New AGESA Parameters May Allow for Higher Ryzen RAM Speeds

Zarathustra[H]

Extremely [H]
Joined
Oct 29, 2000
Messages
39,153
In AMD's latest Community Update titled "Let's Talk DRAM!" they tease some interesting updates coming up in AGESA v1.0.0.6. These updates add 26 new RAM parameters largely targeted towards improving the stability of overclocked RAM, and may allow for speeds as high as DDR4-4000, which is a considerable improvement over the seeming hard cap at 3200 today.

So why should we care? RAM speeds rarely have a huge impact on overall system performance, right? You may remember all the reports regarding Ryzen's inter-CCX module performance from back when AMD first launched the CPU. It is often mentioned in our forums that inter-CCX performance seems to be related to RAM speed. If this holds true, and we can repeatably hit 4000Mhz with RAM after this update, it may alleviate one of the largest criticisms related to AMD's Ryzen platform since launch.

AGESA is an acronym that stands for “AMD Generic Encapsulated System Architecture.” As a brief primer, the AGESA is responsible for initializing AMD x86-64 processors during boot time, acting as something of a “nucleus” for the BIOS for your motherboard. Motherboard vendors take the core capabilities of our AGESA updates and build on them with their own “secret sauce” to create the BIOS files you download and flash. Today, the BIOS files for AMD AM4 motherboards are largely based on AGESA version 1.0.0.4.
 
I'll be interested to see if enabling 2T lets my RAM hit 3200 via XMP (currently I can't get higher than 2933 with extremely loose timings to boot as my board forces 1T).
 
I'm currently stable at 2933, I'll let others be guinea pigs until I update and hopefully hit a stable 3200.
 
My current system may be on its last leg so am watching the ryzen space closely and was hoping to see some of these loose ends tightened up before I have to make a decision on what to get next.

This part seemed a little ambiguous as I understand the current state of the memory speed -> memclk/CCX domain handling:

"Added dividers for memory clocks up to DDR4-4000 without refclk adjustment"

Does this mean AMD themselves just unlinked those two clock domains? I feel like I can't be reading that right since it would be a massive shift. Would you just need to set the memory clock at whatever you wanted then set the CCX refclk independently?
 
I've been able to run at higher speeds and pass stability tests, but I get random kernel panics every couple days at anything above 2133.
 
I had no issues running at 3200 by simply selecting the XMP profile. But I am only running 16GB with two slots filled. I'm hoping these new AGESA will let me be able to have all 4 slots filled without issues.
 
I've been able to run at higher speeds and pass stability tests, but I get random kernel panics every couple days at anything above 2133.

I set my ram by the XMP profile in the Ryzen Master application and it worked well. I supposedly did the same in the bios of this AsRock mobo, but it crashed like you described and would reset the settings. Setting it in Ryzen Master seems to make it "stick" properly and run without issues. Of course that was quite a few bios updates ago so this may not be the same situation today. I just use the Ryzen Master regardless as I've gotten accustomed to using it for checking my ram speeds.
 
I was under the impression that the *final* version of the code was coming out *this* month, not a *beta*. :confused:
 
Yeeessss, bring the ram speed on !!! I am still struggling to get past 2400mhz with DDR4 3000 ram
 
I'm on the fence to move to Ryzen, since my gaming system is still running great ([email protected]), but if this bios enable even high performance, i might just pull the trigger on a 1700x, since microcenter is running a nice deal on them.

Hope someone here do some tests soon!
 
On the Asus Crossfire VI Hero with a Ryzen 1700 - the beta 9943 and 9945 bios are out. For some 4 stick or Hynix configurations, 9945 may be better, though in general, 9943 will give the best results. I can boot and do general use at 3600 MHz with 16-16-16-16-36 1T, but still playing to get torture testing level stability. At stress test stability, I run 3466 MHz at 14-14-14-14-34 1T. These are both using no BCLK overclocking, just the new available memory dividers. With these new bios, you will find learning about ProcODT, CLDO_VDDP, Geardown Mode, secondary and tertiary timing parameters, etc. will help you tune for stability and performance.
 
hoping this fixes my cold boot issues at 3200, 2933 is solid as fuck. I'm running Hynix chips.
 
I've set my ram to 2667, and crap timings.

haven't even bothered to play with the memory yet, as yet new bios changes things bunches in that regard. Perhaps when there is a non beta working bios with this new code, then I will commit some time to actually overclocking my Ryzen chip further. Been rocking 4ghz on all cores @ 1.35 volts since day one in terms of clock speed :)
 
the real question is can amd fix the underlying inter ccx latency cause this may become an issue on the mcm chips.
 
This is great news for Raven Ridge. DDR4 at 4000 mHz will help a lot. For everyone else it will mean a 1% performance boost. (Sorry, It's the truth)
 
This is great news for Raven Ridge. DDR4 at 4000 mHz will help a lot. For everyone else it will mean a 1% performance boost. (Sorry, It's the truth)

I don't know if having memory clock all the way up to 4000 MHz will still allow the infinity fabric to run at half that speed, but if it does then that ought to improve gaming performance even more on those games that were shown to be sensitive to ram speed, and the clock speed was more impactful than timings, so that could net much more than a 1% increase in some games. For general stuff it's a non issue either way, but the area where first gen ryzen was a bit behind the 8ball was gaming, and this could help narrow that gap more.
 
This has been my only complaint about the new Ryzen CPUs. If this gets squared away, AMD would be back in business as far as my $$$ for future builds are concerned. Now lets hope the near future GPUs turn out well also.
 
Supposedly this bios also fixes the IOMMU groups, which allows a Linux Box to run a windows virtual machine that has direct access to the GPU meaning you can game on a windows VM in linux with little to no performance loss

ie:

Looks like a huge pain in the behind and is still experimental right now, but give it 6 months or so and I'm sure it will be the standard method of gaming from linux
 
Supposedly this bios also fixes the IOMMU groups, which allows a Linux Box to run a windows virtual machine that has direct access to the GPU meaning you can game on a windows VM in linux with little to no performance loss

ie:

Looks like a huge pain in the behind and is still experimental right now, but give it 6 months or so and I'm sure it will be the standard method of gaming from linux


I tried to pass through some Nvidia adapters in both ESXi and KVM a while back on my server and was never successful. I hear AMD adapters are easier though.

My current desktop is the first stepping of Sandy-E where VT-d (Intel's name for IOMMU) was disabled due to a bug :(

When I finally get around to upgrading my motherboard/CPU (maybe Ryzen Next?) I might try this again. It would certainly be more convenient than rebooting all the time.
 
Back
Top