AMD RDNA3 GPUs get new toned down specifications, Navi 31 with up to 12288 Stream Processors

Published: May 2nd 2022, 10:24 GMT   Comments

Please note that this post is tagged as a rumor.

Rumored specs of RDNA3 architecture has changed

According to Greymon55, the internal structure of the Navi 3X GPUs might be different from previously reported. 

Not 15360 Stream Processor but 12288 for Navi 31 GPU, the flagship processor of the upcoming RDNA3 series. This means that the 92 TFLOPS claim from a few days ago is either wrong, or there is something else that we do not know about yet.

The Navi 31 with 6 Shader Engines, 12 Shader Arrays and 48 Work Group Processors would ship with up to 12288 Stream Processors, that’s a reduction of 20% in core count compared to previously rumored 15360 cores. The same applies to Navi 32, which instead of 10240 cores would ship with 8192 Stream Processors. For the Navi 31 this means 4096 cores instead of 5192, but it should be noted that lower core count was mentioned by Greymon55 before.

https://twitter.com/Redfire75369/status/1521052789996400640

With the updated Navi 31 specs, the rumored TFLOPS would drop to 73 TFLOPS, and that’s assuming 3.0 GHz GPU clock. We still do not have any leaks or rumors on the Navi 32 and 33 clocks speeds, though.

One should remember that we are still months away from the launch of new 3rd Gen Navi GPUs. This is, however, the first major update to the rumored specs that have been shared for more than a year. So unless there are other reliable sources claiming otherwise, we will assume that those specs are closer to final.

Furthermore, the specs posted below only represent the full GPU configuration, which may not be guaranteed at launch for 7900/7700/7600 series. AMD may choose to lower the specs for each GPU to increase the yields for its Radeon RX 7000 graphics cards.

Rumored Specs of AMD RDNA3 GPUs
VideoCardz.com
AMD NAVI 31
AMD NAVI 32
AMD NAVI 33
Fabrication NodeTSMC 5nm + 6nmTSMC 5nm + 6nmTSMC 6nm
ArchitectureAMD RDNA3AMD RDNA3AMD RDNA3
GPU PackageMulti-Chip-Module (MCM)Multi-Chip-Module (MCM)Monolithic
Estimated GPU Size~800 mm² (or more)TBC~440 mm²
Graphics Dies221
Shader Engines642
Shader Arrays1284
RDNA Workgroups (WGP)60 → 4840 → 3220 → 16
Compute Units120 → 9680 → 6440 → 32
Stream Processors15360 → 1228810240 → 81925120 → 4096
GPU Clock~3.0 GHzTBCTBC
Memory & BusGDDR6 256-bitGDDR6 192-bitGDDR6 128-bit
 Infinity Cache256 – 512MB256 – 384MB128 – 256MB
ReleaseQ3/Q4 2022Q3/Q4 2022Q1 2023
Rumored SKURadeon RX 7900XTRadeon RX 7700XTRadeon RX 7600XT
FP32 Performance~ 92 → 73 TFLOPSTBCTBC
All information is based on rumors

Source: Greymon55, Kepler_L2




Comment Policy
  1. Comments must be written in English.
  2. Comments must not exceed 1000 characters. Comment splitting is not allowed.
  3. Comments deemed to be spam or solely promotional in nature will be deleted.
  4. Discussions about politics are not allowed on this website.
  5. Sharing relevant links is permitted; avoid bypassing the link and word filters. Our team will approve links.
  6. Comments complaining about the post subject or its source will be removed.
  7. Offensive language in comments or usernames result in a ban.
  8. Direct attacks/harassment result in immediate ban.
  9. VideoCardz isn’t and was never sponsored by AMD, Intel, or NVIDIA. Users claiming otherwise will be banned.
  10. Moderators may edit/delete comments without notice.
  11. If you have any questions about the commenting policy, please let us know through the Contact Page.
Hide Comment Policy
Comments