[UPDATE #1] AMD readies AGESA microcode update to improve Ryzen DDR4 memory compatibility (finally!)

[Morgan MLGman]

###[UPDATE 19.05.2017]###

 

https://www.forbes.com/sites/antonyleather/2017/05/18/an-interview-with-amd-the-latest-on-ryzen-memory-support-game-performance-and-ryzen-3s-launch/#695a3e555ff4

 

The update is coming "in late May" according to Robert Hallock - AMD's head of global technical marketing:

 

Quote

There is one thing coming down the pipeline – enthusiasts may be aware of the AGESA BIOS updates we’ve mentioned in the Ryzen community updates. There’s another one coming in late May that will focus on the robustness and compatibility of overclocked memory. Being overclocked, it’s easy to forget that you’re not guaranteed anything, but there are some knobs and buttons we can tweak to improve things here. There will be more information on this at the end of the May.

 

More interesting info from James Prior (product manager for Ryzen):

Quote

I think it’s important to point out that Ryzen is not limited to memory kits with Samsung’s B-die. We’ve got Hynix-based memory kits running at 3,200MHz today in the channel – there are plenty of people doing that. It is true that the B-die offers the best compatibility with our memory controller, especially at higher speeds, but that’s not the only die that works.

This is more down to memory vendors and how they set their timings and different aspects of what they call SPD – this impacts on compatibility. This is all part of reverse-engineering a non-standard protocol. JEDEC is very easy to work with and that all works beautifully with Ryzen. When we look at the auto overclocking profiles, that’s where things get a little bit harder as it’s all tuned for a proprietary standard we don’t have full visibility into.

Depending on the manufacturers of the memory and how they decide to program the memory cells themselves, it can change the compatibility with Ryzen. There are some things we can do here but there are things memory vendors can do as well and this is what we’re doing right now. We’re increasing plug and play compatibility but also going to memory vendors and asking what drove a decision here with this module at these timings and seeing if there are ways we can increase compatibility without impacting on any other products in the market.

 

 

 

[Jito463]

On 5/19/2017 at 5:38 AM, Morgan MLGman said:

This is more down to memory vendors and how they set their timings and different aspects of what they call SPD – this impacts on compatibility.

That's a good point.  People forget that a lot of this RAM was built specifically for the Intel platform, while AMD now has to work backwards to make those Intel specific settings work correctly on their brand new architecture.

 

Of course, it's easier to just accuse AMD of releasing a beta product than acknowledging the difficulties in this.

[Taf the Ghost]

23 minutes ago, Jito463 said:

That's a good point.  People forget that a lot of this RAM was built specifically for the Intel platform, while AMD now has to work backwards to make those Intel specific settings work correctly on their brand new architecture.

 

Of course, it's easier to just accuse AMD of releasing a beta product than acknowledging the difficulties in this.

More or less.  XMP is Intel tech, so they have to reverse-engineer to work with it.  Ryzen handles all of the official JEDEC timings just fine.  It's the overlocked stuff that's taking work.

 

Also, it appears that the 1.0.0.6 code is starting to reach motherboards, so any Ryzen owners should look forward to it soon.

[79wjd]

On 5/21/2017 at 8:54 AM, Jito463 said:

That's a good point.  People forget that a lot of this RAM was built specifically for the Intel platform, while AMD now has to work backwards to make those Intel specific settings work correctly on their brand new architecture.

 

Of course, it's easier to just accuse AMD of releasing a beta product than acknowledging the difficulties in this.

You're arguing semantics. AMD didn't prepare their platform to properly handle what is readily available on the market. Whether or not it's because of Ryzen's memory controller design being inherently lacking or most memory being tailored towards Intel is irrelevant. The fact remains that AMD released a product that couldn't work fully with what was/is available for consumers to buy. 

 

Well, AMD could have easily waited a month or two and ironed out the memory issues as well as providing motherboard manufacturers enough time to release motherboards with a proper bios. But they chose not to. They chose to rush things. They chose to rush a product to market. Intel has done the same thing, but that doesn't make it any less of a beta product.

[Humbug]

one thing seems clear.

we need newer faster JEDEC standards for DDR4 memory upto 3200Mhz.

[Jito463]

1 hour ago, djdwosk97 said:

You're arguing semantics. AMD didn't prepare their platform to properly handle what is readily available on the market. Whether or not it's because of Ryzen's memory controller design being inherently lacking or most memory being tailored towards Intel is irrelevant. The fact remains that AMD released a product that couldn't work fully with what was/is available for consumers to buy. 

 

Well, AMD could have easily waited a month or two and ironed out the memory issues as well as providing motherboard manufacturers enough time to release motherboards with a proper bios. But they chose not to. They chose to rush things. They chose to rush a product to market. Intel has done the same thing, but that doesn't make it any less of a beta product.

Perhaps, but given how cash strapped AMD is, I can forgive them for wanting to get Ryzen to market.  All RAM on the market worked, and some even worked at their rated speeds, it was just a matter of optimizing for the rest.  Intel doing it would be a different story, because they have the cash reserves to delay a launch like this.  Even then, we've seen various issues more severe than this on their launch platforms.

 

The fact is, early adopters always have to deal with some bugs.  There's no way to squash them all in the lab, without an army of beta testers who can track down every bug (and even then, there's a chance of some slipping through).  This is a very minor polish issue, and one that they're currently working to correct.

[79wjd]

32 minutes ago, Jito463 said:

Perhaps, but given how cash strapped AMD is, I can forgive them for wanting to get Ryzen to market.  All RAM on the market worked, and some even worked at their rated speeds, it was just a matter of optimizing for the rest.  Intel doing it would be a different story, because they have the cash reserves to delay a launch like this.  Even then, we've seen various issues more severe than this on their launch platforms.

 

The fact is, early adopters always have to deal with some bugs.  There's no way to squash them all in the lab, without an army of beta testers who can track down every bug (and even then, there's a chance of some slipping through).  This is a very minor polish issue, and one that they're currently working to correct.

Releasing a product that then gets reviewed and shown to have performance issues/quirks isn't the way to make money either. I seriously doubt pushing off the release two months would've made much of a difference. 

 

Minor or obscure bugs, sure, those won't show up until you have a large userbase. But RAM compatibility and BIOS issues that reviewers all found/had were obvious and were easily noticeable by anyone who did any form of testing with Ryzen pre-launch. 

[Paragon_X]

For those interested ASUS has a beta BIOS out with AGESA 1.0.0.6 if i am not mistaken with up to 4000mhz support

[Jito463]

26 minutes ago, djdwosk97 said:

Releasing a product that then gets reviewed and shown to have performance issues/quirks isn't the way to make money either. I seriously doubt pushing off the release two months would've made much of a difference.

The fact is, neither of us knows for sure.  It's possible that AMD had to release, or risk financial issues they may not have been able to recover from.  Perhaps they were getting pressure from some of their funding sources.  That's why I can cut them some slack, because I know AMD as a company has been in financial trouble for some time now.

 

Maybe it's hypocritical, but I see it as simply acknowledging the reality of the situation.

26 minutes ago, djdwosk97 said:

Minor or obscure bugs, sure, those won't show up until you have a large userbase. But RAM compatibility and BIOS issues that reviewers all found/had were obvious and were easily noticeable by anyone who did any form of testing with Ryzen pre-launch. 

I suppose we just have different definitions of what's acceptable.  We'll have to agree to disagree.

[Taf the Ghost]

3 hours ago, djdwosk97 said:

Releasing a product that then gets reviewed and shown to have performance issues/quirks isn't the way to make money either. I seriously doubt pushing off the release two months would've made much of a difference. 

 

Minor or obscure bugs, sure, those won't show up until you have a large userbase. But RAM compatibility and BIOS issues that reviewers all found/had were obvious and were easily noticeable by anyone who did any form of testing with Ryzen pre-launch. 

Ryzen handles the JEDEC standards just fine.  It's the XMP profiles, which is Intel tech and has to be reverse-engineered to be usable (otherwise you're violating IP), that had to worked around.  Some boards handled things better than others at the beginning, but Ryzen supported & handled all of the claimed speeds out of the box.  And even in the Day 1 reviews, a few people were getting 3600 on their RAM speed.

 

This isn't to say that the microcode was in a great state at the beginning, but the launch was in-line with what they stated would be handled for RAM. And they've worked to improve the ability to overclock RAM.  

[Belgarathian]

I see a lot of comments around OC RAM and saying that Ryzen is a Beta product. It's not, other than the VM issues @zMeul pointed out. This is very much a consumer ready product. You guys are forgetting OC is not guaranteed, and that Ryzen is fully compatible with JEDEC DDR4 standards. 

[ARikozuM]

10 hours ago, Belgarathian said:

I see a lot of comments around OC RAM and saying that Ryzen is a Beta product. It's not, other than the VM issues @zMeul pointed out. This is very much a consumer ready product. You guys are forgetting OC is not guaranteed, and that Ryzen is fully compatible with JEDEC DDR4 standards. 

I'd agree with you, but marketing overclock ability and technologies put them in this spot. If they didn't want to be responsible for the overclocking issues, they shouldn't have advertised so many concepts around their Infinity Fabric especially when AMD had to add an asterisk saying "will likely not happen" next to OC. 

[Emberstone]

This is probably a dumb question, but any word on this update reaching the Asus Prime B350 Plus BIOS soon? Impatiently waiting for that BIOS update right now 

[Taf the Ghost]

43 minutes ago, Emberstone said:

This is probably a dumb question, but any word on this update reaching the Asus Prime B350 Plus BIOS soon? Impatiently waiting for that BIOS update right now 

Have to poke around Asus specific forums for that information. 

[Morgan MLGman]

1 hour ago, Emberstone said:

This is probably a dumb question, but any word on this update reaching the Asus Prime B350 Plus BIOS soon? Impatiently waiting for that BIOS update right now 

Apparenty ASUS mobos got the new update recently, check on your mobo's webpage

[Emberstone]

55 minutes ago, Morgan MLGman said:

Apparenty ASUS mobos got the new update recently, check on your mobo's webpage

For the Asus Prime B350 Plus, the last update was almost exactly a month ago 

 

That's the version I'm on now. I hope they didn't forget my board.