PCIE Bifurcation

Thanks a lot for the photo's already spotted some stuff

1) 99% sure it's a PLX chip under the heatsink, you can see all traces go from the PCIe finger to the chip and then from the chip to both PCIe slots.

2) On the same PCB you can find a ICS 9DB883AGILF, which is a bigger brother of the IDT 9DB233 I had selected, so good news :p

Can I request 4 more photos?
a closeup shot of the Chip on the other 2 cards, they are probably the same, on one I can identify the same logo.

And a photo of the back side.
You can see the reference clock signal go upto the first PCIe slot, halfway to the second slot then go into a via, it is probably on the way to the chip on the top left.

Here's a quick collage of my findings
IeauYva.jpg


Edit;
First
Damn you [H] for not saving drafts!!!!!

Second
I've been taking a look at both the 9DB833 and the 9DB233 and trying to figure out what the differences are and why the 833 is used.
833 has
Up to 8 Differential Clock signals available
Each signal can be disabled if nothing is connect, so you could run 1 card in a 2 card riser, I'm not sure what the problem might be if you kept the not used signal on, maybe some weird noise?
More control lines to allow finer adjustments

It seems more like an inventory choice, better have 1 slightly more expensive chip you can use in multiple PCBs and the same control logic then using the right chip for that specific application.

All in all I feel like using the 9DB233 should be fine, I just need more information about how to hook it up, I'm not that big in electronics.
 
Last edited:
Qinx, I’ll take the requested photos tomorrow when I get back home. But don’t you find it odd that none of these splitters worked? Especially the PLX chip one from super micro? That one should support it regardless of the bios supports it or not correct? Otherwise how do cards like the 7990 function in just one 16x slot. I’m guessing it’s something trivial that needs to be linked?

My goal is to have this work perpendicular in the asrock x99. I plan to have two PNY 970 GTX short reference cards (single slot after moving the fan/heatsink) with EK water blocks/backplates in SLI. Dual 92mm Magicool radiators and the 700W SFX from Silverstone. Currently have the 600W one at home and testing. I plan to power it with the 12C Xeon Lower Power one. Hopefully this all works out.
 
Qinx, I’ll take the requested photos tomorrow when I get back home. But don’t you find it odd that none of these splitters worked? Especially the PLX chip one from super micro? That one should support it regardless of the bios supports it or not correct? Otherwise how do cards like the 7990 function in just one 16x slot. I’m guessing it’s something trivial that needs to be linked?

My goal is to have this work perpendicular in the asrock x99. I plan to have two PNY 970 GTX short reference cards (single slot after moving the fan/heatsink) with EK water blocks/backplates in SLI. Dual 92mm Magicool radiators and the 700W SFX from Silverstone. Currently have the 600W one at home and testing. I plan to power it with the 12C Xeon Lower Power one. Hopefully this all works out.

I don't find it that odd with the SuperMicro boards, As I've seen with the one I have here they just do whatever they want with the PCIe cards, as long as they work in the specified serverboards they don't care.

If I where you I'd send an email to SuperMicro techsupport asking if the RSC-R2UG-A2E16-A is fully PCIe complaint if so it should work regardless, that is the power of the PLX chip and the reason why it's expensive.

Also your build is basically what I was planning a while back, before the X99e was even out :p.
BTW I wouldn't use 1 dual 92mm radiator, that is not going to be enough cooling unless your are planning on using industrial fans.
 
I don't find it that odd with the SuperMicro boards, As I've seen with the one I have here they just do whatever they want with the PCIe cards, as long as they work in the specified serverboards they don't care.

If I where you I'd send an email to SuperMicro techsupport asking if the RSC-R2UG-A2E16-A is fully PCIe complaint if so it should work regardless, that is the power of the PLX chip and the reason why it's expensive.

Also your build is basically what I was planning a while back, before the X99e was even out :p.
BTW I wouldn't use 1 dual 92mm radiator, that is not going to be enough cooling unless your are planning on using industrial fans.


If we comply to the pci-e specs and use the PLX, wouldn't we be able to get this to work without a bios update?
 
If we comply to the pci-e specs and use the PLX, wouldn't we be able to get this to work without a bios update?

Using a PLX at least for me isn't an option, you'd need to be an electronics engineer in order to design, build and validate the entire PCB, pricing wouldn't help either as order a couple PLX chips runs you about 110 euros for 1, and ordering 100 makes it 101 each. I can't do that, someone else is ofcourse welcome to.

In my opinion at least for now I'd go with the PCIe bifurcation that needs BIOS support. PLX has some definite benefits if you can cough up the money.

For me it would be
PLX:
Usable on every PCIe 16x slot
Hard to implement, Special equipement and support needed.
high price per chip, 100+

PCIe Bifurcation:
Currently only usable on ASRock X99e-ITX/AC in the future on 100 series boards
Cheap and easy implementation.
Easy to change PCB revisions/different configurations
 
I've done some quick drafting in CAD to get a feeling at what kind of dimensions we are looking at.
If you stick to a SFX PSU, Silverstone has a 700W version coming, I can stay just shy of 10 liters, a little smaller. Rough dimensions would be 35cm*30cm*9cm=9.54 liters

You'd be able to run 2x GTX 980 ti without too much trouble.

KjnA476.jpg
This is a neat concept.

However...
Have you taken the time to draw up a Micro ATX board and 2 GPU's plugged in there vertically without risers? It looks like it would take up the same amount of space. It's a neat technical achievement doing this in ITX, but its starting to look like an exercise in futility. Maybe try drafting up the scenario I mentioned and overlap them or put them side by side with dimensions on each.

If you were really going to do this, I think you would have to watercool the GPU's get them down to 1 slot each.
 
This is a neat concept.

However...
Have you taken the time to draw up a Micro ATX board and 2 GPU's plugged in there vertically without risers? It looks like it would take up the same amount of space. It's a neat technical achievement doing this in ITX, but its starting to look like an exercise in futility. Maybe try drafting up the scenario I mentioned and overlap them or put them side by side with dimensions on each.

If you were really going to do this, I think you would have to watercool the GPU's get them down to 1 slot each.

Considering NOVA is a 17L build that is already as optimal as it might get for mATX I think that at 10L there might be a place for this. But I do agree, going a fully watercooled route might prove a better space saver and I'm looking into that as we speak.
 
Considering NOVA is a 17L build that is already as optimal as it might get for mATX I think that at 10L there might be a place for this. But I do agree, going a fully watercooled route might prove a better space saver and I'm looking into that as we speak.


I wonder if it's at all possible to fit all this into a 5L cube, that's a least my current goal.
 
I wonder if it's at all possible to fit all this into a 5L cube, that's a least my current goal.

5L is very ambitious, my project looks at being around 3.1L. And that is with external Power adapters, custom radiator and waterblocks and a Thin ITX board.

If you want to add another GTX970 have the X99 board and the PSU internal no way you can get 5L. There is also more heat so the Dual 92mm radiator isn't going to be enough

The best I can come up with, and that is with cutting corners for compatibility is 7L+,
 
Edge-case question: When using PCIe bifurcation with a riser card, how is the 75W PCIe slot power distributed? Might the risers that are not working simply not be transferring enough (or any) power for GPUs to initialise? IIRC QuinX, you had a similar issue with your custom riser for the 3L build.
 
Edge-case question: When using PCIe bifurcation with a riser card, how is the 75W PCIe slot power distributed? Might the risers that are not working simply not be transferring enough (or any) power for GPUs to initialise? IIRC QuinX, you had a similar issue with your custom riser for the 3L build.

Doesn't seem to matter, it just needs a valid 12V signal coming in through the PCIe slot.
 
Doesn't seem to matter, it just needs a valid 12V signal coming in through the PCIe slot.

You still need to provide an extra 12V signal over the PCI-E for the second card, correct? I can see where the extra power signal might come for the RSC-R2UG-A2E16-A, but I can't see it for the other two devices.
 
Some clever questions about splitters and bifurcation:

1. I have X99 motherboard with two PLX chips. That means they must be supported in BIOS and bifurcation works with the PCIe slots, slots autoconfigure to be x16 or x8 depending if cards are inserted or not. Now here is the question: If I would use a splitter card with the PLX chip (like the one from Supermicro) in this mobo, is there a chance it will work? Or still there will be piece of software missing in BIOS to accommodate the third PLX chip? (You may ask why my question, it is coming from the fact that the motherboards with two PLX chips have 7 PCIe slots and for my application area the ideal number would be 8).

2. How it is there are professional PCIe expansion systems which are connected to the PCIe bus and apparently operate with every motherboard without dependence on BIOS?

3. There are PCIe x4 ->x1 splitters which seem to work with every mobo, no special BIOS required. How they do it?
 
You still need to provide an extra 12V signal over the PCI-E for the second card, correct? I can see where the extra power signal might come for the RSC-R2UG-A2E16-A, but I can't see it for the other two devices.

12V is passed through to both slots from the original PCIe slot. Most cards work fine. It seems Nvidia and AMD try to pull all power via the PCIe PEG connectors instead of though the PCIe slot.

Some clever questions about splitters and bifurcation:

1. I have X99 motherboard with two PLX chips. That means they must be supported in BIOS and bifurcation works with the PCIe slots, slots autoconfigure to be x16 or x8 depending if cards are inserted or not. Now here is the question: If I would use a splitter card with the PLX chip (like the one from Supermicro) in this mobo, is there a chance it will work? Or still there will be piece of software missing in BIOS to accommodate the third PLX chip? (You may ask why my question, it is coming from the fact that the motherboards with two PLX chips have 7 PCIe slots and for my application area the ideal number would be 8).

PLX chips are completely software independed, that is their strength, you should need any BIOS/software support for it, it is purely a physical solution. However I don't know if "stacking" PLX chips like you suggest works, in theory is should but I've no idea in practice.

2. How it is there are professional PCIe expansion systems which are connected to the PCIe bus and apparently operate with every motherboard without dependence on BIOS?

Probably the a PLX chip being used, remember in industrial applications people don't care if a solution costs 300 bucks if it does what they want it to do.

3. There are PCIe x4 ->x1 splitters which seem to work with every mobo, no special BIOS required. How they do it?

Look at this image on their webpage
x4-pcie-splitter4-host-card.jpg


PLX chip!

PLX CHIPS EVERYWHERE!!!

Also 1 thing to remember
When working with GPUs the reason why using the PLX chip is expensive is because you need to to have all Gen 3 PCie lanes be switch, running 16-In/32-Out PCIe Gen 3 lanes is more intensive then running 4-In/4-Out PCIe Gen 2 lanes.
 
Just wanted to say, this is the reason [H] is the only forum I still read.

Yeah, the quality of all of these responses are really top notch. Where else can a question about splitting PCI-e lanes turn into a discussion about designing a PCB to make it work?
 
Hi, ASROCK support has responded!

http://forum.asrock.com/forum_posts.asp?TID=525&title=pcie-bifurcation-support-on-x99eitx

I am going to try this new bios 1.20E first thing when I get back from work tonight on all 3 of my splitters, wish me luck! Exciting times!

Direct link to beta bios just posted: http://www.asrock.com/mb/Intel/X99E-ITXac/?cat=Beta

Great looking forward to hearing your results.
I've got high hopes that your flexible riser should work now, that is what is closest to what we are looking for. Can you tell us what you paid for that specific riser?

Also, don't forget the photos I asked for :p with those I can try and confirm my other thoughts and if they prove right I should be able to design the PCB.
 
Wow excellent! It looks like they just uploaded that beta BIOS today, I wonder if they added it based on your message.
 
Wow excellent! It looks like they just uploaded that beta BIOS today, I wonder if they added it based on your message.

I PM'ed the ASROCK_TSD personnel who replied to the Z87 thread stating that the X99 had bifurcation support two days ago. One day later, he read my PM but did not reply. Was maybe working on the rom? IDK, but I decided to post a thread on this specific topic for the x99e mobo on their forum last night to escalate:

http://forum.asrock.com/forum_posts.asp?TID=525&title=pcie-bifurcation-support-on-x99eitx

I referred back to the hardforum that many were interested in this topic and encouraged them to support us. Cheesy, I know: "C'mon ASROCK, you guys made such an awesome MOBO, now let us unlock its full potential!"

and now three hours later, we have the beta 1.20E rom. They did it because we love the product and this thread clearly shows interest and community rallying.

Had it not been for the ASROCK x99e board I would not have come back to hardforum. In fact, I was on hiatus for 6 years after college since I joined the working force and just lost interest in tweaking with stuff and instead just opted for things that worked - MacBooks and Mac Minis.

1 month ago when I saw the x99e review on anandtech, my interest in SFF was once again rekindled and I was determined to get Dual 970 into my phoenix build. I did quite a bit of research on my own and bought up everything out there that might work since it didn't seem like anyone else was trying. I did not get the cyclone 437 splitter yet but that was next on my list: http://www.cyclone.com/products/expansion_backplanes/pcie2-437.php but it would have cost a pretty penny, I was quoted about 700 dollars.

Anyways, I was always a big fan of Shuttle XPCs back in the days when it was still a niche market but my god how things have changed 6 years later. In fact, the SFF topic now has more active views than other topics which was quite the contrary 6 years back.

I'll take the pics tonight for Qinx and test out the new 1.20E with those risers. Hope they work and when they do, I have a lot of shopping to do. :)
 
Last edited:
PLX chip!
PLX CHIPS EVERYWHERE!!!

Ok, this I know having two PLX chips in my mobo :D. The crux of my problem is what is the relation of the PLX chip to BIOS support? Some claim that without BIOS support there is no way for bifurcation working. From my examples it may appear that when the PLX chip is present, bifurcation will be done by it and BIOS support is not necessary. If it is so then the Supermicro splitter with the PLX chip should be working with any X99 mobo with PLX chip(s) (unless Supermicro did something to make it working only with their mobos).
 
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!
aVZDNIj.jpg

Do you have a possibility to check if this Supermicro splitter with PLX chip would work in an X99 mobo with one or two PLX chips onboard? These are high-end mobos with many PCIe slots e.g. Asus X99-E WS or ASRock X99 WS-E.
 
Do you have a possibility to check if this Supermicro splitter with PLX chip would work in an X99 mobo with one or two PLX chips onboard? These are high-end mobos with many PCIe slots e.g. Asus X99-E WS or ASRock X99 WS-E.

I don't have the Asus X99-E or the X99 WS-E board. Just the Mini-ITX. But why would you want to try this? Don't those boards already have more than 1 PCI-E slot?
 
Ok, this I know having two PLX chips in my mobo :D. The crux of my problem is what is the relation of the PLX chip to BIOS support? Some claim that without BIOS support there is no way for bifurcation working. From my examples it may appear that when the PLX chip is present, bifurcation will be done by it and BIOS support is not necessary. If it is so then the Supermicro splitter with the PLX chip should be working with any X99 mobo with PLX chip(s) (unless Supermicro did something to make it working only with their mobos).

The super micro board with PLX chip did not work with the x99-itx board. I will try it tonight also with the bios though and that should shed some light.
 
Ok, this I know having two PLX chips in my mobo :D. The crux of my problem is what is the relation of the PLX chip to BIOS support? Some claim that without BIOS support there is no way for bifurcation working. From my examples it may appear that when the PLX chip is present, bifurcation will be done by it and BIOS support is not necessary. If it is so then the Supermicro splitter with the PLX chip should be working with any X99 mobo with PLX chip(s) (unless Supermicro did something to make it working only with their mobos).

That's what I've been saying.

PCIe bifurcation and PLX are totally independent of eachother, two different methods to achieve the same results.

PLX chips work independent of the motherboard, You should be able to stick it into a LGA775 P35 motherboard and it should still work, it should be totally hardware and platform independent.

BUT, and this is the main question and chemist_slime is going to check it with the new BIOS.
The SuperMicro riser might have something special about it and that might be the reason it won't work in any other boards that those designed by SuperMicro for it.

Go to this page:
SuperMicro Riser Cards
Search for RSC-R2UG-A2E16-A

It has Gen3 support, GPU support and Autodetect, it is only supported in 4 systems it seems.

I think that the reason why the PLX riser doesn't work is that SuperMicro has some special commands it might send to it through the SMbus lines present on the PCIe interface, this might be used to enable the riser on boot, hence why it only works on the 4 listed system, but this is an assumption.

The main problem with all the SuperMicro cards is because they have been designed by SuperMicro for use on SuperMicro serverboards, so they don't have to be PCIe complaint.

Technically the AMD R9 295X2 isn't PCIe compliant either, but more so from the power standpoint rather then signaling

Go to:
Meet the Radeon R9 295X2: Cooling & Power Delivery
Search for "Compliant"

So back to the point, PCIe bifurcation support on the X99e is important because it will allow a cheap solution to connecting multiple PCIe devices to a ITX motherboard.
We are talking about SLI and CrossFire, but ofcourse it can be possible to run a GTX 980ti with and Intel 750 1.2TB PCIe SSD.
 
I don't have the Asus X99-E or the X99 WS-E board. Just the Mini-ITX. But why would you want to try this? Don't those boards already have more than 1 PCI-E slot?

These boards have 7 PCI-E slots :eek: But it is never enough, I can live with 7 but ideally I would need 8 slots :cool:.

That's what I've been saying. PCIe bifurcation and PLX are totally independent of each other, two different methods to achieve the same results.

OK, I appreciate your expertise, I just wanted to get this confirmed direct and loud.

Now the question is: Is the Supermicro PLX extender board the only one which is available or somebody else is doing such thing too?
 
These boards have 7 PCI-E slots :eek: But it is never enough, I can live with 7 but ideally I would need 8 slots :cool:.



OK, I appreciate your expertise, I just wanted to get this confirmed direct and loud.

Now the question is: Is the Supermicro PLX extender board the only one which is available or somebody else is doing such thing too?

I think the chances are extremely small that the board you are looking for exists.
Basically the only market that would ever have need for this in a reasonable volume either does what SuperMicro does or directly does it on the serverboard, why bother with a $100 chip on a riser PCB if you can do a $10 implementation on the motherboard.

The only way to get a consumer PLX splitter board is either have someone that has the equipement and knowledge to develop it, or get enough people onboard to be able to outsource it to a company that manufactures a small batch of these boards. Either way you are easily looking at a couple grand in development and validation before you could do a 50 board run or something, and the boards themselves will probably cost $150/$200 bucks each.

Simple case of supply and demand unfortunately.
 
I think the chances are extremely small that the board you are looking for exists.
Basically the only market that would ever have need for this in a reasonable volume either does what SuperMicro does or directly does it on the serverboard, why bother with a $100 chip on a riser PCB if you can do a $10 implementation on the motherboard.

The only way to get a consumer PLX splitter board is either have someone that has the equipement and knowledge to develop it, or get enough people onboard to be able to outsource it to a company that manufactures a small batch of these boards. Either way you are easily looking at a couple grand in development and validation before you could do a 50 board run or something, and the boards themselves will probably cost $150/$200 bucks each.

Simple case of supply and demand unfortunately.

I was just thinking if Supermicro does it other server manufacturers could do it too. There is demand for such cards, people want to put graphics cards into thin server. Making own board with the PLX chip is out of question.
 
I was just thinking if Supermicro does it other server manufacturers could do it too. There is demand for such cards, people want to put graphics cards into thin server. Making own board with the PLX chip is out of question.

It could be that other server manufacturers are doing the same, but think about this.
You are designing a new Serverboard, why invest into a PCIe Riser card that costs $60 dollars to make when implementing PCIe bifurcation right onto the Serverboard might cost you only $10 dollars?

Unless the manufacturer needs way more PCIe lanes that can be achieved via PCIe bifurcation will PLX be an option for them. I'm not saying the ideal PCIe PLX riser card doesn't exist, but chance are very very very low.
 
That's what I mean, why go the PLX route when you can just go ham like that!
 
It's like 4.5 mITX boards, so even with PCIe Bifurcation, this board still wins in efficiency :D

I've single-handedly ruined this thread, I'll shut up now.
 
Last edited:
What was the general idea about making this thing work for mITX?

Since someone noted already two stacked dual slot cards make no sense since its almost as tall as those two cards in mATX without the riser, then what about making sli out of two itx sized 970's single slotted with water blocks and packed into console like case?

That's already a really custom idea so why not make an interesting sandwitch-like dual card water block?

How big of a radiator would we need to dissipate those 2x160W gpu?

What if the whole case was CNC'd radiator with water flowing through it?

Don't let such a nice topic die :D
 
Oh the topic is far from dying, we are just getting started :D

as mentioned a case with 2 dual slot GPUs can clock in at under 10L, that's still 7L under NOVA, 2.6L under the M1 and just 3L above A4-SFX. Sounds like very good performance per liter ratio. imagine a 24 core Xeon with dual GTX 980ti? that's some raw horsepower to have in 10 liters.

Going for 2 single slot cards watercooled you can go a little smaller, but then you replace the 2 aircoolers with a radiator and pump.

I think if we are going to explore this, going for 2 dual slot aircooled cards is the most affordable step to get people on board. afterwards you could always explore the liquid cooler route.
 
What was the general idea about making this thing work for mITX?

Since someone noted already two stacked dual slot cards make no sense since its almost as tall as those two cards in mATX without the riser, then what about making sli out of two itx sized 970's single slotted with water blocks and packed into console like case?

That's already a really custom idea so why not make an interesting sandwitch-like dual card water block?

How big of a radiator would we need to dissipate those 2x160W gpu?

What if the whole case was CNC'd radiator with water flowing through it?

Don't let such a nice topic die :D

The two PNY 970GTX single slot short reference + 2 x EK full water blocks will do exactly what you said above for me. I just need a custom short bridge from them to link the two cards together. All others they have are too far apart and won't fit into my custom case.

Do you know of anyone that makes custom radiators? I actually have a CNC machine on pre-order to build my custom case but I do not know much about custom radiators. Thx.

and yes, as Qinx said, this thread is far from dying! :)
 
The two PNY 970GTX single slot short reference + 2 x EK full water blocks will do exactly what you said above for me. I just need a custom short bridge from them to link the two cards together. All others they have are too far apart and won't fit into my custom case.

Do you know of anyone that makes custom radiators? I actually have a CNC machine on pre-order to build my custom case but I do not know much about custom radiators. Thx.

and yes, as Qinx said, this thread is far from dying! :)

My build H2O-Micro uses the same PNY GTX 970 with custom parts for just about everything and it sits at 3.1L. that is without the external power adapters. Too be honest I think 5L is close to impossible.
Disclaimer: using a small radiator with 6000 rpm fans doesn't count ;)

The Mac Pro is 5.5L and that this as custom engineered as it gets

I'm hoping you start a build log here I'd love to see what you're going to make of it.
 
My build H2O-Micro uses the same PNY GTX 970 with custom parts for just about everything and it sits at 3.1L. that is without the external power adapters. Too be honest I think 5L is close to impossible.
Disclaimer: using a small radiator with 6000 rpm fans doesn't count ;)

The Mac Pro is 5.5L and that this as custom engineered as it gets

I'm hoping you start a build log here I'd love to see what you're going to make of it.

Are you using two PNY 970 GTX's or one? I'm assuming two since we're having this entire PCIE bifurcation discussion in the first place.

My current goal is to use 2 single 92mm radiators with the PNYs underclocked and undervolted. But I'm intrigued with the idea of building the radiator into the case itself, I think that will definitely be interesting and worth exploring, I just need to learn more about it.

How do you plan on making your custom parts? Do you have access to a CNC mill as well?

Build log will come soon when I have all my parts together. Currently I have an ASUS 970 GTX mini and was planning on de-soldering the second DVI port to make it single slot with the EK thermosphere. However, once I discovered that the PNY 970GTX was single slot to begin with after removing the fan, I'll buy two of them this weekend if the PCIE bifurcation works this evening when I get back home to confirm.
 
Are you using two PNY 970 GTX's or one? I'm assuming two since we're having this entire PCIE bifurcation discussion in the first place.

My current goal is to use 2 single 92mm radiators with the PNYs underclocked and undervolted. But I'm intrigued with the idea of building the radiator into the case itself, I think that will definitely be interesting and worth exploring, I just need to learn more about it.

How do you plan on making your custom parts? Do you have access to a CNC mill as well?

Build log will come soon when I have all my parts together. Currently I have an ASUS 970 GTX mini and was planning on de-soldering the second DVI port to make it single slot with the EK thermosphere. However, once I discovered that the PNY 970GTX was single slot to begin with after removing the fan, I'll buy two of them this weekend if the PCIE bifurcation works this evening when I get back home to confirm.

Have a scroll through my build ;)
Watercooled 3 Liter system with GTX970

Basically to get 3.1 liters in volume you need to do:

PNY GTX 970 with custom made GPU waterblock
Thin ITX motherboard with custom CPU waterblock
Custom dual 92mm high FPI radiator
ultra slim 15mm thick fans
2 external AC adapters
Custom reservoir with custom pump attachment
Custom Case

See the word custom mentioned a lot? :p

adding a second 970, and going for regular ITX combined with a 2011-3 CPU and placing the PSU internally means "only" increasing the size by 2 liters isn't going to happen.

BUT, if you do I'll smash my build against the wall for being so lacking.
 
Have a scroll through my build ;)
Watercooled 3 Liter system with GTX970

Basically to get 3.1 liters in volume you need to do:

PNY GTX 970 with custom made GPU waterblock
Thin ITX motherboard with custom CPU waterblock
Custom dual 92mm high FPI radiator
ultra slim 15mm thick fans
2 external AC adapters
Custom reservoir with custom pump attachment
Custom Case

See the word custom mentioned a lot? :p

adding a second 970, and going for regular ITX combined with a 2011-3 CPU and placing the PSU internally means "only" increasing the size by 2 liters isn't going to happen.

BUT, if you do I'll smash my build against the wall for being so lacking.

haha, so why are you trying to get this PCIE Bifurcation to work then, you have no need for it in this build.
 
haha, so why are you trying to get this PCIE Bifurcation to work then, you have no need for it in this build.

Well, 1 month into the build I had the same idea you want to do right now, only the X99e wasn't even out so it was a dead end, I also assumed that the X99e also didn't have PCIe bifurcation support. But now it seems more people have gotten the same kind of idea :p
 
Back
Top