A call to vendors - Direct Drive GPU Needed!

[NaeKuh]

Ok, lately, water has exploded. Im talking almost big bang cosmic creation level exploded.

Its nothing like what it was when i first signed up on this forum.

This means, we are getting a lot of interest in this hobby, and i dont want our high class vendors to lose to some corsair H#0.

Recently on Anandtech especially, ive been getting a lot of inquiries on a gpu only system.

Until recently gpu only loops were kinda taboo, because of the initial cost and sort.

So heres the thing.... I challenge one of our high class vendors to come out with a direct drive GPU, so we can get even a further migration into our field.


Lets also help out vendors design a direct drive GPU.

My initial thoughts on a direct drive GPU that would work, would need to be like the EK bridge.

Imaging a side port option on the GPU block which we can attach a pump directly on top, or next to the gpu's like a bridge, that would give the GPU a direct drive option, and make GPU only loops more sensible.

Your thoughts people?

[86 5.0L]

I've seen those all-in-one units for GPUs, the only problem was the adapter plates for the various GPUs were pricey

[Vapor]

MCR-Drive-B + any block(s) you want?

[NaeKuh]

MCR-Drive-B + any block(s) you want?

120x2 min size also not friendly sideways with barb down, like u would if u top mounted it... otherwise that would of been a viable option.

[Church]

You mean all-in-one LC solution for gpu (IIRC i've seen such .. wasn't it for 480 fermi? IIRC not too effective compared to stock air)? Or exactly pump on gpu foremost? I don't see WHY that's good and must have in corsair H50/70 and i think same about GPUs. On top of chip can be striked out already, as usually there is less clearance below gpu card (especially in SLI setups) then above cpu.

What i really WOULD like to see in LC field related to GPUs - for all vendors to standardise placement of inlets/outlets. So that SLI and alike bridges of one vendor can be reused on other vendor full covers too. AND - to have some replacement tops reaching sideways of card for most common universal GPU WBs .. again, to be able to use SLI/CF water bridges bling with them
Of course, i understand that at least first wish is just in our customer interests but NOT in interests of vendors. These days if you wish to use EK bridge you HAVE to buy EK blocks. It's in EK's interests. Probably same for any other vendor. Good thing almost everybody everywhere is using G1/4 at least. But even then, often problems arise eg. when some block is designed for fittings of same vendor with shorter thread, or if you can fit only some thinner fittings side by side of particular vendor when using bigger tubing sizes (nice to see that in many cases of new products last part of trend is no more).

[NaeKuh]

No lemme explain it...

A Bridge, which would be able to somehow connect to both ports on the block.
This Bridge can be sealed so only 1 card used, or you can have 2 cards attached.

Now the bridge is ported to take a DDC where the inlet would be located on the outlet of the bridge.
this way u can get Tubing going from rad -> inlet on gpu... then it would be pulled to the pump and back @ the rad.

The bridge is basically what im challenging vendors to make.
This way, people who really dont care about the cpu, can keep on upgrading just by replacing the cards per simular bridge maker, and enjoy watercooling backwards.

Also would make multi loops a heck of a lot simular.

[Bodkin]

Sounds like an idea, as gpu do make the most noise. Can you draw up a rough sketch of what you mean with this bridge thing, im finding it hard to visualizes.

[Church]

Hmm. But what advantages placing pump there vs placing it anywhere else? It's just more weight for already weighted down with heavy FCs gpus. Then some standoffs to support that weight will be must heave. I'm a bit skeptical about vibration isolation too, it's not that strongly mounted and as heavy as rad with water, like direct drive rads with pumps. I'm also a bit against overintegrating things. Makes less flexible configurations that sometimes are harder to replace/upgrade/customize.

[NaeKuh]

mmmmm actually now u got me thinking because you brought up weight.. and thats very true, weight does become a problem.....

Hard mounting the pump direct to outlet of the block using a long bridge.
So you can use the pump to support the cards up with the long bridge.... then one probably wouldnt need a direct drive gpu... lol.... gonna have to play with this type of design.

[zeropluszero]

yeh, i think this needs a diagram dude.

[NaeKuh]

well originally i was thinking it would be a good idea, but the notion of Weight changed everything.

The unit would be fairly big, and if u had a hori mount, the pump inlet would be inverted.

So the entire idea now that i think about it more carefully, is kinda moot..


But my other statement i was going on about using a hard bridge to support the weight of the cards is this:


You can probably open the port on top of the GPU extend tubing for a T-line so you wont need a res.

The thing is i want to think of a small package for gpu only loops, which i think will bring in more people to our hobby.
And the only way to do that is to make it simple as possible.

[zeropluszero]

problem is not all cases have floors right below the cards. i think bottom mounted PSU is generally the done thing in cases these days.

[Waterlogged]

well originally i was thinking it would be a good idea, but the notion of Weight changed everything.

The unit would be fairly big, and if u had a hori mount, the pump inlet would be inverted.

So the entire idea now that i think about it more carefully, is kinda moot..


But my other statement i was going on about using a hard bridge to support the weight of the cards is this:


You can probably open the port on top of the GPU extend tubing for a T-line so you wont need a res.

The thing is i want to think of a small package for gpu only loops, which i think will bring in more people to our hobby.
And the only way to do that is to make it simple as possible.

OK Einstein, tell me this. What are you going to do if someone want to put this setup in a case like the RV02?

[CedricFP]

Off topic, but I'd really like to see single rads with built in pumps.

[meanmoe]

weights not an issue, I've seen a configuration like this before where an mcp355 was completely supported by a tube from a FC block. It was in a DD Torture rack and they modified the torture rack with spacers on the side to give a little more width in the MB area. If I can find a link I'll post it. I may have videos somewhere...

EDIT: Why couldn't we just build that with existing components?

[NaeKuh]

EDIT: Why couldn't we just build that with existing components?

great things are brought out about innovation.

And were discussing a solution to it.

To be honest, a 120 can probably handle a gpu fine.. 2 gpu's even while still holding better temps then air.... 3.. err..... i hope you can afford a larger rad, if you can afford tri fire or tri sli + .


Need a tight and compact solution that doesnt require all the messy work a full on loop requires.

OK Einstein, tell me this. What are you going to do if someone want to put this setup in a case like the RV02?

lolol... well help me out here bud.. lol... lets think of a solution.


I can tell ya this.. if we think of a solution, it will probably recruit a ton of people who do F@H. Seeing how they love gpu crunching.

[Utnorris]

Why not a bridge with the pump mounted on top of the bridge sort of like the old Apogee Drive? You could have the ports on the pump housing and have the path go in series from the pump to the first gpu back to the second gpu and then back out the pump housing. Attach a rad with a built in res and you have a nice compact solution.

[Metal Wolf]

Swiftech microdrive add 35x tubes and gpu plate.

Basically hard to improve on that setup. Thought about possible ways to mount the drive to the gpu plate but that just ends up as fail cause it adds even more weight to the graphics card so pump goes elsewhere. Could put it in a reservoir but thats just to complex for a small water system, so that leaves the radiator as the last substantial chunk of material and therefore pump holder.

To drop the price cause a 35x is hella overkill for a single gpu use something similar in design tot eh microdrive series radiator with a less powerful temperature controlled pump. small knob to tell it how sensitive to be to temperature(simple circuit: knob adds more resistance to thermocouple, controller circuit increases voltage to pump).

Drunkgeneering over.

[schoolslave]

Is this another troll thread?

I find it a much better idea to integrate pumps in the radiator, as that tends to be one component that lasts the longest (blocks, especially GPU blocks, get changed all the time).
No need for a Pump/GPU-Block monstrosity.

[Vetalar]

my 2cents: universal GPU block (or just copper base)+interchangable acetal covers with water chanels and copper inserts for hot parts (VRMs VRAM etc.) for different cards and MCR-drive type rad. maybe there is no need for DDC class pump in this setup.

added:
smth like this

[CedricFP]

Is this another troll thread?

I find it a much better idea to integrate pumps in the radiator, as that tends to be one component that lasts the longest (blocks, especially GPU blocks, get changed all the time).
No need for a Pump/GPU-Block monstrosity.

This. I feel like if we can start getting single rads with integrated pumps, to go along with the dual and triple configurations.

I think that the kind of flexibility that would offer for serial loops with redundancy, or stand-alone setups would be great.

Single 120 w/ pump in back exhaust fan, to GPU block. Talk about a very self contained loop.

[Waterlogged]

lolol... well help me out here bud.. lol... lets think of a solution.

Current pumps are too large for what I'd like to do and pumps as small as I'm thinking of would absolutely have to be a PD design to get the job done, which are usually loud...I mean really loud (ppl think a DDC or D5 are loud, they should hear what PD pumps sound like ). It should be something that can be built into one of the GPU block covers yet keep a really low profile, something like the new Antec KÜHLER only...much smaller and much much better.



OH, and CooliT has already kind of been down this road. http://www.performance-pcs.com/catal...e3cfc66c7b6d23

This. I feel like if we can start getting single rads with integrated pumps, to go along with the dual and triple configurations.

I think that the kind of flexibility that would offer for serial loops with redundancy, or stand-alone setups would be great.

Single 120 w/ pump in back exhaust fan, to GPU block. Talk about a very self contained loop.

While I agree completely and believe this is the most realistic option, we are in the minority. The majority of the end users want bigger, better, quieter, faster and single rad drives simply don't fit into their way of thinking.

[Conumdrum]

Guys, the RAD is the largest part. Put the pump there. So if blocks need changed, all you need is a new block.

I don't see any other option. And a rad/pump setup lasts forever. Who knows what the manus of GPU blocks will do in two years.

[zalbard]

The problem is VRM cooling. All cards are very different. And passively cooling VRM is the suck.
Need to come up with some kind of universal VRM cooler first.
Another problem is, one does not gain much OC potential when swapping a GPU heatsink for a better one. Just better noise levels.
But there are Prolimatech / Accelerto / Thermalright / Zalman solutions for it, that are quiet and likely cheaper than a WC solution. Not to mention little to no installation issues (compared to mounting a 120-240 rad somewhere).
Good luck, but I don't think it's a very good idea.