I seriously doubt this. AMD would not give 1 company a sneak peak without giving them all a sneak peak. Also, AMD would have provided an X470 motherboard and a complete kit for testing. Either they pieced things together from the leaked benchmarks or they managed to get their hands on an engineering sample. We'll find out soon enough. The actual NDA isn't until April.
Who said anything about AMD giving them an intentional sneak peak?
New hardware is sent out to many companies and people to get things ready for launch - system integrators, OEM's, Boutique PC builders, motherboard vendors, and many many more. If you have the right contacts, you can get hardware early. I personally know someone who already has the hardware.
And that aprox 20% improvement in L2 latency...crazy what is that?
It's an odd thing really. Latency for the L2 was supposed to be 12 cycles from AMD presentations and documentation, and yet Ryzen shipped at 17 cycles. What's strange is that Threadripper, EPYC, and Raven Ridge all shipped at 12 cycles - and the first two are the same die!
I suspect there was some reliability issue with production silicon that they corrected but already had too many dies to throw away. That's my theory, got nothing official and any attempt to coax the information out of AMD results in marketing spiel.
AMD, nVidia and most likely Intel as well implement reliability calculations in their boost algorithms.
This allows them to reach the highest possible boost speeds in light loads. The silicon is able to withstand significantly higher voltage in low load conditions (i.e. partial utilization, low current and temperature).
If the same voltage levels used at the highest boost speeds would be applied during the full load as well, there would most likely be permanent damage to the silicon.
That's VERY interesting!
Is that why XFR goes up to 1.55V?