Ryzen AGESA 1.0.0.6 exposes more memory overclocking options

Over the past couple months, we've come to understand that memory speed and latency plays a major role in Ryzen CPU performance. Higher speeds and lower latencies are desirable for getting the most out of a Ryzen chip. Outside of a few motherboards with external base-clock generators, however, the memory multipliers for Ryzen CPUs have maxed out at 32. With a 100-MHz default base clock, that means builders have been limited to running overclocked RAM at DDR4-3200 speeds.

Today, that all changes. AMD has announced that its AMD Generic Encapsulated System Architecture, or AGESA, version 1.0.0.6 will expose 26 new RAM tuning settings to tweakers looking to extract the most from their Ryzen systems' RAM. These updates should bring AM4 motherboards' firmware settings for memory tweaking more on par with that of the Intel competition.

Of the wealth of memory overclocking options AGESA 1.0.0.6 will offer, the two most important settings to me are an extended range of memory multipliers (up to 40, or DDR4-4000) and the option to set command rates at one tick or two ticks (1T or 2T). For example, the hot G.Skill DDR4-3866 memory I have here requires a 38.66 multiplier and a 2T command rate, but those settings simply couldn't be dialed in on Ryzen motherboards under the current AGESA (1.0.0.4) without resorting to changes in the base clock that also controls important bus rates like PCI Express. Changing that base clock rate could result in unexpected behavior, so it's a relief that the masses won't have to resort to that approach any longer. 

Of course, AMD is quick to point out that using memory speeds above DDR4-2667 or settings not in accordance with JEDEC values is considered overclocking. Overclocking RAM with these new knobs will void a Ryzen CPU's factory warranty, even if you're using the company's Ryzen Master utility. Still, the temptation of performance gains will probably be too much to resist for many Ryzen owners.

In news that will be music to the ears of another class of users, AMD also says that AGESA 1.0.0.6 will improve support for PCI Express Access Control Services, or ACS. According to the company, these improvements will allow users who run virtual machines on their Ryzen systems to better manage the assignment of PCIe devices within IOMMU groups. In short, the company says this feature will let VM wizards dedicate multiple graphics cards within a Ryzen system to the operating systems of their choice, whether natively installed or virtual, among other benefits.

AMD says motherboard firmware incorporating AGESA 1.0.0.6 will be available starting in "mid-to-late June." The company notes that some motherboard makers have already begun issuing beta firmwares for some products with the new AGESA on board, most notably Gigabyte with its AX370-Gaming 5 and Asus with its Crosshair VI Hero. Folks eager to push Ryzen memory performance to its limits will doubtless welcome this update warmly.

 

Comments closed
    • helix
    • 3 years ago

    Dedicating PCIe cards to a VM would be really neat.
    Is this a firmware-only feature? Can we expect this to show up in most AM4 motherboards already on the market?

    • Toby
    • 3 years ago

    What is going on at Tech Report? I’ve been a big fan of this site since near its inception and it seems the community has evolved into an reaffirming echo chamber for all of AMD’s efforts despite effectiveness. I’m cheering for AMD to provide some meaningful competition as much as the next guy, but every time I wander into the comments I feel like I’m wading through blind loyalty interspersed with tribal hatred to anyone that questions it. The up/downvoting of comments exacerbates the issue; I often see very negative attacking comments on the right hand side of the page because they’ve been “voted up”. I find this sort of thing somewhat depressing.

    Can someone please explain what I’m missing? How did this site get to this point? With the volume of tribalism in society right now, I don’t think I need it in my technology news.

    Thank you!

      • ronch
      • 3 years ago

      Oh TR just became your average club where going against the flow makes you unpopular. Kinda like the governments of the world.

      • LostCat
      • 3 years ago

      Voting is easier than entering conversations if you either don’t want to discuss things with certain people or don’t have anything to add.

      • cynan
      • 3 years ago

      Regarding AMD “bias”, I think you’ve pretty much identified the issue. Enthusiasts are dying for some meaningful competition-fueled innovation (in technology, performance and pricing). More recently in the GPU space, and for the longest time in the CPU space – the two most exciting PC components – this has been relatively non-existent. And the only near-term hope for either is AMD. Suddenly, AMD is coming out full guns blazing (at least according Ryzen, their PR and released road maps), promising to change all of that. Kind of hard to avoid getting carried away at least a little by the hype.

      Also, I think a lot of TR’s membership has been building PCs long enough to remember the good old days when AMD-based systems represented the “enthusiasts” choice. Nostalgia is a powerful drug.

      And then there’s chuckula championing Intel, who has proven able to more than holds his own against an entire army of gerbils when it comes to the comments section 😉

        • LostCat
        • 3 years ago

        chuckie seems to like chasing shadows.

        • chuckula
        • 3 years ago

        How exactly have I “championed” Intel?

        All I have done is accurately call out the clique around here who have insulted anything and everything from Intel for the past 10 years when their holy miracle chip RyZen doesn’t blow the doors off of higher-core count Intel parts by 50%. Because that’s what they’ve been promising us and if an Intel part with RyZen’s performance launched this year there ain’t one single AMD shill around here who would be jumping up and down to praise it. That’s my objective standard, and literally every AMD fanboy on here has failed it miserably.

        Let’s put it another way: Here’s what I would actually do if I was an actually biased “Intel Champion”

        1. Denigrate RyZen in 2017 just as much as a d-bag loser like Bensam123 did to Haswell when it launched in 2013.

        2. Wait four years for Intel to launch an 8 core part on a more advanced node in 2021 with a noticeably higher clockspeed.

        3. Act like the greatest miracle in the history of the universe has happened when an Intel part from 2021 with a more advanced process node and the same core count as an AMD part from 2017 manages to beat the AMD part in some benchmarks some of the time kinda, while also literally flat-out losing in numerous real-world benchmarks [b<]and[/b<] not even being particularly "future proof". Bonus points for pretending to be some sort of hard-core "streamer" gamer while the 4-year-newer Intel part loses to the older AMD part pretty much across the board in games. 4. Oh yeah, and act like repeated additional miracles occur every few weeks when some idiotic "secret setting" is discovered that magically unlocks the Intel part while acting like there's some huge conspiracy and pretending that the chip manufacturer is intentionally trying to gimp its own product.

          • synthtel2
          • 3 years ago

          Who exactly ever claimed Zen was going to beat an Intel part with more cores by 50%? That’s the core of the problem here. If there were actually a clique saying garbage like that all the time, I’d kinda almost understand where you’d be coming from, but I can’t think of any regular commenters like that, much less a whole clique of them. On other sites, sure, but on TR the worst I see is sarcasm, off by <20% errors, and occasional obvious trolls who aren’t going to hang around long regardless.

          The dudes you want to fight aren’t here. If you really want to fight them this badly, why not go somewhere they actually are?

          • raddude9
          • 3 years ago
    • rika13
    • 3 years ago

    The Introlls and AMDbags will fight, but that is good. It means we have competition.

    AMD giving out settings that can break things that they won’t cover, not so much, unless there is a “you are doing something very naughty and we won’t cover it when it breaks” warning.

    • albundy
    • 3 years ago

    void warranty by overclocking ram? who would offer amd solutions with that kind of liability?

    • xDoritox
    • 3 years ago

    4GHz kits are really costly they go upward of $300. Some of these kits cost more than the CPU itself.

    • DPete27
    • 3 years ago

    [quote<]using memory speeds above DDR4-2667 ... will void a Ryzen CPU's factory warranty[/quote<] Wait...really? Is this the same for Intel >2400MHz (Kaby Lake for example)?

    • Jigar
    • 3 years ago

    I know this is asking too much but i would love to see a review of Ryzen 7 1800X running with RAM @ 3800 MHZ or 4000 MHZ.

      • BobbinThreadbare
      • 3 years ago

      Why is his asking for too much?

      Let’s see some benchies baby

        • raddude9
        • 3 years ago
          • chuckula
          • 3 years ago

          Whenever TR runs benchmarks of a RyZen part they get yelled at to turn off half the chip, get some new RAM, turn core parking on/off/sideways, turn off hyperthreading (no wait, turn ON hyperthreading), etc.

          If they had run more benchmarks, this announcement would be just another reason to invalidate any benchmarks that they had published in the first place.

            • Concupiscence
            • 3 years ago

            I’m also pretty sure diminishing returns kick in after DDR4-3200, outside of mostly theoretical benchmarks or beating the crap out of I/O. When I sell off most of my other hardware and switch to an 8 core Ryzen box next month, I’m reusing 16 gigs of DDR4-2400 because it just won’t make that much of a difference for my use cases.

            • derFunkenstein
            • 3 years ago

            It doesn’t help that AMD is cranking out firmware updates that they SWEAR will improve performance so much, the old benchmarks aren’t worth running.

            I mean, yeah, I bought a Ryzen rig, but at this point I can’t blame Jeff for skipping it. Performance “right now” supposably won’t be the same as performance in 2 weeks. It’s not a finished product, so don’t bother reviewing until it is.

            • BobbinThreadbare
            • 3 years ago

            I wouldn’t mind quickies that just test 2-3 things and show us the progression of performance with updates

            • derFunkenstein
            • 3 years ago

            I wouldn’t mind it but I can’t blame them for not doing it.

            • BobbinThreadbare
            • 3 years ago

            I’m not blaming anyone for anything, just asking for things.

            If I get them? Great. If not? I’ll live.

            • anotherengineer
            • 3 years ago

            Looks like to improve compatibility more than anything, but if performance increases, well I guess that’s a bonus.

            Just annoying when you plug in a couple sticks of ram and they don’t work :/ but I have seen that from both platforms. 10 BIOS updates later…………….improved stability/memory compatibility, not only amd mobos
            [url<]http://www.gigabyte.us/Motherboard/GA-Z170X-UD5-rev-10#support-dl[/url<] I think Ryzen for now is a finished product, it's the motherboard BIOS's that are not. But will they ever be?

            • derFunkenstein
            • 3 years ago

            Well, at least in TR’s reviews they are yet to get memory to run at DDR4-3200. The Ryzen 7 review had DDR4-3800 running at 2400 speeds, and the Ryzen 5 review got that speed up to 2667. We’ve not see memory hit Ryzen’s previous theoretical max (my guess is that it’s always been because of Command T1 vs T2) so 1.0.0.6 might fix that. It’s just more and more triage of a product that was released too soon.

            • Jeff Kampman
            • 3 years ago

            This is incorrect; the Ryzen 7 review used DDR4-2933 for Ryzen CPUs and the Ryzen 5 gaming tests used DDR4-3200.

            • derFunkenstein
            • 3 years ago

            Hmm. Where did I get that idea?

            Not a question for you to answer. Maybe it was another review.

            All the same it’s still not like you need to benchmark beta hardware in depth. If the lack of a completed review is keeping people from buying, I say good. They should wait.

            • anotherengineer
            • 3 years ago

            I think CPUs should be reviewed at whatever their official max. designed/supported memory clock is. If it is 2667, then bench at that. If it is 2133 or 2400 then bench at that. And there are the benches at the designed/recommended timings. If a cpu can/will run memory at it’s designed memory speed, I would not call it ‘beta hardware’ because of that reason alone.

            And bench again for whatever frequencies you can achieve, but put it under the overclocking benchmarks.
            -OC CPU benches @ stock ram speeds or
            -OC Ram benches @ stock cpu ram speeds or
            -OC Ram & CPU benches
            A good way to see how much gains can be had either way, but yes, definitely a lot more work.

            • cegras
            • 3 years ago

            I’m pretty sure intel parts have been picked apart in detail in the same fashion, especially when HT was first introduced, and even more so when they still had MCM that talked over FSB.

            Is it not their job to do this?

    • just brew it!
    • 3 years ago

    The possibility of functional IOMMU support on consumer-level gear is an interesting development. While it has theoretically been supported on AMD CPUs for a while, motherboard support has been flaky on non-enterprise platforms. Perhaps this added focus on IOMMU will push motherboard vendors to get it right this time around.

      • ptsant
      • 3 years ago

      Worked nicely for me on the FX8350 with the Asus Crosshair V. It should also work well on the Ryzen, because IOMMU v2.5 is a requirement for HSA and AMD want to push for its adoption.

      I have no idea how the MB vendors handled support, but I am mildly optimistic.

        • just brew it!
        • 3 years ago

        Ahh, that’s good to know. Last time I looked into this (maybe 2-3 years ago?) word still seemed to be that proper IOMMU support on consumer gear was iffy at best, even for boards that advertised IOMMU support and had the BIOS setting for it.

          • ptsant
          • 3 years ago

          I haven’t yet tested anything because frankly I don’t have much use for it right now. The option is buried deeply in the Crosshair VI BIOS (under Advanced/AMD CBS/NBIO Common Options).

          This is what the latest linux kernel says on my machine (Ryzen 1700X/Crosshair VI):
          [quote<] [ 1.270269] AMD-Vi: Found IOMMU at 0000:00:00.2 cap 0x40 [ 1.270270] AMD-Vi: Extended features (0xf77ef22294ada): [ 1.270270] PPR NX GT IA GA PC GA_vAPIC [ 1.270272] AMD-Vi: Interrupt remapping enabled [ 1.270273] AMD-Vi: virtual APIC enabled [ 1.270373] AMD-Vi: Lazy IO/TLB flushing enabled [ 1.271036] perf: AMD NB counters detected [ 1.271039] perf: AMD LLC counters detected [ 1.271585] perf: amd_iommu: Detected. (0 banks, 0 counters/bank) [/quote<]

        • MOSFET
        • 3 years ago

        I never had any problems either (actually, and still don’t) with Vishera platforms, always the Asus 990FX Pro R2, about 5 different boards.

        On another note, I think Ryzen and Threadripper are excellent, memorable names. Much better than Celeron and Xeon.

    • ronch
    • 3 years ago

    As we continue to unleash the power of Ryzen, we realize just how powerful this architecture is!! Don’t be surprised if it ends up being twice as fast as anything from Intel!

    Heh.

    Edit – er, I was being sarcastic.

      • D@ Br@b($)!
      • 3 years ago

      🙂

    • DancinJack
    • 3 years ago

    AMD could have avoided a lot of headache by just waiting a couple months to launch Ryzen. I still just do not understand why they HAD to get it out when they did.

    That said, there are some good features included here. Awful name, but good content.

      • LostCat
      • 3 years ago

      True, but I haven’t had any problems since the 1004 update.

        • 5UPERCH1CK3N
        • 3 years ago

        I’m still having cold boot issues even with a beta update and the latest beta BIOS, so I’m hoping for a better memory training process with this upcoming AGESA update.

      • ronch
      • 3 years ago

      Because they want their fans to help them with testing. Go on, keep on testing guys. Tell me when everything’s ironed out so I can go out and buy a copy. 🙂

        • raddude9
        • 3 years ago
        • derFunkenstein
        • 3 years ago

        You were the most obnoxious of all the AMD phanatix posting comment after comment about how Ryzen was really going to stick it to [s<]Sitwell[/s<] Intel.

          • ronch
          • 3 years ago

          Yes I applaud Ryzen but it doesn’t mean I won’t criticize its weak points. Unlike some folks who are all over it and will look the other way when talking about its weak points.

    • slowriot
    • 3 years ago

    [quote<]If you’re the kind of user that just needs (or loves!) virtualization every day, then AGESA 1.0.0.6-based firmware will be a blessing for you thanks to fresh support for PCI Express® Access Control Services (ACS). ACS primarily enables support for manual assignment of PCIe® graphics cards within logical containers called “IOMMU groups.” The hardware resources of an IOMMU group can then be dedicated to a virtual machine.[/quote<] Taken from the AMD site on this update. This is a major update for anyone whose been researching Ryzen for virtualization purposes. I'll definitely wait to see some independent testing, but prior to this update the PCIe lane groupings on the Ryzen chips produced tons of headaches when you tried to pass a GPU through to a VM. It was a complete show stopper for me to use Ryzen in my main rig (I use a Linux host and then a Windows VM for gaming).

      • RAGEPRO
      • 3 years ago

      [quote<]In news that will be music to the ears of another class of users, AMD also says that AGESA 1.0.0.6 will improve support for PCI Express Access Control Services, or ACS. According to the company, these improvements will allow users who run virtual machines on their Ryzen systems to better manage the assignment of PCIe devices within IOMMU groups. In short, the company says this feature will let VM wizards dedicate multiple graphics cards within a Ryzen system to the operating systems of their choice, whether natively installed or virtual, among other benefits.[/quote<] Taken from the Tech Report post on this update. This is a major update for anyone whose been researching Ryzen for virtualization purposes. I'll definitely wait to see some independent testing, but prior to this update the PCIe lane groupings on the Ryzen chips produced tons of headaches when you tried to pass a GPU through to a VM. It was a complete show stopper for slowriot to use Ryzen in his main rig (he uses a Linux host and then a Windows VM for gaming).

      • shank15217
      • 3 years ago

      Can’t you just disable ACS for pci pass through?

    • Srsly_Bro
    • 3 years ago

    Sup?

      • chuckula
      • 3 years ago

      [quote<]Sup?[/quote<] The memory frequencies, didn't you read the story?

        • morphine
        • 3 years ago

        Wazaaaaa?!

        • Srsly_Bro
        • 3 years ago

        After I posted! This is exciting news. I need to go read some Ryzen memory scaling articles.

        BRB, bro.

      • ronch
      • 3 years ago

      Sup yo!

Pin It on Pinterest

Share This