• Now's your chance win big! Join our community and get entered to win a RTX 2060 GPU, plus more! Join here.

    Pi Cast Episode 3 streams live on Tuesday, August 4th at 2:30 pm ET (7:30 PM BST). Watch live right here!

    Catch Scharon on the Tom's Hardware Show live on Thursday, August 6th at 2:00 pm ET (7:00 PM BST). Click here!

News AMD Confirms Zen 3 Compatibility on B550, X570 Motherboards

Apr 1, 2020
52
19
35
0
"For pre-500 motherboards, vendors could easily weed out support for the older Ryzen 1000 or 2000-series lineups to make space for Zen 3 via an optional firmware update. Existing owners may not complain if the compromise offered them a cheap path to upgrade to Zen 3. "


This is a question I hope TomsHardware, as well as all of the other reputable review sites, pose to all of the AIBs, and post an article about that. They all need a statement on record as to why they will not remove Ryzen 1000 and 2000 series support and make a BIOS which will accept Ryzen 3000 and 4000 series chips on 128Mbit boards, mostly the X370 and X470 series. We need to know if this is due to an explicit order from AMD, or if they simply want to sell more motherboards. The way AMD makes it sound in their announcement, it's a direct order.
 

Chris Fetters

Honorable
Dec 6, 2013
26
8
10,535
0
"For pre-500 motherboards, vendors could easily weed out support for the older Ryzen 1000 or 2000-series lineups to make space for Zen 3 via an optional firmware update. Existing owners may not complain if the compromise offered them a cheap path to upgrade to Zen 3. "


This is a question I hope TomsHardware, as well as all of the other reputable review sites, pose to all of the AIBs, and post an article about that. They all need a statement on record as to why they will not remove Ryzen 1000 and 2000 series support and make a BIOS which will accept Ryzen 3000 and 4000 series chips on 128Mbit boards, mostly the X370 and X470 series. We need to know if this is due to an explicit order from AMD, or if they simply want to sell more motherboards. The way AMD makes it sound in their announcement, it's a direct order.
This is literally no different than when the 300 series lost official support for 3rd Gen (look at that graph again buddy). The OEMs can do whatever they want (just look at A320, where AMD was absolutely explicit they weren't supporting).
 

JarredWaltonGPU

Senior GPU Editor
Editor
Feb 21, 2020
338
273
560
0
This is a question I hope TomsHardware, as well as all of the other reputable review sites, pose to all of the AIBs, and post an article about that. They all need a statement on record as to why they will not remove Ryzen 1000 and 2000 series support and make a BIOS which will accept Ryzen 3000 and 4000 series chips on 128Mbit boards, mostly the X370 and X470 series. We need to know if this is due to an explicit order from AMD, or if they simply want to sell more motherboards. The way AMD makes it sound in their announcement, it's a direct order.
I'm guessing a big part of it is that updating the BIOS code of all older motherboards would be time consuming. Sure, a company can earn some good will by doing that, but ASRock, Asus, Gigabyte, MSI, etc. would really love people to upgrade to a new motherboard and CPU rather than trying to make a new CPU work in an old motherboard and potentially causing more support headaches.

Put another way, lots of people complain about Intel making new chipsets that aren't backward compatible with older CPUs, and likewise making new CPUs that aren't backward compatible with old chipsets. However, I don't think I've ever heard a motherboard company complain about Intel's approach to planned platform obsolescence.
 
Reactions: TCA_ChinChin

TCA_ChinChin

Honorable
Feb 15, 2015
413
102
11,090
29
Personally, I much prefer shorter platform cycles if it means I don't have to worry about what supports what. Not a fan of A supports B with caveats 1, 2, 3c, 4a, 4d, 5 and 6c.
New stuff works with new stuff with no worry. If you want new stuff to work with old stuff then it requires some reading.

Edit: If you don't want complicated choices, then you don't have to make them. Some people like making those choices though.
 

Deicidium369

Upstanding
Mar 4, 2020
388
59
290
5
Personally, I much prefer shorter platform cycles if it means I don't have to worry about what supports what. Not a fan of A supports B with caveats 1, 2, 3c, 4a, 4d, 5 and 6c.
I never upgrade - I replace. The i9900K / Dual 2080Ti will be replaced completely with Rocket Lake S and Ampere. Nothing will be retained from the old system. Licenses for Office Pro & Visio will be transferred to new system - but OS, Malwarebytes, etc will stay on the system they were installed on.
 

InvalidError

Titan
Moderator
I never upgrade - I replace. The i9900K / Dual 2080Ti will be replaced completely with Rocket Lake S and Ampere.
Bold statement on the CPU side. I'd wait to see if Willow Cove's IPC gains can offset its likely significant clock frequency deficit from getting back-ported from 10nm+++ (10nm++ is the process Intel had hoped to have for Ice Lake / Sunny Cove years ago, Willow Cove would be one step after that) to 14nm++++++ first!
 
Reactions: mitch074

mitch074

Distinguished
The way I read it is that it depends on the amount of flash BIOS available on the motherboard. Most vendors already made smaller UEFI systems to accommodate Zen 2's original AGESA, and then AMD unified their AGESA the UEFI images sizes got quite a bit smaller. For those who already did the work, adding support for Zen 3 may not be so much of a stretch.
 

InvalidError

Titan
Moderator
For those who already did the work, adding support for Zen 3 may not be so much of a stretch.
Since AGESA contains some chipset-related stuff, AMD has the final say on how many chipsets it will ultimately include in its Zen 3/3+ update. Some chipsets may get included though not officially supported like AMD did with Zen 2 and 300-series.
 

Phaaze88

Illustrious
Ambassador
Oh wow... this royally sucks.
I guess accommodating 4 different sets of cpus was too much, huh?

Well, I think we all know how AM5 is gonna roll: Ryzen 5xxx, 6xxx, and 7xxx - or dare I say the Intel route, with only 5xxx and 6xxx...
At least they're not charging an organ for their products, right?
 

mitch074

Distinguished
Oh wow... this royally sucks.
I guess accommodating 4 different sets of cpus was too much, huh?

Well, I think we all know how AM5 is gonna roll: Ryzen 5xxx, 6xxx, and 7xxx - or dare I say the Intel route, with only 5xxx and 6xxx...
At least they're not charging an organ for their products, right?
AM5 is supposed to support DDR5 - changes in pinout and voltages, that does warrant a socket change.
 

mitch074

Distinguished
?
I was already aware of that. Just talking about cpu compatibility here. They're not going to try and stretch AM5 as far as they did with AM4.
Who knows. It does seem the main problem with older chipsets is the amount of flash ROM dedicated to UEFI, nothing about the pinout or power delivery. Oh well, I guess I'll keep my X2700 a few years still - it's not as if I'm short on CPU power.
 

InvalidError

Titan
Moderator
Who knows. It does seem the main problem with older chipsets is the amount of flash ROM dedicated to UEFI, nothing about the pinout or power delivery.
I wouldn't be surprised if AMD was under some pressure from motherboard manufacturers who don't want to have to keep updating years old motherboards they aren't making any more money from. If one board manufacturer adds support, then it becomes a PR nightmare for all other manufacturers to not do so.
 

mitch074

Distinguished
I wouldn't be surprised if AMD was under some pressure from motherboard manufacturers who don't want to have to keep updating years old motherboards they aren't making any more money from. If one board manufacturer adds support, then it becomes a PR nightmare for all other manufacturers to not do so.
Neither would I - and that's why there's a chance we may yet see expanded compatibility.
 

JarredWaltonGPU

Senior GPU Editor
Editor
Feb 21, 2020
338
273
560
0
I wouldn't be surprised if AMD was under some pressure from motherboard manufacturers who don't want to have to keep updating years old motherboards they aren't making any more money from. If one board manufacturer adds support, then it becomes a PR nightmare for all other manufacturers to not do so.
This, totally. It's not just pressure from the mobo makers; maintaining backward and forward compatibility is a nightmare for AMD as well. the 300-series chipsets and 1st Gen CPUs were a bit flaky -- first of a new generation of Ryzen CPUs and boards. The first six months was a mess of compatibility. I had four different 300-series boards fail on me over time. Trying to support future CPUs had to have taken a ton of time and effort, with only partial success.

The 400-series boards were better, but I still had at least three different boards go belly up on me over the past couple of years. Really, it was just what we really wanted from Gen1 boards and CPUs -- refinements and some microcode updates and tweaks to make the platform as a whole feel less raw.

500-series and 3rd Gen CPUs changed a bunch of stuff. While it was possible to make things backward compatible with earlier CPUs, doing proper validation would basically double the time required for Q&A on each motherboard.

Put another way:
Asus makes six X370, three B350, and three A320 boards (just looking at the site -- there may be others that aren't listed). That's 12 "Gen1" boards.
There are six more X470 and nine B450 boards. 15 "Gen2" boards.
And right now there are 12 X570 boards, while we wait for B550.
That's at least 39 different AM4 motherboards from just one company.

Now, imagine trying to properly validate each and every board with the 10 Gen1 Zen CPUs and APUs, six Gen2 Zen+ CPUs and APUs, and currently nine Gen3 Zen 2 CPUs -- with potentially several additional models that are only available in select parts of the world.
That's at least 25 Ryzen CPUs, but there are probably closer to 30 once you add in revisions like the 1600AF.

Validating every board across every CPU means at least 975 potential combinations -- and then there's RAM compatibility testing, USB devices, M.2 SSDs, etc. Just getting a platform to work well with one generation of hardware takes effort. Cross-generation compatibility can double or triple the effort required, and then the companies are stuck supporting products that are no longer for sale with new CPUs.

If I were using a Gen1 Ryzen CPU and motherboard, and felt the need to upgrade, I'd be more inclined to just sell the old system and buy a completely new PC. Or give it to a family member or a child that needs a PC. Otherwise it's like putting new wine in old skins.
 

mitch074

Distinguished
This, totally. It's not just pressure from the mobo makers; maintaining backward and forward compatibility is a nightmare for AMD as well. the 300-series chipsets and 1st Gen CPUs were a bit flaky -- first of a new generation of Ryzen CPUs and boards. The first six months was a mess of compatibility. I had four different 300-series boards fail on me over time. Trying to support future CPUs had to have taken a ton of time and effort, with only partial success.
I'm not so sure about that - I built a handful of Zen 1 systems with these early chipsets, but none failed. Afterwards, I was happy to discover I had bought the "stable" models (with proper VRMs and traces and such), as I only bricked one - and it was my fault (foot caught in PSU cable while updating BIOS). All of these boards did have very long boot times, bad stability problems under Windows etc. And it was all solved with subsequent UEFI, Windows and chipset driver updates. Most of these systems still run today, and they fulfill their gaming and productivity work daily without fail - from a junior programmer's workstation to a demo setup for a compact gaming system, they're all still Zen 1 with B350 chipsets, and they've been running daily (if not 24/7) since 2017-2018.
It IS sad to think that they wouldn't bear Zen3; I console myself thinking that, if an upgrade does become needed, Zen2 ain't too bad still, and a simple CPU upgrade 4 years down the line keeping the machine still relevant is still mighty impressive.

And then I think about that smart guy who told me in 2017, "spend 200 bucks more and get a 7700k, it'll last longer - AMD is sh*t, Intel is king". And I look at that old B350 board that supports thermal envelope management (i.e. I can restrict a 90W CPU to use only 65W) showing on the latest BIOS : "added support for 3900X and 3950X". Mwahahahaha.
 

InvalidError

Titan
Moderator
Most of these systems still run today, and they fulfill their gaming and productivity work daily without fail
I'd hope they're all still working today, I'd be mighty pissed if one of my PCs died before 5+ years old. I still have a perfectly working 12 years old Core2 in my living room and a 16 years old P4 that I keep around due to node-locked academic licenses that are still useful every now and then.
 

JarredWaltonGPU

Senior GPU Editor
Editor
Feb 21, 2020
338
273
560
0
I'm not so sure about that - I built a handful of Zen 1 systems with these early chipsets, but none failed. Afterwards, I was happy to discover I had bought the "stable" models (with proper VRMs and traces and such), as I only bricked one - and it was my fault (foot caught in PSU cable while updating BIOS). All of these boards did have very long boot times, bad stability problems under Windows etc. And it was all solved with subsequent UEFI, Windows and chipset driver updates. Most of these systems still run today, and they fulfill their gaming and productivity work daily without fail - from a junior programmer's workstation to a demo setup for a compact gaming system, they're all still Zen 1 with B350 chipsets, and they've been running daily (if not 24/7) since 2017-2018.
It IS sad to think that they wouldn't bear Zen3; I console myself thinking that, if an upgrade does become needed, Zen2 ain't too bad still, and a simple CPU upgrade 4 years down the line keeping the machine still relevant is still mighty impressive.

And then I think about that smart guy who told me in 2017, "spend 200 bucks more and get a 7700k, it'll last longer - AMD is sh*t, Intel is king". And I look at that old B350 board that supports thermal envelope management (i.e. I can restrict a 90W CPU to use only 65W) showing on the latest BIOS : "added support for 3900X and 3950X". Mwahahahaha.
Granted, I swap hardware components for testing more often than ... just about anyone other than a hardware reviewer. But I can absolutely say that I had way more problems with early AM4 boards than anything I've used from Intel. In fact, off the top of my head, I can't think of the last time I had an Intel board fail on me. I've still got at least two or three of the dead AMD boards sittings around somewhere. Anyway, could have been caused by using higher spec RAM maybe (DDR4-3200 CL14), but these weren't budget boards -- MSI and Gigabyte models that were at least $175 at launch.

One of the GB X470 boards died after a BIOS update and took my 2700X CPU with it -- not during the BIOS update, but after. It was super weird. I flashed the BIOS, did some testing, swapped CPUs. Booted and I think I went into the BIOS settings because of the CPU swap. "New CPU detected." Saved and rebooted, and the system never came back. I tried to get it working for hours, finally gave up and tried a different motherboard. That's when I discovered the CPU was dead.

Again, I can't think of the last time I had an Intel CPU die on me when just running stock. The 2700X wasn't overclocked, and it had just booted, but somehow saving and exiting the BIOS glitched out I guess. RIP 2700X.
 

Deicidium369

Upstanding
Mar 4, 2020
388
59
290
5
Bold statement on the CPU side. I'd wait to see if Willow Cove's IPC gains can offset its likely significant clock frequency deficit from getting back-ported from 10nm+++ (10nm++ is the process Intel had hoped to have for Ice Lake / Sunny Cove years ago, Willow Cove would be one step after that) to 14nm++++++ first!
Well I assume the Willow Cove IPC gains will be at last as good as the Sunny Coves. The +++ thing makes you look foolish and is based on nothing. I don't care what Intel planned on having years ago - just like AMD planned on having competitive products that would gain a respectable amount of market share -which they haven't. I do not expect the frequency to be in the 5GHz+ range...

As far as back ported - what exactly about that is negative? You think it's somehow less than? Take your AMD wagon and go home.

Will be Rocket Lake and most likely NVlink dual Ampere. You buy what you want and with less ++++ since you think that is a negative.
 

spongiemaster

Upstanding
Dec 12, 2019
304
116
360
0
Well I assume the Willow Cove IPC gains will be at last as good as the Sunny Coves.
Based on what? Single threaded performance improvement is not a goal for Willow Cove by Intel's own admission. Probably not going to see much more than about 5% over Sunny Cove. Golden Cove is next architecture targeting IPC gains.

 

ASK THE COMMUNITY