[SOLVED] PCIe 4.0 riser cables

Bassman999

Prominent
Feb 27, 2021
481
99
440
Ive built an SFF PC in a NR200 and realized after no post that the included riser cable is a useless 3.0 version.
Im on my backup PC right now sourcing the right cable in PCIe 4.0.
From what I can see LINKUP has a monopoly and they sell the only ones?
They dont have the right size and hoping someone knows of another source.

Thanks Guys!
 
Solution
according to this ( https://www.elektronik-kompendium.de/sites/com/0904051.htm ) source for PCI-E 4.0 the total length is reduced to 8 to 12 inch, leaving about 3 to 7 inch for cabling

That's 20-30cm left for the riser.

About 5 inches is used up by the motherboard, and with Gen3 can be upto 20 inch total, leaving a 15 inch possible riser. (38cm)

But there are considerations. Shielded cables are viable longer than cheap unshielded as they don't suffer electrical/rfi interference, especially when dodging around components stuffed in a pc. So a 50cm shielded will work for Gen4 mostly although there might be some small % of signal loss due to length.

Afaik and can readily find, there isn't an exact measurement for any Gen...

InvalidError

Titan
Moderator
If your CPU has an IGP or you have a spare 3.0 card, you could go in BIOS and limit PCIe to 3.0.

If you haven't installed the latest BIOS on your new motherboard, you could also give that a shot, there is a chance a newer version may have tweaks to get it working on 4.0. There is a chance this step will be required regardless of whatever other riser you may get.

In principle, PCIe should automatically downgrade the speed to whatever is necessary for stability, so you shouldn't need to manually do anything to get it working, albeit at a degraded speed.
 
  • Like
Reactions: Bassman999

Bassman999

Prominent
Feb 27, 2021
481
99
440
If your CPU has an IGP or you have a spare 3.0 card, you could go in BIOS and limit PCIe to 3.0.

If you haven't installed the latest BIOS on your new motherboard, you could also give that a shot, there is a chance a newer version may have tweaks to get it working on 4.0. There is a chance this step will be required regardless of whatever other riser you may get.

In principle, PCIe should automatically downgrade the speed to whatever is necessary for stability, so you shouldn't need to manually do anything to get it working, albeit at a degraded speed.

Im going to have to try the 1050 ti I have in the Dell I'm using right now to see if I can get into bios with that.

I wasnt sure if

Ryzen 5900x so no integrated graphics unfortunately.
 

Bassman999

Prominent
Feb 27, 2021
481
99
440
A 1050Ti is only 3.0 capable.
So buying a 4.0 adapter is a waste of money if you already have a 3.0 riser cable.
Unless it is a bad cable it is not the problem. Check it with your other computer also.
The 1050 is just my Dell GPU.
I have an RX 6800 and RTX 3080 that will be (one of them) will be in my new build. The 3080 isnt here yet, but with the 6800 fans spin up and it lights up then doesnt an no post.
But hopefully I can use the 1050 ti to figure out the bios setting to make it usable with the new GPUs till I can get a proper riser
 
The RX 6800 and the RTX 3080 will both run on 3.0 motherboards and riser cables just fine. With no performance penalties.
So unless the cable is bad it is not the problem.
Ryzen systems can take several minutes on first boot to post. Especially if you have fast memory with very few slower Jedec profiles.
The memory controller has to learn what memory settings will work for initial post.
 
  • Like
Reactions: Bassman999

Bassman999

Prominent
Feb 27, 2021
481
99
440
The RX 6800 and the RTX 3080 will both run on 3.0 motherboards and riser cables just fine. With no performance penalties.
So unless the cable is bad it is not the problem.
Ryzen systems can take several minutes on first boot to post. Especially if you have fast memory with very few slower Jedec profiles.
The memory controller has to learn what memory settings will work for initial post.
Everything was fine till i used the riser. Been using the PC for 2 weeks straight with GPU directly connected to the MB. But 4.0 to 3.0 cable to 4.0 card isnt working.
 

Bassman999

Prominent
Feb 27, 2021
481
99
440
Update:
I took out the RX 6800 and installed the 1050 ti straight to MB and it booted fina as expected.
I went inti bios and turned off auto and put to PCIe 3.0.
Installed riser cable and RX 6800 NO GO.
Took out 6800 and installed 1050 to riser and no issue.
So I took out 1050 and riser and put RX 6800 back in and no issues.
RX 6800 doesnt want to play with this cable at all.
I ordered a PCIe 4.0 riser from Amazon in 10cm which is the shortest even though 5 cm is the needed length.
Fingers crossed it works and that I can make that longer than stock ribbon not impede the fan on the bottom mounted (I know dont do that lol) radiator.
If everything works right Ill get an AIO with pump in the rad to be safer.
I did stress test in this config nd no worse that side rad, but not the same since using a tiny gpu in that test.

Riser is scheduled o arrive Monday
 

Karadjgne

Titan
Ambassador
Bit-wit had the same issue in a build, 6900xt no-joy. Dude changed everything from the psu, gpu (to nvidia which worked), risers, got nothing out of that card. Until he talked with a bud at AMD who told him to change the cable from the gpu to monitor. It's got to be a Vesa certified DP 1.4 cable or HDMI 2.1, it can't be the standard DP 1.2.

Edit: found the vid.
View: https://youtu.be/IQQKlAY5OfM
 
Last edited:
  • Like
Reactions: Bassman999

Karadjgne

Titan
Ambassador
There's a somewhat sizable difference between DP 1.4/HDMI 2.1 and DP 1.2/HDMI 2.0 as far as the cables go. Dp and hdmi both rely on handshakes to/from the monitor to enable them, but very little is said about how those handshakes occur or where. If you think back to old vga with the D-Subs, for a monitor they were 15 pin, but the 5v #9 pin was removed. If you used that same hookup to a TV, it didn't work at all, you needed that #9 pin to complete recognition for resolution by the TV itself. No handshake.

After watching Bit-wit's video, it's entirely probable there's something in the DSC (Display Stream Compression) or extra bandwidth from the new cables that's effectively doing the same thing, transmitting on pin #9 from the gpu, but there's nothing coming out the other end as pin #9 is effectively not there on a standard DP 1.2 cable or HDMI 2.0 cable which aren't DSC capable. Same channels electrically, so physically they are the same, it's the physical limits that have changed.

No handshake.
 
  • Like
Reactions: Bassman999

Bassman999

Prominent
Feb 27, 2021
481
99
440
Bit-wit had the same issue in a build, 6800xt no-joy. Dude changed everything from the psu, gpu, risers, got nothing out of that card. Until he talked with a bud at AMD who told him to change the cable from the gpu to monitor. It's got to be a DP 1.4 cable, it can't be the standard DP 1.2.
I had it working direct from the MB with HDMI.
Hopefully I didnt damage the GPU getting it out of the MB.
 

Bassman999

Prominent
Feb 27, 2021
481
99
440
Yes I have heard AMD is more troublesome to work with as well.
A guy in reddit said the LINKUP cable didnt work, but that post was a few months old and they have updated their SW since so maybe thats not an issue anymore.
Ordered the LINKUP cable anyway.
Im using certified HDMI cable. No DP yet as I an currently still connected t a smart tv with HDMI only.
 

kanewolf

Titan
Moderator
Ive built an SFF PC in a NR200 and realized after no post that the included riser cable is a useless 3.0 version.
Im on my backup PC right now sourcing the right cable in PCIe 4.0.
From what I can see LINKUP has a monopoly and they sell the only ones?
They dont have the right size and hoping someone knows of another source.

Thanks Guys!
The problem with riser cables and PCIe 4.0 or 5.0 in the future is that the faster signaling means that the total bus length has to be shorter.
This company, which sells risers http://www.adexelec.com/tq-14 says the max PCIe 4.0 length is 6 inches. Your 10CM is right at the limit for PCIe 4.0
 
  • Like
Reactions: Bassman999

Bassman999

Prominent
Feb 27, 2021
481
99
440
The problem with riser cables and PCIe 4.0 or 5.0 in the future is that the faster signaling means that the total bus length has to be shorter.
This company, which sells risers http://www.adexelec.com/tq-14 says the max PCIe 4.0 length is 6 inches. Your 10CM is right at the limit for PCIe 4.0
Its crazy the length issue, but fortunately im in an SFF
What would we do without prime 2 day delivery? Id have to wait 5-8 days for a delivery of a part that might now work, then pay to ship it back 🤣
 

Karadjgne

Titan
Ambassador
according to this ( https://www.elektronik-kompendium.de/sites/com/0904051.htm ) source for PCI-E 4.0 the total length is reduced to 8 to 12 inch, leaving about 3 to 7 inch for cabling

That's 20-30cm left for the riser.

About 5 inches is used up by the motherboard, and with Gen3 can be upto 20 inch total, leaving a 15 inch possible riser. (38cm)

But there are considerations. Shielded cables are viable longer than cheap unshielded as they don't suffer electrical/rfi interference, especially when dodging around components stuffed in a pc. So a 50cm shielded will work for Gen4 mostly although there might be some small % of signal loss due to length.

Afaik and can readily find, there isn't an exact measurement for any Gen riser limits, as different companies use different techniques in manufacturing and different materials, which affect transmission. Fibre cabling can be longer than metallic wiring for instance. As can larger awg wires and shielded/seperated wires.

Only conclusion I can make is that as long as there's no hard bends (use the right end connections) the shortest shielded risers will be the best, far more compatible than longer or unshielded.
 
  • Like
Reactions: Bassman999
Solution

Bassman999

Prominent
Feb 27, 2021
481
99
440
That's 20-30cm left for the riser.

About 5 inches is used up by the motherboard, and with Gen3 can be upto 20 inch total, leaving a 15 inch possible riser. (38cm)

But there are considerations. Shielded cables are viable longer than cheap unshielded as they don't suffer electrical/rfi interference, especially when dodging around components stuffed in a pc. So a 50cm shielded will work for Gen4 mostly although there might be some small % of signal loss due to length.

Afaik and can readily find, there isn't an exact measurement for any Gen riser limits, as different companies use different techniques in manufacturing and different materials, which affect transmission. Fibre cabling can be longer than metallic wiring for instance. As can larger awg wires and shielded/seperated wires.
Im not sure what design the LINKUP is, but it looks like the common PCIe 3.0 ones from the pictures.
Ive seen hem advertising they can be bent and folded without compromising them.