AMD RDNA 3 'Navi 31' MCM GPU said to use up to 384 MB of Infinity Cache in 3D V-Cache

AMD RDNA 3 “Navi 3X” GPU with double cache per compute unit and shader array

AMD listed the latest information on the size of the GPU cache for the upcoming RDNA 3 “Navi 3X” in a Linux patch.

AMD’s Next-Generation RDNA 3 GPUs for Navi 3X Lineup Double Compute Unit and Shader Array Cache Size

Aaron Liu of AMD published on the FreeDesktop Linux repository and contributed by Coelacanth-Dream and Kepler_L2we’ll see the first details about the cache size in upcoming RDNA 3 GPUs such as the recently leaked Navi 31, Navi 32, and Navi 33 chips.

Speaking of details, the AMD RDNA 3 (GFX11) GPU lineup will feature double the L0 vector cache per compute unit or CU, as well as double the GL1 data cache (RDNA L1 cache per shader array or SA. According to the new info, the vector register file per SIMD will increase from 128KB to 192KB on RDNA 2, the L0 vector/texture cache will increase from 16 KB to 32 KB per CU, and the GPU L1 data cache per shader array will increase from 128 The KB is increased to 256 KB and the L2 data cache will remain the same as in RDNA 2.

Also listed are the cache sizes for AMD’s Navi 33 and Phoenix APUs, which will also feature the RDNA 3 graphics core, but in a monolithic package. The L0 vector/texture size was increased from 16 KB to 32 KB, and the L1 data cache (graphics) was increased from 128 KB to 256 KB. Register file size remains unchanged on Navi 33 GPUs and Phoenix APUs.

cache information Yellow Carp (Rembrandt) RDNA 3 (GFX11 Navigation 31/32) Phoenix (GC 11.0.1, GFX1103)
L0 vector register file per SIMD 128KiB 192KiB 128KiB
L0 vector data (per CU) 16KiB 32KiB 32KiB
L1 Scalar Institute (according to WGP) 32KiB 32KiB 32KiB
L1 scalar data (per WGP) 16KiB 16KiB 16KiB
GL1 Date (per SA) 128KiB 256KiB 256KiB
L2 data 2048KiB (2MiB) 2048KiB (2MiB) 2048KiB (2MiB)
L3 (Mall) Not applicable Yes Not applicable

Coelacanth-Dream also said that all RDNA 3 “Navi 3X” GPUs come with VODP (dual-issue Wave32) instructions, WMMA (Wave Matrix Multiply Accumulate) support, and that the performance of each WGP has been greatly improved. The increased GL1 cache is said to improve pixel processing performance, and it’s one of the many changes AMD is bringing to its RDNA 3 Navi 3x GPU lineup.

AMD has confirmed that its RDNA 3 GPUs will launch later this year with a massive performance boost. The next-generation GPUs in the Radeon RX 7000 series will provide more than 50 percent performance-per-watt improvement over existing RDNA 2 GPUs, said David Wang, senior vice president of engineering at the company’s Radeon Technologies Group. Some of the key features of the RDNA 3 GPUs that AMD highlighted include:

  • 5nm process node
  • Advanced Chiplet Packaging
  • Re-architected computing unit
  • Optimized graphics pipeline
  • Next Generation AMD Infinite Cache
  • >50% Perf/Watt vs RDNA 2
amd-radeon-rx-7000-rdna-3-navi-3x-gpu-official-presentation-_2
amd-radeon-rx-7000-rdna-3-navi-3x-gpu-official-presentation-_1

AMD will re-architect the compute units in RDNA 3 to provide enhanced ray tracing capabilities. While there’s no mention of what those features are, if we’re guessing, we’d say it’s definitely talking about the performance and set of advanced features at the core of the RDNA 3 GPU used in the Radeon RX 7000 graphics card. AMD Radeon RX 7000 graphics cards will launch later this year and deliver a giant leap in gaming performance, so stay tuned for more information in the coming weeks.

AMD RDNA 3 Navi 3X GPU Configuration (Preliminary)

GPU name Navigation 21 Navigation 33 Navigation 32 Navigation 31 Navigate 3X
code name Sienna Cichlid Hotpink Bonefish Mai Nas Plum bonito pending
GPU process 7nm 6nm 5nm/6nm 5nm/6nm 5nm/6nm
GPU package Monolithic Monolithic MCM (1 GCD + 4 MCD) MCM (1 GCD + 6 MCD) MCM (pending)
GPU chip size 520mm2 203mm2 (GCD only) 200mm2 (GCD only)
425mm2 (with MCD)
308mm2 (GCD only)
533mm2 (with MCD)
pending
shader engine 4 2 4 6 8
GPU WGP 40 16 30 48 64
SP per WGP 128 256 256 256 256
Compute unit (per chip) 80 32 60 96 128 (per GPU)
256 (total)
core (per chip) 5120 4096 7680 12288 8192
cores (total) 5120 4096 7680 12288 16,384
memory bus 256 bits 128 bits 256 bits 384 bit 384 bits x2?
memory type GDDR6 GDDR6 GDDR6 GDDR6 GDDR6
Memory Capacity up to 16 GB up to 8 GB up to 16 GB up to 24 GB up to 32 GB
memory speed 16-18Gbps pending pending 20Gbps pending
memory bandwidth 512-576 GB/sec pending pending 960 GB/sec pending
Unlimited cache 128MB 32MB 64MB 96/192MB pending
Flagship SKUs Radeon RX 6900 XTX Radeon RX 7600XT? Radeon RX 7800XT?
Radeon RX 7700XT?
Radeon RX 7900XT? Radeon Pro
pending 330W ~150W ~250W ~350W pending
emission Q4 2020 Q4 2022? Q4 2022? Q4 2022? 2023?

Leave a Comment

Your email address will not be published.