AMD RDNA 3 GPUs to Feature Scaled-Down Specs Compared to What We’ve Known, According to New Rumors

In a new leak roundup that pretty much no one expected to see, especially today, we’re now learning that AMD might be looking at toning down the specs of its RDNA 3 lineup by a sizable margin. Just a few hours ago, we looked at new leaks pertaining to both NVIDIA and AMD‘s next-gen GPUs and how they are supposedly neck-in-neck in terms of computer power and clock speeds, but that seems to be outdated now.

For the past year or so, all leaks and rumors have unanimously pointed to one thing: Navi 31, the flagship GPU from AMD’s RDNA 3 linup, will comprise of 15,360 Stream Processors. This information was the basis of today’s leak claiming the 92 TFLOPs of FP32 compute power for Navi 31. Now, we have fresher information that contradicts our previous presumptions, and it shakes up things quite a bit.

Navi 33 and Navi 32 toned down

According to multiple leakers including the trusty Greymon55, Redfire, and 3DCenter.org, AMD has reduced the Stream Processor count of at least two of the three RDNA 3 GPUs. Not only that, but the number of WorkGroup Processors also seems to have been affected by this recent tone-down. New leaks claim that Navi 31 will ship with just 12,288 Stream Processors instead of 15,360 and we’ll see only 48 WorkGroup Processors inside the GPU instead of 60.

AMD RDNA 3 concept render | Videocardz

That is a 20% reduction in core count, a figure that is large enough of itself, but also brings down the theorical computer performance, aka the recently-leaked TFLOP count, of the GPU. But as I said, more than one GPU is part of this situation, which brings us to our second candidate: Navi 32. According to newly-surface info, Navi 32 will feature only 8,192 Stream Processors and not 10,240 as was previously rumored.

You can see in the tweet below the exact changes made to the RDNA 3 lineup. It should be noted that Navi 33, which has also seemingly seen a reduction from 5,192 Stream Processors to just 4,096, already had conflicting reports regarding its core count. While most leaks pointed towards the former, Greymon55 was insistent on the latter, 4096 core count already.

As I mentioned before, this reduction in core count will lead to a corresponding reduction in how many teraflops of computer power the GPUs have. 3DCenter.org thankfully already did this calculation for us, showing how a 20% fall in core count will convert the previous 92 TFLOPs claim to just 73 TFLOPs, a number that is around 30 TFLOPs slower than NVIDIA’s next-gen Ada Lovelace flagship GPU.

We don’t have any leaks on the clock speeds of Navi 32 or Navi 33 so we can’t figure out their computer power just now, but the 3Ghz clock speed, as seen in the tweet above, was leaked for the Navi 31 by Greymon, hence we’re able to calculate that one. Moreover, this also confirms that there have been no changes to the target frequencies of these GPUs, which is a good sign.

Some salt is in order

As is customary with these things, you should always take leaks and claims with a heft grain of salt as they’re rumors at the end of the day, after all. The GPU market is about to enter an influx right now which means leaks are pouring in left and right, everyday. Just this bit of information on how quickly rumors changed within hours should be enough evidence to make you wary of these reports.

AMD and NVIDIA are still months away from their respective launches—both are targeting Q3 2022—which means that more accurate leaks will start coming out closer to the release than right now. Though, the sources attached above are some the most trusted in the game and they stuck by their word for nearly a year before today’s sudden switch-up. 

NVIDIA and AMD are both set to release their next generation of desktop GPUs by the third quarter of this year.

 

This is the first major update from rumor town pertaining to RDNA 3, so we’ll have to wait for further leaks more sources to confirm this claim. It should also be noted that these claims are only representative of the full-die specification, which means the non-XT Radeon RX 7900/7700/7600 series would actually get even slighly worse specs than this. Ultimately though, if this means that AMD can increase the yield and squeeze out more GPUs to combat the chip drought, there is surely a silver lining hiding somewhere in there.

Huzaifa Haroon
Born and raised around computers, Huzaifa is an avid gamer and a keyboard enthusiast. When he's not solving the mysteries of technology, you can find him scrutinizing writers, striving to inform the curious.
Back to top button