Best PCIE 1x risers

A forum for discussing FAH-related hardware choices and info on actual products (not speculation).

Moderator: Site Moderators

Forum rules
Please read the forum rules before posting.
MeeLee
Posts: 1375
Joined: Tue Feb 19, 2019 10:16 pm

Best PCIE 1x risers

Post by MeeLee »

I bought 3 PCIE 1x risers from amazon, and they're the best from all the ones I've ever used.
These are ribbon based PCIE risers, not USB riser.
So they don't have the latency issues USB risers have.
On an RTX 2080 I get 1,48-1,52M PPD, vs ~1,28-1,43M PPD on a USB style riser in Linux.

They're also PCIE 3.0 compatible.

If you're running anything below an RTX 2080 Ti, via PCIE 3.0 1x slot, these are the cables to use:
Image

Too bad I purchased the last one on Amazon. They're going for $12.

My RTX 2080 Ti does seem to be constrained by the 1x bandwidth.
On a 4x/8x/16x slot it gets 2,3M PPD, on a PCIE 3.0 1x slot with this riser, it only gets 1,9-2,1M PPD.

An alternative to this riser, are these cheaper version 1x to 16x riser cables:
Image
https://www.amazon.com/Riser-Ribbon-Ext ... 07Q2SCVDL/

They come with dual ribbon cables, but the max ribbon length, if you look around, is 20cm or below (7,5").
These are good if the card is mounted in close proximity to the motherboard PCIE slot.
And on older PCIE 2.0 interfaces you can double up these cables for longer length using standard 1x risers:
Image

On PCIE 3.0 I've had my system error out, when doubling these cables.
Even after each individual passed the check, they still gave errors. But only when running at PCIE 3.0 speeds.
ari2asem
Posts: 21
Joined: Sat Jun 08, 2019 1:07 am

Re: Best PCIE 1x risers

Post by ari2asem »

I use this ones

https://www.highflow.nl/hardware/videok ... black.html

pricey, but also quality. right now using them in 2 compute rig, 1 rig for FAH, 1 rig for gpugrid.net
bruce
Posts: 20910
Joined: Thu Nov 29, 2007 10:13 pm
Location: So. Cal.

Re: Best PCIE 1x risers

Post by bruce »

Has anybody connected a USB riser to an actual USP port, or does it have to use the PCIe 1x connection on the MB?
I have a M/B with two 16x slots plus a 1x slot but it's between the 16x slots so I can't use it if I have a double-wide GPU in the 16x slot ... unless I can find a right-angle 1x male that's extremely low to fit under the GPU.
foldy
Posts: 2061
Joined: Sat Dec 01, 2012 3:43 pm
Hardware configuration: Folding@Home Client 7.6.13 (1 GPU slots)
Windows 7 64bit
Intel Core i5 2500k@4Ghz
Nvidia gtx 1080ti driver 441

Re: Best PCIE 1x risers

Post by foldy »

No! USB risers are not really USB, they just use the USB cable electrical wires.

The black risers are shielded the grey ones not.

For riser power it is not recommended to use SATA power plugs but only molex power plugs. Because for GPU with high watt usage that is too much for SATA and the plugs can burn and short circuit then. https://i.imgur.com/Xg2wvF1.png

There should not be any difference between pcie x1 riser using ribbon or USB cable as USB is only used as electrical wire.

@MeeLee: Did you compare TPF for the same work units? I guess you had different projects when comparing USB or ribbon risers.

Recommended x1 risers by miners: (but throw away the sata power plug)

https://www.amazon.com/MintCell-6-Pack- ... 5685f41790

https://www.amazon.com/Ubit-Express-Gra ... 55b29385f3

https://www.amazon.com/MintCell-Multi-P ... uage=en_US
MeeLee
Posts: 1375
Joined: Tue Feb 19, 2019 10:16 pm

Re: Best PCIE 1x risers

Post by MeeLee »

ari2asem wrote:I use this ones

https://www.highflow.nl/hardware/videok ... black.html

pricey, but also quality. right now using them in 2 compute rig, 1 rig for FAH, 1 rig for gpugrid.net
Yes, I've seen them too.
In fact, I own some 8x risers from that brand, and those riser cables are well soldered, and you can pull the cables, or stuff em through a small case opening smaller than the ribbon cable, without really damaging the cables!
They're really well made for tough situations, or risers that often need to be adjusted.

Though I try to plug my GPUs straight in the full size GPU slots on the motherboard, and use risers for the 1x slots.
That way I have best performance.

Should a case arise where I have no other option than to raise such GPU, I would raise them only a little; and recommend these:
https://www.amazon.com/gp/product/B076HM5HDB/
Image

They are of lower quality, but since the lanes are doubled up (2 ribbon cables per riser), I haven't yet had any issues with them.
On PCIE 2.0 you can even double them up, basically giving you up to 40cm (15") of length.
On PCIE 3.0 a single one is all that works. Doubling them up gives errors on some motherboards.

That's the only issue with the cheaper ones, you don't have them beyond 20 cm (7,5") online.

For $15 you're better off (especially when you're not really installing/uninstalling them often) than for $40-85.
bruce wrote:Has anybody connected a USB riser to an actual USP port, or does it have to use the PCIe 1x connection on the MB?
I have a M/B with two 16x slots plus a 1x slot but it's between the 16x slots so I can't use it if I have a double-wide GPU in the 16x slot ... unless I can find a right-angle 1x male that's extremely low to fit under the GPU.

@Bruce:
For low profile risers, where PCIE 1x slots are inaccessible due to a GPU blocking them, I use the one below.
Just use the motherboard riser piece, and USB cable, but discard the riser board on the other end; and use a full size (16x) riser board (the one that connects to the GPU) instead.
The 1x or 4x riser boards that come with it are useless for GPUs (unless you have a GPU that has a 4x slot, they sell them too)
The motherboard 1x riser plug fits just under a GPU blocking the slot!

https://www.amazon.com/gp/product/B07N38Y799
Image



foldy wrote: For riser power it is not recommended to use SATA power plugs but only molex power plugs. Because for GPU with high watt usage that is too much for SATA and the plugs can burn and short circuit then. https://i.imgur.com/Xg2wvF1.png

There should not be any difference between pcie x1 riser using ribbon or USB cable as USB is only used as electrical wire.
@Foldy,

The problem with SATA cables powering risers is twofold.
1- My experience is that the SATA cables that use only 2 wires (one +12V, one ground) are the ones that cause most errors.
I've not had any card burn out with them, however, I do have cards that hesitate, and stop the WU (freeze) due to a lack of power.
It appears most cards work fine with 2 SATA wires up to 128Watts. Any higher and errors occur, and WUs stop.
I've ran many weeks these cards, and it doesn't seem to matter what RTX card I put in there. 2060, 2070, 2080; 128W is the limit for these cards before they crap out due to riser voltage errors.
I can only presume, once a card uses more than xx amount of watts (125-129), it'll start pulling more than 20W on the riser.
An online document states a maximum of 1,5A on these plugs, which is 18-20W.

2- SATA cables with 3 wires (Yellow, Black, Black) or more (+red and/or Orange) usually are ok.
I've ran many GPUs for many hours at 170-225 Watts just fine with them.
These types come with 2 ground wires, that should alleviate a 'ground lift' issue common with ground wires that are connected to circuits under load.
Most PSUs seem to keep the 12V pretty stable, but have trouble keeping the ground wire at 0V (it usually rises to +2V or more under load, causing the overall voltage difference between the leads to be less than 10V, and the cards to crap out).
This (ground lift) is the main cause for hardware errors.
Ground lift, is also why an 8pin PCIE plug is actually a 6 pin plus 2 more grounds; allowing it double the power, just for adding 2 extra ground wires (rather than adding a +12V and ground).
The Windows and Linux driver is pretty quick to recover from these errors, but FAH client does not.

Another issue to keep in mind, is that there often are 3 or 4 sata connectors on a single PSU lead.
What people do, is try to connect more than 2 riser boards on such a lead, and then it craps out on them.
I usually connect up to 1 riser per lead (even if there are 3 or 4 Sata plugs on them), or try to divide 2 boards with one lead, and then split an additional 4 pin molex power lead to each riser board (powering each GPU riser end with both SATA and MOLEX on the riser cards accepting MOLEX and SATA connectors like pic below).

Image


You can't go over 75 Watts per SATA/MOLEX lead coming from the PSU, as each lead has only 2-3 ground wires. But even using 2x risers (35W each), some power spikes can destabilize some GPUs enough for the GPU to stop folding.
Perhaps the reason I don't see fire, is because I 'power cap' the cards, rather than increasing the power draw over stock (overclocking).
It may make it possible to use methods gamers can't use.

One solution to ground lifts due to power spikes, is to solder a larger capacitor on the riser board, to catch these 'voltage fluctuation', however, I have little time to test or research this now.
I will want to do this in the near future when my second server is online.


I've ran many WUs on USB riser boards, but I still feel that the ones with ribbon cables are faster and better than the USB counterparts.
While the slowdown may not be 'latency' like I thought, I've done some research into the matter, and found that:
- USB risers use USB 3.0 cables that have 9 wires, with at least 8 individual pins, where of 4 data, 2 clock, 1 ground, and possibly 1x +12V (not confirmed), and the last one I don't really know, but expect it to be a -3.3V wire or possibly non-assigned, couldn't find any info on that one).
- PCIE 1x has 18 pins per side, or 36 pins in total; Some are multiples (like +12V, GND, +3V), and some aren't really used by risers or motherboard for folding (Hot plug presence detect, Power Good, Link Reactivation, Reserved...; these aren't really necessary connections)
Consolidate most of those doubles and remove unused ones, and a PCIE 1.0 ribbon has about 13 individual wires. This is 4 to 6 more than in a USB riser cable.
But what I think makes the biggest difference, is that:
PCIE 1.0 has 5x +12V, and 9x ground pins, (vs 1x ground + a maximum of 1x 12V on USB risers) which is why I believe ribbon cables to be superior to usb risers, and why the overall score measured in 2 days average is higher with ribbon cables than with USB.
This may not be the case with mini/micro PCIE ribbons, like the ones found on mobile devices, where the ribbon does not meet the requirements to push 25/35Watts, but it does with the standard PCIE ribbon cables.

If you look at the difference between PCIE 1x and 16x, aside from data channels, you'll find mostly an added (more advanced) reference clock added at PCIE 4x, and the rest from 4x onward, is just ground wires.
Seeing that each data channel is actually a differential pair (https://en.wikipedia.org/wiki/Differential_signaling), it means that the data channel doesn't need the ground to return transmit.
It uses a +3V to, and -3V return wire, which means that the ground wire is just for power purposes.

Ribbon connectors can draw the same amount of power from a FDD plug, as Risers can from a SATA/MOLEX connector (up to 35W), but they don't have to; as they're already getting 25 out of 35 Watts from the PCIE 1x slot on the motherboard.
A ribbon cable with 2 pin jumper plug, allows up to 1 amp (or 12Watts) of power draw; added to the 25W from the PCIE 1x slot, supports the full 35W of PCIE 4x. So these cables can use a 2 wire SATA cable.
And one PSU sata lead can easily power 4 of these type of risers.


USB risers need to get nearly all their power from external connectors, as some sites say the USB wire doesn't carry +12V at all, though others claim that for risers there should be 1x +12V lead coming from the motherboard.
I personally doubt that, as when a Sata connector is unplugged from the riser board, the GPU goes dead, and doesn't try to pull the 35W from the single 12V USB lead, or you'd be seeing 'magic smoke'...
Last edited by MeeLee on Fri Jul 12, 2019 10:01 am, edited 9 times in total.
MeeLee
Posts: 1375
Joined: Tue Feb 19, 2019 10:16 pm

Re: Best PCIE 1x risers

Post by MeeLee »

Seems like the 1x to 16x ribbon cables I purchased are discontinued, and replaced by 1x to 1x open end risers, for $15.
They don't provide power for the GPU (25W max, vs 35W for PCIE 4x), so you'd have to either see if a modern RTX card can run fine from 1x, or if you'll need to find a way to get more power to the GPU through the 2 copper connection points on the GPU riser board (underneath the riser plug; not the copper mounting holes; in picture below).
Image
https://www.amazon.com/GODSHARK-Extende ... 07RWGC441/

I'm not sure if these are ok for folding.
I would probably recommend these $6 ones over the 1x ones of above without the extra power, if 20cm/7.5" length is long enough:
https://www.amazon.com/Womail-Extention ... 073NYTQ7V/
Image

I'm not sure if the 1x slot above is capable of providing more than 25 Watts of power, even if an additional 12V line was soldered on there, as most of the PCIe 4x lanes (minus the PCIE 1x lanes) are there to provide power for the GPU.
Last edited by MeeLee on Fri Jul 12, 2019 8:10 am, edited 3 times in total.
antropofob
Posts: 59
Joined: Mon Aug 22, 2011 8:03 am

Re: Best PCIE 1x risers

Post by antropofob »

Using ribbon riser without power connector=playing with fire.
MeeLee
Posts: 1375
Joined: Tue Feb 19, 2019 10:16 pm

Re: Best PCIE 1x risers

Post by MeeLee »

antropofob wrote:Using ribbon riser without power connector=playing with fire.
If you mean, connecting a 1x riser to a modern (RTX) GPU, then, perhaps yes, perhaps no.
If you mean, connecting an open ended, unpowered 4x riser cable (from a motherboard's 4x, 8x or 16x port) to a modern GPU, then no. There shouldn't be any issue.
Most modern GPUs, including an RTX 2080 ti will run fine from a 4x port from the motherboard, without needing extra power.

I've looked around, but so far I've not found any article testing 1x PCIE ports on modern graphics cards.
I think by now, engineers should have forseen that some people will want to run modern graphics cards from PCIE 1x ports, and perhaps the card will detect this (like it can detect being on a 4x port) and can be ran with a 10W lower TDP from a single 1x slot without external power.
But I'm not sure if it actually will cause any hardware damage, or if it's safe either.
I'd say I'd try, but I don't have enough hardware to do that with.

Edit:
After some looking around, in diagrams, I found this website here: https://pinouts.ru/Slots/pci_express_pinout.shtml
I found that PCIE 16x slot, has more data and ground cables, but not any more +12V cables than a PCIE 1x slot.
In theory, a GPU should work on an open ended, unpowered PCIE 1x slot.
It probably would benefit adding a ground wire, and while you're at it, it wouldn't harm connecting a lead to a +12V wire as well, as most plugs come with +12V and GND (SATA/MOLEX/FDD connectors).
Power wise, a GPU should run from an unpowered PCIE 1x port, and draw at best another 10 or so Watts from an external source; but the main issue is a good ground connection. Adding that, will most surely make the GPU work well, and alleviate the lack of more ground wires in the PCIE slot.

One thing you do need to be aware of, is that for multi PSU systems, the powered ribbon riser needs to be connected to the same PSU as the motherboard.
With a USB riser card, this isn't necessary (as far as I know). There a secondary PSU can power the riser and/or GPU independently from the motherboard.
foldy
Posts: 2061
Joined: Sat Dec 01, 2012 3:43 pm
Hardware configuration: Folding@Home Client 7.6.13 (1 GPU slots)
Windows 7 64bit
Intel Core i5 2500k@4Ghz
Nvidia gtx 1080ti driver 441

Re: Best PCIE 1x risers

Post by foldy »

x1 riser USB cables is same speed as ribbon cables because both use 9 data lanes.

Sata connectors have the problem that its wires are very narrow to each other and on high power load can break and short circuit - only certified for 54 watts.

Molex connectors don't have that problem and certified for 154 watts like pcie 8-pin plug.

That is why millions of miners use the molex or 6/8-pin-plug powered usb x1 risers
MeeLee
Posts: 1375
Joined: Tue Feb 19, 2019 10:16 pm

Re: Best PCIE 1x risers

Post by MeeLee »

The 9 data lanes would contradict what I've read elsewhere.
Usb3.0 only has 9 lanes where of 1 for sure is ground, so it can't use 9 data lanes.

They also don't use the same protocol of USB 3, but a trimmed down pcie protocol, most likely without +12v lanes.
foldy
Posts: 2061
Joined: Sat Dec 01, 2012 3:43 pm
Hardware configuration: Folding@Home Client 7.6.13 (1 GPU slots)
Windows 7 64bit
Intel Core i5 2500k@4Ghz
Nvidia gtx 1080ti driver 441

Re: Best PCIE 1x risers

Post by foldy »

Yes it was luck that the USB 3 cable has enough lanes to transport the pcie x1 signals. And with its good signal and shielding made USB cable a perfect choice for x1 risers. The power is taken from separate power connector so there is no power taken from mainboard. This is perfect for having a mainboard running e.g. 8 GPUs as all power comes from power supplies and not through mainboard so there is no heat issue.
bruce
Posts: 20910
Joined: Thu Nov 29, 2007 10:13 pm
Location: So. Cal.

Re: Best PCIE 1x risers

Post by bruce »

In other words, older/smaller GPU which do not feature a direct power connection cannot use the USB cabled riser and must use a separate 12v connection. The ribbon cables will provide power for GPUs that use power from the MB.

The PCIe standard detects the width (how many data lanes are connected) and a GPU that works on 16x or 8x simply works slower on narrower connections. The driver will send/receive all of the data through whatever connection you have. Bandwidth depends on both the number of data lanes and the connection speed.

If 9 lanes @ PCIe @ 1.0 speed happens to be enough to get the data to/from a "slow" GPU when it needs it, fine. Running 3.0 speeds tends to compensate for the narrower connection, but the amount of slowdown will depend on a comparison of how much data your WU needs compared to the speed and width of the connection.
MeeLee
Posts: 1375
Joined: Tue Feb 19, 2019 10:16 pm

Re: Best PCIE 1x risers

Post by MeeLee »

GPUs without top connector, will need to get their power from an external power supply on the USB riser board; while powered ribbon risers are using 'some' of the maximum power (35W) the motherboard can provide per slot; not all of it.
But in both cases (USB and powered Ribbon) will provide the full 35 watts to the GPU.
USB risers use 16x slot, so they could (in theory) provide up to 75Watts, but most risers are of unknown companies, and each company can do as they please.
So no guarantees that USB risers provide a full 75W to the GPUs, but it is possible.
Unpowered PCIE 1x Ribbon risers, only offer up to 20W per GPU, 4x risers offer 35W, and 8x or 16x risers are rated up to 75W.
The powered part of risers are supposed to take care of the extra juice the GPU needs.
Most GPUs are 4x and 8x compatible, meaning they'll run fine at 35W.
Not sure about 1x (20W) or older cards without top header pins (eg: 6 pin molex connector).
I would presume older cards, I'd look at what they're rated at. Older 75W GPUs might not work (or work well) in a PCIE 4x slot.





Shielding USB VS RIBBON:
Since the length of ribbon cables is very short, outside interference is very little.
Shielding matters for longer lengths.
But sub 20cm/7,5" length ribbons don't need shielding.


There are many benefits why going for USB risers instead of ribbon risers (especially when limited to 20cm (7,5") ribbon lengths.
However, the ribbon riser I've purchased is just as flexible, and long (3ft) as a USB risers, which negates most of the flexibility benefits USB risers have.


But they are good for reducing the stress on the motherboard capacitors, when you're running a lot of PCIE cards.
Like 10 or more!
But for 6GPUs or less it's not a real issue.
Motherboards are loaded with high quality capacitors.
Also, a lot of people hashing/mining use PCIE splitters, splitting PCIE 1x slots into 4. This works really bad (or not at all) for folding.
In such case, USB risers are the only option.


Powered ribbon risers should get only very little power from the motherboard. The motherboard capacitors also help it run more stable, than relying on riser capacitors (which often are cheap and under performing, especially when ran from a SATA 2 lead connector).
You're relying on the motherboard of presumably a known manufacturer, instead of the risers of an unknown Chinese company.
Powered ribbon risers by themselves also don't consume any power, as at best, they're just copper lines with a cap soldered on it.
Riser boards do consume (some) power, though little.




There's very little information out there on USB riser pin-out data, but I've found some info:

More forum users report, that 7 of the 9 leads are known, but the 2 remaining wires is unknown to most, or not mentioned what they are.
My best bet would be to trace back the wires from the USB header of an actual PCIE 1x riser piece, and compare it with the PCIE pin-out data.

Since I don't have access to a pcie 1x riser piece right now, I can only deduct the following:
- Since USB cables are limited in the amount of leads, duplicates of PCIE channels will be erased, like,
- All +12V will be removed from the equation,
- Only 1 Ground wire will remain, only one +3V will remain
- Wake up, hot plug and swap, and power configuration data lines will be erased (risers don't have these options)
- Some test wires will be erased (only used for testing purposes on PCIE cards, not necessary for risers).
- Reserved wires will be removed.
- Paired wires will either both be removed, or both be present. Can't delete 1 lead of a pair.

Which leaves these 9 remaining wires:
Image
Original PCIE 1x data comes from this source.
Thick red lines scrap duplicates, non functional, or obvious unused items.
Thin red lines scrap most likely unused pins for USB (I can be wrong about this); based on their meaning and my knowledge on how risers operate. (eg: they have no recover option, no hot swapping possible, no test wires....

These most likely will be the ones in a USB riser, which includes:
- 2 receiver channels, and
- 2 transmitter channels for data
Which is only 4 channels, the same as PCIE, or ribbon cables.

The remaining wires are probably going to be:
- +3V (for powering riser chips and LED),
- PowerGood, to indicate to the OS that there's sufficient power to power the device. There's a possibility they'll use '11. wake link reactivation' instead.
- Reference clock +
- Reference clock -
- Ground
antropofob
Posts: 59
Joined: Mon Aug 22, 2011 8:03 am

Re: Best PCIE 1x risers

Post by antropofob »

How come those teeny tiny wires used for ribbon risers do not melt/catch fire when subjected to 75W of power draw?
As opposed to big fat cables found on Sata power connectors that are somehow not recommended due to being prone to melting/catching fire?
foldy
Posts: 2061
Joined: Sat Dec 01, 2012 3:43 pm
Hardware configuration: Folding@Home Client 7.6.13 (1 GPU slots)
Windows 7 64bit
Intel Core i5 2500k@4Ghz
Nvidia gtx 1080ti driver 441

Re: Best PCIE 1x risers

Post by foldy »

Because SATA connector splits each power lane into 3 small narrow pins https://imgur.com/Xg2wvF1

It may work great but if you have bad luck then it can break.
Post Reply