Friday, March 12th 2021

AMD Fixes Intermittent USB Connectivity Issues on 500 Series Chipsets, BIOS Update Arrives in April

AMD has four weeks ago acknowledged that there was a problem with 500 series motherboard chipsets. The problem has occurred with a few chipset functions like USB connectivity, USB 2.0 audio crackling (e.g. DAC/AMP combos), and USB/PCIe Gen 4 exclusion. To fix these problems, consumers were forced to either put up with problems or lower the PCIe standard from Gen 4 to Gen 3 and switch USB protocol revision from 3.0 to 2.0. This of course wasn't the ideal solution, especially for bandwidth-heavy applications. Users have submitted many reports to AMD, and the company appears to have found a root cause of these issues. AMD has published a Reddit thread, that reports that the company found a solution to the problem and that we are going to see a fix for it in a form in AGESA BIOS update.
AMD Reddit
AMD has prepared AGESA 1.2.0.2 to deploy this update, and we plan to distribute 1.2.0.2 to our motherboard partners for integration in about a week. Customers can expect downloadable BIOSes containing AGESA 1.2.0.2 to begin with beta updates in early April. The exact update schedule for your system will depend on the test and implementation schedule for your vendor and specific motherboard model. If you continue to experience intermittent USB connectivity issues after updating your system to AGESA 1.2.0.2, we encourage you to download the standalone AMD Bug Report Tool and open a ticket with AMD Customer Support.
Source: AMD Subreddit
Add your own comment

107 Comments on AMD Fixes Intermittent USB Connectivity Issues on 500 Series Chipsets, BIOS Update Arrives in April

#76
qubit
Overclocked quantum bit
This would really piss me off. I'm currently running Old Reliable here in the form of a 2700K and Z68 chipset.
Posted on Reply
#77
ssdpro
TheLostSwede
So what is the virtual pitch fork squad going to bash AMD for now?

Obviously an annoying issue for those that experienced it, but it was clearly not a hardware flaw as some seemed to suggest.
Exactly. AMD is new to computing and USB is complicated. USB protocols are incredibly demanding and it takes time for engineers to find a way to allocate enough computing power from the cripple AMD cores.
Posted on Reply
#78
JMccovery
TheLostSwede
It feels like a lot of people like to throw some crap out on the internet because they think they know something.
I spent over a decade of my life as a tech journalist and I started back in the days when you had 20-odd motherboard manufactures to choose from and 6-7 chipset vendors and at least four x86 CPU manufacturers.
I remember testing the first DDR 133MHz capable chipset from VIA with a stick of some crap RAM they sent along. I could never get that stick of RAM working and they could never figure out why.
Oh man, I remember those days... Trying my hardest to get two and three sticks of PC-133 SDRAM running in an Athlon TBird + Biostar M7VKD v1.0 system. Would get one stick running fine, then two; tried for three (1.5GB was an amazing amount back then), wouldn't boot, set to 100MHz, still wouldn't boot, cleared CMOS, could only get a single stick to work at 133MHz. Somehow killed the board, v2.0 replacement ran all three at 133MHz with no sweat.
Posted on Reply
#79
windwhirl
ssdpro
Exactly. AMD is new to computing and USB is complicated. USB protocols are incredibly demanding and it takes time for engineers to find a way to allocate enough computing power from the cripple AMD cores.
I don't know if you're being sarcastic or serious.
Posted on Reply
#80
Vanny
I heard this issue surfaces the most when you try to use a VR headset with a 500 chipset AMD board...

Well 3900X here with a ROG Strix B550-F, no issues when using my Rift S. All my USB ports are filled up. AGESA 1.2.0.0, latest bios. In fact, no issues at all ever since I built this thing months ago as my first AMD rig.

Not to mention my 3900X took a 4x8GB 3600 16-16-16-36 kit like a champ. Was even able to tighten all timings and the frequency some before I ran into IF limitations for my silicon. But let's bash ONE company and forget that both of them make mistakes.
Posted on Reply
#81
TheLostSwede
ssdpro
Exactly. AMD is new to computing and USB is complicated. USB protocols are incredibly demanding and it takes time for engineers to find a way to allocate enough computing power from the cripple AMD cores.
So funny :rolleyes:

It's not the first time AMD has had issues with USB, their own USB 3.0 implementation was crap and as the only journalist that ever did a comparison between all the USB 3.0 controllers that were at the market at the time, I tried to ask AMD about it. Got in touch with their engineer (he was attending an AMD event in Taiwan), who pretty much told me to get lost...
JMccovery
Oh man, I remember those days... Trying my hardest to get two and three sticks of PC-133 SDRAM running in an Athlon TBird + Biostar M7VKD v1.0 system. Would get one stick running fine, then two; tried for three (1.5GB was an amazing amount back then), wouldn't boot, set to 100MHz, still wouldn't boot, cleared CMOS, could only get a single stick to work at 133MHz. Somehow killed the board, v2.0 replacement ran all three at 133MHz with no sweat.
There has been a ton of really, really crappy motherboards and motherboard makers over the years.
Worst one ever was PC Chips, as no-one else got close to selling as much crap as they did.
Worked in a computer shop at one point, 6/10 brand new boards were DOA. The owner bought them as people wanted to buy £200 computers from him...
Posted on Reply
#82
Zareek
TheLostSwede
AMD put out some terrible chipsets, the 750 chipset was a turd for example. I mean, not even MSI (or possibly Biostar, too long ago) wanted to put their name on the first board with the 750 chipset.
Then Asus finally brought one out and things improved a bit, but AMD would've been screwed back in the day if it wasn't for Nvidia and VIA and to a lesser extent ALi and SiS.
See, I didn't see the AMD 750 that way. MSI screwed it up and ASUS did it right. I had one friend with the MSI AMD750 (K7 Pro I think) board. It was so bad, it randomly wouldn't play nice with Creative sound cards or with soft-modems. I had another friend who had an ASUS AMD 750 board it was amazing, zero issues. Looking back maybe the ASUS board had the VIA southbridge. I still say a quality motherboard was and probably still is the biggest factor in AMD versus Intel platforms.
Posted on Reply
#83
JMccovery
TheLostSwede
There has been a ton of really, really crappy motherboards and motherboard makers over the years.
Worst one ever was PC Chips, as no-one else got close to selling as much crap as they did.
Worked in a computer shop at one point, 6/10 brand new boards were DOA. The owner bought them as people wanted to buy £200 computers from him...
Best thing I remember about PC Chips was a dirt cheap Socket 7 board that I bought at a local computer shop; it didn't have a keyed socket, so I managed to put a K6-2 350 in the wrong way. Magic Smoke was released.

Would've been highly upset, yet it was only a $50 CPU on a $30 motherboard.
Posted on Reply
#84
TheLostSwede
Zareek
See, I didn't see the AMD 750 that way. MSI screwed it up and ASUS did it right. I had one friend with the MSI AMD750 (K7 Pro I think) board. It was so bad, it randomly wouldn't play nice with Creative sound cards or with soft-modems. I had another friend who had an ASUS AMD 750 board it was amazing, zero issues. Looking back maybe the ASUS board had the VIA southbridge. I still say a quality motherboard was and probably still is the biggest factor in AMD versus Intel platforms.
The early boards were terrible, trust me, I tested enough systems based on that platform at the time. The ones with the VIA southbridge were better for sure, but only marginally so. The whole Slot-A platform was pretty wonky to be honest, things only improved when they moved to Socket-A.
JMccovery
Best thing I remember about PC Chips was a dirt cheap Socket 7 board that I bought at a local computer shop; it didn't have a keyed socket, so I managed to put a K6-2 350 in the wrong way. Magic Smoke was released.

Would've been highly upset, yet it was only a $50 CPU on a $30 motherboard.
They had all sorts of bizarre hardware. Remember we got some boards that didn't even come with suitable drivers on the CD that shipped with the boards.
They had these onboard CPUs that I have no idea who made for them, but they were x86 compatible and were paired with a terrible onboard graphics chip that sometimes worked, sometimes not.
The day they disappeared off the market was a good day.
Posted on Reply
#85
Patriot
I think these hiccups showing up later on is a result of AMDs io controllers being so flexible, vendors can have much wider configurations throughout their board skus.
With freedom comes, omg the endless testing required to support it.
chrcoluk
I wouldnt consider pcie 3.0, usb 2.0, gigabit lan ancient. Maybe you are a cutting edge type of guy/girl, but things dont become useless the moment something new comes along.

Imagine the grief AMD would got if this chipset didnt support the techs you mentioned, instead of people having a workaround, they would have just been SOL.

Many vendors still make devices using these tech's, people still have older monitors etc, one of the great things about PC's is been able to mix and match hardware.

Also if I am not mistaken a reason boards still include usb2 ports is for compatibility with booting, making sure keyboard/mouse works in bios etc.
Pcie gen 3 is the youngest at 11yrs old... the others are 20-22yrs old.
Those are ratification dates, Consumer use subtract 1-3 years, pretty ancient in the pc world.

I don't see the need to kill off gen3/usb2.0 as they have their uses. Gbit eth needs to be replaced with 2.5/5/10 plz.
It's time to make that economical.
Posted on Reply
#86
Raiden85
Fouquin
Yeah like nForce 680i. Man I remember when forums were filled with "My motherboard killed my RAID array/DRAM!" due to the FSB bug on those boards.

Or i820 and the MTH issue which led to a market wide recall of all the chips and boards using it.

Or i850s PCI write delay issue that limited bus bandwidth, or i850's RDRAM null cycle issue where it would prematurely terminate IIO to RDRAM and cause the whole system to lock up.

Or Northwood and the entire SNDS debacle. $400 CPUs falling like flies left and right.

Or how about Mobile 965 Express's IGPU issues.

Or H67/P67's Rev.B SATAII PLL defect?

Maybe want an asterisk on the "always" in your statement there. :)
Or we can go all the way back to the early Pentium with the Pentium FDIV bug that affected the FPU, I'd say Intel has had more serious issues than AMD over the years :)

en.wikipedia.org/wiki/Pentium_FDIV_bug
Patriot
I don't see the need to kill off gen3/usb2.0 as they have their uses. Gbit eth needs to be replaced with 2.5/5/10 plz.
It's time to make that economical.
Gbit ethernet has needed killing for years, absolutely ridiculous it's taken this long to start seeing 2.5Gb and up getting more popular at last. When you have onboard motherboard wifi connecting at 2.4Gbps, phones at 1.2Gbps and even 5G that can have some really fast speeds, seeing a pathetic 1Gbps wired connection on high end motherboards is just embarrassing.
Posted on Reply
#87
TheLostSwede
Patriot
I think these hiccups showing up later on is a result of AMDs io controllers being so flexible, vendors can have much wider configurations throughout their board skus.
With freedom comes, omg the endless testing required to support it.


Pcie gen 3 is the youngest at 11yrs old... the others are 20-22yrs old.
Those are ratification dates, Consumer use subtract 1-3 years, pretty ancient in the pc world.

I don't see the need to kill off gen3/usb2.0 as they have their uses. Gbit eth needs to be replaced with 2.5/5/10 plz.
It's time to make that economical.
2.5Gbps is already economical. You can get a dumb 5-port switch for less than $200, although that really need to be more like $50 for it to really take off. Most new mid-tier and up motherboards seem to have it as standard now, as the cost difference between a 2.5Gbps PHY and a 1Gbps PHY is no more than 50 cents.
The issue with 5 and 10Gbps is that the technology is quite different and so far, no-one has managed to make a proper consumer friendly chip for either standard. As long as we're looking at $25-30 for a controller that draws 5W+, it's never going to be a mainstream product.
Raiden85
Gbit ethernet has needed killing for years, absolutely ridiculous it's taken this long to start seeing 2.5Gb and up getting more popular at last. When you have onboard motherboard wifi connecting at 2.4Gbps, phones at 1.2Gbps and even 5G that can have some really fast speeds, seeing a pathetic 1Gbps wired connection on high end motherboards is just embarrassing.
I think you're overestimating the actual speed of wireless connection. All those speeds are mostly marketing. Yes, you might see a sync rate of that, but it doesn't mean you'll get anything near those speeds.
You'd need two of something like this thing, to get above Gigabit speeds and no client device has that many antennas. Also, it can't do Gigabit speed in terms of upload speeds.
www.smallnetbuilder.com/wireless/wireless-reviews/33226-wi-fi-6e-preview-with-asus-gt-axe11000-rog-rapture
Posted on Reply
#88
TheGuruStud
RJARRRPCGP
I remember it being like that in the T-bird era, with Soyo. Despite I was able to get a fix quickly, I wonder if that's why the Soyo motherboard division went under.
My first Athlon system: (T-bird 900 Mhz) It kept crashing and then it looked like it possibly was only randomly crashing when idle. The fix was to lower the Vcore, when at stock frequency. Ironically, running a load didn't noticeably make the crashes increase.
My Tbird system was rock solid for the years I used it. IIRC, I had a Tyan board and never updated bios. Those were the good ol days. Tbird, XP, and athlon 64. Intel blaster time after time, but it means nothing with criminal gorilla.
Posted on Reply
#89
TumbleGeorge
TheLostSwede
The issue with 5 and 10Gbps is that the technology is quite different and so far, no-one has managed to make a proper consumer friendly chip for either standard. As long as we're looking at $25-30 for a controller that draws 5W+, it's never going to be a mainstream product.
Do not swear by this assumption. In my country, so far only limited in the capital, from the end of 2020 began to offer Internet access in versions with 5 and 10Gb / s at very low prices, so apparently Internet providers obviously have different views on the issue, and obviously access to inside information on how prices will develop in the near future and the implementation of 10Gb / s Internet controllers in motherboards.
Posted on Reply
#90
TheLostSwede
TumbleGeorge
Do not swear by this assumption. In my country, so far only limited in the capital, from the end of 2020 began to offer Internet access in versions with 5 and 10Gb / s at very low prices, so apparently Internet providers obviously have different views on the issue, and obviously access to inside information on how prices will develop in the near future and the implementation of 10Gb / s Internet controllers in motherboards.
Sorry, but what are you on about?
I was discussing Ethernet controllers, i.e. chips, not internet access and ISPs.
Posted on Reply
#91
chrcoluk
Sadly I dont consider a dumb switch at 200usd as a cost effective investment, that is still way too expensive to be considered mainstream, and whilst at wholesale a 2.5gbs chip might be only 50 cents more expensive than a 1gbit chip, the retail pricing doesnt reflect that.

In my home network I have 3 gbit managed switches, I was able to source two of them at about 20usd each. The other I already owned for years. If we ramp that up to 200usd per switch, there is no way I can justify 600usd to upgrade my lan from gbit to 2.5gbit, especially when only once in a blue moon I even get close to saturating gbit of lan throughput.

The upgrade would also probably entail upgrading my storage devices as well, as single spindles would be saturated.

Bear in mind as well wifi rated speeds dont mean real world speeds.

When I can buy 2.5gbit lan cards for 10usd, 100usd motherboards have it included, and 2.5gbit managed switches for no more 20usd used (50usd new), then we can talk about it been mainstream.
Posted on Reply
#92
TheLostSwede
chrcoluk
Sadly I dont consider a dumb switch at 200usd as a cost effective investment, that is still way too expensive to be considered mainstream, and whilst at wholesale a 2.5gbs chip might be only 50 cents more expensive than a 1gbit chip, the retail pricing doesnt reflect that.

In my home network I have 3 gbit managed switches, I was able to source two of them at about 20usd each. The other I already owned for years. If we ramp that up to 200usd per switch, there is no way I can justify 600usd to upgrade my lan from gbit to 2.5gbit, especially when only once in a blue moon I even get close to saturating gbit of lan throughput.

The upgrade would also probably entail upgrading my storage devices as well, as single spindles would be saturated.

Bear in mind as well wifi rated speeds dont mean real world speeds.

When I can buy 2.5gbit lan cards for 10usd, 100usd motherboards have it included, and 2.5gbit managed switches for no more 20usd used (50usd new), then we can talk about it been mainstream.
Well, that's you and your usage. I have a two port 10Gbps switch, 10Gbps Ethernet in my PC and NAS and I can saturate that link when I copy data between the two.

I was mistaken about the price of the switch, it's less than $110.
QNAP also has a managed eight port 2.5Gbps switch with two 10Gbps SFP+ ports for $250 and $300 with an additional pair of 10Gbps copper ports.

Managed Gigabit switches don't cost $50 new on this planet, so I don't know where you live. That you can buy something for $20 second hand, has no relevance.
Also, 99% of consumers don't need a managed switch.
Posted on Reply
#93
TumbleGeorge
TheLostSwede
Sorry, but what are you on about?
I was discussing Ethernet controllers, i.e. chips, not internet access and ISPs.
Yes but if has not controllers has not offers. Depending each other. Without doubt soon will have many mainstream motherboards with onboard 10G NICs. ISP with 5-10Gb/s(via fiber to the home) said that next 3-4 years will spread it's network in all country area.
Posted on Reply
#94
TheLostSwede
TumbleGeorge
Yes but if has not controllers has not offers. Depending each other. Without doubt soon will have many mainstream motherboards with onboard 10G NICs. ISP with 5-10Gb/s(via fiber to the home) said that next 3-4 years will spread it's network in all country area.
You clearly have no idea what you're talking about. Motherboard makers aren't about to add either of those options, as it simply costs them too much.
But I guess you could invent a better chip and sell it to them for $1.50 or less, then it'll happen.
Posted on Reply
#95
BSim500
Raiden85
When you have onboard motherboard wifi connecting at 2.4Gbps, phones at 1.2Gbps and even 5G that can have some really fast speeds, seeing a pathetic 1Gbps wired connection on high end motherboards is just embarrassing.
Wireless marketing labels based on placing wireless devices 1mm apart inside a Faraday cage under lab conditions aren't even close to real world throughput though between rooms. Ethernet = You're usually guaranteed to get 90% of rated speed assuming a non-faulty half decent cable (eg, +900Mbps minimum for Gigabit on Cat6) as well as complete immunity to slowdowns / disconnects in Wi-Fi congested areas (apartment blocks, etc). Wi-Fi & Powerline technology = actual transfer rate is often barely 1/4 to 1/3rd of BS marketing figures. Typical example review for an "866Mbps" device = "Buffalo’s adapter performed much better than most of its competitors when the client was in my home theater (35 feet from the router), delivering TCP throughput of 227 megabits per second". Also Ethernet is Full Duplex, Wi-Fi only Half Duplex. 1Gbps Ethernet is also more energy efficient than many 2.5Gbps chips, something that's more important for the bulk of laptops that don't have Internet speeds anywhere near 2.5Gbps but always need more battery life. Gigabit Ethernet to 2.5 / 5 / 10Gbps was always going to be a slow & gradual transition.
Posted on Reply
#96
chrcoluk
TheLostSwede
Well, that's you and your usage. I have a two port 10Gbps switch, 10Gbps Ethernet in my PC and NAS and I can saturate that link when I copy data between the two.

I was mistaken about the price of the switch, it's less than $110.
QNAP also has a managed eight port 2.5Gbps switch with two 10Gbps SFP+ ports for $250 and $300 with an additional pair of 10Gbps copper ports.

Managed Gigabit switches don't cost $50 new on this planet, so I don't know where you live. That you can buy something for $20 second hand, has no relevance.
Also, 99% of consumers don't need a managed switch.
you just covert via installing openwrt and bam you have a managed switch.

Your last line is relevant though, because 99% of users dont need 2.5gbit lan. I do get though that for specific niche uses such as maybe high speed NAS there is some demand for it, but if I do a quick google, I see people having similar thoughts to me, they look into it, see the price and are like "hell no".

They just need to get the prices down to no more than 1/4 of where they are now.
Posted on Reply
#97
TheLostSwede
chrcoluk
you just covert via installing openwrt and bam you have a managed switch.

Your last line is relevant though, because 99% of users dont need 2.5gbit lan. I do get though that for specific niche uses such as maybe high speed NAS there is some demand for it, but if I do a quick google, I see people having similar thoughts to me, they look into it, see the price and are like "hell no".

They just need to get the prices down to no more than 1/4 of where they are now.
Sorry, do you know what a switch is? OpenWRT is a router operating system.
A managed switch is something entirely different and OpenWRT doesn't allow you to magically add features that the switch IC in the router doesn't support.

Most people don't know what Ethernet is, final.
BSim500
Wireless marketing labels based on placing wireless devices 1mm apart inside a Faraday cage under lab conditions aren't even close to real world throughput though between rooms. Ethernet = You're usually guaranteed to get 90% of rated speed assuming a non-faulty half decent cable (eg, +900Mbps minimum for Gigabit on Cat6) as well as complete immunity to slowdowns / disconnects in Wi-Fi congested areas (apartment blocks, etc). Wi-Fi & Powerline technology = actual transfer rate is often barely 1/4 to 1/3rd of BS marketing figures. Typical example review for an "866Mbps" device = "Buffalo’s adapter performed much better than most of its competitors when the client was in my home theater (35 feet from the router), delivering TCP throughput of 227 megabits per second". Also Ethernet is Full Duplex, Wi-Fi only Half Duplex. 1Gbps Ethernet is also more energy efficient than many 2.5Gbps chips, something that's more important for the bulk of laptops that don't have Internet speeds anywhere near 2.5Gbps but always need more battery life. Gigabit Ethernet to 2.5 / 5 / 10Gbps was always going to be a slow & gradual transition.
You don't actually test wireless devices in a faraday cage, but rather in an anechoic chamber, or something like the chamber from octoScope with one of their multipath emulators.
octoscope.com/English/Products/Ordering/index.html

You're right that the marketing speeds aren't what you'll actually see in terms on throughput though, although as per the link I posted above, it seems like some equipment can now reach 1,500Mbps download speed over wireless, albeit my guess is that it would be inside the same room, but you should still be able to get Gigabit speeds a room away from the router. That said, there are no client devices that can reach these speeds, so it's a bit moot.

Gigabit only needs Cat 5e and if you don't get 950Mbps+ or more like 980Mbps today, you either have a devices that can't cope with the Gigabit speed, a busted cable or some really poor network drivers.

I doubt most people will notice that Wi-Fi is half duplex, since there are always some pauses in the transmission during which time data can be sent. That said, you're obviously correct that Ethernet is far superior to Wi-Fi if speed is the main target application.

2.5 and 5Gbps are actually quite new technologies and the transition to 2.5Gbps seems to be happening quite fast, mostly because Realtek and Intel made some affordable 2.5Gbps network controllers and PHYs, which meant the motherboard makers slapped them on their boards as an upsell/feature. We just need more affordable switches now, as 2.5Gbps doesn't need any better cabling than Cat 5e, which is a big bonus vs. 5 and 10Gbps.

Some pictures from back in the day when I was setting up the octoScope gear when I worked for a router maker. Note that the router pictured was just to make sure everything was working as expected before we started testing our gear.



Posted on Reply
#98
chrcoluk
I do know yes :) and yes openwrt cannot do magic, you obviously have to buy hardware that already has the capability but that capability may not be already unlocked out of the box. But I think we derailing this, we ended up going from broken USB connectivity to talking about if 2,5gbit lan is ready for mainstream.
Posted on Reply
#99
TheLostSwede
chrcoluk
I do know yes :) and yes openwrt cannot do magic, you obviously have to buy hardware that already has the capability but that capability may not be already unlocked out of the box. But I think we derailing this, we ended up going from broken USB connectivity to talking about if 2,5gbit lan is ready for mainstream.
I didn't start it, I simply wanted to clear up a few things.
Posted on Reply
#100
Tom Sunday
Turmania
It is true that Intel is more reliabe as a system package all together.
Thanks for your comment on the system package. I have always looked at both Intel and AMD only for the big picture. A bit of an investor view perhaps! Not specific CPU offerings and performances. Simply said AMD has some way to go as to full CPU maturity and associated hardware compatibility.

According to a new report from South Korea, AMD is now looking to increase its CPU and APU production as its current chip supplier, TSMC, is unable to do so. Hence, AMD is rumored to outsourcing and split even more production. Unlike Intel who has their own foundry-factories. Wall Street is not too happy with AMD as they love certainty, instant product availability and subsequent sales. Besides the engagement of a second CPU supplier for AMD will take additional tool-up time and prepping the factories. Bad for availability and sales.

Intel plans to report its earnings for the first quarter of 2021 on April 22, 2021 promptly after close of market. Let's see then what the Intel boys have to say to their institutional investors and of course the revolutionary and or the all new improved Alder Lake product coming soon or in the months ahead.
Posted on Reply
Add your own comment