Question B550 chipset, so AMD joins the dark side after all.

[Shivansps]

I just read the article...

So let me get this straight, this chipset is coming out like a year later, they did not even bother to add CPU PCI-E 4.0 uplink support or to increase the number of sata ports that is ALREADY a problem on every 6 sata B450 motherboard (NVME x4 disables the 2 SOC Sata, thus 6 sata B450 mbs losses 2 sata if NVME is used), and they even dare to futher reduce backguard compatibility?

I was not expecting for the PCI-E lanes FROM the chipset to be 4.0, but only USB 3.2 G2, no more satas, CPU link still 3.0 and the PCI-E lanes 3.0 is beyond disappointing.

 

[DrMrLordX]

But this whole thing of spending months and months to try to make PCI-E 4.0 traces to work on a chipset that has no PCI-E 4.0 I/O and dont even need those traces makes no sence to me. It is just a glorified PCI-E I/O hub device and little more.

Traces produce signal interference. That's why routing is so critical. Remember that someone had to produce the entire reference platform for B550 - like I said, these things don't exist in a vacuum. It was at least partially ASMedia's job to do all the trace routing to/from the CPU socket in such a way that signal interference wouldn't screw up anything else, including the chipset of their design (and vice versa). If you really want to get into the technicals you'll have to discuss it with people from the mobo design world who work on multi-layer PCBs. In this case, ASMedia's job was to produce something cheap enough that it could be sold at price points at or near B450 and B350. PCIe 4.0 itself supposedly requires a 6-layer board (though x570 has a few 4-layer boards). More layers = higher baseline costs.

And i know about references designs as well, that means nothing, there is plenty of examples of OEMs doing things whiout references designs, like when AMD blocked the CPU core unlock on Phenoms, most OEM developed their own options

That particular issue had nothing to do with reference hardware designs. That was a software/firmware issue. You're pulling the rest of it out of thin air. Next you're going to tell me that SkyOC required significant PCB redesign. Here's a hint: it didn't!

it was just part of the controversy of the 300 series chipset at that time, nothing special. BTW, A320 is the only 300 series chipsets that is not EOL yet, so i was kinda right to say it was more important than some people belived.

. . . aaaaaaaand we see the real problem here. You still think A320 should have supported Matisse. It was a controversy in the minds of a few. We wasted pages of forum posts trying to point out that A320 was a Bristol Ridge chipset for bargain systems. Some things never change.

First, timing was never in question here, it was about if you can do a PCI-E 4.0 mb with proper traces to use the lanes from the CPU to the main PCI-E and NVME with a b450. There is no information to say this is not possible.

I'm sorry, but that's just crazy.

The chipset is not much diferent from any other PCI-E device and is not connected to those lanes either.

It doesn't have to be connected for signal interference to cause key motherboard components to malfunction. If the traces get too close to one another in any dimension, signal integrity can break down.

PCIe 4.0 is NOT simply an extension of PCIe 3.0 | Keysight Community

The New Design and Test Challenges If you plan on leveraging the work you previously did in your PCIe 3.0 design, you are mistaken. A doubling of the

community.keysight.com

But if timing is the question, lets remember Phenom CPU core unlock, thats a fine example of AMD blocking something

Those cores often didn't work correctly when unlocked. They locked those cores and sold them as cheaper CPUs because they didn't pass validation for a higher-end product. It was a simple firmware switch to reactivate those cores. Nobody had to reroute traces on the motherboard to make that happen.

Where do you get this stuff???

 

[DrMrLordX]

Not after 2020, but I thought they had promised it would be supported in 2020.

That was my recollection as well.

AMD massively overdelivered comparing to "industry" standard

Overall I agree. They've mishandled the pre-launch situation with Zen3/Vermeer though. Lots of people thought that AMD would continue to overdeliver - and now, apparently, they aren't. I still think there's more going on than AMD just being meanie-heads. OEMs need to make a buck, and they're used to being able to sell new batches of boards every year or two when Intel was their workhorse. Now they're having to rely on AMD boards to keep up their sales to DiY buyers. Why do you think x570 got as expensive as it did? Those boards had to replace sales of Z390 and Z370 that never happened.

 

[piokos]

For OS and programs, NVME. For data storage in volumes, sure thats where sata comes in. I even have that om my main box. But more than 4 sata ports is just not needed anymore. I mean I am not sure the limit, but I have seen 10 tb drives !

As I said: 4 SATA ports are enough for almost all consumers, so AMD can easily afford that compromise.

But the fact that it's not *needed* doesn't mean it can't be utilized. I use 5 SATA ports (and 1 M.2, so 6 drives total). And it's an mITX system.
Why? Because I can.
When I approach a storage limitation, I add a new drive - not replace one.
SSDs don't use a lot of power, don't take a lot of space and don't make any noise.

I have 6 SATA ports, so that's the limit, but if I had an ATX system, I'd keep adding SSDs for as long as they fit. No reason to throw a working one away.

And it lets me save a lot of money, because I can hit a target capacity buying smaller drives.
And B550 is supposed to be a value-oriented platform.

 

[piokos]

I am sure this is a nightmare they just have to back out of and again doesn't actually go against what they said.

While AM4 longevity clearly isn't bad, let's just digest what you said (and what, noticeably, many people agree with).

What if they kept using AM4, but CPUs would require chipsets from the same generation?
What if a CPU would require a particular chipset? Like Ryzen 3, 5 only on "B..." and Ryzen 5,7 only on "X..." ?
What if Zen2 chips would already be "AM5" but AMD would keep making revised Zen+ AM4 chips for few more years?
All of that would also comply to the initial claim of supporting AM4 until 2020.
A claim that was so vague that we had to endure 3 years of flame wars about platform longevity.

And the reality is as follows:
Essentially, all 3 chipset generations support the CPU generation they launch with and one that follows.
I.e. you have a single opportunity to upgrade.

That's better than what Intel offers, but not exactly a revolution many people foreseen.
Yes - X470, B450 and X570 also support 1 generation back, but that's hardly an "upgrade path".
Previous generation is always fully replaced and - at least until now - always with significant gains.
The only exception: people who buy an expensive X570 and decide to use it with a 4-core Zen+ APU until Zen2 APUs arrive.
I'm sure they exist, but it's not exactly a huge niche.

In the end, changing sockets is a forced process anyway - with enough space and pin headroom, a socket could last for a decade. It has nothing to do with CPU compatibility.
Sockets are changed so that it's easier for consumers to choose parts (and harder to make a mess).

Intel has been using the same physical socket since 2009: 1156, 1155, 1150, 1151 (*3) and 1200.
They could call it "Socket H" and make a funky CPU compatibility slide that would dwarf that one from AMD.

 

[coercitiv]

Lots of people thought that AMD would continue to overdeliver - and now, apparently, they aren't.

As I said in a previous post, it's bad idea to claim a competitive advantage and leave the interpretation to consumers, we will chose that which is in our advantage.

When AMD comes out and says AM4 will be a long-life socket, when they differentiate themselves from Intel on this, I will not start documenting on required UEFI flash sizes and OEM software support & sales requirements. I will not do AMDs homework even as an enthusiast. What I will do instead is to look at Intel and their 2 gens per socket (present gen + next gen) and assume AMD will deliver at least 3 gen support. Last gen chipset on the socket will obviously carry exceptions, and the mix of future/backwards compatibility is also up to them. (although "future" is obviously a minimum 1x requirement).

They also had the opportunity to be specific in advance about the scope of their AM4 chipset support, or simply not claim a competitive advantage and reap lesser PR rewards.

 

[gdansk]

The more I think about this the more disappointed I am with AMD.

1. B550 should have launched last year with X570. Instead people bought B450 boards for 3000 series CPUs not knowing they were a dead end. Some refreshed B450 boards gave the impression they would not be a dead end.
2. BIOS sizes limitations would not impact all B450/X470 boards. MSI's Max series, for example, have larger ROM than some X570 boards.
3. B550 should support previous generation processors to the same extent as X570 (i.e. back to the 2000 series).

I find it unlikely AMD will backtrack. I believe them when they say it isn't easy to have 1000 to 4000 series support on a single chipset. But it seems like it should be possible (though not with a single BIOS version).

 

[DamZe]

As much as I see the point in them "going back on their promise", original Ryzen owners who bought a X370 or B350 board had the ability to upgrade their IPC up 30 percent from Ryzen gen1 to 3, which is something Intel owners could only dream of.

AMD is in the business of making money, and so are their board partners. I personally expected that they would most likely limit compatibility going forward on some AM4 platforms.

The worst thing is that AMD didn't release B550 boards last year, which is a very calculated attempt to force people to chose between going budget friendly (B450) or high end (X570) with all the bells and whistles for your 3000-series Ryzen chips.

Overall I think 3 years of solid upgrade paths for early adopters is still a "great bang for the buck", considering what Intel was doing for the last 10 years.

 

[amrnuke]

What has been the timeline of AMD revealing new chipsets?

B550 comes at a time that makes me wonder if they were going to leave the mid-range mobo market up to B450 but then realized that there is some kind of significant flaw running a Zen3 on a 4x0 or 3x0 board and have had to rush out a mid-range option mid-cycle (and even the lower end A520 line rumored).

 

[piokos]

As much as I see the point in them "going back on their promise", original Ryzen owners who bought a X370 or B350 board had the ability to upgrade their IPC up 30 percent from Ryzen gen1 to 3, which is something Intel owners could only dream of.

X370 and B350 don't support Zen2. It's even on the graph that started this discussion.

First gen chipsets officially only support 3000-series APUs, which are Zen+ (like Ryzen 2000-series).
AMD uses this naming on purpose and it clearly works.

AMD is in the business of making money, and so are their board partners.

Absolutely. No one says this is a bad practice or that AMD explicitly said ALL AM4 CPUs will work on ALL AM4 motherboards.
The issue is that they gave us very general information of long-term support and people interpreted it in different ways. People who like AMD read this as "full compatibility". This is one of the themes that dominated CPU discussions on forums like this one.

I personally knew that they would most likely limit compatibility going forward on some AM4 platforms.

Well, maybe you did. Good for you. Many people were lured by the optimistic scenario - including popular reviewers.

Overall I think 3 years of solid upgrade paths for early adopters is still a "great bang for the buck", considering what Intel was doing for the last 10 years.

Intel isn't focusing on DIY clients, so they didn't even try to fake anything. Yes, they replace sockets all the time.
But AMD is focused on DIY clients...

 

[maddie]

The more I think about this the more disappointed I am with AMD.

1. B550 should have launched last year with X570. Instead people bought B450 boards for 3000 series CPUs not knowing they were a dead end. Some refreshed B450 boards gave the impression they would not be a dead end.
2. BIOS sizes limitations would not impact all B450/X470 boards. MSI's Max series, for example, have larger ROM than some X570 boards.
3. B550 should support previous generation processors to the same extent as X570 (i.e. back to the 2000 series).

I find it unlikely AMD will backtrack. I believe them when they say it isn't easy to have 1000 to 4000 series support on a single chipset. But it seems like it should be possible (though not with a single BIOS version).

Just missed making this recommendation for some friends.

 

[DrMrLordX]

X370 and B350 don't support Zen2. It's even on the graph that started this discussion.

"Unofficially" they do. As in, I can't think of a single B350 or X370 motherboard without Matisse support. Are you implying that OEMs will patch in support for Vermeer in the same way?

 

[coercitiv]

"Unofficially" they do. As in, I can't think of a single B350 or X370 motherboard without Matisse support. Are you implying that OEMs will patch in support for Vermeer in the same way?

It's official support, only "selective". As it stands now, this won't repeat with 400 series unless AMD changes their stance, which is to not provide OEMs with the necessary AGESA updates.

 

[piokos]

It's official support, only "selective". As it stands now, this won't repeat with 400 series unless AMD changes their stance, which is to not provide OEMs with the necessary AGESA updates.

View attachment 20796

"Selective Beta BIOS" is not proper support - they would just give it a "✓" otherwise.
What does it even mean? An experimental implementation (public beta)?

 

[DrMrLordX]

@piokos

10 months later, we can clearly see that it means that B350 and X370 get the same updates as B450 and X470, but on a relaxed schedule and with some features sometimes missing in UEFI.

 

[VirtualLarry]

Honestly, I wonder if the motivation driving this lack of compatibility, is either:
1) OEMs just want to make more money than they have been, or
2) Some relatively unforeseen hardware incompatibility, which is likely also why B550 chipset and boards were delayed.

Could #2 be the real reason for this lack of continuous compatibility? Some sort of technical glitch, that was relatively unsurmountable, and if they had allowed it, then there might be scores of AM4 / Ryzen rigs out there, with "compatibility glitches", not unlike the "teething pains" we went through with early-adopter Zen1 and Zen2 boards and immature AGESA code(s), and whatnot.

AMD doesn't want their platform to be known as an "unstable one", they don't need that (dis-)reputation. Not now.

So, perhaps this lack of completely continuous compatibility with AM4 socket chipset(s) and CPU(s), is a function of that reticence to appear unstable, because the combinations at this point that would need to be coded for, and tested on, are approaching "too large" numbers for the mobo OEMs.

Maybe AMD will "relent" down the line, and release AGESA codes for back-rev chipset boards, to support Zen3, but I'm not holding my breath. If they wouldn't let B450 boards, support PCI-E Gen 4.0 for the CPU-driven PCI-E slot, then they're probably not going to enable back-rev compatibility either.

 

[chrisjames61]

I currently have the following X370 boards. I may have some I am forgetting. MSI X370 XPower Gaming Titanium, ASUS Prime X370-Pro, ASUS ROG Crosshair VI Hero, ASRock X370 Taichi, MSI Performance Gaming B450I Gaming Plus, ASUS ROG Strix B450-I Gaming, B450 I AORUS PRO WIFI. I think they had a good run. Even the B450 boards. Regardless of if they take the next generation of processors or not hardly makes them obsolete. The 3000 series of processors are beasts and will remain so for a good while. I personally will want a new board when the time the 4th gen cpu's hit my local Micro Center's shelves.

 

[Shivansps]

Honestly, I wonder if the motivation driving this lack of compatibility, is either:
1) OEMs just want to make more money than they have been, or
2) Some relatively unforeseen hardware incompatibility, which is likely also why B550 chipset and boards were delayed.

Could #2 be the real reason for this lack of continuous compatibility? Some sort of technical glitch, that was relatively unsurmountable, and if they had allowed it, then there might be scores of AM4 / Ryzen rigs out there, with "compatibility glitches", not unlike the "teething pains" we went through with early-adopter Zen1 and Zen2 boards and immature AGESA code(s), and whatnot.

AMD doesn't want their platform to be known as an "unstable one", they don't need that (dis-)reputation. Not now.

So, perhaps this lack of completely continuous compatibility with AM4 socket chipset(s) and CPU(s), is a function of that reticence to appear unstable, because the combinations at this point that would need to be coded for, and tested on, are approaching "too large" numbers for the mobo OEMs.

Maybe AMD will "relent" down the line, and release AGESA codes for back-rev chipset boards, to support Zen3, but I'm not holding my breath. If they wouldn't let B450 boards, support PCI-E Gen 4.0 for the CPU-driven PCI-E slot, then they're probably not going to enable back-rev compatibility either.

About the compatibility, OEM wants more compatibility, that means more sales for them, and less angry customers yelling, so #1 is no.

If i had to bet, ill say that ASmedia failed to get PCI-E 4.0 cpu link working on the B550, and finally settled for 3.0 as the reason for the delay, it could be they wanted to do b450 stock clearing beforehand, considering they planned to cut support this makes more sence now, but this is just my guess. Also they still need to support everything on X570 so, i dont think that hardware incompatibility is the issue.

About the 400 series forward compatibility, im really not suprised, they tried to pull this move with the 300 ones when Matisse launched, it was the huge backlash that stopped it.

About the 500 series retro-compatibility, there is no excuse, when X570 launched every 14mn Ryzen were EOL (Ryzen 1000 and APU 2000), they still went back on that add added unofficial support with Agesa 1.0.0.4.
Now B550 there is no reason for it to have less compatibility than X570, it just makes things very messy. Also they are cutting support to products that arent EOL. Raven2 on 14nm is not EOL (3000G), 12nm Ryzen arent EOL (1200AF-1600AF-2600-2700), 12nm APU arent EOL (3200G, no info on 3400G). They are putting the blame on bios size, what ill admit it is a issue, still think of this, a old cheapo Asus Prime A320M-K supports from the A6-9500 to the R9 3950X on the same bios, so the cheaspest of the AM4 board already has a bios large enoght to support the entire AM4 product stack.

My recomendation is that people have to make some noise, they will eventually backtrack. The main reason is the CPU support is a f* mess, i can assure you OEMs does not want that, but they are no mages, AMD can just AGESA block and thats it.

 

[Markfw]

<snip>Also they still need to support everything on X570 so, i dont think that hardware incompatibility is the issue.<snip>

Now why would it need to everything on x570 ? its a bargain motherboard, its cheaper, so it will have less features than an x570.

You can't have your cake and eat it too as the saying goes.

 

[Shivansps]

Now why would it need to everything on x570 ? its a bargain motherboard, its cheaper, so it will have less features than an x570.

You can't have your cake and eat it too as the saying goes.

Because they cant cut compatibility now, they launched X570 with 12nm CPU&APU support, you cant take away that now to match B550. So they are forced to support these CPUs on x570. What they can take away is the 14nm CPU&APU support they added unofficially with 1.0.0.4.

 

[mopardude87]

I went x570 to get the most out of this build, so far no regrets.

 

[jpiniero]

Remember that Zen 3 is the end of the line. Maybe they have a bunch of old chipsets they have to unload, and they are hoping to steer buyers of Zen 1/+ to that to clear inventory.

 

[maddie]

About the compatibility, OEM wants more compatibility, that means more sales for them, and less angry customers yelling, so #1 is no.

If i had to bet, ill say that ASmedia failed to get PCI-E 4.0 cpu link working on the B550, and finally settled for 3.0 as the reason for the delay, it could be they wanted to do b450 stock clearing beforehand, considering they planned to cut support this makes more sence now, but this is just my guess. Also they still need to support everything on X570 so, i dont think that hardware incompatibility is the issue.

About the 400 series forward compatibility, im really not suprised, they tried to pull this move with the 300 ones when Matisse launched, it was the huge backlash that stopped it.

About the 500 series retro-compatibility, there is no excuse, when X570 launched every 14mn Ryzen were EOL (Ryzen 1000 and APU 2000), they still went back on that add added unofficial support with Agesa 1.0.0.4.
Now B550 there is no reason for it to have less compatibility than X570, it just makes things very messy. Also they are cutting support to products that arent EOL. Raven2 on 14nm is not EOL (3000G), 12nm Ryzen arent EOL (1200AF-1600AF-2600-2700), 12nm APU arent EOL (3200G, no info on 3400G). They are putting the blame on bios size, what ill admit it is a issue, still think of this, a old cheapo Asus Prime A320M-K supports from the A6-9500 to the R9 3950X on the same bios, so the cheaspest of the AM4 board already has a bios large enoght to support the entire AM4 product stack.

My recomendation is that people have to make some noise, they will eventually backtrack. The main reason is the CPU support is a f* mess, i can assure you OEMs does not want that, but they are no mages, AMD can just AGESA block and thats it.

I can't seem to make sense of your 1st line.

How would having individuals not needing to buy a new board to upgrade CPU generations result in more sales. I can see if some mainboard manufacturers could do that, then they would get more sales at the expense of the rest. If all had to follow the same guidelines then no, sales would decrease if no one needed to buy new boards for several generations.

 

[Muadib]

I went x570 to get the most out of this build, so far no regrets.

I did the same, and I agree with you. With DDR5 on the horizon, I wasn't expecting much. If I can get another cpu series on this motherboard, I'm grateful.

 

[Insert_Nickname]

Wait, no I don't. When it's obvious the OEMs didn't even want to support Matisse on B450 anyway. The only thing AMD "forced" anyone to do was support Matisse on every AM4 chipset except A320.

Of course the forum's one A320 fan pitched a fit when that happened too. Reeeeeeeee

Meanwhile, some A320 boards -did- get Matisse support. Eventually. My own Asus A320M-K included.

https://www.asus.com/Motherboards/PRIME-A320M-K/HelpDesk_CPU/

Official 3000 series support. I don't know how they crammed everything in there but it apparently supports everything back to Bristol Ridge too.

 

[Glo.]

Honestly, I wonder if the motivation driving this lack of compatibility, is either:
1) OEMs just want to make more money than they have been, or
2) Some relatively unforeseen hardware incompatibility, which is likely also why B550 chipset and boards were delayed.

Could #2 be the real reason for this lack of continuous compatibility?

Could it be both reasons?