AMD: Zen 4, Speculation, Rumours and Discussion


We followed up with AMD to ask if the XDNA AI engine is removed entirely or disabled, and the company responded that it is disabled. That means the chips are likely identical at the silicon level. In the end, there will only be one way to truly suss out the differences: benchmarking them as they hit our labs. For sure, making a chip work in a handheld the size of a Steam Deck (in the case of the upcoming Asus ROG Ally, it's actually smaller) is going to take some work, as you don't get the amount of space or cooling that you do in a laptop.

We've also yet to see what the Ryzen AI engine from the XDNA architecture is capable of. That may give the laptops an edge, though it's unlikely current games would take advantage of it, which makes a case for removing it from Z1 and Z1 Extreme.
 
Maaaaaaaaaaaaaaan the ol' Pencil Mod for unlocking all those extra pipes on the R300!!! Miss those days of the 9500 non-pro, unlocked and overclocked!

Pretty sure my softmodded + hard modded (voltage) 9500np held the 2nd place world record spot for 3DMark back when those things were in vogue. Wonder if I can find any of those ancient threads on Overclockers.net....
 

We've done proper measurements, but it's in Finnish.
Despite GN claims, at least ROG Crosshair X670E Hero adheres to AMDs 1.3V limit just like it should. Apparently GN has been measuring the voltages from all the wrong places or something. HWInfo & direct measurement under socket gives right numbers
 
Just reinforces my feelings to never adopt a new platform in the first few months. :p It'll probably be OK, but every now and then firmware issues bite you in the ass.

Regards,
SB
 

We've done proper measurements, but it's in Finnish.
Despite GN claims, at least ROG Crosshair X670E Hero adheres to AMDs 1.3V limit just like it should. Apparently GN has been measuring the voltages from all the wrong places or something. HWInfo & direct measurement under socket gives right numbers
Is it possible due too different manufacturing batches?
 
Is it possible due too different manufacturing batches?
Don't think so.
They need to check they're actually reading the correct value, it's CPU VDDCR_SOC Voltage (SVI3 TFN) in HWInfo.
If I understood it finally correctly, you can't actually measure it with outside hardware reliably, even from socket it's higher than it is when it hits the CPU.
 
Last edited:
That would mean then that basically any other value other than the auto option in the LLC is now a placebo in pratical terms.
 
Can I pencil it?

As it’s part of the Meteor Lake SoC tile, the VPU will be present in all Meteor Lake SKUs. Intel will not be using it as a feature differentiator, ala ECC or even integrated graphics; so it will be a baseline feature available to all Meteor Lake-based parts.

I hope it won't be necessary to pencil the upcoming 7x40U APUs and Ryzen AI will be available on all Phoenix SKUs.
 
Meteor lake made me want to wait for 2024 to look for laptops.

If they release i5 class / ryzen 5 class with max igpu cores, then I would prefer that than i7 class / ryzen 7 class.
 
Back
Top