Question How much information do we have on 9070 XT performance?

[Scar Mountain]

Someone glancing at publicly-available information on the RTX 5000 series - Nvidia releases and official specs, leaked benchmarks, etc. - as of early January would have been able to precisely predict the performance improvement over the RTX 4000 series. So far that's held for the 5090 and 5080, and looks very likely to hold for the 5070 Ti and non-Ti.

In other words, the information sources were reliable across the board, more or less.

Information on the AMD 9000 series is sparser and shakier across all sources, it seems to me. Is there anything of comparable reliability or promise for the 9070 XT as the Nvidia content was? Can we discern with any similar confidence whether this piece of hardware will be more like a 7900 GRE or more like a 4080 Super? Or do we just have to wait another couple weeks? For example, looking at specs leaks alone, as reflected on Techpowerup, the 9070 XT should be less capable across the board than a 7900 XT, besides the hard overclock and new layout. If the available data on the 9070 XT could be analyzed as analogous data was for the 5000 series, what would we expect?

Feel free to link me to other posts if everything has already been considered to exhaustion.

 

[Gururu]

My understanding is that it is a 5070 competitor.

 

[aburak621]

According to the "Moore's Law Is Dead"s leaks on YouTube, AMD are internally "targeting" 3% faster raster performance than 4080 Founders Edition. As for ray tracing performance there were Cyberpunk tests around the 4070 TI Super. You can look at the timestamped videos yourself:

View: https://youtu.be/bZ6NeSGad4I?si=2YiM3a4BmN1rVgxf&t=475

View: https://youtu.be/udG8y64A8eo?si=HNwP1Y2f-orjpIsC&t=358

 

[Scar Mountain]

What is this channel's track record (ngl they give a bit of a hackish vibe)? On the 5000 series, some alleged leaks were reportedly from sources that had delivered credible information in the past, and those leaks were also aligned with Nvidia's implicit claims (e.g. pixel counting) and official specs. All those sources combined were proven precise within a month.

In the AMD case, the leaked specs commonly cited don't align well with the most optimistic performance claims or rumors, such as those presented by Moore's Law is Dead. They would fit better with the 9070 XT being an enhanced 7900-class. For example, if we went by collations such as this one, the lower-end of projections for the 9070 XT are most plausible. That's why I wonder if there's anyone who is sorting, assessing, and justifying the available information based on its quality, rather than just presenting this or that as-is. So I would want to know: how reliable is MLID, and how reliable have the sources he is citing been against known results in the past?