AMD has confirmed the first key details of its next-generation RDNA 3 'Radeon RX' GPUs which will be coming this year with a 50% performance per watt uplift.
AMD RDNA 3 'Radeon RX 7000' GPUs To Deliver Massive >50% Performance Per Watt Uplift With New 5nm Chiplet & Infinity Cache Architecture
AMD confirmed that its RDNA 3 GPUs will be coming later this year with a huge performance uplift. The company's Senior Vice President of Engineering, Radeon Technologies Group, David Wang, said that the next-gen GPUs for Radeon RX 7000 series will offer over 50% performance per watt uplift vs the existing RDNA 2 GPUs. Some of the key features of the RDNA 3 GPUs highlighted by AMD will include:
- 5nm Process Node
- Advanced Chiplet Packaging
- Rearchitected Compute Unit
- Optimized Graphics Pipeline
- Next-Gen AMD Infinity Cache
- >50% Perf/Watt vs RDNA 2
In the information published by AMD, the company highlighted a few key features of its RDNA 3 GPUs that will power the next generation of Radeon RX graphics cards. The RDNA 3 GPU will be based on a 5nm process node and utilize an advanced chiplet packaging that delivers increased performance per watt. Furthermore, the GPU will house a range of new technologies such as a brand new and rearchitected Compute Unit, an optimized graphics pipeline, and the next-gen of Infinity Cache.
AMD RDNA 3 Navi 31 GPUs For Radeon RX 7900 Series
The AMD Navi 31 GPU, the flagship RDNA 3 chip, would power the next-gen enthusiast cards such as the Radeon RX 7900 XT graphics card. We have heard that AMD will drop CU (Compute Units) in favor of WGP (Work Group Processors) on its next-gen RDNA 3 GPUs. Each WGP will house dual CU (Compute Units) but with twice the SIMD32 clusters as opposed to just 2 on each CU within RDNA 2. Rumors are that AMD has the option to select between Samsung and TSMC for the 6nm die.
According to the latest information, the AMD Navi 31 GPU with RDNA 3 architecture is expected to offer a single GCD with 48 WGPs, 12 SAs, and 6 SEs. This will give out a total of 12,288 stream processors which is lower than the previous count. This will also drop the overall compute performance unless AMD goes crazy with over 3.0 GHz clock frequencies on its flagship part. The Navi 31 GPU will also carry 6 MCD's which will feature 64 MB Infinity Cache per die and are also likely to carry the 64-bit (32-bit x 2) memory controllers that will provide the chip with a 384-bit bus interface.
The Navi 31 (RDNA 3) MCDs will be linked to a single GCD via a next-generation Infinity Fabric interconnect and feature 256-384 MB of Infinity Cache. Considering the latest memory design, next-generation Navi 31 GPUs can be equipped with up to 24 GB of VRAM, the same memory capacity as the upcoming and existing flagship from NVIDIA (RTX 3090 / RTX 4090). The newer memory spec can be seen as a reduction over the previously rumored 32 GB memory but it's definitely going to help tone down the pricing of the card and have parity with NVIDIA while Infinity Cache is in the play can deliver competitive performance in higher-resolution gaming.
AMD RDNA 3 Navi 32 GPUs For Radeon RX 7800 Series
The AMD Navi 32 GPU is also one of the two MCM GPUs featured in the RDNA 3 lineup. The GPU will feature a single GCD (Graphics Compute Dies) and four MCDs (Multi-Cache Die). The die is very similar to the flagship Navi 31 GPU but has one less Shader Engine across each die. The AMD Navi 32 GCD is expected to utilize TSMC's 5nm process node while the MCD will be based on a 6nm process node.
The GCD will be equipped with 4 Shader Engines and each Shader Engine has 2 Shader Arrays (4 in total). Each Shader Array is composed of 4 WGPs (8 per SE / 32 in total) and each WGP features 8 SIMD32 units with 32 ALUs (32 SIMD32 per SA / 64 per SE / 128 in total). These SIMD32 units combine to make up 8192 cores in total.
The Navi 32 (RDNA 3) MCDs will be linked to the GCD via a next-generation Infinity Fabric interconnect and feature a rumored 256 MB of Infinity Cache. Each MCD should also feature 2 memory connect links (32-bit). That's a total of 8 32-bit memory controllers for a 256-bit bus interface. The graphics chip should provide up to 16 GB GDDR6 memory, the same amount as the existing Navi 21 GPUs but 33% higher memory capacity than the Navi 22 GPUs.
This would bring the performance of the AMD Radeon RX 7700 series above the RX 6800 and RX 6900 series, offering one big leap in graphics horsepower in games. This would also pump up the power numbers as the RX 6700 XT is currently rated at 230W but that could go up to 270-300W.
AMD RDNA 3 Navi 33 GPUs For Radeon RX 7700 Series
As for specifications, if Navi 33 is similar to Navi 21, then we can expect a total of 80 compute units which will pack a total of 5120 stream processors if AMD keeps the per CU core count to 64 SPs as the existing chips. Now what's interesting is that a previous rumor had hinted that the Navi 31 GPU was also going to feature 80 compute units but would be based on the new RDNA 3 IP. It seems like the 80 CU rumor could also be true but that won't be the full specifications the chip will offer.
AMD Navi 31 chips will end up with an MCM solution and feature dual 80 Compute Unit chiplet dies, and each would feature 5120 cores. This would round up to a maximum of 160 Compute Units & a total of 10,240 stream processors. If that ends up being the case (& it's looking more so than ever that it would be), then we can definitely expect an 80 CU Navi 33 part. The Navi 32 SKU could end up with around 120-140 CUs and that too could be a multi-chip-module design.
This would make up for a very hefty GPU design and we can also expect it to be based on a brand new process node such as TSMC's 5nm. AMD has already patented an active bridge chiplet solution for its next-generation GPUs which comes with an integrated cache and interconnects the multiple dies featured on the Navi 3X (RDNA 3) GPUs, you can read more on that here.
AMD RDNA 3 Navi 3X GPU Configurations (Preliminary)
|GPU Name||Navi 21||Navi 33||Navi 32||Navi 31||Navi 3X|
|Codename||Sienna Cichlid||Hotpink Bonefish||Wheat Nas||Plum Bonito||TBD|
|GPU Package||Monolithic||Monolithic||MCM (1 GCD + 4 MCD)||MCM (1 GCD + 6 MCD)||MCM (TBD)|
|GPU Die Size||520mm2||203mm2 (Only GCD)||200mm2 (Only GCD)|
425mm2 (with MCDs)
|308mm2 (Only GCD)|
533mm2 (with MCDs)
|SPs Per WGP||128||256||256||256||256|
|Compute Units (Per Die)||80||32||60||96||128 (per GPU)
|Cores (Per Die)||5120||4096||7680||12288||8192|
|Memory Bus||256-bit||128-bit||256-bit||384-bit||384-bit x2?|
|Memory Capacity||Up To 16 GB||Up To 8 GB||Up To 16 GB||Up To 24 GB||Up To 32 GB|
|Memory Speed||16-18 Gbps||TBD||TBD||20 Gbps||TBD|
|Memory Bandwidth||512-576 GB/s||TBD||TBD||960 GB/s||TBD|
|Infinity Cache||128 MB||32 MB||64 MB||96/192 MB||TBD|
|Flagship SKU||Radeon RX 6900 XTX||Radeon RX 7600 XT?||Radeon RX 7800 XT?|
Radeon RX 7700 XT?
|Radeon RX 7900 XT?||Radeon Pro|
|Launch||Q4 2020||Q4 2022?||Q4 2022?||Q4 2022?||2023?|