AMD Radeon RX 7900 XT turns out to be a completely different RDNA 3 ‘Navi 31’ GPU-powered beast in the latest rumors from Greymon55. The doctor has reported that based on new information, AMD has created a much more powerful design than previously expected.
The AMD Radeon RX 7900 XT, the flagship RDNA 3 ‘Navi 31’ GPU-powered graphics card, could offer over 3 GHz clock speeds and almost 100 TFLOP performance
Over the last few days, there have been several observations of the AMD RDNA 3 (GFX11) GPUs that have begun to appear in the LLVM project in the Linux operating system drivers and development tools. It appears that AMD is working on at least four new GPUs so far, which have been confirmed through their ID leak, which is as follows (Credits: @ Kepler_L2):
- GFX (1100) – Possible Navi 31 GPU
- GFX (1101) – Possible Navi 32 GPU
- GFX (1102) – Possible Navi 33 GPU
- (GFX 1103) – Possible Phoenix APU
Although these IDs do not reveal much information such as specifications or configurations, they tell us that preliminary support is being added and soon these GPUs will start testing within the respective platforms for optimized performance. But this topic is not about the leaked IDs, it’s about the latest rumor surrounding the flagship RDNA 3 GPU, the Navi 31, which will power the Radeon RX 7900 series’ cards.
According to the latest rumor from Greymon55, it seems that the clock speeds have experienced a big boost from 2.5 GHz to now hovering around and even above the 3 GHz frequency mark. What this will mean is that the flagship chip with 15360 stream processors will be able to deliver nearly 100.0 TFLOPs of FP32 computing power (92 TFLOPs to be exact). It’s without adding the OC limits, which will definitely help it push closer to the already close 100 TFLOPs barrier.
≈92T FP32
– Greymon55 (@ greymon55) April 30, 2022
3GHz + 7900xt
– Greymon55 (@ greymon55) April 30, 2022
But it seems that the latest reports about NVIDIA making a major change to its Ada Lovelace GPU architecture to the GeForce RTX 40 series have also been taken into account, and leakers expect that Ada Lovelace flagships will not only be the first to break 100 TFLOPs GPU Calculate barrier, but also deliver clock speeds similar to AMD’s RDNA 3, which could be thanks to the shift from Samsung’s 8nm to TSMC’s bleeding-edge 4N node, which is an optimized variant based on the TSMC N5 process ( 5nm).
AMD RDNA 3 Navi 31 GPUs for the Radeon RX 7900 Series
The AMD Navi 31 GPU, the flagship RDNA 3 chip, would power the next generation of enthusiast cards like 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 their next generation RDNA 3 GPUs. Each WGP will house dual CU (Compute Units), but with twice the SIMD32 clusters as opposed to only 2 on each CU within RDNA 2.
According to Linux drivers, it appears that Compute Units still exist and WGPs are still DualCUs. The big change goes back to 4x SIMD per. CU as GCN, but now with RDNA’s 32 ALU design. pic.twitter.com/hAzGs9M9E6
– Kepler (@ Kepler_L2) April 30, 2022
The Navi 31 GPU configuration shown here has two GCDs (Graphics Core Die) and a single MCD (Multi-Cache Die). Each GCD has 3 Shader Engines (6 in total) and each Shader Engine has 2 Shader Arrays (2 per SE / 6 per GCD / 12 in total). Each Shader Array is composed of 5 WGPs (10 per SE / 30 per GCD / 60 in total), and each WGP has 8 SIMD32 units with 32 ALUs (40 SIMD32 per SA / 80 per SE / 240 per GCD / 480 in total). These SIMD32 devices total 7,680 cores per second. GCD and 15,360 cores in total.
A preliminary block diagram of AMD’s next-generation RDNA 3-based Navi 31 GPU that will power the flagship Radeon RX 7900 XT graphics card. (Image credits: Olrak)
Navi 31 (RDNA 3) MCD will be connected to the two GCDs via a next generation Infinity Fabric connection and has 256-512 MB Infinity Cache. Each GPU must also have 4 memory connection links (32-bit). There are a total of 8 32-bit memory controllers for a 256-bit bus interface.
There are also rumors of 3D Stacking on the flagship Navi 31 GPU, and the launch is currently expected at the end of 4. quarter 2022, although it is likely that we may see a paper launch with actual availability starting in Q1 2023.
Upcoming flagship AMD, Intel, NVIDIA GPU specifications (preliminary)
| GPU name | AD102 | Navi 31 | Xe2-HPG |
|---|---|---|---|
| Code name | Ada Lovelace | RDNA 3 | Battlemage |
| Flagship SKU | GeForce RTX 4090 Series | Radeon RX 7900 Series | Arc B900 series |
| GPU process | TSMC 4N | TSMC 5nm + TSMC 6nm | TSCM 5nm? |
| GPU package | Monolithic | MCD (Multi Chiplet Die) | MCM (Multi-Chiplet Module) |
| GPU dies | Mono x 1 | 2 x GCD + 4 x MCD + 1 x IOD | Quad-Tile (tGPU) |
| GPU Mega Clusters | 12 GPCs (Graphics Processing Clusters) | 6 Shader motors | 10 Render Slices |
| GPU Super Clusters | 72 TPC (Texture Processing Clusters) | 30 WGPs (per MCD) 60 WGPs (total) |
40 Xe cores (per tile) 160 Xe cores (total) |
| GPU clusters | 144 stream multiprocessors (SM) | 120 calculation units (CU) 240 calculation units (total) |
1280 car VE (per tile) 5120 Xe VE (total) |
| Kerner (Per Die) | 18432 CUDA Cores | 7680 SPs (Per GCD) 15360 SPs (total) |
20480 ALUs (total) |
| Highest clock | ~ 2.85 GHz | ~ 3.0 GHz | TBD |
| FP32 Compute | ~ 105 TFLOPs | ~ 92 TFLOPs | TBD |
| Memory type | GDDR6X | GDDR6 | GDDR6? |
| Memory capacity | 24 GB | 32 GB | TBD |
| Memory bus | 384-bit | 256-bit | TBD |
| Memory speeds | ~ 21 Gbps | ~ 18 Gbps | TBD |
| Cache subsystems | 96 MB L2 Cache | 512 MB (Infinity Cache) | TBD |
| TBP | ~ 600W | ~ 500W | TBD |
| Launch | Q4 2022 | Q4 2022 | 2023 |
News source: Videocardz