Riser cables are a serious problem with AMD Ryzen cpus and motherboards

[sixer9682]

I found this out the hard way with my new build and wanted to pass along some helpful information.

Everybody loves vertical mounts, they look great, it's a different layout in the case, and for watercooling it opens up different tubing options. But with the Ryzen cpus and AMD's AM4 motherboards the riser cables, as I found out, are creating a lot of headaches and are a pain in the ass to work with. Worse yet, AMD knows there is a problem with their AM4 platform and riser cables, more about that below, and they could care less.

So what's the problem? It wasn't one until the new gpus became available late last year. Previously you could use a riser cable to vertically mount your gpu and it worked just fine, was auto-detected by the motherboard and all the user had to do was plug and play. With the Nvidia RTX 3000 series and AMD's 6000 series that does not apply any longer. Plug and play is broken on AM4 platforms.

With any new gpu and using a riser cable, which are 99% gen3 right now, if you use a riser cable, an AM4 motherboard, and a new gpu you get "no signal" on your monitor. Why? Because the cable is gen3 PCIE and you have to manually set in the bios for a gen3 slot. That's no big deal, right? Wrong. If you do not have an old gpu to work with there is no way to get into the bios to make the changes. Even if you have an old gpu, you first have to plug that in, make the bios changes, take out the old gpu, and then install the new one with the riser cable.

This is especially frustrating for bios updates because with an update, all settings go back to default, meaning your gpu is no longer detected wit the riser cable and you're back to "no signal".
This is doubly frustrating when you are water cooling.

So why not get a gen4 riser cable you ask? Because there are currently 2 companies that sell them in the US right now and they're $80-90+ for the cable and the supply is extremely low meaning good luck finding one. Even then you're not guaranteed that default auto detect bios settings are going to work so you have to change them for PCIE to gen4....and we're right back where we started with "no signal" if you're trying to use a new gpu and the AM4 platform.

Worse yet, AMD became aware of the problem late December and posted this gem of a reply https://www.amd.com/en/support/kb/faq/pa-270 that basically says "fuck it" they don't care.

Intel platforms currently do not have these problems, but I think that's because they're still using old tech for cpus and motherboards for 14nm.
At any rate, if you're building an AMD rig, this is a huge pain in the ass that AMD is ignoring and it is causing end users a lot of frustration.

More about this here: https://hardwarecanucks.com/video-cards/theres-a-problem-with-riser-cables/

 

[smokerob79]

shit with some riser cables starting fires i will not use them regardless of what looks good....i skipped doing a ITX build over the riser cable problems and i really want 1 of the tiny slim cases

 

[JoeLeb]

Thanks for posting the helpful information @sixer9682

You've sure had your share of "growing pains" with your latest build. Hope you don't have to rip your tubing apart again.

Just to add to your post about some things to look out for with AMD, I was reading about some people having their USB peripherals randomly losing connection as well. They're currently trying to recreate the issue and hopefully find a fix for those affected. I only mention it in case that problem creeps into your build too.

Here's more information about the USB issue, for anyone interested. https://www.tomshardware.com/news/amd-investigating-usb-connectivity-issues-with-ryzen-pcs

Good luck @sixer9682

 

[5I5O]

yeah riser cables are freakin expensive.. even i thought about using one.. but fuck that!

 

[smokerob79]

@JoeLeb ....i starting having the USB issues on my ASUS TUF x570 wifi with a 3700x CPU.....i just updated bios and now everything works fine for over 2 months.....for some reason the 5 series has had a FUCK TON of bios updates over the last 6 months...like my mother getting 1 every month....the good part is every time it really fixes issues.....like the motherboard not booting to a M.2 drive with wifi and blutooth on when i first got the system....

on the riser cables i also found people saying intel is having the same problem with some cards on gen 4.0 able mothers.....

 

[sixer9682]

@JoeLeb ....i starting having the USB issues on my ASUS TUF x570 wifi with a 3700x CPU.....i just updated bios and now everything works fine for over 2 months.....for some reason the 5 series has had a FUCK TON of bios updates over the last 6 months...like my mother getting 1 every month....the good part is every time it really fixes issues.....like the motherboard not booting to a M.2 drive with wifi and blutooth on when i first got the system....

on the riser cables i also found people saying intel is having the same problem with some cards on gen 4.0 able mothers.....

Damn; if Intel is having the same problem there's a root cause error that both should be able to fix. The thing that bothers me is AMD pretty much told everyone to fuck off they aren't interested in fixing the issue.

I've got my gpu installed vertically with a riser cable currently and ordered a gen4 vesion. We'll see how that works, but if it doesn't fix the issue then the gpu is going to be mounted horizontally. I just didn't want to do this because the 3080 FE is so short that the IO ports for the waterblock sit directly beneathe my memory which would just cause awkward tubing layouts. It might come to that though if the gen4 cable doesn't fix the issue. Right now I can't even update the bios with the cable I have in there because I'll lose all video signal if I do.

 

[smokerob79]

i am still wonder what the fuck was changed from gen 3 to 4 that would make the cabling be affected.....the reason i am coming back to cabling is the simple fact all of these people regardless of system say it runs fine without the riser and the cards run 4.0 when on the mother direct, so it has to be the cabling or the socket/pinset used in the risers causing the problem....wondering if distance from GPU to CPU is a problem or if the cabling in a (claimed gen 4) riser is more shielded or of a heaver gauge wiring.....i still think the problem is coming back to the riser cables and real world AMD cant do shit about it as they are not making them....


as with anything in computers, you get the system back to the basics and then add parts 1 at a time to find the problem...card and mothers work when they are connect directly together, add riser cable shit happens....problem has to be riser cable.....

and i fell you on the tubes being fucked up, nvidia did crazy PCBs with the new founders cards and memory stick have only gotten taller in the last 10 years....full coverage water blocks are the only way to go but man it can make shit line up funny sometimes....i wonder if slim 90's hardline connectors would help you getting around the memory???....hope its only 1 tube you have to get a 90 for.....i have seen people link 2 90's to get around memory before so remember you can do it if needed

 

[JoeLeb]

You mentioned Intel was having a problem with Gen 4 pcie @smokerob79

To my understanding, Intel Cpu's that support gen 4 pcie aren't out until March or April.

 

[sixer9682]

i am still wonder what the fuck was changed from gen 3 to 4 that would make the cabling be affected.....the reason i am coming back to cabling is the simple fact all of these people regardless of system say it runs fine without the riser and the cards run 4.0 when on the mother direct, so it has to be the cabling or the socket/pinset used in the risers causing the problem....wondering if distance from GPU to CPU is a problem or if the cabling in a (claimed gen 4) riser is more shielded or of a heaver gauge wiring.....i still think the problem is coming back to the riser cables and real world AMD cant do shit about it as they are not making them....


as with anything in computers, you get the system back to the basics and then add parts 1 at a time to find the problem...card and mothers work when they are connect directly together, add riser cable shit happens....problem has to be riser cable.....

and i fell you on the tubes being fucked up, nvidia did crazy PCBs with the new founders cards and memory stick have only gotten taller in the last 10 years....full coverage water blocks are the only way to go but man it can make shit line up funny sometimes....i wonder if slim 90's hardline connectors would help you getting around the memory???....hope its only 1 tube you have to get a 90 for.....i have seen people link 2 90's to get around memory before so remember you can do it if needed

Yeah the tubing can be done, but what a PIA to basically be limited to the underside ports. The nice thing about the vertical mount is different routing options for rigid tubing, for me anyway.

 

[smokerob79]

@JoeLeb i fucked up.........what i found on the intel stuff was coming back to cheap riser cables that were included with ITX cases....a lot of them were probably bad cables that got pasts quality control

after researching this (on the AMD side) with 4.0 being a massive jump in bandwidth i think enough of a signal is getting to the motherboard that it sees 4.0 and tries to run it when the cabling cant handle the bandwidth.....motherboard makers had a fuck ton of problems implementing gen 4.0 and had to redesign how they made the mothers....so few true gen 4 riser cables are out there because its hard to make 1 with enough shielding to run a distance of even 6 inches....some makers are trying to make in line amplifiers to make longer riser cables but none have come to market.....its like going from cat 5 to cat 8 Ethernet cabling but the network port has no way of knowing it cant run cat 8 speeds over the cat 5 cabling

another kicker....some 3.0 risers worked with 4.0 on both sides of the cable but then had others that were the SAME MODEL not work.....its so fucked up that you could have 1 work and when you go to order a second riser cable to do the same thing it wont work at all.....you will have 2 identical cables in your hands and 1 works when the other does not.....

now add in the fact that some of these riser cables have started FUCKING FIRES IN PEOPLES COMPUTERS....and I myself will not use 1 when its connecting the most expensive thing in my PC case....and that sucks as i really wanted to build a ITX system....i am over full sized ATX.....


after a couple of hours of looking in to this more the other day i found this video that was made back in Dec....wish i had found the video first....oh well

 

[sixer9682]

Yeah, it's crazy. I'm seriously considering just getting rid of the vertical mount, though it looks fantastic, and saving myself a headache just by putting the card horizontally but directly in the pcie slot. I hate to do it because the veritical mount really shows off the beauty of the block, but right now I think it's more of a hassle than it's worth.

 

[RainMotorsports]

@JoeLeb i fucked up.........what i found on the intel stuff was coming back to cheap riser cables that were included with ITX cases....a lot of them were probably bad cables that got pasts quality control

after researching this (on the AMD side) with 4.0 being a massive jump in bandwidth i think enough of a signal is getting to the motherboard that it sees 4.0 and tries to run it when the cabling cant handle the bandwidth.....motherboard makers had a fuck ton of problems implementing gen 4.0 and had to redesign how they made the mothers....so few true gen 4 riser cables are out there because its hard to make 1 with enough shielding to run a distance of even 6 inches....some makers are trying to make in line amplifiers to make longer riser cables but none have come to market.....its like going from cat 5 to cat 8 Ethernet cabling but the network port has no way of knowing it cant run cat 8 speeds over the cat 5 cabling

another kicker....some 3.0 risers worked with 4.0 on both sides of the cable but then had others that were the SAME MODEL not work.....its so fucked up that you could have 1 work and when you go to order a second riser cable to do the same thing it wont work at all.....you will have 2 identical cables in your hands and 1 works when the other does not.....

now add in the fact that some of these riser cables have started FUCKING FIRES IN PEOPLES COMPUTERS....and I myself will not use 1 when its connecting the most expensive thing in my PC case....and that sucks as i really wanted to build a ITX system....i am over full sized ATX.....


after a couple of hours of looking in to this more the other day i found this video that was made back in Dec....wish i had found the video first....oh well

The same model of any given cable has a spec it has to meet and the worst ones to pass QC are still above spec for 3.0. One cable might have 1 wire with a few extra microfractures that at an specific frequency it doesn't work but so long as it works for 3.0 signalling that will never matter. Since its a computer we say fuck it and try but we would never buy a fuel line that has a 50% chance of handing 1 bar higher pressure than its rated for and complain that sometimes it works and sometimes it blow gasoline all over the exhaust and kills the people in the backseat. The controller is going to negotiate the fastest mode a devices is compatible with so long as it doesn't fail out of running it. It has no clue what's in the pathway. The assumption is always going to be that a system will be built with appropriate components. There has been a bit of a failure on the fallback side of this.

As someone who upgrades 1 to 2 times a decade and isn't currently in danger of PCIE 4 issues I thought this was very well known. Like covered to death in everything from 12 years olds bitching on facebook to probably 10 videos from GN.

Unfortunately it's likely that proper risers will remain expensive.

 

[smokerob79]

@RainMotorsports ....meeting 3.0 spec wont help if the card is defaulting to 4.0 and staying there because it has the feedback to say its 4.0 on both ends.....how do you make the motherboard understand it has a cable in between that is 3.0??? real world why would a motherboard turn off 4.0 if its getting a 4.0 signal???

 

[RainMotorsports]

@RainMotorsports ....meeting 3.0 spec wont help if the card is defaulting to 4.0 and staying there because it has the feedback to say its 4.0 on both ends.....how do you make the motherboard understand it has a cable in between that is 3.0??? real world why would a motherboard turn off 4.0 if its getting a 4.0 signal???

But meeting 3.0 spec is all it is supposed to do. While AMD failed right out of the box to support a situation they knew was going to happen. It's still a user created issue. A user calls me up and tells me "Hi, I have installed an incompatible part between 2 otherwise compatible devices and I am upset". I would response with "I would be upset if I was stupid as well". Once again it's computer hardware so we give it a shot and it if it works cool if it doesn't... For many standards you will see things work for a little bit and crap out and this is not because it almost works. Things like error correction in a given standard might prop up a connection until encounters a serious issue. Crosstalk might not become an issue until a certain condition is met, etc. I am not much for electrical engineering unfortunately.

This is also why we give "X works fine" people shit because X works fine sometimes but that's not how you go giving advice to other people who are spending money. I prefer to give advice under "You can try but it might not work".

 

[sixer9682]

But meeting 3.0 spec is all it is supposed to do. While AMD failed right out of the box to support a situation they knew was going to happen. It's still a user created issue. A user calls me up and tells me "Hi, I have installed an incompatible part between 2 otherwise compatible devices and I am upset". I would response with "I would be upset if I was stupid as well". Once again it's computer hardware so we give it a shot and it if it works cool if it doesn't... For many standards you will see things work for a little bit and crap out and this is not because it almost works. Things like error correction in a given standard might prop up a connection until encounters a serious issue. Crosstalk might not become an issue until a certain condition is met, etc. I am not much for electrical engineering unfortunately.

This is also why we give "X works fine" people shit because X works fine sometimes but that's not how you go giving advice to other people who are spending money. I prefer to give advice under "You can try but it might not work".

That would have been a better response than AMD gave on their statement.

 

[smokerob79]

But meeting 3.0 spec is all it is supposed to do. While AMD failed right out of the box to support a situation they knew was going to happen. It's still a user created issue. A user calls me up and tells me "Hi, I have installed an incompatible part between 2 otherwise compatible devices and I am upset". I would response with "I would be upset if I was stupid as well". Once again it's computer hardware so we give it a shot and it if it works cool if it doesn't... For many standards you will see things work for a little bit and crap out and this is not because it almost works. Things like error correction in a given standard might prop up a connection until encounters a serious issue. Crosstalk might not become an issue until a certain condition is met, etc. I am not much for electrical engineering unfortunately.

This is also why we give "X works fine" people shit because X works fine sometimes but that's not how you go giving advice to other people who are spending money. I prefer to give advice under "You can try but it might not work".

so you both are still blaming AMD when they did not make the PCI-E 4.0 standard.....sounds about right.....the reality is AMD did it first and other makers would have the same problems if they had the balls to do something first....AMD cant change the standard of USB any more then it can change the PCI-E interface....this will never be a AMD problem when its a problem with how the standard was put in place....

 

[sixer9682]

so you both are still blaming AMD when they did not make the PCI-E 4.0 standard.....sounds about right.....the reality is AMD did it first and other makers would have the same problems if they had the balls to do something first....AMD cant change the standard of USB any more then it can change the PCI-E interface....this will never be a AMD problem when its a problem with how the standard was put in place....

IDK where you got that from. Nobody is blaming AMD, all I said was riser cables don't work with AMD and new gpus and they do not. I also pointed out AMD's statement.

They might not be able to fix the problem directly but AMD is one of the only 2 players in desktop cpu makers and they have a lot of influence.
My point is they could use that influence to fix the problem but they have basically said "fuck it".

 

[JoeLeb]

I saw this on Twitter today regarding the latest "guidance" on the AMD USB issue, in case any of their suggestions are helpful. It was posted an hour ago and mentions a few things to do, in case one of them sticks.

https://www.tomshardware.com/news/a...flow&utm_source=twitter.com&utm_medium=social

If the link doesn't work for some reason, here are the main takeaways.

1.) Verify that your motherboard is updated to the latest BIOS version and configured using optimized/factory default settings. Check your motherboard manufacturer’s website for BIOS update and download instructions.
2.) Check if your Windows 10 is on the latest build and fully up to date. For information on updating Windows 10, please refer to Microsoft article: Update Windows 10
3.) Ensure that the Ryzen chipset driver from AMD is installed and up to date. Latest Ryzen chipset driver version is 2.13.27.501 and can be download here.

If you continue to experience USB connectivity problems after following the suggestions above, you may consider using either of the following workaround:

1.) Set PCIe mode from Gen4/Auto to Gen 3 in the BIOS
2.) Disable Global C-State in the BIOS.

Hope you get that beauty of a build purring like a happy kitten @sixer9682

 

[sixer9682]

I saw this on Twitter today regarding the latest "guidance" on the AMD USB issue, in case any of their suggestions are helpful. It was posted an hour ago and mentions a few things to do, in case one of them sticks.

https://www.tomshardware.com/news/a...flow&utm_source=twitter.com&utm_medium=social

If the link doesn't work for some reason, here are the main takeaways.

1.) Verify that your motherboard is updated to the latest BIOS version and configured using optimized/factory default settings. Check your motherboard manufacturer’s website for BIOS update and download instructions.
2.) Check if your Windows 10 is on the latest build and fully up to date. For information on updating Windows 10, please refer to Microsoft article: Update Windows 10
3.) Ensure that the Ryzen chipset driver from AMD is installed and up to date. Latest Ryzen chipset driver version is 2.13.27.501 and can be download here.

If you continue to experience USB connectivity problems after following the suggestions above, you may consider using either of the following workaround:

1.) Set PCIe mode from Gen4/Auto to Gen 3 in the BIOS
2.) Disable Global C-State in the BIOS.

Hope you get that beauty of a build purring like a happy kitten @sixer9682

I got a gen4 riser cable that, thankfully, detects the gpu on "auto" settings. I was finally able to update the bios and get all the memory working at XMP rated speeds and timings. I also installed the single 12pin PSU cable I got from EVGA for the 3080FE. FINALLY everything is humming along and working wonderfully.

http://www.3dmark.com/spy/18849472

 

[JoeLeb]

Yay, now she'll run as fine as she looks. What a beauty!