RDNA 3 Reportedly More Than Doubles the Core Count of the RX 6000 Series GPUs

AMD Big Navi teaser
(Image credit: AMD)

Reported Stream Processor configurations of RDNA 3 have been shared online by popular hardware leaker @Greymon55 on Twitter. Big grains of salt, as these are super light on details, but his "reports" indicate that RDNA 3 will feature a substantial core count buff over current RDNA 2 graphics cards, with shader counts more than doubling compared to RDNA 2 and Radeon RX 6000 series cards.

According to Greymon55 — and correspondingly @Redfire75369 who interpreted Greymon's post — AMD's largest Navi 31 die will feature a whopping 12,288 streaming processors housed inside 48 WGPs (workgroup processors). RDNA 3's midrange Navi 32 die will receive 8192 SPs housed inside 32 WGPs. Last, the entry-level Navi 33 will get 4096 SPs in 16WGPs.

These numbers are substantially greater compared to RDNA 2, where the top Navi 21 die peaked at 5120 SPs, Navi 22 at 2560 SPs, and Navi 23 at 2048 SPs. It's possible AMD will do something similar to Nvidia's Ampere where the number of shaders per functional unit (WGP). Note that a single WGP in RDNA contains two Compute Units (CUs).

Another large change compared to RDNA 2 is the much greater core count difference between the lower end Navi 33 and midrange Navi 32 dies, which have increased to a two-fold gap. Navi 22 for reference topped out at 40 CUs in the RX 6700 XT while Navi 23 had up to 32 CUs in the RX 6600 XT. This could be a larger performance gulf between entry-level and midrange RDNA 3 products if we're assuming other specs such as infinity cache size and video memory speeds remain the same.

Keep in mind that "leaks" of this nature are often more of an educated shot in the dark rather than based on actual details. While Navi 31 should launch by the end of this year, Navi 23 is only seven months old and isn't due to be replaced for about a year. Still, we do expect improved performance and core counts, and other rumors suggest we may see a doubling to tripling of the Infinity Cache size, along with higher clock speeds brought along by TSMC's newer 5nm and 6nm nodes — possibly up to 3GHz or more. RDNA 3 could be a colossal upgrade over RDNA 2.

As a result of the various improvements, performance rumors for the top Navi 31 die suggest it could be 3x to 4x faster than the Radeon RX 6900 XT in FP32 compute, according to Videocardz (again, large servings of salt). Bear in mind that FP32 performance doesn't translate perfectly into gaming performance, so we shouldn't see a whopping 4x improvement in FPS. However, we still expect a big improvement in gaming performance with RDNA 3 due to the core counts alone.

Obviously, these results are based on leaks and rumors and are not in any way official. These core specifications are subject to change, or at least, future rumors could pull different numbers out of a hat. Previously reported stream processor counts for Navi 31 and 32 were 20% higher, for example. There's also no word on TBP (typical board power), though with Nvidia supposedly going as high as 600W, we would expect AMD to increase power use as well.

AMD's initial Radeon RX 7000-series graphics cards based on RDNA 3 should be arriving sometime later this year. Based on previous launches, we suspect they'll land in the October to December time frame. As the actual launch draws closer, we'll likely see numbers start to hone in on the actual specs. Right now, preliminary specifications of RDNA 3 suggest it will be a monstrously powerful architecture from AMD.

Aaron Klotz
Freelance News Writer

Aaron Klotz is a freelance writer for Tom’s Hardware US, covering news topics related to computer hardware such as CPUs, and graphics cards.

  • tamalero
    I hope these numbers are not fantasy. Just like Nvidia changed their technology and their "double the cuda cores" were BS and were actually now counted twice.
    Reply
  • ddcservices
    tamalero said:
    I hope these numbers are not fantasy. Just like Nvidia changed their technology and their "double the cuda cores" were BS and were actually now counted twice.
    If AMD is really going the MCM route with RDNA 3, then AMD can put together 2-4 dies worth without fab process yields being an issue. That will allow for larger GPUs without the price going up significantly for only a 5-10 percent performance increase.

    Other than MCM, I have a hope that AMD put in more engineering for each "core" when it comes to performance, so 3 times the core count for only double the price of the 6900XT would be nice.
    Reply
  • hotaru.hino
    ddcservices said:
    If AMD is really going the MCM route with RDNA 3, then AMD can put together 2-4 dies worth without fab process yields being an issue. That will allow for larger GPUs without the price going up significantly for only a 5-10 percent performance increase.

    Other than MCM, I have a hope that AMD put in more engineering for each "core" when it comes to performance, so 3 times the core count for only double the price of the 6900XT would be nice.
    If AMD goes the MCM route, I'm curious to know how they'll overcome the issues on what's essentially CrossFire on a single card.
    Reply
  • hotaru251
    hotaru.hino said:
    how they'll overcome the issues on what's essentially CrossFire on a single card.
    this.
    a card can be amazing spec wise but if it suffer crossfire/sli's latency/dropepd frames....it is basically dead at launch.
    Reply
  • gg83
    Will we continue to increase wattage? I thought die shrinks were supposed to increase performance while decrease power usage?
    Reply
  • hotaru.hino
    gg83 said:
    Will we continue to increase wattage? I thought die shrinks were supposed to increase performance while decrease power usage?
    This was an observed thing called Dennard Scaling, much like Moore's Law, where the power density of an IC remains constant even though the transistor density goes up. However this broke down in around the early 2000s.

    Basically everything you might've heard about IC manufacturing to improve efficiency or whatnot has gone downhill in either the 2000s or 2010s.
    Reply
  • spentshells
    tamalero said:
    I hope these numbers are not fantasy. Just like Nvidia changed their technology and their "double the cuda cores" were BS and were actually now counted twice.
    while being weaker
    Reply
  • spentshells
    Just before the next launch right at the end they will drop a card between the 6800 and the 6750XT and I will buy it.
    Reply
  • Umfriend
    "It's possible AMD will do something similar to Nvidia's Ampere where the number of shaders per functional unit (WGP)." - That's some wild speculation, even if unintelligible.
    Reply
  • hotaru.hino
    tamalero said:
    I hope these numbers are not fantasy. Just like Nvidia changed their technology and their "double the cuda cores" were BS and were actually now counted twice.
    No, NVIDIA's definition of what a "CUDA core" remained consistent. Since the only compute spec that seems to matter is FLOPS, NVIDIA has only counted the units that perform floating point units. For the longest time (probably since GeForce 8), each shader unit was FP and INT capable. Pascal had 128 units per SM. When Turing came around, NVIDIA split it between 64 FP and 64 INT units. For Ampere, there are 128 FP units, but only 64 of them are capable of INT workloads and are partitioned off from the other 64FP units

    The reason they did this was so an SM could do concurrent INT and FP workloads. But again, since the only compute spec that seems to matter is FLOPS, the FP units are what matter more.

    spentshells said:
    while being weaker
    I mean, if you want to consider half of the shader units in an SM losing their ability to do INT workloads, which is still a minority of workloads by a wide margin, sure, I guess.
    Reply