		355 W
  1. ^ Approximate die size of all active dies (one GCD and up to six MCD).[33]
  2. ^ a b c Boost values (if available) are stated below the base value in italic.
  3. ^ Texture fillrate is calculated as the number of Texture Mapping Units multiplied by the base (or boost) core clock speed.
  4. ^ Pixel fillrate is calculated as the number of Render Output Units multiplied by the base (or boost) core clock speed.
  5. ^ Precision performance is calculated from the base (or boost) core clock speed based on a FMA operation.
  6. ^ Unified shaders : Texture mapping units : Render output units : Ray accelerators : AI accelerators and Compute Units (CU)

Mobile

Model
(Code name) 		Release
date 		Architecture
& fab 		Chiplets Transistors
& die size 		Core Fillrate[a][b][c] Processing power[a][d]
(TFLOPS)[e] 		Infinity
Cache 		Memory TDP Interface
Config[f] Clock[a]
(MHz) 		Texture
(GT/s) 		Pixel
(GP/s) 		Half Single Double Size Bandwidth
(GB/s) 		Bus type
& width 		Clock
(MT/s)

Radeon RX 7600S
(Navi 33)[34] 		Jan 4, 2023 RDNA 3
TSMC N6 		Monolithic 13.3×109
204 mm2 		1792:112:64:28:56
28 CU 		1500
2200 		168.0
246.4 		96.00
140.8 		21.50
31.54 		10.75
15.77 		0.336
0.493 		32 MB 8 GB 256 GDDR6
128-bit 		16000 75 W PCIe 4.0
×8
Radeon RX 7600M
(Navi 33)[35] 		1500
2410 		168.0
269.9 		96.0
154.2 		21.50
34.55 		10.75
17.28 		0.336
0.540 		90 W
Radeon RX 7600M XT
(Navi 33)[36] 		2048:128:64:32:64
32 CU 		1500
2615 		192.0
334.1 		96.00
167.0 		24.58
42.84 		12.29
21.42 		0.384
0.669 		288 18000 120 W
Radeon RX 7700S
(Navi 33)[37] 		1500
2500 		192.0
320.0 		96.0
160.0 		24.58
40.96 		12.29
20.48 		0.384
0.640 		100 W
Radeon RX 7800M
(Navi 33)[38] 		Sep 11, 2024 28.1×109
346 mm2 		3840:240:96:60:120

60 CU


2145
560.4 		?
71.73
35.87
1.204 		48 MB 12 GB 432 GDDR6

192-bit

180 W ?
Radeon RX 7900M
(Navi 31)[39] 		Oct 19, 2023 RDNA 3
TSMC N5 (GCD)
TSMC N6 (MCD) 		1 × GCD
4 × MCD 		57.7×109
529 mm2 		4608:288:192:72:144
72 CU 		1825
2090 		525.6
601.9 		350.4
401.3 		67.28
77.05 		33.64
38.52 		1.051
1.204 		64 MB 16 GB 576 GDDR6
256-bit 		PCIe 4.0
×16
  1. ^ a b c Boost values (if available) are stated below the base value in italic.
  2. ^ Texture fillrate is calculated as the number of Texture Mapping Units multiplied by the base (or boost) core clock speed.
  3. ^ Pixel fillrate is calculated as the number of Render Output Units multiplied by the base (or boost) core clock speed.
  4. ^ Precision performance is calculated from the base (or boost) core clock speed based on a FMA operation.
  5. ^ GPUs based on RDNA 3 have dual-issue stream processors so that up to two shader instructions can be executed per clock cycle under certain parallelism conditions.
  6. ^ Unified shaders : Texture mapping units : Render output units : Ray accelerators : AI accelerators and Compute units (CU)

Workstation

Main article: Radeon Pro W7000 series

Desktop workstation

Model
(Code name) 		Release Date
& Price 		Architecture
& fab 		Chiplets
(active) 		Transistors
& die size[a] 		Core Fillrate[b][c][d] Processing power[b][e]
(TFLOPS) 		Infinity
Cache 		Memory TDP Bus
interface
Config[f][g] Clock[b]
(MHz) 		Texture
(GT/s) 		Pixel
(GP/s) 		Half Single Double Size Bandwidth
(GB/s) 		Bus type
& width 		Clock
(MT/s)
Radeon Pro W7500
(Navi 33)[40][41] 		Aug 3, 2023
$429 USD 		RDNA 3
TSMC N6 		— 13.3×109
204 mm2 		1792:112:64
28:56:28 CU 		1500
1700 		168.0
190.4 		96.0
108.8
24.37
12.19
0.381 		32 MB 8 GB 172 GDDR6
128-bit 		18000 70 W PCIe 4.0
×8
Radeon Pro W7600
(Navi 33)[40][42] 		Aug 3, 2023
$599 USD 		2048:128:64
32:64:32 CU 		1720
2440 		220.1
312.3 		110.0
156.2
39.98
19.99
0.625 		288 130 W
Radeon Pro W7700
(Navi 32) 		Nov 13, 2023
$999 USD 		RDNA 3
TSMC N5 (GCD)
TSMC N6 (MCD) 		1 × GCD
4 × MCD 		28.1×109
~346 mm2 		3072:192:96
48:96:48 CU 		1900
2600 		364.8
499.2 		182.4
249.2 		56.54 28.3 0.884 64 MB 16 GB 576 GDDR6
256-bit 		190 W PCIe 4.0
×16
Radeon Pro W7800
(Navi 31)[43][44] 		Apr 13, 2023
$2499 USD 		57.7×109
~531 mm2 		4480:280:128
70:128:70 CU 		1855
2499 		519.4
699.7 		237.4
319.8
90.50
45.25
1.414 		32 GB 260 W
Radeon Pro W7900 Dual Slot
(Navi 31)[45][46] 		Jun 19, 2024
$3499 USD 		1 × GCD
6 × MCD 		6144:384:192
96:192:96 CU 		1855
2495 		712.3
958.0 		356.1
479.0
122.6
61.32
1.916 		96 MB 48 GB 864 GDDR6
384-bit 		295 W
Radeon Pro W7900
(Navi 31)[43][47] 		Apr 13, 2023
$3999 USD
  1. ^ Approximate die size of entire MCM package that consists of single GCD (Graphics Compute Die) and six MCDs (Memory Cache Die).
    Radeon Pro W7800 has only four active MCDs, inactive one is for structural support and heat dissipation.
  2. ^ a b c Boost values (if available) are stated below the base value in italic.
  3. ^ Texture fillrate is calculated as the number of Texture Mapping Units multiplied by the base (or boost) core clock speed.
  4. ^ Pixel fillrate is calculated as the number of Render Output Units multiplied by the base (or boost) core clock speed.
  5. ^ Precision performance is calculated from the base (or boost) core clock speed based on a FMA operation.
  6. ^ Unified shaders : Texture mapping units : Render output units : Ray accelerators : AI accelerators and Compute units (CU)
  7. ^ GPUs based on RDNA 3 have dual-issue stream processors so that up to two shader instructions can be executed per clock cycle under certain parallelism conditions.

Integrated graphics processing units (iGPUs)

Model Launch Codename Architecture
& fab 		Die
size 		Core Fillrate[a][b][c] Processing power[a][d]
(GFLOPS) 		Cache TDP
Config[e][f] Clock[a]
(MHz) 		Texture
(GT/s) 		Pixel
(GP/s) 		Half
[FP16] 		Single
[FP32] 		Double
[FP64] 		L0 L1 L2
Radeon 740M Apr 2023 Phoenix RDNA 3
TSMC N4P 		178 mm2 4 CUs
256:16:8:4 		2,500 40.0 20.0 5,120 2,560 80.0 64 KB 512 KB 2 MB 15–30 W
Radeon 760M 8 CUs
512:32:16:8 		1,000
2,600 		32.0
83.2 		21.3
55.5 		4,096
10,649 		2,048
5,324 		64.0
166.4 		128 KB 1 MB 35–65 W
Radeon 780M 12 CUs
768:48:24:12 		1,000
2,800 		40.0 20.0 6,144
17,203 		3,072
8,601 		192
537.6 		192 KB 1.5 MB
Ryzen Z1 Jun 13, 2023 4 CUs
256:16:8:4 		2,500 40.0 20.0 5,120 2,560 80.0 64 KB 512 KB 9–30 W
Ryzen Z1 Extreme 12 CUs
768:48:24:12 		2,800 134.4 67.2 17,203 8,600 268.8 192 KB 1.5 MB
  1. ^ a b c Boost values (if available) are stated below the base value in italic.
  2. ^ Texture fillrate is calculated as the number of Texture Mapping Units multiplied by the base (or boost) core clock speed.
  3. ^ Pixel fillrate is calculated as the number of Render Output Units multiplied by the base (or boost) core clock speed.
  4. ^ Precision performance is calculated from the base (or boost) core clock speed based on a FMA operation.
  5. ^ Compute Units (CUs)
    Stream Processors : Texture mapping units : Render output units : Ray accelerators
  6. ^ GPUs based on RDNA 3 have dual-issue stream processors so that up to two shader instructions can be executed per clock cycle under certain parallelism conditions.

References

  1. ^ Smith, Ryan (June 9, 2022). "AMD's 2022-2024 Client GPU Roadmap: RDNA 3 This Year, RDNA 4 Lands in 2024". AnandTech. Retrieved April 8, 2023.
  2. ^ Walton, Jarred (June 9, 2022). "AMD GPU Roadmap: RDNA 3 With 5nm GPU Chiplets Coming This Year". Tom's Hardware. Retrieved April 8, 2023.
  3. ^ Wickens, Katie (August 31, 2022). "AMD's Lisa Su confirms chiplet-based RDNA 3 GPU architecture". PC Gamer. Retrieved April 8, 2023.
  4. ^ "AMD Unveils World's Most Advanced Gaming Graphics Cards, Built on Groundbreaking AMD RDNA 3 Architecture with Chiplet Design". AMD (Press release). Las Vegas, NV. November 3, 2022. Retrieved April 8, 2023.
  5. ^ James, Dave (June 24, 2022). "AMD suggests a Ryzen-like design for RDNA 3 chiplets would be 'a reasonable inference'". PC Gamer. Retrieved April 8, 2023.
  6. ^ Alcorn, Paul; Walton, Jarred (June 23, 2022). "Into the GPU Chiplet Era: An Interview With AMD's Sam Naffziger". Tom's Hardware. Retrieved April 8, 2023.
  7. ^ a b Brosdahl, Peter (November 22, 2022). "AMD Lead Engineer Sam Naffziger Explains Advantages of RDNA3 Chiplet Design". The FPS Review. Retrieved April 8, 2023.
  8. ^ Walton, Jarred (June 5, 2023). "AMD RDNA 3 GPU Architecture Deep Dive: The Ryzen Moment for GPUs". Tom's Hardware. Retrieved April 29, 2024.
  9. ^ Ridley, Jacob (November 14, 2022). "AMD's Infinity Links is the unsung hero of RDNA 3 and chiplet gaming GPUs". PC Gamer. Retrieved April 29, 2024.
  10. ^ a b "AMD Explains the Economics Behind Chiplets for GPUs". TechPowerUp. November 14, 2022. Retrieved April 8, 2023.
  11. ^ Klotz, Aaron (January 29, 2023). "AMD GPU Appears to Leave Room for Future 3D V-Cache". Tom's Hardware. Retrieved April 8, 2023.
  12. ^ Ridley, Jacob (January 30, 2023). "Tiny spots on AMD's RDNA 3 GPU hint at massive cache potential". PC Gamer. Retrieved April 8, 2023.
  13. ^ a b Walton, Jarred (November 14, 2022). "AMD RDNA 3 GPU Architecture Deep Dive: The Ryzen Moment for GPUs". Tom's Hardware. Retrieved April 8, 2023.
  14. ^ Gula, Damien (November 3, 2022). "AMD's RDNA 3 GPUs are Way Cheaper Than the RTX 4090". Gizmodo. Retrieved April 8, 2023.
  15. ^ Walton, Jarred (June 15, 2024). "AMD RDNA 3 and Radeon RX 7000-series GPUs: Everything we know". Tom's Hardware. Retrieved July 20, 2024.
  16. ^ Walton, Jarred; Alcorn, Paul (June 23, 2022). "Into the GPU Chiplet Era: An Interview With AMD's Sam Naffziger". Tom's Hardware. Retrieved July 20, 2024. We asked whether AMD would include some form of tensor core or matrix core in the architecture, similar to what both Nvidia and Intel are doing with their GPUs. He responded that the split between RDNA and CDNA means stuffing a bunch of specialized matrix cores into consumer graphics products really isn't necessary for the target market, plus the FP16 support that already exists in previous RDNA architectures should prove sufficient for inference-type workloads.
  17. ^ Vasishta, Aaryaman (January 10, 2023). "How to accelerate AI applications on RDNA 3 using WMMA". GPUOpen. Archived from the original on January 10, 2023. Retrieved August 14, 2023.
  18. ^ Walton, Jarred (December 15, 2023). "Stable Diffusion Benchmarks: 45 Nvidia, AMD, and Intel GPUs Compared". Tom's Hardware. Retrieved July 20, 2024.
  19. ^ Olšan, Jan (November 7, 2022). "AMD RDNA 3 details: architecture changes, AI acceleration, DP 2.1". HWCooling. Retrieved April 8, 2023.
  20. ^ a b "Microbenchmarking AMD's RDNA 3 Graphics Architecture". Chips and Cheese. January 7, 2023. Retrieved April 29, 2024.
  21. ^ Shilov, Anton (May 4, 2022). "First Details About AMD's Next Generation Video Engine Revealed". Tom's Hardware. Retrieved April 10, 2023.
  22. ^ Klotz, Aaron (December 12, 2022). "AMD's Radeon RX 7900 AV1 encoder is almost on par with Intel Arc and Nvidia's RTX 40 series". TechSpot. Retrieved April 8, 2023.
  23. ^ Taylor, Adam (December 14, 2022). "Tested: With RDNA 3, AMD Radeon is finally useful for content creators". PCWorld. Retrieved April 8, 2023.
  24. ^ Sag, Anshel (November 14, 2022). "AMD's New Radeon RX 7900XTX And 7900XT Put The Pressure On NVIDIA". Forbes. Retrieved April 8, 2023.
  25. ^ "AMD Radeon RX 7600 Specs". TechPowerUp. Retrieved May 24, 2023.
  26. ^ "AMD Radeon RX 7600 XT Specs". TechPowerUp. Retrieved January 8, 2024.
  27. ^ Mujtaba, Hassan (January 8, 2024). "AMD Radeon RX 7600 XT 16 GB GPU Introduced: Faster 2048 Core RDNA 3 Chip, Double The VRAM of RTX 4060 At $329". Wccftech. Retrieved January 8, 2024.
  28. ^ "AMD Radeon RX 7700 XT Specs". TechPowerUp. Retrieved August 25, 2023.
  29. ^ "AMD Radeon RX 7800 XT Specs". TechPowerUp. Retrieved August 25, 2023.
  30. ^ Walton, Jarred (February 26, 2024). "AMD Radeon RX 7900 GRE review: the lowest binned Navi 31 variant is now globally available, starting at $549". Tom's Hardware. Retrieved March 1, 2024.
  31. ^ "AMD Radeon RX 7900 XT Specs". TechPowerUp. Retrieved November 4, 2022.
  32. ^ "AMD Radeon RX 7900 XTX Specs". TechPowerUp. Retrieved November 4, 2022.
  33. ^ "AMD Unveils World's Most Advanced Gaming Graphics Cards, Built on Groundbreaking AMD RDNA 3 Architecture with Chiplet Design". AMD (Press release). April 21, 2023.
  34. ^ "AMD Radeon RX 7600S Specs". TechPowerUp. Retrieved January 16, 2023.
  35. ^ "AMD Radeon RX 7600M Specs". TechPowerUp. Retrieved January 16, 2023.
  36. ^ "AMD Radeon RX 7600M XT Specs". TechPowerUp. Retrieved April 20, 2023.
  37. ^ "AMD Radeon RX 7700S Specs". TechPowerUp. Retrieved January 16, 2023.
  38. ^ Discuss, btarunr (September 11, 2024). "AMD Intros Radeon RX 7800M Mobile Graphics Based on "Navi 32" Silicon". TechPowerUp. Retrieved November 12, 2024.
  39. ^ "AMD Radeon RX 7900M Specs". TechPowerUp. Retrieved November 15, 2023.
  40. ^ a b "New AMD Radeon PRO W7000 Series Workstation Graphics Cards Deliver Advanced Technologies and Exceptional Performance for Mainstream Professional Workflows" (Press release). AMD. August 3, 2023. Retrieved August 4, 2023.
  41. ^ "AMD Radeon PRO W7500 Specs". TechPowerUp. Retrieved August 4, 2023.
  42. ^ "AMD Radeon PRO W7600 Specs". TechPowerUp. Retrieved August 4, 2023.
  43. ^ a b "AMD Unveils the Most Powerful AMD Radeon PRO Graphics Cards, Offering Unique Features and Leadership Performance to Tackle Heavy to Extreme Professional Workloads" (Press release). AMD. April 13, 2023. Retrieved April 13, 2023.
  44. ^ "AMD Radeon PRO W7800 Specs". TechPowerUp. Retrieved April 13, 2023.
  45. ^ "AMD Unveils Next-Gen "Zen 5" Ryzen Processors to Power Advanced AI Experiences" (Press release). AMD. June 2, 2024. Retrieved June 3, 2024.
  46. ^ Smith, Ryan (June 2, 2024). "AMD Slims Down Compute With Radeon Pro W7900 Dual Slot For AI Inference". www.anandtech.com. Retrieved June 3, 2024.
  47. ^ "AMD Radeon PRO W7900 Specs". TechPowerUp. Retrieved April 13, 2023.
Retrieved from "https://en.wikipedia.org/enwiki/w/index.php?title=RDNA_3&oldid=1235633992"