AMD's Robert Hallock has confirmed that the upcoming AMD Ryzen 4000 Zen 3 processors will be compatible with Socket AM4 as long as the motherboard features an X570 or B550 chipset. Zen 3 will not support older chipsets owing to lower space on the EEPROM of these motherboards. Future prospects of Socket AM4 depend on the industry's I/O innovation.
[...] In a blog post, Hallock confirmed that current AMD X570 and B550 chipset motherboards will support Zen 3 processors after a BIOS update. However, Zen 3 processors will not be compatible with any chipset prior to X570 or B550. This means end of the road for all those who have X470, B450 and below chipset boards. Hallock says that this decision had to be taken as due to BIOS capacity limitations on older platforms.
We've seen AMD taking a similar stance with Zen 2 as well by removing drop-in support for motherboards that have just a 16 MB EEPROM. X570 motherboards have a 32 MB EEPROM thereby enabling larger a AGESA[*] codebase to be comfortably accommodated.
Wikipedia explains that AGESA:
AMD Generic Encapsulated Software Architecture (AGESA), is a procedure library developed by Advanced Micro Devices (AMD), used to perform the Platform Initialization (PI) on mainboards using their AMD64 architecture. As part of the BIOS of such mainboards, AGESA is responsible for the initialization of the processor cores, memory, and the HyperTransport controller.
See also: AMD axes Zen 3 support on 400-series motherboards: Is AMD pulling an Intel?
AMD will use the AM4 socket through its 'Zen 3' CPUs, but it will drop older chipset support
B450 and X470 chipsets won't support AMD Ryzen 4000 processors
Hardware Unboxed: No AMD Zen 3 Support on 400 and 300 Series Motherboards
AMD Zen 3 Based Ryzen 4000 'Vermeer' Desktop CPUs Will Be Compatible With Existing AM4 (X570, X470, B550, B450) Motherboards, Confirmed By XMG (from April 16, fake news or specific to the motherboard manufacturer?)
(Score: -1, Offtopic) by Anonymous Coward on Tuesday May 12 2020, @11:41PM (6 children)
I want to use your chip on my Apple II motherboard.
(Score: 2) by Kitsune008 on Wednesday May 13 2020, @02:25AM
Fsck you, Zombie Jobs.
Just take your past glory, and shove it where the sun don't shine. At this point in time, I don't think your turtleneck will impede this operation, despite your head being up your ass.
"PPC, or nothing!" should be your mantra, you traitor, you heretic. WTF?!?!? AMD?
Ghahhh! I hope you are strangled by that turtleneck. You are a total disgrace and complete disappointment to the Reality Distortion Field, and it's High Priests.
The only appeasement you can offer at this point is to commit seppuku. </sarcasm>
BTW, why do you want to use an AMD chip in your Apple II?
(Score: 2) by richtopia on Wednesday May 13 2020, @04:17AM (3 children)
In all seriousness, I'm really frustrated by this. I went to AMD in part because of their good history of upgradability. But even today, the B550 chipset is not available to consumers! I specifically bought a MAX series board from MSI which has the 32MB of EEPROM in an effort to make a transition to Zen 3 easier.
Maybe the Ryzen 9 CPUs will drop in price with the Zen 3 which will provide an upgrade path. Still, I'm disappointed.
(Score: 2) by takyon on Wednesday May 13 2020, @04:50AM (2 children)
There was some speculation that chips like the 3950X will take even longer to drop in price since Zen 2 is (apparently) the final upgrade for many motherboards.
Meanwhile, if you wanted to build a new Zen 3 system, you might be smacking yourself 12-18 months later when Zen 4 arrives on totally incompatible socket AM5, even if that involves buying expensive new DDR5 memory.
It does seem like your motherboard and others with 32 MB EEPROM could still get support for Zen 3, despite AMD's blog post. We'll just have to wait and see how bad this backlash gets.
[SIG] 10/28/2017: Soylent Upgrade v14 [soylentnews.org]
(Score: 0) by Anonymous Coward on Thursday May 14 2020, @12:34AM (1 child)
There is a silver lining; current Zen 2 chips are massive overkill for most applications. Gaming, compute, business, and CAD/CAM systems will age well over the next five years or so. By the time Zen 2 feels "slow" we'll be on AM6/DDR6 (and Intel will still be at 14nm+++++++++++++++++ of course).
(Score: 2) by takyon on Thursday May 14 2020, @05:49AM
Zen 2's slight gaming deficiency is related to latency (memory, cross-CCX) and clock speed (AMD is currently ahead in IPC). They will get a chance to address this and pull ahead of Intel with Zen 3 (by single digit FPS!!!!1). If Zen 3 can actually consistently beat the last Intel 14nm++++++++ CPUs in gaming, that might get a few more people to jump on, even if it is the end of the socket.
Gaming core counts are going to be pushed up to 8, and arguably 16+ [soylentnews.org] (could be hype), by the next-gen consoles. So the 4-6 core Zen 2 chips could suffer a bit in 2-5 years, even if they are mostly excellent and overkill.
Intel is going to aggressively push towards "7nm" by around 2022, which should be the equivalent of at least TSMC's "5nm". So the process node advantage AMD has enjoyed with Zen 2/3 will not be around forever.
AM5 should include at least Zen 4 and Zen 5 (both confirmed to exist), and likely Zen 6. The big transition will be to monolithic 3D chips, which could improve performance by orders of magnitude. That might happen in the late 2020s on a new socket.
[SIG] 10/28/2017: Soylent Upgrade v14 [soylentnews.org]
(Score: 2) by Bot on Wednesday May 13 2020, @12:24PM
We don't want to run on apple II, mr. 50c army guy. We want to speculate on whether AMD is capitalizing on its tech supremacy by intentionally removing compatibility or whether there are unavoidable technical issues. From a theoretical POV it would have been better to build some failsafe mode for the CPU to fit older boards at reduced performance, which would make people upgrade eventually all the same.
Anyway your post was frist, with a troll title and a good defense of AMD, so, well played.
Account abandoned.
(Score: 3, Insightful) by RamiK on Tuesday May 12 2020, @11:50PM (11 children)
I don't know what's worse, them straight face lying over obvious planned obsoleteness or them actually telling the truth and their "platform" really reached the point where it takes the space of fully functional 4 openwrt images to get the system to a the boot menu...
compiling...
(Score: 1, Funny) by Anonymous Coward on Tuesday May 12 2020, @11:55PM (1 child)
It's written in LISP you see...
(Score: 3, Informative) by RamiK on Wednesday May 13 2020, @02:16AM
As much as I enjoy LISP jokes, LibreCMC comes with two to three different LISP and LUA interpreters, a linux kernel, userland, package manager and plenty of networking daemons all rolled up into a fully functional router image under 4MB (= 32Mb): https://librecmc.org/librecmc/downloads/snapshots/v1.5.1-core/targets/ar71xx/generic/ [librecmc.org]
compiling...
(Score: 3, Informative) by shortscreen on Wednesday May 13 2020, @01:10AM (5 children)
I haven't kept up with the latest PC hardware but I suspect they are talking 16/32 megabits (not bytes) here. How much of that is taken up by UEFI crap?
The AGESA 3.x module used on socket AM2/AM3 boards is only 35KB uncompressed (and is stored compressed inside the ROM) but it mainly does CPU identification, not configuration of bus/memory which is handled by separate modules. If the entire ROM is now 4x as big, I find it hard to imagine that all of that is taken up by anything to do with CPU initialization.
Oh well, AMD was big on maintaining socket compatibility in the past, with some of their Athlon 64 chipsets still being recycled on AM3+ boards many years later, and new CPUs being able to run in years-old boards. Now that they are pushing the performance envelope again maybe they don't care so much.
(Score: 4, Informative) by RamiK on Wednesday May 13 2020, @02:02AM (4 children)
No they're talking about 32MB not Mb. It's already common place in some high-end desktop models with 128MB (not Mb) seen in SBCs, routers and servers.
Ironically, MSI's B450M Mortar Max has an 32MB ( = 256Mb).
compiling...
(Score: 2) by shortscreen on Wednesday May 13 2020, @03:14AM (2 children)
I see. Bloaters have been busy these past years.
(Score: 3, Interesting) by RamiK on Wednesday May 13 2020, @04:08AM (1 child)
Well, to be fair, as nodes shrink the EEPROM factories simply aren't making smaller capacity storage anymore... And it's not all bad. BIOS and UEFI drivers were always very sketchy so when projects like LinuxBoot [linuxboot.org] come along taking advantage of the extra space to get the same functionality but with better tested code, it's not without its merits... But yeah. The bloaters pretty much run the place nowadays.
compiling...
(Score: 0) by Anonymous Coward on Wednesday May 13 2020, @08:10AM
Not to mention that the EEPROM stores its own operating system that operates between the hardware and the user's OS. Sure, it can get out of the way so the overhead isn't too bad, but it is still there in the EEPROM. Plus, many EEPROM updates use an A/B image or "seamless" system now. So you have to have room for two complete boot images now. And that is in addition to the space needed for read and write wear leveling.
(Score: 2) by takyon on Wednesday May 13 2020, @04:01AM
Yeah, I'm guessing AMD will not actually care if vendors can update older motherboards to work with Zen 3 (see XMG link). They just want to curb enthusiasm early before there are a flood of compatibility complaints.
[SIG] 10/28/2017: Soylent Upgrade v14 [soylentnews.org]
(Score: 0) by Anonymous Coward on Wednesday May 13 2020, @03:14PM (2 children)
AMD platforms last practically forever. I am pretty surprised they are sticking with AM4 for the next CPU revision, it's already pretty much the worst thing about Ryzen. I guess they are not going to do a new socket until DDR5 memory forces it. Intel releases new incompatible sockets with almost every CPU, even if the new socket doesn't offer any new functionality.
AMD really, really needs an enthusiast/workstation socket comparable to LGA 2066. AMD's high end processors are great, but they're breathing through a straw.
(Score: 2) by takyon on Wednesday May 13 2020, @03:30PM
Zen 2 (Ryzen 3000 desktop) = AM4, 16 cores, DDR4
Zen 3 (Ryzen 4000 desktop) = AM4, 16 cores, DDR4
Zen 4 (Ryzen 5000 desktop) = AM5, 24 cores? DDR5
PCIe 5.0 was rumored for Zen 4, but they will probably kick that can down the road for desktop users.
More memory channels? DDR5 supports 2 memory channels per module [rambus.com].
[SIG] 10/28/2017: Soylent Upgrade v14 [soylentnews.org]
(Score: 0) by Anonymous Coward on Thursday May 14 2020, @03:24AM
200 usd for a 1950x 12 core, with 40 PCIe channels, and another 250-400 for the motherboard.
You can pay more for other chips, or get the newer incompatible TRx board for Zen... 3? If you need even bigger chips.
Plus there are always single processor Epyc boards if you really want their biggest chips. Even if not officially supported I'm sure they can be overclocked just like the lower end models.
(Score: 1, Informative) by Anonymous Coward on Wednesday May 13 2020, @12:08AM
How about the "support" for the AMD Platform Security Processor?
https://libreboot.org/amd-libre.html [libreboot.org]
(Score: 3, Interesting) by zion-fueled on Wednesday May 13 2020, @12:28PM
All you have to do is edit the code and remove older processor support. Make the board support only Zen3 and the code will shrink. I've had to remove intel ME or edit bios on almost all computers I buy now. Latest one I added menus to be able to do p-state overclock.
(Score: 0) by Anonymous Coward on Wednesday May 13 2020, @03:06PM (1 child)
How many different pwnware from how many different government agencies do you need to bundle? Use the same pwnware for all and just bundle their public keys.
(Score: 2) by takyon on Wednesday May 13 2020, @03:31PM
1 pwn per thread, 1024 cores, 2048 threads.
[SIG] 10/28/2017: Soylent Upgrade v14 [soylentnews.org]