AMD’s Radeon RX 6000 range of graphics cards, which will soon be revealed, has been hit with a major leak showing purported details of various GPUs, including the flagship Big Navi.
Before we begin, we should firmly bear in mind that these are just nuggets from the rumor mill, so we need to be careful about jumping to conclusions. However, the info seems much more solid than whispers on the GPU grapevine often are, having been drawn straight from the macOS Big Sur 11 beta, complete with a lot of details.
This was highlighted on Reddit (as spotted by Tom’s Hardware), and the details spilled on Big Navi (Navi 21) back up a leak from earlier in the week (from the same Redditor in fact), but add some juicy details on the clock speeds.
The GPU will allegedly have 80 compute units (CUs) and 5,120 stream processors (SPs – given that AMD’s RDNA 2 architecture still carries 64 SPs per CU). The new info on clock speeds has them pegged much higher than current-gen Radeon GPUs – as expected – but according to this leak, one variant of Big Navi clocks up to 2.2GHz.
The mention of reaching these speeds – or possibly even a touch higher – has certainly stoked excitement on Reddit, but of course we can’t really know how fast Big Navi will be until the impact of AMD’s new architecture is taken into account in the performance equation.
AMD has promised a 50% increase in terms of performance per Watt for RDNA 2, and if that’s realized, on top of all these other rumored specs – including an alleged 16GB memory loadout, versus only 10GB on the 3080 – we’re looking at a beefy flagship.
Tom’s observes that single-precision performance for Navi 21 hits 22.5 Tflops, which is certainly an impressive figure on the face of it, although it does fall considerably behind the RTX 3080 at 29.8 Tflops. This hardly represents the full picture when it comes to performance in actual gaming, though, and of course Big Navi has the RTX 3070 beat (at 20.4 Tflops).
Obviously we can’t get carried away here with purported specs anyway, and exactly how Big Navi will compete against the top Nvidia graphics cards will depend a lot on exactly how AMD pitches the pricing, and how aggressive the GPU maker is in that respect. As previous rumors have indicated, though, it’s looking like the card may nestle somewhere between the RTX 3070 and 3080 in performance terms.
Of course, further bear in mind that Nvidia supposedly has a reply tucked away up its sleeve for AMD’s next-gen launch (namely a new spin on the RTX 3080 with double the RAM, according to fairly strong rumors at this point).
Navy Flounder
This leak also details further models in the RX 6000 range, and the next graphics card down from Big Navi, known as Navi 22 (Navy Flounder), will be equipped with 40 CUs (2,560 SPs). That’s the same as the existing RX 5700 XT, but of course there’s that major architectural uplift to take into account, plus the fact that the boost clock of Navi 22 is pegged much higher at 2.5GHz.
At a rated 12.8 Tflops, this indicates performance close to a third faster than the 5700 XT, delivered within a much tighter power envelope (170W versus 225W). This graphics card should be good for 4K gaming at a more affordable level.
Navi 23, the next step down and the entry-level offering, is also mentioned in the leak as having 32 CUs (2,048 SPs), but no information on clock speeds has been dug up in this case, which seems to indicate that this GPU could be further off than Navi 22 and 21.
Interestingly, a Navi 31 graphics card is also present, which has certainly raised some eyebrows as you can imagine. It’s based on RDNA 3 and could be a planned refresh of Big Navi (it has the same 80 CUs as Navi 21).
AMD’s next-gen graphics cards launch on October 28, and the firm has hinted that stock levels will be much stronger than what we’ve seen with Nvidia’s recent RTX 3000 releases. So no matter how powerful the new GPUs turn out to be, at least you theoretically stand a better chance of actually being able to buy one.
Source: techradar.com
