An Experimental Challenge For Anyone Using A Particular Type Of System/2x-nVidia SLI

[BonzaiDuck]

This shouldn't be an irritation to forum members: If you have no interest in the thread -- avoid it!

Here's the scenario:

You have a midtower case with good airflow. You use a heatpipe cooler for the CPU. Maybe!-- You have the cooler ducted to the rear exhaust. Airflow CFM intake is such that Intake >= Exhaust.

AND -- you have 2x SLI with a pair of "Twin-Frozr" cards: dual-fans with heatpipes poking out of the assembly, as distinguished from the ubiquitous barrel-fan reference design. You do not water-cool the graphics cards.

Here's the challenge: Remove the PCI-cover-plates between/above each graphics cards and verify that you get a reduction of 1C to 3C in temperature for the hottest (upper) card under some consistent stress-test -- either FurMark or Heaven BenchMark, for instance. If there is actually a slight difference, it will still show up in the highest temperatures for both cards.

I THINK I recorded a difference, but at this point, I'm even a bit suspicious of my own imagination.

WHY? Because adding that ventilation, if it really poses some slight difference or improvement, means that a bigger difference or improvement can be possible -- maybe with an extra fan, maybe with only passive cooling "enhancement."

 

[ClockHound]

Sounds logical not, suspicious.

Not unlike this poster, air molecules are lazy by nature - hence why we have to push them around with fans - even then they will always seek the path of least resistance. With all that annoying intake fan pressure, the molecules want to find an exit - the easier the better.

That opening an easier exit path for the slacker molecules, one that happens to pass by the two towers of GPU heat, where they pickup/bump/push some heated GPU air hangers-on with them on the easy way out is just a bonus benefit.

Also with your ducted cpu cooler, they can't just join the exhaust fan entourage and leave the case in a nice dignified stream. No, they'd have to sit by the GPUs, get hot and bothered, then wait for their turn to get sucked into the CPU cooler fan to get bumped around some stupid fins and finally exhausted into some less turbulent climes. Or worse. Wait to be pushed out some little crack without much ceremony or company.

However, when other slacker air molecules get wise to your exploitation of their inherent inertia, they might stage a protest - block a path or take a run at your exposed nasal cavities. Or they may just ignore your controlling airflow intent and just go with the flow wherever it leads. :biggrin:

 

[YBS1]

They used to have pci bracket exhaust fan/blowers. I've not seen them in a while, but then I've not been looking either. I assumed they'd have an annoying pitch but possibly they are low speed fans.

 

[BonzaiDuck]

They used to have pci bracket exhaust fan/blowers. I've not seen them in a while, but then I've not been looking either. I assumed they'd have an annoying pitch but possibly they are low speed fans.

The 2x (that's "double") fans on my 2x (now "quadruple" fans) GTX 970's spin up to a maximum 2,100 to 2,200 RPM. Not "noiseless," but just barely detectible.

Adding a fan somewhere is a second or third-best option. I'd looked briefly at the blower fans you mention, but they're criticized for noise, and there were other aspects for which I wasn't eager. I'll look at some of them again.

But just adding a fan may miss something.

My basic idea is to cover both cards with "three sides of a box", forcing all the air through either (a) the lowest card's 2x fans and out the 2x PCI slots (one above each card), or (b) through 2x the 2x fans and out the rear PCI's successively. With (b), there'd have to be a "slot" opening for the pressurized case air to enter the front-most side of the box (between the cards). Either way, the top card would have air coming from the bottom card and the lower of the PCI slots (which also would exhaust some air from the lower card), or from the opening in the box as well as the low-card's fan-exhaust with possibility of drawing some fresh air from the lower slot.

As you see, the second card complicates things. Use of the open PCI slot above a single "twin -frozr" graphics card would definitely be effective by constructing a three-sided cover to force all the exhaust air out the rear.

By allowing only the air-intake from the lowest card, and possibly leaving the PCI slot covered above it, the hot (upper) card get's already-heated air, but all the air is still forced through narrow channels -- moving faster with greater pressure and therefore increased effectiveness (but for it's temperature). Ideally, you'd want to isolate the two cards with the respective open PCI slots above them. But there's not a lot of room for that.

Another option: With the single three-sided box idea, fit a 92x15mm fan to the front panel (sitting above SATA ports), forcing air through both slots and feeding more air to the upper card. The air from the lowest card would have only one place to go -- "up" -- mixing with this 92mm fan's exhaust, and out the PCI slot below the top card.

So the more I think about this, the three-sided box idea may be best in its most minimal form. The analog: In the heat of the summer, you don't wear full-length pants -- you wear cutoffs or Bermuda shorts.

In other words, only the upper card needs a three-sided box -- forcing all its exhaust air out the PCI slot above it. Maybe, this construction could also cover the gap between the cards on the "long" side, leaving the space open at the front of the cards.

Whatever I choose to do, "Simple is Best." I'll spend a lot more time thinking about it before I sit down with pieces of Lexan or a slab of foam-board, and the appropriate knife for the selected material.

 

[BonzaiDuck]

Clockhound mentioned something which I think adequately describes the real dilemma.

Air is not being exhausted fast enough for these non-reference-design cards, and warmer air builds up in a case BECAUSE of the messy "Twin-Frozr" warm air -- ESPECIALLY because it has to "wait" to be exhausted through the CPU cooler (which is ducted.)

I'm actually beginning to think there's a solution for this which would actually keep the motherboard VRM components cooler. And the central feature of the solution is the opened PCI-slots.

 

[heymrdj]

Use the Supermicro workstation idea, a cooler mounted directly behind the cards externally on the chassis.

 

[BonzaiDuck]

Use the Supermicro workstation idea, a cooler mounted directly behind the cards externally on the chassis.

You mean a FAN mounted behind the cards externally, don't you?

One of my ideas was to construct an art-board box that secures outside those PCI plates and holds the fan. Then I realized that my monitor and HDTV cables are plugged in there! So I'm thinking to construct the box with slots for the cables, big enough the cover the DVI/HDMI plugs and hold the fan -- possibly directed upward.

Then there's the possibility of mounting a 40mm fan INSIDE the case at the PCI plate. But it might be better to just get a slot cooler with barrel fan the exhausts out the PCI plate. Even so, you'd still benefit modestly from a duct-box covering the back of the top graphics card. And if the duct-box is a nice precise fit, you'd hope that the exhaust from the Twin-FRozr fans would push air out through that duct.

With the slot-cooler/barrel-fans, I looked again at those the other day, and you'd want to unclip the PCI plate to reverse the fan and its housing so that it sucks air from the back of the motherboard.

I've got nothing better to do than play with these ideas, trivial as the parts and solutions seem. Truth is, howsoever long it takes to come up with an elegant and simple design, it could be effortlessly replicated with each and every succeeding PC build. Unless, of course, I decide that the cards for the next one will sport dual full-cover water-blocks.

Come to think of it -- do you have a link with pictures for this "Supermicro workstation?" It now occurs to me that some minor modification of a slot cooler/barrel-fan might have the fan and housing sticking outside the case, so that the fan still pulls air from the case interior. Why wouldn't that be possible? If it is, then no fretting about monitor cables and so forth.

Little suckers are bound to be louder than you want, though . . .

 

[heymrdj]

Yes

http://www.supermicro.com/products/system/4U/7048/SYS-7048GR-TR.cfm

 

[BonzaiDuck]

Yes

http://www.supermicro.com/products/system/4U/7048/SYS-7048GR-TR.cfm

That's definitely the way I imagined it. But you can see the problem of the monitor cables, although I did have a solution idea for that.