One loop or two?

[marcusmiller360]

I have an i7 940 @ 4 gigahertz, two GTX 295s and the northbridge on the Gigabyte motherboard is built for water cooling.

Currently, the loop looks like this:
D5 pump ->i7 ->MCR320 ->northbridge ->MCR220 ->GTX295 ->MCR320 ->GTX295 ->MRC320 ->reservoir ->pump. The fans are 800 rpm Yate Loons.

The goal is, as always, quite cool. I'm wondering though if this should really be one loop. My thinking is that any heat not dissipated by the last rad in the loop is going to get passed on to processor and raise its temps considerably over running two loops.

Am I just worrying or should I split this up?

Oh the rest: 1/2 inner diameter tubing, compression fittings (with the GPU's having 90degree adapters on the blocks), a Swiftech GTZ block and the blocks on the GPU's are XSPC blocks.

Edit: Correction: the blocks on the GPU's are from Swiftech as well.

[Conumdrum]

Hat dissipated by eash rad isn't the issue. It's the full loop and how eqch part affects the others. The diff in any loop is less than 2C anywhere, unless your loop is really screwed up.

Look at your total heat load. Remeber a CPU needs the lowest Delta T temps, the GPU doesn't care as much, 12 C DT is fine on a GPU.

TWO GTX 295's can put out over 250 watts of heat EACH. Time to use the charts at Martins to look at the heatload and Delta T temps.

Then do the same on your i7. An i7 at high clocks can do 250 watts. You want low water versus ambient air temps under load (Delta T) on a CPU. Under 5C is good.

Martins, you do...... have the link bookmarked?

You need big raddage, you need a BIGG case or external area, have fun!

You need a 120x4 for the GPUs and in a fully seperate loop a minimum 120x2 for the CPU, if your a big overclocker on the CPU. a 120x3 for JUST the CPU.

TWO fully seperate loops unless you can fit TWO Feser Monstas. Then two pumps for a single loop.

[Spawn-Inc]

wow, get some pics ASAP interested in where you placed the rads.

if you have the space, i might split that up for the cpu's sake.

also i would remove the northbridge, it's got aluminum in there.

[marcusmiller360]

Hat dissipated by eash rad isn't the issue. It's the full loop and how eqch part affects the others. The diff in any loop is less than 2C anywhere, unless your loop is really screwed up.

Look at your total heat load. Remeber a CPU needs the lowest Delta T temps, the GPU doesn't care as much, 12 C DT is fine on a GPU.

TWO GTX 295's can put out over 250 watts of heat EACH. Time to use the charts at Martins to look at the heatload and Delta T temps.

Then do the same on your i7. An i7 at high clocks can do 250 watts. You want low water versus ambient air temps under load (Delta T) on a CPU. Under 5C is good.

Martins, you do...... have the link bookmarked?

You need big raddage, you need a BIGG case or external area, have fun!

Yep, I've been to Martin's, and the MCR320 will dissipate 301 watts at 1000rpm (@ 10 degree delta T) according to him. Skinnee says it will do 356 watts at 1000rpm for him. According to Martin, at 800rpm I am just under 250 watts. According to Skinnee, I am a little over.

Which is fine for the GPU, but that also means a ten degree delta on the CPU, feeding into the Northbridge. That sounds a little too hot...

But I also find it silly to run a dedicated northbridge loop and CPU loop. Thats three pumps, three res.

Argh, the slow process of loop design.

Also, according to Martin's tests, to get a 5 degree delta on the CPU with it pouring out 250 watts, I'd need a TFC X-Changer 480 PLUS a MCR320. The X-Changer dissipates 150 watts at a 5 degree delta, and the MCR about 120 watts.

So..... wait. That makes no sense. That would mean I'd feed water 5 degrees warmer than ambient into the Northbridge, which on a double rad will probably leave the same temperature, which will enter the first GPU, which will hit a ten degree delta leaving the second MCR320, which will again be fed more heat by the second GPU and then have some sucked out by the next 320....

So it'd still feed water that is more than 10 degrees Celsius over the ambient temperature into CPU.

So then, I'd need (in the hypothetical setup) the TFC X-Changer AFTER the GPU's in order to cool the water down more so that it can go back to the processor cool. In fact, that makes it sound like I would almost need two of those.

But if I split up into two loops, I'm going to be passing the CPU water, chilled to a lower temperature, into the CPU again, and the GPUs, well, they can take the extra heat anyway.

As far as cases go, its being custom assembled at a machine shop a friend owns, and will be built to spec from either stainless or aluminum, whatever she has once we decide on we are going to do.

Also, Conundrum, did your team win today or something? Just a few misspells in your post left me wondering if I am drinking with you or against you, or drinking alone.

[marcusmiller360]

wow, get some pics ASAP interested in where you placed the rads.

if you have the space, i might split that up for the cpu's sake.

also i would remove the northbridge, it's got aluminum in there.

it does? Crap, he told me it was copper. Well.... so now we need a Northbridge cooler for the gigabyte board that comes with a waterblock.

[Spawn-Inc]

which board is it? the last time i checked there is mixed metals for their i7 board. so unless it's changed in the last 4 months its still mixed.

[marcusmiller360]

Pretty sure its this one:
http://www.gigabyte.us/Products/Motherboard/Products_Overview.aspx?ProductID=2960

As it appears to be the only one on their site that features a waterblock. On the site it does say the heatpipe is copper, it'd only make sense its an all copper design right? Well... actually, I guess it doesn't, it is a vendor board it could just be marketing.

I really wish he has his cell phone on, I'd ask him to look at the box.

[Spawn-Inc]

sure looks like aluminum to me, but i sent a email to gigabyte to find out.

[rge]

The waterblock on GB ex58ex is copper on inside, just nickel coated on outside. But all the water will touch is copper. They are 3/8 barbs. We had this discussion last year when i got mine...hicookie who is with gigabyte chimed in in this thread http://www.xtremesystems.org/forums/showthread.php?p=3409746#post3409746

I have i7 OC 4.4 and 1 gtx 295 with 2x360 rads. I have used 1 loop and 2 loops, have pics/temps of both on xtreme. I have 5C delta water to cpu, either way, but using 1300 to 1600 rpm fans. But really made little difference with 1 loop or 2 on mine, because gtx 295 downclocks itself to 300core and puts out ~60-70W idle. So while stressing cpu, gpu puts out little heat. While stressing gpu or gaming, cpu does not get that hot, typically less than 50% load. Only running furmark with prime does cpu temps suffer with 1 loop, but no real reason to do that.

EDIT: just to see, I just ran prime for 30 mins with fans cut down to 800rpms (though I do have push/pull).
800rpms = 8.5C delta air to water with gtx 295 and i7 OCed to 4.4 in single loop with 2x360 rads equivalent.
1300 rpms = 5.5C delta air to water (tested in past)
1600 rpms = 4.8C delta air to water (tested in past)
I get same performance cpu wise with gtx on mcr220 rad and cpu on 4x120 (pa120.3 + mcr 220).
Those temps are with fan filters on + air resistance with internal rads...which when I compared in past is about 30% worse temps than martins figures with external rads.

If running 2 loops, just put gpus and NB in 1 loop and cpu in other. also the NB waterblock sits on top of air cooler for NB, kind of screwy setup, so wb will only drop NB 5-6C and not much more. With 800 rpm fans you will definitely need some surface area for both gpu and cpu, also remember those figures in martins are for external rads, if yours are internal you will lose 20-30% performance from air resistance.

[Spawn-Inc]

i know this is late, but so was gigabyte... they said

http://img33.imageshack.us/img33/7042/gigaalu.jpg