Fractal Ridge Case PCIe Riser Has Trouble Running in PCIe Gen 4 Mode, Company Outs Workaround

[btarunr]

The Fractal Design Ridge SFF tower-type case won critical praise including from us, for its unique design, well-planned interiors, and room for even triple-slot graphics cards with a little adjustments. The case relies on a PCIe riser cable to maintain its SFF form, since the graphics card has to be oriented vertically. The riser included with the case meets PCI-Express 4.0 standards, but end-users started experiencing problems running their latest-generation PCIe Gen 4 graphics cards with this case, with the problem being localized to the riser cable. Fractal investigated this issue, and confirmed the issue.

Apparently, the PCIe riser included with the Fractal Ridge, while rated for PCIe Gen 4, has a design flaw that affects signal integrity. The riser is found to be only stable with PCIe Gen 3 or lower. The company is recommending a workaround for end-users while it works on a solution: to confine PCIe to Gen 3 mode using the motherboard's UEFI setup program (BIOS setup program), in which you can restrict the x16 PEG slot to Gen 3 mode. "We are developing a solution to enable full PCIe 4.0 compatibility, but until that is ready, we will update our listings to reflect that only PCIe 3.0 compatibility is guaranteed," the company said in a statement.



In this above images from our review of the Fractal Ridge, you'll see that Fractal's PCIe Gen 4 riser is a wacky contraption of two PCBs, and the PCIe signal has to change hands twice before reaching your graphics card. The first, larger, rectangular PCB connects to your motherboard's PCIe slot, and turns it to a 90° angle; while the second shorter PCB slots into this one, and extends the slot to your graphics card, at a 0° angle. A simpler design choice would've been to use flexible PCIe Gen 4-rated cables, and a single point of exchange.

Our GeForce RTX 4090 PCI-Express Scaling article should give you a comprehensive look at how much performance you stand to lose by running an RTX 4090—currently the fastest graphics card—in PCIe Gen 3 mode. You lose about 2% performance at 4K Ultra HD when averaged across 25 games from our test bench, so not by much.

The complete statement by Fractal Design follows.

The PCIe riser card supplied with Ridge has been discovered to be incompatible with some hardware configurations when run in PCIe 4.0 mode.

After extensive testing, we have determined that while most systems will run PCIe 4.0 using the Ridge riser card without issue, those experiencing issues with stability will need to run the system in PCIe 3.0 mode for now. This can be enabled in the motherboard BIOS settings.

We are developing a solution to enable full PCIe 4.0 compatibility, but until that is ready, we will update our listings to reflect that only PCIe 3.0 compatibility is guaranteed.

Currently affected users, please contact our customer service team at ridge.support@fractal-design.com to receive assistance and continuous updates with the latest information until we have a riser replacement available. The customer service team can also help those requiring guidance in switching to PCIe 3.0 mode.

We thank you for your understanding and apologize for any inconvenience.

View at TechPowerUp Main Site

 

[Mussels]

I've had personal experience with fractals warranty after a USB-C port melted out of nowhere, they didnt even care about proof of purchase for that one or age of warranty since it failed so potentially catastrophically
They'll handle this pretty well, IMO

Likely coming out with a V2 riser as an optional purchase, but send one free to anyone who has proof of purchase

 

[TheDeeGee]

I wonder if Gen5 riser cables will be possible, seeing how difficult it already is to get the Gen4 spec going for them.

 

[maxfly]

There are so many crappy riser cables out there its ridiculous. It's literally a crapshoot. Despite what individual reviews may say you don't know what you're going to end up with.

It's great to see Fractal acknowledging the issue and planning a fix. Hopefully sooner rather than later.

 

[Haile Selassie]

Recall imminent. It either passes EM testing for Gen4 signaling per design or it doesn't.

 

[Bwaze]

A lot of people had problems with riser cables and PCIe 4 in various forums - either they abandoned the riser option, changed them to some cable that worked, or, hilariously, accepted they work in PCIe 3 mode.

 

[Assimilator]

On the plus side, at least these don't catch fire.

I wonder if Gen5 riser cables will be possible, seeing how difficult it already is to get the Gen4 spec going for them.

It's only difficult once you start optimising for cost; NZXT and now Fractal optimised too far.

 

[Bjørgersson]

I've had personal experience with fractals warranty after a USB-C port melted out of nowhere, they didnt even care about proof of purchase for that one or age of warranty since it failed so potentially catastrophically
They'll handle this pretty well, IMO

Likely coming out with a V2 riser as an optional purchase, but send one free to anyone who has proof of purchase

I don't think it'll be an optional purchase, but rather a replacement for current owners and the default part for future batches.

 

[TheinsanegamerN]

On the plus side, at least these don't catch fire.


It's only difficult once you start optimising for cost; NZXT and now Fractal optimised too far.

No, its difficult regardless of cost. PCIe5 is very latency sensitive and a riser is the antithesis of the type of tight tracing needed for PCIe5.

 

[trsttte]

I don't think it'll be an optional purchase, but rather a replacement for current owners and the default part for future batches.

Something they could do is sell the current non compliant version as pcie3.0 with a discount to soften the blow. Not great but at least they recoup some money. Though, from how the company reacted to past problems they'll suck up and pay for the mistake (which makes sense to keep a good reputation).

It's only difficult once you start optimising for cost; NZXT and now Fractal optimised too far.

In this particular case they over engineered their riser with 2 different pcb's each with a pair of connectors instead of just going with the standard flat cables everyone else uses.

 

[Assimilator]

No, its difficult regardless of cost. PCIe5 is very latency sensitive and a riser is the antithesis of the type of tight tracing needed for PCIe5.

No, it's not difficult, it's already been achieved over 2 years ago.

 

[Punkenjoy]

No, it's not difficult, it's already been achieved over 2 years ago.

I am not sure if current PCI-E Gen 4 solution have Retimers, but if that was so easy, we would already have PCI-E Gen 5 solution around.

I suspect that the current retimers are just too costly and are probably aimed at datacenter solutions rather than desktop.

 

[Assimilator]

I am not sure if current PCI-E Gen 4 solution have Retimers, but if that was so easy, we would already have PCI-E Gen 5 solution around.

I suspect that the current retimers are just too costly and are probably aimed at datacenter solutions rather than desktop.

It's only difficult once you start optimising for cost

I love quoting myself.

 

[Punkenjoy]

I love quoting myself.

Going to the moon is only difficult once you start optimizing for cost.

Give them unlimited budget and they will be there quickly.

That do not mean anything. If something is really expensive, it's not "easy". It's not a jewelry or another kind of luxury item.

That do not mean that in the future it will be, but that is not the case right now.

 

[Assimilator]

Going to the moon is only difficult once you start optimizing for cost.

Give them unlimited budget and they will be there quickly.

LMAO, you evidently know nothing about how America got to the Moon in the 1960s. I'll give you a hint: effectively limitless funding for NASA.

 

[LabRat 891]

I'm no EE, but I do have some 'hands-on' (non-scientific) experiences in dealing w/ EMI / Signal Integrity; this particular riser appears to be designed rather poorly.
Every additional (inter)connection adds impedance, and un-shielded (and/or improperly laid-out) traces on inexpensive riser cards 'like to' pick up interference from other circuits/devices. In this case, using 'riser cards' and a 'flex riser' was absolutely antithetical towards Gen4 Signal Integrity.
(I wonder how 'cost-effective' a riser using AMMC / W.FL / MHF III micro-coax leads and connectors would be? IIRC I've ran across 'industrial PC' risers built using microwave/RF interconnects)

side note:
I find it highly amusing that as 'serial' buses get faster and more 'lanes' are added, that we're back to the same problems that led the industry *away* from 'parallel' buses:
Signal timing and integrity.
(Also, that 'digital' interfaces are more and more reliant on complex analog/RF-like modulation; QAM, etc.)

 

[Punkenjoy]

LMAO, you evidently know nothing about how America got to the Moon in the 1960s. I'll give you a hint: effectively limitless funding for NASA.

Some apollo mission crews would have prefer that it was just a question of money and that it was just easy if you had unlimited funds. But hey, I don't know anything about that.

Still took them a decade. And will take even more to get there now with more modern tech.

But hey, it's just a matter of cost like you say. They are being cheap and anyone with unlimited money would be there right away.

that just prove my point that if you need unlimited money (or a lot of money) for something, it's not easy.

 

[Aleksandar_038]

I'm no EE, but I do have some 'hands-on' (non-scientific) experiences in dealing w/ EMI / Signal Integrity; this particular riser appears to be designed rather poorly.
Every additional (inter)connection adds impedance, and un-shielded (and/or improperly laid-out) traces on inexpensive riser cards 'like to' pick up interference from other circuits/devices. In this case, using 'riser cards' and a 'flex riser' was absolutely antithetical towards Gen4 Signal Integrity.
(I wonder how 'cost-effective' a riser using AMMC / W.FL / MHF III micro-coax leads and connectors would be? IIRC I've ran across 'industrial PC' risers built using microwave/RF interconnects)

side note:
I find it highly amusing that as 'serial' buses get faster and more 'lanes' are added, that we're back to the same problems that led the industry *away* from 'parallel' buses:
Signal timing and integrity.
(Also, that 'digital' interfaces are more and more reliant on complex analog/RF-like modulation; QAM, etc.)

Ins't that a proof that moving "away" from the "parallel" buses was right solution? Cables already becoming very complex, I can not imagine how achieving similar speed would look like with "parallel" buses... Isn't lage-age SCSI bit the dust due to extremely complex and expensive cabling? I read it somewhere, not that I had any experience with it, that is why I am asking.

 

[Mussels]

There are so many crappy riser cables out there its ridiculous. It's literally a crapshoot. Despite what individual reviews may say you don't know what you're going to end up with.

It's great to see Fractal acknowledging the issue and planning a fix. Hopefully sooner rather than later.

Even good ones break

All of mine have failed except one, they're flexible but also so fragile - if you have to add and remove them from a case more than just the once, they degrade - mine all went from 4.0 stable, to needing 3.0 to be stable, to no display after maybe 5 installations (the whole rebuilding watercooling looops shenanigans)


my only intact one is in an ITX Case where it's mandatory - you cant use the PC without it, so it never got removed to get worn out and the cable length was exactly right since they knew exactly the distance for the required slot

 

[Assimilator]

Ins't that a proof that moving "away" from the "parallel" buses was right solution? Cables already becoming very complex, I can not imagine how achieving similar speed would look like with "parallel" buses... Isn't lage-age SCSI bit the dust due to extremely complex and expensive cabling? I read it somewhere, not that I had any experience with it, that is why I am asking.

Yes, the problem with parallel buses is timing, i.e. keeping all the separate signals in sync.

 

[Camm]

Fractal will absolutely give us a PCIE Gen 4 riser in due time. That said, I'm overall just suspect about their electronics full stop. I've had two front USB ports on the Fractal Torrent short, along with the fan controller shorting & killing any fan plugged ino that header. In Fractals defence they've sent me new fan controllers & front panels without question, but I think I will probably stay away from Fractal for my next case.

 

[Mussels]

Fractal will absolutely give us a PCIE Gen 4 riser in due time. That said, I'm overall just suspect about their electronics full stop. I've had two front USB ports on the Fractal Torrent short, along with the fan controller shorting & killing any fan plugged ino that header. In Fractals defence they've sent me new fan controllers & front panels without question, but I think I will probably stay away from Fractal for my next case.

In their defense, i've had every single case i've ever owned have issues with their front USB 3 ports. Very few have been quality overall, i've got some fussier flash drives that only work in one single USB3.0 front panel in the house, but zero issues on any rear ports.

Another common one is my NVME drive on USB 3.2, that often gets stuck at USB 2.0 on front ports, but repeated attempts eventually get the speed up - it's not just a fractal issue