Monday, July 8th 2019

BIOS ROM Size Limitations Almost Derail AMD's Zen2 Backwards Compatibility Promise

AMD succeeded in delivering on its backwards-compatibility promise for the 3rd generation Ryzen processors on motherboards based on AMD 300-series and 400-series chipsets. This promise was very close to being derailed suggests a community thread on MSI forums. According to MSI representatives active on the forum, the capacity of the SPI flash EEPROM chip that stores the motherboard UEFI firmware is woefully limited to cram in the AGESA ComboAM4 1.0.0.3a microcode on many of its motherboards.

The company had to make several changes to its UEFI BIOS package that's currently being circulated as a "beta," to accommodate support for 3rd generation Ryzen processors along with AGESA ComboAM4 1.0.0.3a. First, it had to kick out support for A-series and Athlon processors based on the 28 nm "Bristol Ridge" silicon. Second, it had to [and this is a big one], kick the RAID module, breaking SATA RAID on many of its motherboards. Third, it had to replace its feature-rich Click BIOS 5 setup program with a barebones "GSE Lite" Click BIOS program, which lacks many of the features of the original program, and comes with a dull, low-resolution UI. This program still includes some essential MSI-exclusive features such as A-XMP (which translates Intel XMP profiles to AMD-compatible settings), Smart Fan, and M-Flash.
The scary part? Many other motherboard brands appear to be using 16-megabyte EEPROMs on their older socket AM4 motherboards. These companies are bound to run into similar ROM capacity issues unless they keep their UEFI setup programs lightweight. Motherboards based on the latest X570 chipset feature 32-megabyte EEPROMs. The AMD X570 chipset lacks support for not just "Bristol Ridge," but also first-generation Ryzen "Summit Ridge" and "Raven Ridge" processors.

We recommend that unless you literally possess a 3rd generation Ryzen processor, do not update the BIOS of your older socket AM4 motherboard. You may risk losing features and break your RAID volumes. Find out the latest version of BIOS that has the classic AGESA PinnaclePI 1.0.0.6 microcode, and use that instead.
Source: MSI
Add your own comment

88 Comments on BIOS ROM Size Limitations Almost Derail AMD's Zen2 Backwards Compatibility Promise

#26
spectatorx
tony359To be honest I’ve always thought that all that graphic on the BIOS was not necessary. I don’t mind a basic one as long as it has all the features I need. I don’t use RAID on my Ryzen but that would be a bummer. The MSI page doesn’t mention RAID missing though. Maybe it’s just the beta?
It does:
* If you already set RAID for your system, please don`t update these beta BIOS as they are not ready for raid function.
The way this sentence is written i would think they removed RAID support in beta bios but they MAY restore it in final builds.
Posted on Reply
#27
Mamya3084
MSI hasn't even added support for my B450 motherboard. Was going to update the HTC PC with a 3700.
Wonder if this will be an issue come time of the threadripper 3 release....
Posted on Reply
#28
GoldenX
I have an Olivetti Pentium MMX pc, with mouse support on BIOS, and that's not a 16MB BIOS chip.
I would prefer a de-bloated blue with white text UEFI but with full support for all AM4 chips than some stupid flashy UEFI.
Posted on Reply
#29
Aerpoweron
Gigabyte also had to throw out Bristol Ridge support on the B350 and X370 i own.

X370 Gaming K5

AB350M D3H

I only have one Bristol Ridge, and it is an interesting CPU. But broken here and there. It is ok that they kick out support, only a handful of people used this cpu. And i can still run it on my Asus X370 if i like :)
Posted on Reply
#30
Bitgod
I had planned on using a MSI B450 Gaming Pro Carbon AC with a Zen 2 since that was a mobo that was largely favored the past few weeks for a cheap setup. Now I'm on the fence on using it or returning it and getting a low-end X570. Ugh. On the one hand, the 570 would "just work". On the other, I really don't want to go back to a time of having a small fan on the mobo.
Posted on Reply
#31
turbogear
This is really sad to see backwards compatibility limited and support for some CPUs being removed due to bios chip size. :(

I checked on my ASUS C7H. Luckily my board has 32MB flash size.

Though to mention that the new bios from ASUS C7H is bugged. On 2406, the USB support is not working properly. None of the mice I have works inside the bios.
Also have trouble with DOCP support for my 16GB G.Skill Trident Z RGB DDR4-3200 CL14.
I went back to 2304.
Posted on Reply
#32
IceShroom
spectatorxBasically this means two things:
1. UEFI is bloated huge crap.
2. Motherboards manufacturers cheap on bios modules and use 16MB instead of 32MB.

I'm lucky and my x470 motherboard has 256Mb module which is 32MB, i just checked it right after reading this news. I've never been a fan of fancy GUIs in UEFI and i would assume these take a lot of space. I totally would prefer classic look with added mouse support for these who do not know how to use keyboard to navigate in bios.
Typecal mobo manufacture. Instead spending money on biger UEFI chip, they spend that money on RGB. Also mobo manufacturer need to clean up those flashy design and animation from URFI.
Posted on Reply
#33
londiste
Considering how widespread this is and that most 400-series motherboards come with a 32 MB chip this looks more like AMD not communicating the required size properly at first.
Posted on Reply
#34
btarunr
Editor & Senior Moderator
Re: those talking about price difference:

According to DigiKey, an 8-pin 128Mbit (16 MB) SPI flash ROM chip of the kind used on motherboards starts at $1.31 a piece when bought in bulk reels. A 256Mbit (32 MB) chip in the same form-factor costs $2.04 a piece. Motherboard makers have the bargaining power to bring those prices down even further. Even if not, they're literally 70 cents apart. Motherboard designers could have cut out some RGB crap to make room for a 256Mb chip. They probably wanted forced-obsolescence by telling AMD "it can't be done" when the time came, but failed at that when AMD replied "it can be done, just delete support for older processors and make your setup program smaller."
Posted on Reply
#35
silentbogo
I think they are trying to make a viable excuse to sell more boards. First of all, cheaping-out less than $1 on SPI flash is their fault, not ours or AMDs. Second of all, I very much doubt that fitting a 3200 byte section is such a big problem - AGESA microcode is already peppered over 3 different places in the firmware with latest updates (same for pretty much all manufacturers, cause copy-pasting requires no effort).
Posted on Reply
#36
bug
efikkanI'm just surprised they didn't foresee this, I thought this platform was planned to last for several cycles. It shouldn't be too hard to estimate the space required for firmware for several new CPU lineups.

Now, what about the CPUs coming in the next two-three years?
The X570 boards only support Zen+ and Zen2. That should answer your question.

Now, if instead of UEFI anyone had the the wisdom to write a BIOS that sticks to its job... www.coreboot.org/
Posted on Reply
#37
TheLostSwede
News Editor
spectatorxBasically this means two things:
1. UEFI is bloated huge crap.
2. Motherboards manufacturers cheap on bios modules and use 16MB instead of 32MB.

I'm lucky and my x470 motherboard has 256Mb module which is 32MB, i just checked it right after reading this news. I've never been a fan of fancy GUIs in UEFI and i would assume these take a lot of space. I totally would prefer classic look with added mouse support for these who do not know how to use keyboard to navigate in bios.
1. You forget that we've gone from a VERY basic non graphical UI, to a clicky, flashy UI with UEFI, that takes up space, even more so when they went from 640x480 to 1080p or higher resolution in the UI for the UEFI. This takes up space. Most of the clicky, flashy parts aren't needed though, but they make it look nice and it makes it easier for people to do things in the UEFI, or so the board makers seem to think. I never use the mouse in the UEFI, as I find it harder.

2. Well, there's a cost difference, so low cost boards will use low cost parts, pretty logical, no?
bugThe X570 boards only support Zen+ and Zen2. That should answer your question.

Now, if instead of UEFI anyone had the the wisdom to write a BIOS that sticks to its job... www.coreboot.org/
We're way past the point of no return for this, UEFI is here to stay and it's coming to embedded ARM devices too.
This is what has been decided by higher powers and as a user, we have no choice.
UEFI is supposed to be more secure, smarter and better, but I'm not sure this is really true...
btarunrRe: those talking about price difference:

According to DigiKey, an 8-pin 128Mbit (16 MB) SPI flash ROM chip of the kind used on motherboards starts at $1.31 a piece when bought in bulk reels. A 256Mbit (32 MB) chip in the same form-factor costs $2.04 a piece. Motherboard makers have the bargaining power to bring those prices down even further. Even if not, they're literally 70 cents apart. Motherboard designers could have cut out some RGB crap to make room for a 256Mb chip. They probably wanted forced-obsolescence by telling AMD "it can't be done" when the time came, but failed at that when AMD replied "it can be done, just delete support for older processors and make your setup program smaller."
Yes, but ramp this up to a million products and that's a million bucks someone could've pocketed...
This is what people in general forget when they talk component costs. This is also why companies try to save a cent on things, as it's all about the quantity, not the individual component cost.
You're absolutely right about the RGB crap though, as removing all the useless LEDs would most likely have saved even more money. But the LEDs are there for marketing purposes, no? More important that having a solid product that will last...
GoldenXI have an Olivetti Pentium MMX pc, with mouse support on BIOS, and that's not a 16MB BIOS chip.
I would prefer a de-bloated blue with white text UEFI but with full support for all AM4 chips than some stupid flashy UEFI.
Urgh, Olivetti made the worst PCs...
The mouse BIOS was not unique to them though, it was actually found on a lot of boards.
What you miss though, is that it only worked with PS/2 and maybe serial mice, not USB mice.
It's not problem getting the blue and white UEFI though, it's actually common on a lot of non consumer devices that use UEFI, as well as most notebooks.

ami.com/en/products/bios-uefi-firmware/aptio-4/
TotallyWhy not have a set of good BIOSes instead of one that is compromised? 1 for 1000/2000 CPUs and another for 3000 CPUs?
It wouldn't quite work, as if you then want to go to the 3000-series, how would you go around updating the UEFI? Your older CPU would instantly not work with the board after you updated, which might cause some problems. But otherwise it makes sort of sense.
Posted on Reply
#38
tony359
TotallyWhy not have a set of good BIOSes instead of one that is compromised? 1 for 1000/2000 CPUs and another for 3000 CPUs?
It'd be great but imagine then having to develop, update, test and release TWO sets of BIOS for each MB - and remember that those are free, we "paid" for them when we purchased the mobo.
Posted on Reply
#39
londiste
Two sets of BIOSes will require some type of dual-BIOS switching or Flashback functionalities to be reasonably useful. Getting a different supported CPU to get the BIOS upgraded on the motherboard you just purchased to support a CPU you also just purchased is annoying enough already, making this standard operating procedure is not a good idea.
Posted on Reply
#40
bug
londisteTwo sets of BIOSes will require some type of dual-BIOS switching or Flashback functionalities to be reasonably useful. Getting a different supported CPU to get the BIOS upgraded on the motherboard you just purchased to support a CPU you also just purchased is annoying enough already, making this standard operating procedure is not a good idea.
Worse, two sets of BIOSes will open you up to badmouthing because having to trade one feature for another. Not to mention lawyers will have a field day dragging you to court.
Posted on Reply
#41
efikkan
SteevoI had a windows 95 machine years ago with mouse support in the BIOS. It's too many fancy looking backgrounds and the tables for control and support.
You mean winbios?
It doesn't take many lines of assembly to implement basic mouse support.
But the fancy BIOS GUIs is beside the point, that's not the problem. The size of that bloatware should be known to the motherboard maker, the problem is when adding support for new CPUs needing new firmware, which grows every time there is a new lineup or even possibly if they expand the current lineup.
bugThe X570 boards only support Zen+ and Zen2. That should answer your question.

Now, if instead of UEFI anyone had the the wisdom to write a BIOS that sticks to its job... www.coreboot.org/
Sorry, I must have misunderstood. I thought this platform was going to last longer.

I do like Coreboot, but I don't think it will gain any adoption. UEFI is a bloated piece of crap, but that's not the worst part of it; it's also a security nightmare. It's only a matter of time before someone figures out a way to break into it remotely.
Posted on Reply
#42
bug
efikkanSorry, I must have misunderstood. I thought this platform was going to last longer.
It's why I always took that with a grain of salt: despite AMD's best intentions, they simply cannot know the requirements of future CPUs. And neutering future CPUs for the sake of backwards compatibility would be even worse. So something's got to give.
efikkanI do like Coreboot, but I don't think it will gain any adoption.
I'm not hold my breath either. I was just pointing out there are still sane people at work, somewhere.
efikkanUEFI is a bloated piece of crap, but that's not the worst part of it; it's also a security nightmare. It's only a matter of time before someone figures out a way to break into it remotely.
Honestly, I'd be really surprised if interested parties haven't already ;)
Posted on Reply
#43
btarunr
Editor & Senior Moderator
TheLostSwedeYes, but ramp this up to a million products and that's a million bucks someone could've pocketed...
This is what people in general forget when they talk component costs. This is also why companies try to save a cent on things, as it's all about the quantity, not the individual component cost.
You're absolutely right about the RGB crap though, as removing all the useless LEDs would most likely have saved even more money. But the LEDs are there for marketing purposes, no? More important that having a solid product that will last...
Yes, they chose RGB to market over spending $0.7 more on a better SPI chip. Guess what? Enthusiasts will now prefer those brands that chose 256Mb chips way back when designing their 300-series/400-series boards (eg: ASUS, ASRock), over brands that cheaped out (eg: MSI, GbT). Also, those buying a Zen2 processor with a brand new X470 chipset motherboard will avoid boards with 128Mb ROMs like cancer. I'm sure some based nerd on Reddit will compile a list of motherboards with ROM sizes to help others out.
Posted on Reply
#44
tony359
btarunrI'm sure some based nerd on Reddit will compile a list of motherboards with ROM sizes to help others out.
Looking forward to that indeed.
Posted on Reply
#45
hojnikb
This is odd. Looking at my gigabyte B450M-DS3H, it has 16MB bios chip (according to bios bin file anyway) but seems to retain all the features even after updating to zen2 support.

Unless gigabyte forgot to mention that in their description (haven't tried the new bios yet).
Posted on Reply
#46
spectatorx
btarunrYes, they chose RGB to market over spending $0.7 more on a better SPI chip. Guess what? Enthusiasts will now prefer those brands that chose 256Mb chips way back when designing their 300-series/400-series boards (eg: ASUS, ASRock), over brands that cheaped out (eg: MSI, GbT). Also, those buying a Zen2 processor with a brand new X470 chipset motherboard will avoid boards with 128Mb ROMs like cancer. I'm sure some based nerd on Reddit will compile a list of motherboards with ROM sizes to help others out.
Well, as already one user here mentioned his asrock x370 taichi which is asrock's high end motherboard of 3xx line up has 16MB chips too and i did check and can confirm, specification lists 16MB module. I'm lucky and x470 taichi has 32MB.
Posted on Reply
#47
tony359
They seem to have forgotten to mention on BIOS 31 for the GA-AX370-Gaming K5 so I'd be tempted to think they forgot for yours as well - obviously I hope I am wrong!

Posted on Reply
#48
londiste
spectatorxWell, as already one user here mentioned his asrock x370 taichi which is asrock's high end motherboard of 3xx line up has 16MB chips too and i did check and can confirm, specification lists 16MB module. I'm lucky and x470 taichi has 32MB.
It is not lucky. BIOS flash chip being small was a big enough deal already when Zen+ came out, there was definitely a very clear push in getting 32MB chips to 400-series motherboards.
Posted on Reply
#49
Midland Dog
dont mean to literally crap on the entire thesis of this, but arent bios chips easily swapable?
Posted on Reply
#50
EarthDog
Midland Dogdont mean to literally crap on the entire thesis of this, but arent bios chips easily swapable?
On some boards you can drop new ones in, sure. That isn't close to a majority though. Most are soldered on the board.
Posted on Reply
Add your own comment
Nov 21st, 2024 11:52 EST change timezone

New Forum Posts

Popular Reviews

Controversial News Posts