PCIE Bifurcation

I've been in contact with John. Nice guy and quick response times.

So I guess my question would be the following: How can you get the raw computation power from those monsters *if* for arguments sake the bandwidth is capped? Doesn't every application rely on the amount of bandwidth available?

Yea, that's a good question. All I know is that even going from PCI-E 8x -> 16x, there wasn't a doubling of performance either or vice versa 1/2 the performance. You can take a look at this analysis here for PCI-E scaling:

https://www.techpowerup.com/reviews/NVIDIA/GTX_980_PCI-Express_Scaling/15.html

I think it really depends on what you're using your system for but for what I use it for it's mainly floating point analysis and matrix manipulation and it's more CPU bound. i.e CPU can't pre-process the data fast enough to feed the GPU.
 
I've been in contact with John. Nice guy and quick response times.

So I guess my question would be the following: How can you get the raw computation power from those monsters *if* for arguments sake the bandwidth is capped? Doesn't every application rely on the amount of bandwidth available?

Btw, are you planning on cutting off one of the DVI connectors from each of the 1070 and water cooling them? The N case only has 2 extension slots.
 
Btw, are you planning on cutting off one of the DVI connectors from each of the 1070 and water cooling them? The N case only has 2 extension slots.

I plan on mounting the gpu's with a bracket sideways and vertical where the power supply would normally be. This makes the GPU DVI ports a non issue. The power supply will be horizontally mounted where the fan on the back of the case would be. At least thats the plan so far. I'll be using Display port extension cables to get the ports to the back of the case and installed right where the second port is on the top next to the power cord socket. Not to mention I'm going to try having 2 additional 5 port USB 3.0 cards right under that. Its gonna be a project for sure.

I'm using a ID Cooling Hunter Duet dual all in one cooler. Typically its for one cpu and one gpu, but I think I'm going to double up on the GPU's and see if I can fit a small Corsair all in one in there as well for the CPU. If its physically possible, I'll get it all in there. I don't have the patience to do my own custom loop for everything in addition to everything else planned.

Thanks for the PCI scaling link. Interesting indeed.
 
Last edited:
I plan on mounting the gpu's with a bracket sideways and vertical where the power supply would normally be. This makes the GPU DVI ports a non issue. The power supply will be horizontally mounted where the fan on the back of the case would be. At least thats the plan so far. I'll be using Display port extension cables to get the ports to the back of the case and installed right where the second port is on the top next to the power cord socket. Not to mention I'm going to try having 2 additional 5 port USB 3.0 cards right under that. Its gonna be a project for sure.

I'm using a ID Cooling Hunter Duet dual all in one cooler. Typically its for one cpu and one gpu, but I think I'm going to double up on the GPU's and see if I can fit a small Corsair all in one in there as well for the CPU. If its physically possible, I'll get it all in there. I don't have the patience to do my own custom loop for everything in addition to everything else planned.

Thanks for the PCI scaling link. Interesting indeed.

Cool! Asetek used to make a GPU/CPU combo as well called the 760 GC but that's no longer available. You should take a look at the Asetek 92mm CPU cooler here:

http://www.ebay.com/itm/Asetek-545L...521196?hash=item3f616dbc2c:g:aS0AAMXQMTlRaG9Z

It fits the back of the NCase since it's also a 92mm and is fairly compact. I bought one but didn't use it yet. I'm building my own custom case to house the hardware. Doesn't it seem interesting that once you have an idea it's always changing! I started off with the NCase but because it didn't fit all that I wanted my plans have expanded!

Edit: You'd also need the narrow ILM as well if you were to get a corsair AIO. Asetek also sells it here which I grabbed as well:

http://www.ebay.com/itm/Asetek-Liqu...742554?hash=item3f686a641a:g:thsAAOSweW5VWhRL

I'm not part of Asetek, I swear! lol
 
As an eBay Associate, HardForum may earn from qualifying purchases.
Cool! Asetek used to make a GPU/CPU combo as well called the 760 GC but that's no longer available. You should take a look at the Asetek 92mm CPU cooler here:

http://www.ebay.com/itm/Asetek-545L...521196?hash=item3f616dbc2c:g:aS0AAMXQMTlRaG9Z

It fits the back of the NCase since it's also a 92mm and is fairly compact. I bought one but didn't use it yet. I'm building my own custom case to house the hardware. Doesn't it seem interesting that once you have an idea it's always changing! I started off with the NCase but because it didn't fit all that I wanted my plans have expanded!

Edit: You'd also need the narrow ILM as well if you were to get a corsair AIO. Asetek also sells it here which I grabbed as well:

http://www.ebay.com/itm/Asetek-Liqu...742554?hash=item3f686a641a:g:thsAAOSweW5VWhRL

I'm not part of Asetek, I swear! lol

Thanks for that, I don't think the 92mm Asetek is going to work due to its radiator thickness. I'm definitely open to it though if it can possibly fit because it should provide better cooling than a Corsair H60 for example. My hunch is though that I'll have to sacrifice the cooling ability for smaller size.

Shouldn't the bracket included with the Asrock x99ITX hold the Asetek or other cooler down just the same the way its doing it for my ID Cooling cooler? Its not perfect as far as looks are concerned, but it just uses force to clamp it down.
 
As an eBay Associate, HardForum may earn from qualifying purchases.
Thanks for that, I don't think the 92mm Asetek is going to work due to its radiator thickness. I'm definitely open to it though if it can possibly fit because it should provide better cooling than a Corsair H60 for example. My hunch is though that I'll have to sacrifice the cooling ability for smaller size.

Shouldn't the bracket included with the Asrock x99ITX hold the Asetek or other cooler down just the same the way its doing it for my ID Cooling cooler? Its not perfect as far as looks are concerned, but it just uses force to clamp it down.

Not the asetek ones typically, the bracket wouldn't fit.
 
Soooo....

I installed Windows 10 on a separate HDD just to see the results compared to Windows 7. My score increased over 4000 points in Firestrike (stock settings - no GPU overclocking). Too bad I despise Windows 10 with a passion. The most notable difference in the monitoring section if you compare both of my scores is that Graphics test one fully utilizes my GPU in Windows 10 and pushes up the usage to the max. In Windows 7 the GPU usage drops massively in that test, hence the lower score. The curves are very similar otherwise. This basically murders my theory that there was some bandwidth cap that I was hitting. Shoot for the sky Chemist_Slime, shoot for the sky!

Perhaps it is noteworthy to mention that HWinfo in Windows 10 recognizes that the cards are operating at 5.0GT/s, whereas HWinfo in Windows 7 recognized the cards as running at 2.5GT/s.

xdXWfoe.jpg


9EjLzkz.jpg



VOwHqW4.jpg
 
Last edited:
It seems to also recognize your vram as 8Gb as well

How did you know that without seeing my GPU-z on Windows 10? ;) It was GPU-z that detected my VRAM as 4GB in Windows 7, however, you happen to be absolutely correct! Windows 10 properly detects the correct amount of VRAM, so that could make all the difference in the scores. I think a good question for this finding is "WTH?". Why would Windows 7 not detect my VRAM properly?

There is no weird stuttering in Battlefield 4 on Windows 10 by the way. Everything appears to be running pretty much perfect. I guess I'll just have to play my games on Windows 10 and do everything else on Windows 7


XLzYwMd.gif
 
Last edited:
How did you know that without seeing my GPU-z on Windows 10? ;) It was GPU-z that detected my VRAM as 4GB in Windows 7, however, you happen to be absolutely correct! Windows 10 properly detects the correct amount of VRAM, so that could make all the difference in the scores. I think a good question for this finding is "WTH?". Why would Windows 7 not detect my VRAM properly?

There is no weird stuttering in Battlefield 4 on Windows 10 by the way. Everything appears to be running pretty much perfect. I guess I'll just have to play my games on Windows 10 and do everything else on Windows 7


XLzYwMd.gif

I saw it in your previous post on HWINFO. So which splitter and extender cables are you using exactly for this setup? seems like you also tried a few like me.
 
So many pages but no time: can someone summarize? Looks like you all were able to get this working, what did it take?
 
Hi team, it's been a second but I had a chance to run some compute benchmarks (CompuBench OpenCL) so I figured I'd check back.

First off, someone asked about the riser I was using. It's the Lian Li PW-PCI-E38-1. It's $80, so not cheap, but not as crazy expensive as the 3M risers. It seems it's out of stock everywhere now, though you could try https://www.amazon.com/Lian-Li-Accessory-PW-PCI-E38-1-Shield-Retail/dp/B01AV0R5TM

One potential problem I see with this riser is that the connection to the graphics card has a 90-degree bend. I'm inspired by jb1's dual-nanos-with-one-sideways-in-a-DAN-A4 idea, and I think this bend is going to get in the way, so I may be stuck reverting to an expensive 3M cable to be able to snake it through the case.

As far as benchmarks, I ran CompuBench on both Nanos to see the difference between the one connected to the PCIE slot and the one connected to the riser cable. As you'd expect, the cable doesn't introduce hardly any penalty. This is with the +50% maximum power setting on the AMD drivers, so above stock.

Nano in PCIE slot:

Through_slot.PNG


Nano through riser cable:

Through_cable.PNG


I'm not sure that this even counts as a penalty - it may be within margin-of-error. FYI, these scores compare very nicely to typical Nano scores, so everything looks pretty good so far.

I'm looking forward to seeing if I can cram all the parts lying out on my desk into the A4 when it arrives!
 
As an Amazon Associate, HardForum may earn from qualifying purchases.
Hi team, it's been a second but I had a chance to run some compute benchmarks (CompuBench OpenCL) so I figured I'd check back.

First off, someone asked about the riser I was using. It's the Lian Li PW-PCI-E38-1. It's $80, so not cheap, but not as crazy expensive as the 3M risers. It seems it's out of stock everywhere now, though you could try https://www.amazon.com/Lian-Li-Accessory-PW-PCI-E38-1-Shield-Retail/dp/B01AV0R5TM

One potential problem I see with this riser is that the connection to the graphics card has a 90-degree bend. I'm inspired by jb1's dual-nanos-with-one-sideways-in-a-DAN-A4 idea, and I think this bend is going to get in the way, so I may be stuck reverting to an expensive 3M cable to be able to snake it through the case.

As far as benchmarks, I ran CompuBench on both Nanos to see the difference between the one connected to the PCIE slot and the one connected to the riser cable. As you'd expect, the cable doesn't introduce hardly any penalty. This is with the +50% maximum power setting on the AMD drivers, so above stock.

Nano in PCIE slot:

View attachment 10933

Nano through riser cable:

View attachment 10934

I'm not sure that this even counts as a penalty - it may be within margin-of-error. FYI, these scores compare very nicely to typical Nano scores, so everything looks pretty good so far.

I'm looking forward to seeing if I can cram all the parts lying out on my desk into the A4 when it arrives!


Just to keep track, which bifurcation splitter card are you using?
 
As an Amazon Associate, HardForum may earn from qualifying purchases.
That's a good question. I'm using the RSC-R2UT-2E8R.

But this is a problem because this splitter's PCB is definitely too long to fit in the Dan A4.

Last I checked, the flexible Amerirack splitter, the only other one I've seen tested on this thread, was no longer being sold.

Do you know of another flexible (or low-profile) splitter?
 
That's a good question. I'm using the RSC-R2UT-2E8R.

But this is a problem because this splitter's PCB is definitely too long to fit in the Dan A4.

Last I checked, the flexible Amerirack splitter, the only other one I've seen tested on this thread, was no longer being sold.

Do you know of another flexible (or low-profile) splitter?

Breathless said he bought one from them without issue. Did you try to contact them recently?
 
A quick update everyone: With the latest 3.63 bios (thanks breathless ), I am now able to run bifurcation successfully again with a 2950 V4 Broadwell 105W TDP and Dual 970 GTXs 150W TDP/each with the ameri-rack Gen3 splitter on Ubuntu 14.04. I'm currently running both GPUs and the CPU on a single loop with an EK Coolstream XE 60mm thick, 120mm radiator, single Gentle Typhoon 2150RPM PWM and an Alphacool Eisbaer. At full tilt running FAH, temps hover around 61-62C on the GPUs, and around 58C on the CPUs. I will be stress testing this weekend running a 295x2 500W on the loop and my 2699 V4 which has a TDP of 165W for a total combined TDP of 665W. I also have in hand another Coolstream 60mm XE to run it sandwich style with another GT in the middle. If this pans out, I'll be swapping out the 295X2 for two Titan XPs (will have to single slot mod) which also has a combined TDP of 500W, hence why I'm testing with the 295x2. I'll also eventually swap out the Eisbaer for a modified Swiftech Apogee II with an EK low-profile cooling top. Then off to the races I go cramming it into a 3D printer proto-case that if looks good, will have it CNC'ed. Some pics for your viewing pleasure is attached.

Of note on Wnidows 10: I could not run this without getting the BSOD "video_tdr_failure" each time I load the drivers even on a clean install. If anyone has any tips on this, your feedback would be appreciated.

TsJagJe.jpg
7klqKCU.jpg
ftFgubk.jpg
XJTRNPI.jpg
 
Last edited:
That's great chemist_slime ! I registred on the forum just for your project as I had a similar one in mind :)

I would like indeed to fit two GTX 970 into a 1U or 1.3U rack mount chassis for GPU rendering purposes.

I just purchased at a good price an Asrock Fatal1ty Z170 Gaming-ITX/AC, and an Amerirack ARC1-PELY423-C5V3 splitter.
For the record, I had no issue whatsoever to purchase it directly from Amerirack, whose sales manager have been very responsive.

If you have time, could you please advise the total dimensions of both GTX 970, with and without the EK-FC Terminal DUAL Parallel 1-Slot ?

BTW, are you cooling both GPUs and CPUs with a single 120 mm radiator ?

Just to sum things, besides chemist_slime, here are the other successful case studies :

MaximumBurrito :

ASRock Fatal1ty Z170 Gaming-ITX/ac with bifurcation BIOS support enabled
RSC-R2UT-2E8R
Lian-Li risers

Breathless

ASRock X99E-ITX/ac with 3.63 bios
RSC-R2UG-A2E16-A
3M Risers
 
That's great chemist_slime ! I registred on the forum just for your project as I had a similar one in mind :)

I would like indeed to fit two GTX 970 into a 1U or 1.3U rack mount chassis for GPU rendering purposes.

I just purchased at a good price an Asrock Fatal1ty Z170 Gaming-ITX/AC, and an Amerirack ARC1-PELY423-C5V3 splitter.
For the record, I had no issue whatsoever to purchase it directly from Amerirack, whose sales manager have been very responsive.

If you have time, could you please advise the total dimensions of both GTX 970, with and without the EK-FC Terminal DUAL Parallel 1-Slot ?

BTW, are you cooling both GPUs and CPUs with a single 120 mm radiator ?

Just to sum things, besides chemist_slime, here are the other successful case studies :

MaximumBurrito :

ASRock Fatal1ty Z170 Gaming-ITX/ac with bifurcation BIOS support enabled
RSC-R2UT-2E8R
Lian-Li risers

Breathless

ASRock X99E-ITX/ac with 3.63 bios
RSC-R2UG-A2E16-A
3M Risers

Not a problem. The height of the terminal is 25mm, the bottom of the card slotted into the ameri-rack to the top of the block without the terminal is 120mm. So with the terminal it's around 145mm. The stacked card width is roughly 40mm and the length of the card is around 170mm. Hope that helps!

Edit: Yes, cooling both GPUs and CPU on a single loop 120mm. I just swapped out the 120mm EK Coolstream XE 60mm thick rad for the Alphacool Nexxos Monsta 86mm thick rad and temps have dropped 5-6 C at full tilt to 64-65. With push and pull, it drops another 4 C to around 61. My 295x2 EK waterblock comes next friday so i'll be testing that in lieu of the 970x2 for TDP reasons and they add another 200W to the loop for a total TDP of 665W on the loop. If temps are stable around 75 - 77 i'll be content. Otherwise I have an Alphacool 92mm x 45mm rad i'll add to the loop to see if i can hit those goals. If this works out, then it's case time!
 
Last edited:
If you want two GPUs in 1U, there are plenty of solutions for that already. You can pick up Supermicro 2011 barebones that offer that for about $200 on eBay (case/board/PSU).
 
Not a problem. The height of the terminal is 25mm, the bottom of the card slotted into the ameri-rack to the top of the block without the terminal is 120mm. So with the terminal it's around 145mm. The stacked card width is roughly 40mm and the length of the card is around 170mm. Hope that helps!

Thanks a lot ! Just one last dimension if you don't mind :) therminal's height without the fittings.
http://i305.photobucket.com/albums/nn203/hagoun/Sans titre-1.jpg


If you want two GPUs in 1U, there are plenty of solutions for that already. You can pick up Supermicro 2011 barebones that offer that for about $200 on eBay (case/board/PSU).

Thanks but my project is for personal use. I alreadry have most of the hardware :)
I ended up purchasing this chassis (which looks a bit more attractive than regular ones :):

http://www.istarusa.com/istarusa/products.php?model=D-107#.WE0lsfnhCUk
 
Hi - quick update on my side, my experience is pretty similar: With Ubuntu 16.10 and using the an i6800k, 3.63 Bios and an Amerirack ARC1-PERY423-C10 splitter I can use two GTX-1080s on the x99e-itx/ac perfectly well. Wasted too much time trying to get Windows 10 to work, making various hypotheses such as broker splitters/signal cross-talk etc but the reality is the splitter is just fine, the drivers are broken. As soon as two GPUs are activated in win10, the display locks up. The ASRock guys in Taiwan managed to replicate this and had the exact same behaviour BUT they were able to use dual GTX-980s just fine! So I logged a ticket with NVidia support but after a lot of back and forth in order to get through to someone knowledgeable I got "Sorry we do not support custom risers" ...

Long story short, Windows drivers just don't work for me for this setup and looks like I am not the only one... Any more insights or ways to replicate/debug would be useful (I don't even get a BSOD, just a mouse cursor freeze in Win10...)

A quick update everyone: With the latest 3.63 bios (thanks breathless ), I am now able to run bifurcation successfully again with a 2950 V4 Broadwell 105W TDP and Dual 970 GTXs 150W TDP/each with the ameri-rack Gen3 splitter on Ubuntu 14.04. I'm currently running both GPUs and the CPU on a single loop with an EK Coolstream XE 60mm thick, 120mm radiator, single Gentle Typhoon 2150RPM PWM and an Alphacool Eisbaer. At full tilt running FAH, temps hover around 61-62C on the GPUs, and around 58C on the CPUs. I will be stress testing this weekend running a 295x2 500W on the loop and my 2699 V4 which has a TDP of 165W for a total combined TDP of 665W. I also have in hand another Coolstream 60mm XE to run it sandwich style with another GT in the middle. If this pans out, I'll be swapping out the 295X2 for two Titan XPs (will have to single slot mod) which also has a combined TDP of 500W, hence why I'm testing with the 295x2. I'll also eventually swap out the Eisbaer for a modified Swiftech Apogee II with an EK low-profile cooling top. Then off to the races I go cramming it into a 3D printer proto-case that if looks good, will have it CNC'ed. Some pics for your viewing pleasure is attached.

Of note on Wnidows 10: I could not run this without getting the BSOD "video_tdr_failure" each time I load the drivers even on a clean install. If anyone has any tips on this, your feedback would be appreciated.
 
Hi - quick update on my side, my experience is pretty similar: With Ubuntu 16.10 and using the an i6800k, 3.63 Bios and an Amerirack ARC1-PERY423-C10 splitter I can use two GTX-1080s on the x99e-itx/ac perfectly well. Wasted too much time trying to get Windows 10 to work, making various hypotheses such as broker splitters/signal cross-talk etc but the reality is the splitter is just fine, the drivers are broken. As soon as two GPUs are activated in win10, the display locks up. The ASRock guys in Taiwan managed to replicate this and had the exact same behaviour BUT they were able to use dual GTX-980s just fine! So I logged a ticket with NVidia support but after a lot of back and forth in order to get through to someone knowledgeable I got "Sorry we do not support custom risers" ...

Long story short, Windows drivers just don't work for me for this setup and looks like I am not the only one... Any more insights or ways to replicate/debug would be useful (I don't even get a BSOD, just a mouse cursor freeze in Win10...)

Glad I'm not the only one with the driver issue! I'm hoping someone can eventually fix this issue but breathless was able to run his without problems on both windows 10 and windows 7. He's using the 1070! I'll know more when I eventually upgrade to the Titan XP to see if that works or not. I read somewhere online that it might be the Physx part of the NV control panel but I haven't had time to test this yet though.
 
Hi there,

Just to let you know that I managed to get it working with the Amerirack ARC1-PELY423-C5V3 bifurcated riser !

- PCIe bifurcated riser : ARC1-PELY423-C5V3
- MB : ASRock Fatal1ty Z170 Gaming-ITX/ac with PCIe riser enabled in BIOS
- i5 6400 with 2x8 Go of Corsair Vengeance LPX DDR4 2400
- GPUs : Gainward GTX 970 Phantom & Gigabyte GTX 750Ti
- OS : Win 10.

I used two flexible risers to plug the GPUs into the ARC1-PELY423-C5V3 riser :

- one 3M riser for the GTX 970
- one Liheat riser for the GTX 750 Ti.

I did not test it for long but it worked great. No display glitches under Windows desktop, while browsing or under Engine Vally that I ran for 15 minutes.

This is great to know that chained risers do not cause any issue ! That's was Amerirack sales representative told to me prior to the purchase.
Glad it is confirmed !

Please refer to the pictures enclosed :

Sans%20titre-1_2.jpg

00000.png


Just for fun, some pictures of the messiest setup ever !
I had to borrow some pieces from my current build. Also took the opportunity to test the Picobox X7-ATX-500 dual brick !

P1013490.jpg

P1013492.jpg


Some close-up pictures of the bifurcated riser :

ARC1-PELY423-CxV3-1.jpg

ARC1-PELY423-CxV3.jpg
 
Last edited:
Hi there,

Just to let you know that I managed to get it working with the Amerirack ARC1-PELY423-C5V3 riser !

- PCIe bifurcater : ARC1-PELY423-C5V3
- MB : ASRock Fatal1ty Z170 Gaming-ITX/ac with PCIe riser enabled in BIOS
- i5 6400 with 2x8 Go of Corsair Vengeance LPX DDR4 2400
- GPUs : Gainward GTX 970 Phantom & Gigabyte GTX 750Ti
- OS : Win 10.

I used two flexible risers to plug the GPUs into the ARC1-PELY423-C5V3 riser :

- one 3M riser for the GTX 970
- one Liheat riser for the GTX 750 Ti.

I did not test it for long but it worked great. Please refer to picutures enclosed :

Sans%20titre-1_2.jpg

00000.png


Just for fun, messiest setup ever !

IMG_2594.jpg
Can't wait to see the finished product! I'm happy to see it's been working for you in windows 10 as well.
 
Can't wait to see the finished product! I'm happy to see it's been working for you in windows 10 as well.

I think that we've seen this working before with GTX 970's in Windows 10, correct? The issues have been with GTX 1070/80's in Windows 10 using this riser combination, correct?

Regardless, happy to see that this is working for you, nice job!!
 
Hi everyone!

Signed up because i've been working on this for a while now - Here are my results:

I am using the ASROCK X99E-ITX/ac with 3.63 bios, dual R9 Nanos and my DIY Riser specifically designed for the Raven RVZ-01 case.

My intent was to create a 2 Gamers 1 ITX System. (Virtualization) This is working spendidly.
So far I tried Battlefield 1 Singleplayer: 1440P Ultra 90FPS + 4K Medium 70FPS @ ~80% CPU.
(5820k @ 4.0GHz. Nanos Power-Limit+15%)
Crossfire is working as well. I didnt do much testing though.

My Riser:
I am using a IDT-9DBL02 Clock Buffer. I picked it because it has lower external component count and complies to PCIe Gen4
It is not powered, but the Asrock Mainboard seems to be able to provide enough juice to its PCIe Slot.
Maybe I will Design a powered one at one point.

I have two of them PCB's left because OSH Park always makes 3... so if anyone is interested...
I would also be willing to make the riser in a different height if someone would be interested...


Cheers!

Chris



Riser.jpg RVZ01.jpg 20161229_183955.jpg 20161229_185628_HDR.jpg 20170114_155344_HDR.jpg
 
Hi everyone!

Signed up because i've been working on this for a while now - Here are my results:

I am using the ASROCK X99E-ITX/ac with 3.63 bios, dual R9 Nanos and my DIY Riser specifically designed for the Raven RVZ-01 case.

My intent was to create a 2 Gamers 1 ITX System. (Virtualization) This is working spendidly.
So far I tried Battlefield 1 Singleplayer: 1440P Ultra 90FPS + 4K Medium 70FPS @ ~80% CPU.
(5820k @ 4.0GHz. Nanos Power-Limit+15%)
Crossfire is working as well. I didnt do much testing though.

My Riser:
I am using a IDT-9DBL02 Clock Buffer. I picked it because it has lower external component count and complies to PCIe Gen4
It is not powered, but the Asrock Mainboard seems to be able to provide enough juice to its PCIe Slot.
Maybe I will Design a powered one at one point.

I have two of them PCB's left because OSH Park always makes 3... so if anyone is interested...
I would also be willing to make the riser in a different height if someone would be interested...


Cheers!

Chris

Wow! Just curious, what radiator and pump are you using for this? What kindof temps are you getting? That looks like a lot of hardware on a single 120mm radiator, if that's what you're using!
 
Wow! Just curious, what radiator and pump are you using for this? What kindof temps are you getting? That looks like a lot of hardware on a single 120mm radiator, if that's what you're using!

During my BF1 testing CPU was sitting at around 80 Celsius and GPU's at around 60, with that Nocuta NF12 at 60% PWM.
That Fan packs a Punch and makes some serious Noise at 100%. At 60% it's quite OK... less then a stock nano.
Radiator is 45mm thick 120mm dont remember the Brand or Model... Cheap though... It was the only one i had around that did fit.

CPU Temps can be lowered I hope:
I upped my core voltage to 1.25 Volts for that quick overclock to 4Ghz... I have now found out that it does 4 Ghz at 1.1V using much less Amps.. ;)
Also I am pretty sure I mounted my CPU heatsink 90 Degree off... lets see if that helps.

Water Temperature Sensore Coming in soon.


Cheers
 
During my BF1 testing CPU was sitting at around 80 Celsius and GPU's at around 60, with that Nocuta NF12 at 60% PWM.
That Fan packs a Punch and makes some serious Noise at 100%. At 60% it's quite OK... less then a stock nano.
Radiator is 45mm thick 120mm dont remember the Brand or Model... Cheap though... It was the only one i had around that did fit.

CPU Temps can be lowered I hope:
I upped my core voltage to 1.25 Volts for that quick overclock to 4Ghz... I have now found out that it does 4 Ghz at 1.1V using much less Amps.. ;)
Also I am pretty sure I mounted my CPU heatsink 90 Degree off... lets see if that helps.

Water Temperature Sensore Coming in soon.


Cheers

Great work! Is there any reason why you're using an X8 -> X16 riser instead of making the board support x16 on both slots from the beginning? Perhaps a limitation of the chip? Also, is there any way to adapt the design of your riser to just make it a solid vertical extender from the motherboard's single x16 without an angle. I currently have the ameri-rack simply because it's the only one out there that worked well with somewhat suitable dimensions for my build. I'm also running a single 120mm radiator albeit being 86mm in thickness and from my current temps I might need to add a second one.
 
Is there any reason why you're using an X8 -> X16 riser instead of making the board support x16 on both slots from the beginning?

Dimensions of that case require a second riser.... god knows why they did that... makes no sense to me at all. I went for the x8 slot because it made routing much easier and improves signal quality if i dont have to route my traces around them unused pins. (Those are high speed signal traces after all, despite of what some chinese "RISERS" might suggest)
I couldnt find mechanical x16 / electircal x8 slots anywhere. That would have been my first choice. Also if i would want to fit a x16 card directly i could just dremel out the rear part of that x8 slot,
All that has nothing to do with that clock buffer though.

I am not quite sure about the Layout and case u are using...
In theory I could imagine a PCB with 2 upward facing Slots and one downward facing slot that would connect via a 2nd PCB to your Mobo... That however would require new routing. (==hours of pain)

C.
 
Dimensions of that case require a second riser.... god knows why they did that... makes no sense to me at all. I went for the x8 slot because it made routing much easier and improves signal quality if i dont have to route my traces around them unused pins. (Those are high speed signal traces after all, despite of what some chinese "RISERS" might suggest)
I couldnt find mechanical x16 / electircal x8 slots anywhere. That would have been my first choice. Also if i would want to fit a x16 card directly i could just dremel out the rear part of that x8 slot,
All that has nothing to do with that clock buffer though.

I am not quite sure about the Layout and case u are using...
In theory I could imagine a PCB with 2 upward facing Slots and one downward facing slot that would connect via a 2nd PCB to your Mobo... That however would require new routing. (==hours of pain)

C.

Ah, I see, thanks for the clarification!
 
Dimensions of that case require a second riser.... god knows why they did that... makes no sense to me at all.
Are you talking about the small straight extension riser that fits between the right angle riser and the video card? That's needed because otherwise the bracket on the card will run into the motherboard's rear I/O shield, and also to allow some room for the removable GPU holder.
 
That's needed because otherwise the bracket on the card will run into the motherboard's rear I/O shield, and also to allow some room for the removable GPU holder.
Thanks... looking again this now makes perfect sense to me.
Although pcie standard only allows for 2 connections, not 3.
PCIe gen4 will even only allow 1 connector afaik. So in the future those riser combinations may really run into signal problems... I even had some rare signal errors with the original Raven Riser.

I think that we've seen this working before with GTX 970's in Windows 10, correct? The issues have been with GTX 1070/80's in Windows 10 using this riser combination, correct?
Maybe this is a power delivery problem.... all those risers I have seen are not powered, essentially pulling twice the juice off the mainboard...
 
I've attached below some images of the 3 PCI-E risers that I've used in conjunction with the Asrock x99e-iTX that did not work. Hopefully some of the images will help. I have yet to take off the heatsink on the Supermicro one, will do that tomorrow in the evening when I get back from work.

I have also PM'ed the tech support person who replied to the PCI-E bifurcation support on the x99e motherboard. He has read the PM but has yet to respond. Asrock support has not yet responded either. I will post results when they do. Let's do this!

MinH0ZV.jpg


qBe76G7.jpg


aVZDNIj.jpg


G8NqMsz.jpg


FKR88Ev.jpg


DYBObj1.jpg
Could you tell me does the supermicro RSC-R2UT-2E8R work? I ask this cuz this splitter fits better than the other two in my sketch.
Thanks
 
Back
Top