That's why I asked him, since he claims they were leaked.
Likely just confusion over the name, since the 7600 is cast-iron confirmed now on everything except launch price MSRP. Reviewers have the samples, benchmark leaks are popping up all over the place, and we've seen multiple vendors announce and publish models and specs ahead of Thursday's official launch.
The 7600XT results all over the web are just inaccurate rumour-mill misreporting the RX7600 we now know to be officially launching on Thursday. Navi32 is apparently 60CU (so 3840 shader cores), confirmed by a leak from AMD themselves via AMD's own ROCm development team. 60CU seems far too big to be used for the 7600XT. Knocking it down to 45CU and 192-bit still makes it 41% faster than the RX7600 which is a huge performance (and price) gap. I suspect the harvested part with 42-45CU@192bit w/12GB VRAM is eventually going to be the 7700-series and the 56-60CU@256bit w/16GB VRAM will be the 7800-series.
Bear in mind that Navi32's original released date was supposed to be Q3 2023 and in an interview with AMD's Sam Naffziger (
the chiplet guy)* He mentioned to Steve Burke of GN that Navi31's chiplets weren't stable at the clockspeeds AMD originally intended, so they were forced to downclock the 7900XT and 7900XTX. Direct from AMD's chief RDNA3 chiplet engineer, is confirmation that they are delaying Navi32 to fix those issues in hardware, because they couldn't fix it in software for Navi31.
That means that Navi32's is extremely unlikely to make its original Q3 2023 launch date, so we're more likely expecting no new RDNA3 launches faster than the RX 7600 until Christmas.
* edit - might have actually been a different GN inteview with Mike Mantor from RTG instead, but it was one of those November 2022 7900XTX launch coverage videos.