Flexible PCIe risers

Kmpkt

Innovation through Miniaturization
KMPKT
Bronze Supporter
Feb 1, 2016
3,352
5,826
So I've seen a lot of comments about this riser and that riser having poor data transmission over the past several months. Is there a software benchmark for determining this? Right now I'd just be inclined to do the "does it work" test, but that obviously isn't optimal.
 
  • Like
Reactions: zovc

zovc

King of Cable Management
Editorial Staff
Jan 5, 2017
852
601
I just did a quick look around for anyone who did benchmarks on riser cables and this is what I came up with. There were a few other posts that either linked to this one or more or less cited its results.

Allegedly, all cables are more or less created equal unless you're getting EMI.

It's a very small sample size, but the fact that even daisy chaining to 500cm didn't really affect performance is pretty persuasive to me. I feel like this is arguing that if you have performance issues it must be either EMI, not enough power over the cable, or a dud of a cable.
 
Last edited:
  • Like
Reactions: Ceros_X and Phuncz

aquelito

King of Cable Management
Piccolo PC
Feb 16, 2016
887
995
Yep. Sent them an email yesterday to know what's going on...

I did receive a my 4x to 16x powered riser this week and will try to test them this week-end.
 
  • Like
Reactions: Ceros_X

zovc

King of Cable Management
Editorial Staff
Jan 5, 2017
852
601
Hey guys, I'm looking for a 16x to 8x riser cable. I found one on Newegg but it doesn't specify if it's 'powered' or not, and it doesn't have an external adapter to draw power from.

Also, it'd be better if the power connector was a SATA header rather than a Molex, if that's possible.

@aquelito, How did your riser turn out? I probably wouldn't suffer that much for using a 4x adapter if I had to.
 

Kmpkt

Innovation through Miniaturization
KMPKT
Bronze Supporter
Feb 1, 2016
3,352
5,826
I'd honestly use vary.technology. Their site is obviously going through some growing pains, but the product is good and it ships quickly.
 
  • Like
Reactions: Ceros_X

aquelito

King of Cable Management
Piccolo PC
Feb 16, 2016
887
995
I just briefly tested it : no glichtes, no issues, PCIe Gen 3 !

They have 8x to 16x powered risers, Molex or Sata. Just specify which terminal you want in the comments on the product page.
Do not hesitate to contact them by email, very responsive staff.
 

blubblob

SFF Lingo Aficionado
Jul 26, 2016
94
123
Would it be possible for one of you who bought a riser from vary.technology to provide some pictures or even better measurements that detail the thickness and stiffness of the cable? That would be much appreciated.

It seems like there are only renders on their site. the single picture in this thread is the only real picture i could find of them using google. :D (That might be due to their site being unresponsive/not displaying any images for me though.)

I'm interested in whether the cable is multilayered (which usually results in buckling problems with excessive bends, especially those that don't even themselves out (like U bends, as opposed to S bends)).
 
  • Like
Reactions: zovc

CC Ricers

Shrink Way Wielder
Nov 1, 2015
1,798
1,917
It would be great to give vary.technology more support, make more people aware of them and help their business. These guys seem to know what they're doing and offer a ton of options. For anyone looking to make a build with the GPU laid out back to back, it's a more affordable alternative to the 3M risers.
 
  • Like
Reactions: zovc

zovc

King of Cable Management
Editorial Staff
Jan 5, 2017
852
601
I definitely think I'm going to be trying to give them a chance in the near future. I ordered a riser just before realizing they exist, so in the meantime I'm going to try to make it work. But actually being able to get a 'made-to-order' cable sounds phenomenal.
 

Chrizz

Average Stuffer
Jan 23, 2017
74
80
I found a video that bend-tests the 3m cable. The tested cable wasn't the PCI-e cable though, and the guy mentioned they used an unconventional frequency, even though he mentioned earlier that problems usually arise at specific frequencies. Also, the camera continuously zoomed in on the oscilloscope which doesn't help reading it. At 2:50 there is a drop in amplitude on the oscilloscope, I don't know what would happen in a real life scenario if this was a PCI-e cable connected to a GPU.
I guess the main point of the video was to show that it still works with an insane amount of bends, which makes me more confident that this cable will work with the two bends and one twists I will need.
 

RogueGiant

What's an ITX?
Apr 27, 2017
1
0
Has anyone ever tried an AMD RX480 (Reference Design) on a PCIe Riser cable?
I want to install one of these graphic cards (water cooled) with a riser cable but I am concerned about the PCIe power draw from the Motherboard.
Thanks.
 

EdZ

Virtual Realist
Gold Supporter
May 11, 2015
1,578
2,106
Do the 3M risers really allow for the sharp bends and twists that they show in their promotion video?
From the spec sheet:
Static (one-time) fold Bend Radius: 1.0 mm Bend Types: 45°, 90°,180° (flat fold)
Meets signal integrity specifications (impedance, s- parameters) after a onetime 1 mm minimum bend radius fold
The Twin-axial cable thickness is 1.25mm, so it can indeed fold completely flat back on itself, as long as you do not unfold and refold the cable in this manner.
 
  • Like
Reactions: Phuncz and yawacool

msystems

Airflow Optimizer
Apr 28, 2017
276
438
I am using the Sintech, I chose it because it appears to be shielded and the length was only 30mm which I figured would be helpful. All of my tests show framerate is within margin of error. Also the cable is very stiff. After bending it, the shielding kind of popped out at the ends. If you get this one, try not to bend the cable where it attaches to the board..only bend it in the middle to start. The best way I found to bend it was using two ballpoint pens.
 
Last edited:

msystems

Airflow Optimizer
Apr 28, 2017
276
438
I want to address the potential power issues in detail with these risers.

The main question is: When is it OK to use an unpowered riser? (and when is it not?)

The graphic from this thread illustrates what exactly an unpowered riser is lacking (See the 'Rated power' section below):

So we know a typical unpowered riser does not have thick enough wire to deliver the PCI-E x 16 spec of 75w to the card. (err, well except for this thing)

However this is not necessarily a problem that the riser cannot deliver the full current. If there is a Pci-E cable attached, it usually solves this issue. Even if this means pushing the draw from the 6/8 Pin Pci-E cable past it's artificial 75w / 150w spec. That's because those cables can actually deliver much more than that and it's up to the PSU to enforce the actual limit on their power draw. See this thread for example.

Have learned several interesting facts about the PCI-E plugs and power after researching it this past week, that is worth sharing

The PCI-E connecter has 6 pins. They are not all used for power. For the 6 pin connector, Pins 1 & 3 are 12V & each can carry 8 Amps. Pin 2 by spec is not connected, although some PSU manufacturers do add a 12V line there. Pins 4 & 6 are Com return lines. Pin5 is Com for sensing. Using 2 lines, you get 12V*8A*2= 192Watts, much over the required 75Watts. With an 8 pin PCI-E connector, 2 Com lines are added (4&8) not a 12V & Com. There, Pins 1,2,3 are 12V, Pin 4 is a Com for the 8 pin connector sensing, Pin 5,7,8 are Com return lines, & pin 6 is for the 6 pin connector sensing. Using that config, 12V*8Amp*3=288 Watts, much over the required 150Watts.

Since Mar 2005, the molex pins are required to be "HCS" rather than "Std", which each carry a max of 11Amps. So properly made, an 8 pin PCI-e can supply 12V*11Amps*3lines=396Watts of power for the graphic cards.

So as long as the load distribution can be altered to the PCI-E connector, there shouldn't be an issue. To really get an idea of whether an unpowered riser is a good idea or not though, we need to go through each possible scenario to examine the power demands on both the riser and the Pci-E connector:



1) Very Low Wattage card (<35w). Maybe a specialty card or some type.
- Seems like it will work. Maybe the question is why would you be running this since modern integrated graphics seem to annihilate this category of GPU.

2) Low-Wattage card (35w-75w) with NO Pci-E 6-pin (Some GTX 1050 tis, Fermi budget cards, ect)
- Ok, obviously this just seems like a very very bad idea and it seems like it would cause major issues whenever the card tries to do anything except idle. These cards can and do pull all the way up to 75W.

3) Mid-Wattage card (75w) with 6-pin (other GTX 1050 tis)
- This is where it starts to get interesting. Although the 6-pin PCI-E can supply up to 75w, you have to wonder how much is being drawn through the riser. According to benchmarks without a riser the load was more like a 50/50 split between the motherboard and the PCI-E 12v. See graphic below:


As you can see the draw was 35.6w peak from the motherboard. However with the riser installed, note that the 50/50 proportion no longer applies because, if possible, the card will draw more power from the PCI-E line as needed due to resistance from the riser. The draw would clearly be within spec. Therefore, should not be an issue.

4) Unpowered riser, Upper Mid-Wattage card (120-150w) with single Pci-e 6-pin (RX480, GTX 1060, ect)

The RX480 can draw 150w, but a lot of these cards have only a 6-pin adapter. So it actually needs the whole 75w from the mainboard. Just by design, this is is extremely bad and here is what Toms said about it:

AMD’s Radeon RX 480 draws an average of 164W, which exceeds the company's target TDP. And it gets worse. The load distribution works out in a way that has the card draw 86W through the motherboard’s PCIe slot. Not only does this exceed the 75W ceiling we typically associate with a 16-lane slot, but that 75W limit covers several rails combined and not just this one interface.

That's an AVERAGE of 82w from mainboard and peak 155w. There are documented riser issues with this card, even without a riser actually. Using this card with an unpowered riser will push your 6-pin Pci-E draw way out of spec, we can estimate that it will be pulling 130w avg., instead of 75w.

The GTX 1060 has a similar setup, but not as severe since it is only a 120w TDP card:

The Pci-E connector should be able to handle this even though it will be pushed out of spec. I did a few searches didn't find many issues with this card and risers.


5) Unpowered riser, High-Wattage card (150-180w) with single Pci-E 8-pin (GTX 1070, GTX 1080)
-For a 150w GTX 1070 card, the entire demand should be easily available from the Pci-E line and should have no issuesl. However for a 180w card, an additional 30w maximum is needed from the motherboard, in theory.

Tests of a GTX 1080 without a riser show the following breakdown:


There is a little mainboard power draw there, but it should be no real issue up to the rated 180w TDP because it is almost within the 8 pin Pci-E spec. Actually I have a 1080 and have not noticed any issues so far with an unpowered riser so I assume the Pci-E connector is picking up all the slack. Now if the above test was run with a riser, we would expect to see 150-170w average on the Pci-E wire with up to 300w peaks which is only a bit out of spec. Overclocking however could put the Pci-E connector further out of spec.

6) Unpowered riser, Very High Wattage card (180w-250w) with two 8-Pin Pci-E (OC GTX 1080, GTX 1080TI, Titan XP)
-With two 8-Pin Pci-E cables, these cards can draw 300w without even needing anything from the motherboard and shouldn't have any power issue.


Summary

As long as the rated TDP power can be entirely supplied through the PCI-E inputs, there shouldn't be any issues with a given card
, but more testing is needed on cards where a significant (>30w) part of the TDP power was designed to be supplied through the motherboard (and therefore the riser). For those cards, it seems like it will depend on both the PSU's ability to supply the required current deficit and whether the card's circuitry will allow this re-distribution of power draw. The 6/8-pin Pci-E connector 75w and 150w limits do not seem to be strictly enforced by PSUs. Those are just a guideline that is up to the graphics card to self-enforce. However cards where more than 35 watts of motherboard power is actually needed (due to circuitry design) should be definitely avoided without a powered riser.

At this time, the RX480 seems to have the worst problem.
 
Last edited: