AGESA 1.0.0.7 incoming !

"AGESA 1007 comes with support for Raven Ridge APUs. AMD has also changed the entire BIOS base structure so we have to do a lot of work to port everything to the new version, which may result in further bugs. The advantage is that it makes it easier to support future CPUs (Raven Ridge, Pinnacle Ridge). The cold boot fix will be implemented as soon as we have a recent AGESA version which supports it."

Edited by Rares - 16 Oct 2017 at 5:39pm

SoniC wrote: So means Threadripper will be left without the update as it is mainly targeting AM4 platform? Pity as the memory compatibility on the X399 is far from perfect...

can you share the original link that announces this?

I hope all of you will take a look at my latest response from AMD concerning TR and AGESA:

http://forum.asrock.com/forum_posts.asp?TID=6271&KW=agesa&PID=37940&title=amd-response-to-agesa-bios-query#37940

Enjoy, John.

EDIT: Google sees many references:

http://www.guru3d.com/news-story/amd-preps-agesa-1-7-firmware-to-support-raven-ridge-apus.html


Edited by MisterJ - 16 Oct 2017 at 11:33pm

Wondering how much of the memory compatibility that AMD has learned on the Ryzen platform can or will be applied to the x399 Threadripper. Also does threadripper also have the lower clocks on the memory when all 8 dimms are fully populated cause Ill be investing about 3k in a month or so to be fully decked out to 128 gb.

JUANNY wrote: Wondering how much of the memory compatibility that AMD has learned on the Ryzen platform can or will be applied to the x399 Threadripper. Also does threadripper also have the lower clocks on the memory when all 8 dimms are fully populated cause Ill be investing about 3k in a month or so to be fully decked out to 128 gb.

High memory over clocks are always more difficult to achieve as the capacity of memory used increases. That is simply basic PC memory reality and applies to both AMD and Intel. You can see that in any memory QVL/Support list. The highest memory over clocks are only available with one DIMM per channel, and the capacity of the DIMMs at the highest speeds are reduced compared to lower speed over clocks.

Memory over clocks basically depend on three things, the capability of the processor's IMC, the layout of the DIMM slots on a mother board, and the model of memory being used. The model of memory depends upon the memory chips being used, the extent of the binning of the memory chips, and if the memory is sold (purchased) in a matched set. The programming for the memory in the UEFI/BIOS is also very important, but cannot compensate for the limitations of the other three things.

The greater load on the IMC with more memory makes it more difficult for the memory controller to work at higher speeds. We can increase the voltage(s) used by the memory controller, but that has limits, and the IMC also is affected by the "silicon lottery" affect.

The more memory slots being used, the more interference between the signal traces on the board occurs, and it is more challenging to layout the memory signal traces. Trying to keep the signal traces all the same length or as close as possible matters, due to the very high frequencies involved.

How is that done with four memory channels, and two memory slots per channel? The slots closer to the CPU have their trace lengths increased to match more closely with the slots further away. But that also reduces the speed potential of the closer slots. Balancing one trade-off for another is difficult engineering.

Do we see memory over clocking on very high capacity enterprise mother boards? No because higher speeds and stability cannot be guaranteed with very high capacities. Guaranteed being the keyword there.

With the price of DDR4 memory of any capacity being ridiculous currently, high capacity kits (64GB+) at high rated over clocks are priced beyond ridiculous levels. So to purchase 128GB of memory, we buy two 64GB kits, or possibly four 32GB kits. We take the chance of using unmatched DIMMs in multiple kits. The results can be disappointing.

Using 128GB of memory in X399 systems has been problematic for many users, using boards made by any manufacture. Keep in mind that the responsibility for the ability to use 128GB of memory is shared by all the involved parties, not just the mother board manufacture.