Friday, July 19th 2019
AMD AGESA 1.0.0.3ABA Buggy, Company Pulls it from Motherboard Vendors
The latest version of AGESA ComboAM4 microcode that enables 3rd generation Ryzen support on AMD 400-series chipset motherboards has been deemed buggy and pulled from motherboard vendors. AGESA ComboAM4 1.0.0.3ABA (not to be confused with 1.0.0.3AB that's being widely distributed), was originally released to fix an application crash noticed with "Destiny 2." The microcode inadvertantly destabilizes PCI-Express on motherboards, with users of ASUS motherboards complaining of stability issues with the latest BIOS updates that include 1.0.0.3ABA.
Peter "Shamino" Tan from ASUS commented that the company was under a tight schedule to push 1.0.0.3ABA out as BIOS updates, and didn't have the time to properly validate it. "We just got told to pull (was undergoing validation prior) 1003 ABA version," he said, adding the root cause of the problem being "that PCIE speed of BXB-C downgraded from gen4 to gen2,..." He comments "so its not surprising that bugs emerge since the source has hidden bugs that only gets unraveled with thorough testing. combine that with trying to get firmwares out in a tight time frame, kinda damn if you do (release firmware quickly) and damn if you dont (dont release firmware quickly) situation." It's interesting to note that in their BIOS update change-logs, quite a few motherboard vendors omit the full version string of AGESA. You may encounter ComboAM4 1.0.0.3AB being referred to simply as "AGESA ComboAM4 1.0.0.3."
Peter "Shamino" Tan from ASUS commented that the company was under a tight schedule to push 1.0.0.3ABA out as BIOS updates, and didn't have the time to properly validate it. "We just got told to pull (was undergoing validation prior) 1003 ABA version," he said, adding the root cause of the problem being "that PCIE speed of BXB-C downgraded from gen4 to gen2,..." He comments "so its not surprising that bugs emerge since the source has hidden bugs that only gets unraveled with thorough testing. combine that with trying to get firmwares out in a tight time frame, kinda damn if you do (release firmware quickly) and damn if you dont (dont release firmware quickly) situation." It's interesting to note that in their BIOS update change-logs, quite a few motherboard vendors omit the full version string of AGESA. You may encounter ComboAM4 1.0.0.3AB being referred to simply as "AGESA ComboAM4 1.0.0.3."
35 Comments on AMD AGESA 1.0.0.3ABA Buggy, Company Pulls it from Motherboard Vendors
But as it is AMD and they present objectively more value, no one complains as much. They are easier to defend. You know how it works.
I already said on another thread that we are having a nightmare on the store with this release. Most B450 motherboards have huge problems. They either don´t boot and you have to clear CMOS, when they boot you have a 65w TDP chip hitting 1,5v just by watching an youtube video and you need to manually change 3 or 4 voltage values to calm things down, just to see your CPU not hitting even close to the turbo boost, with 1 core only!!!
Then you can imagine the amount of customers that are complaining and returning their chips/mobos and the amount of work and troubleshoot the staff has, because you have a "warranty" that the new shiny AMD CPUs work on every motherboard chipset... Total nightmare.
This was not an auto update, or forced update. Very few people ever installed this update, and had issues from it.
It was recalled, before it caused major issues.
AMD are smashing out updates incredibly fast to iron out bugs on a newly launched platform, if you arent fixing a bug you simply have no need to be on these beta BIOSes in the first place.
While Zen2 is not the same architecture as Zen(+).
Ryzen 3000 can run on 2 generations-old x370 / B350 MBs which is optimized for a different architecture.
Can a 9900k run on a z270 MB?
Someone did that with BIOS / physical modifications, you can check how buggy it is.
For example, let me get back to Intel's FDIV bug. Floating point division is one of the more expensive operations you can do at hardware level. So instead of figuring out everything from scratch at every computation, the CPU does what you did in school: it "learns" the results of the most common computation by heart. Just like you learned the multiplication table, it too knows a table of common results, but for division. The FDIV bug meant this table was incorrectly written in silicon, thus some operations would return wrong results. Because this was all written in silicon, Intel had to physically replace affected CPUs. Nowadays, this is written in microcode and is fixable by simply giving the CPU an updated table to work with.
I don't know how that works for RDRAND, because I have seen no technical documentation describing the bug. Just its symptoms.
:shadedshu::shadedshu: