GPUs and CPU watercooled setup

Spooky Neo · Post by **Spooky Neo** » Sun Jun 12, 2005 3:26 pm

Hey guys!

I need your opinion about the setup I'm planning to buy. I don't want to take off all my fans of my system, but I want my CPU and GPUs (planning Crossfire solution) to be watercooled because these are the pars that heat the most. Plus, if I can get rid of fans on these parts, that would be cool

Also, the fact that I'm building on watercooling solution for GPUs and CPU is that I don't want a large radiator. I'ld rather have a big flow (like 2 pumps) instead of a big radiator.

Anyway, here's the setup :

Swiftech MCW6000-64 A64 CPU Waterblock
Swiftech MCW50 GPU Waterblock (2x for Crossfire)
HW Black Ice Xtreme (120mm fan) radiator
C-Systems CSP 750 Mark II (12v DC) pump
Swiftech 5.25" Resevoir

You think that would cool ok? I know it's hard to know, but bottom line, my case is 120mm intake and 120mm exhaust on the back. I also got a fanduct retractable and I'll use it to extract air from my PCI slots.

The rear exhaust will be like that CPU->Radiator->REAR CASE->120mmFAN (push INTO the radiator so trough the case)

I don't want the best results as possible, but something silent and something that will cool enough to run my cpu (currently A64 3800+) and my future ATI crossfire's cards (R520 + X850XT probably) stable even after 3hours of Battlefield or GTA

Thanks guys

Neo

edit : changed the layout of my radiator with my case (bold text)

hazindu · Post by **hazindu** » Thu Jun 16, 2005 6:13 am

contrary to popular belief, that radiator will work just fine for a multiple block configuration. The pump should work fine too, although you may want something with a little more head pressure just to be safe.

Spooky Neo · Post by **Spooky Neo** » Sat Jun 18, 2005 7:30 am

hazindu wrote:contrary to popular belief, that radiator will work just fine for a multiple block configuration. The pump should work fine too, although you may want something with a little more head pressure just to be safe.

what do you mean? More GPH? I can buy 2 of them because the company sell a pack of 2 also...so one pump could go to the cpu and the other to gpus...

and I decided to change my radiator to the Xtreme one. Still 120mm, but not the pro

From what I've read over other forums and reviews, the Xtreme would fit better for a multiple waterblocks configuration. But they said that it will be tight using 3 waterblocks on only one 120mm radiator. what do you think?

peteamer · Post by **peteamer** » Sat Jun 18, 2005 8:44 am

Spooky Neo wrote:what do you mean? More GPH?

No is the short answer.

"Head pressure" refers to the ability of the pump to 'push' against pressure/'back pressure'.

All items in your W.C. loop will restrict/resist the flow of water, some more.. some less. The head pressure refers to ability to 'push' against the restriction.

Bigger head... more ability to 'win' over resistance.

Pumps may be described as xyz meters/feet (etc.) of head. This tells you how 'high' it can pump water before the flow 'stalls' and therefore becomes useless... and so gives you a reference between pumps as to their ability to still provide flow in variously restrictive 'loops'. (Note: the closer you are to the 'Max Head' the lower the flow (GPH) will be though.)

Pete

ATWindsor · Post by **ATWindsor** » Sat Jun 18, 2005 10:07 am

According to my calculations flow should almost never be a problem, but i might be wrong. Even a real flow of 100l/h should be more than enough, so I would guess bad cooling because of low flow is very rare.

AtW

Spooky Neo · Post by **Spooky Neo** » Sat Jun 18, 2005 10:41 am

so which pump do you guys recommend to me? The C-SYstem pump or another one?

Thanks for the explanations Pete

pdf27 · Post by **pdf27** » Sun Jun 19, 2005 11:21 am

ATWindsor wrote:According to my calculations flow should almost never be a problem, but i might be wrong. Even a real flow of 100l/h should be more than enough, so I would guess bad cooling because of low flow is very rare.

Blocks will still cool at low flow, but the cooling performance falls off dramatically. See http://www.procooling.com/html/pro_testing.php for an illustration of how this happens for common water blocks.

DougG · Post by **DougG** » Sun Jun 19, 2005 11:43 am

Thats a cool link!
So basically anything above 1.2 GPH is a waste?

A small decent pump is the way to go.

I never had a C-MarkII, I have a MAG, and an AC Ehiem pump (other system). I had a DDC, often stopped and I sent it back.

peteamer · Post by **peteamer** » Sun Jun 19, 2005 12:37 pm

Spooky Neo wrote:Thanks for the explanations Pete

Spooky Neo, you're more than welcome.

Glad it was understandabubble...

"Low Flow" :

My (very limited) experience: My Eheim 1048 is pushing through ~4M Eheim 12mm ID pipe, 2 X 10mm fittings, 4 X 6mm fittings 1.5M+ of 6mmID pipe and Innovotak XX-Flow blocks (CPU + GPU... not known for being 'free-flow' ), and I'm getting a water temps of ~5C above ambient with an overclocked Athlon XP2400+ (ThoroughBred) using only a 1000mm WACC as cooling (Passively).

A 'very' bad mix that results in a flow rate that my tear ducts could compete with...

Pete

ATWindsor · Post by **ATWindsor** » Sun Jun 19, 2005 8:50 pm

pdf27 wrote:
ATWindsor wrote:According to my calculations flow should almost never be a problem, but i might be wrong. Even a real flow of 100l/h should be more than enough, so I would guess bad cooling because of low flow is very rare.
Blocks will still cool at low flow, but the cooling performance falls off dramatically. See http://www.procooling.com/html/pro_testing.php for an illustration of how this happens for common water blocks.

I didn't have time to read the entire ting, but I'll do later today, but those numbers doesn't seem rigth, the seem to indicate that a flow rate of about 5 l/h gives a 10 c increase in temperature. FOr that to happen you need to output 3500 Watts of heat.

AtW

HammerSandwich · Post by **HammerSandwich** » Mon Jun 20, 2005 9:31 am

DougG wrote:So basically anything above 1.2 GPH is a waste?

ATWindsor wrote:...those numbers doesn't seem rigth, the seem to indicate that a flow rate of about 5 l/h gives a 10 c increase in temperature.

I think you're both making the same mistake. The Procooling page shows flow PER MINUTE.

DougG · Post by **DougG** » Mon Jun 20, 2005 12:11 pm

Opps sorry, yes I mean 1.2 gpm

pdf27 · Post by **pdf27** » Mon Jun 20, 2005 1:45 pm

ATWindsor wrote:I didn't have time to read the entire ting, but I'll do later today, but those numbers doesn't seem rigth, the seem to indicate that a flow rate of about 5 l/h gives a 10 c increase in temperature. FOr that to happen you need to output 3500 Watts of heat.

AtW

Those numbers are litres/minute not litres/hour (as has already been stated).

I'm also pretty sure you're making a more fundamental mistake, but can't be certain from your posts to date. The reason the CPU temperature is dropping for increased flow rate is nothing to do with increased mass flow of water through the block (the radiator, heat load and ambient air temperature/flow remain constant during the testing). The reduced temperatures are due to higher levels of turbulence within the block and so better heat transfer between the metal of the block and the water. As such temperature calculations of the type you are trying to do are meaningless - pH's results give a delta-T between CPU core and water, while you need the delta-T between water inlet and outlet (usually a very small number) to do the sort of calculation you are doing. Incidentally, this is exactly why low flow rates ARE a problem.

Oh, and remember that pumps and blocks both have pressure head/flowrate curves (see the pump comparisons article on procooling for details). This means to get your desired 1.2 GPM figure you can't just use a pump capable of 1.2 GPM with zero head (the condition for which most pumps quote their flow) but need one which can put out 1.2 GPM with the sort of pressure head you would get within your system (I'd guess around 1m as a rectal extraction, but so far as I know the data to confirm that simply doesn't exist).

chylld · Post by **chylld** » Mon Jun 20, 2005 3:11 pm

i haven't witnessed a setup like what you plan to build, but in my opinion, i severely doubt that that radiator would be able to handle all that heat.

firstly, given that you're not using an extremely powerful pump, neither the waterblocks nor the radiator will benefit a great deal from the increased turbulence caused by higher flow rates. same the other way - the flow rate isn't so slow it'll hamper heat transfer.

so with flow rate aside, the cpu and 2 gpu's under load will be pumping somewhere between 150 and maybe even 300 watts of heat into the system. i don't know the specifics, but i'm pretty sure this is well outside the capacity of the little BIP radiator!

i think what you'll find is that the system will run extremely hot, since the radiator is being given more heat than it can possibly remove. i wouldn't be surprised if you found the temperatures to rise and rise until the system crashes.

then there's the issue of noise. to remove 150-300 watts of heat from a BIP (if it's at all possible) would require a gargantuan amount of airflow, which means a ridiculous amount of noise. obviously this isn't a path you want to take, so the best way to fix things is to get a radiator with more surface area, i.e. larger cooling capacity. a larger radiator would also allow the use of 2 or more fans running slower and thus quieter.

if i were you, i wouldn't attempt to cool that system with anything less than a BA heatercore, which is like a black ice extreme 2 with a much denser fin arrangement. how you would mount this in your case such that it receives fresh air and doesn't stick out like an eyesore is up to your own ingenuity!

so in my opinion the BIP is severely insufficient for the cooling task you have at hand, but of course i could be totally wrong - who knows, maybe you just have to build it and see!

Spooky Neo · Post by **Spooky Neo** » Mon Jun 20, 2005 6:18 pm

chylld wrote:i haven't witnessed a setup like what you plan to build, but in my opinion, i severely doubt that that radiator would be able to handle all that heat.

firstly, given that you're not using an extremely powerful pump, neither the waterblocks nor the radiator will benefit a great deal from the increased turbulence caused by higher flow rates. same the other way - the flow rate isn't so slow it'll hamper heat transfer.

so with flow rate aside, the cpu and 2 gpu's under load will be pumping somewhere between 150 and maybe even 300 watts of heat into the system. i don't know the specifics, but i'm pretty sure this is well outside the capacity of the little BIP radiator! i think what you'll find is that the system will run extremely hot, since the radiator is being given more heat than it can possibly remove. i wouldn't be surprised if you found the temperatures to rise and rise until the system crashes.

then there's the issue of noise. to remove 150-300 watts of heat from a BIP (if it's at all possible) would require a gargantuan amount of airflow, which means a ridiculous amount of noise. obviously this isn't a path you want to take, so the best way to fix things is to get a radiator with more surface area, i.e. larger cooling capacity. a larger radiator would also allow the use of 2 or more fans running slower and thus quieter.

if i were you, i wouldn't attempt to cool that system with anything less than a BA heatercore, which is like a black ice extreme 2 with a much denser fin arrangement. how you would mount this in your case such that it receives fresh air and doesn't stick out like an eyesore is up to your own ingenuity!

so in my opinion the BIP is severely insufficient for the cooling task you have at hand, but of course i could be totally wrong - who knows, maybe you just have to build it and see!

this is what I thought...BIP! loll

well my idea now is to get an Xtreme radiator instead of a Black Ice PRO. Yeah I'ld to fit it as much as I can with the case and this is a reason why I don't want a big radiator. I'm planning a SLK3000B case so the 120mm rear chassis was perfect for the Black Ice Xtreme I.

Got any ideas? The big reason why I wanna fix the heatsink to my computer is I move my computer at least 4 times a year and an external radiator not fixed to the case would be inacceptable.

chylld · Post by **chylld** » Mon Jun 20, 2005 6:28 pm

well a BIX (black ice extreme) is much better than a BIP, but the problem is you'll only be effectively able to fit 1 fan onto it. you can fit 2 in a push-pull configuration (as you can with the BIP) but i don't think that will significantly increase performance.

with a slow, radial fan, the best configuration is to pull air through it and have a shroud spacing the fan about an inch away from the radiator. however i've fitted a BIX to the rear of my slk3700bqe before and it's such a tight fit you'd need stuff hanging out the back

here's a pic of my old wc system:

that was with the fan mounted directly to the radiator, pulling air from inside the computer through the radiator and out the back. the temperatures were ridiculous, higher than aircooling my cpu with a 7000cu. shroud, fresh air and pull are 3 very important factors, all of which i failed with in my setup lol

i think the best place for a bix in a typical case is at the lower front. you'll have to remove the drive cage, but realistically it's the only simple solution that gives it fresh air.

another option would be to have the radiator hanging off the back of the case (not literally), with just the shroud and fan inside the case. (or shroud outside as well). you'd get fresh air this way, and provided you can make holes for the tubes and make a baffle to direct psu exhaust air away from it, it should work well.

ATWindsor · Post by **ATWindsor** » Mon Jun 20, 2005 9:28 pm

pdf27 wrote:
ATWindsor wrote:I didn't have time to read the entire ting, but I'll do later today, but those numbers doesn't seem rigth, the seem to indicate that a flow rate of about 5 l/h gives a 10 c increase in temperature. FOr that to happen you need to output 3500 Watts of heat.

AtW
Those numbers are litres/minute not litres/hour (as has already been stated).

I'm also pretty sure you're making a more fundamental mistake, but can't be certain from your posts to date. The reason the CPU temperature is dropping for increased flow rate is nothing to do with increased mass flow of water through the block (the radiator, heat load and ambient air temperature/flow remain constant during the testing). The reduced temperatures are due to higher levels of turbulence within the block and so better heat transfer between the metal of the block and the water. As such temperature calculations of the type you are trying to do are meaningless - pH's results give a delta-T between CPU core and water, while you need the delta-T between water inlet and outlet (usually a very small number) to do the sort of calculation you are doing. Incidentally, this is exactly why low flow rates ARE a problem.

Oh, and remember that pumps and blocks both have pressure head/flowrate curves (see the pump comparisons article on procooling for details). This means to get your desired 1.2 GPM figure you can't just use a pump capable of 1.2 GPM with zero head (the condition for which most pumps quote their flow) but need one which can put out 1.2 GPM with the sort of pressure head you would get within your system (I'd guess around 1m as a rectal extraction, but so far as I know the data to confirm that simply doesn't exist).

Yeah, my calculations i sbased on 5 l/m but i wrote 5 l/h, sorry about that.
You are right in your assumption that i was calculating the diffrencese between water in and out. But if the temperature is the diffrence between CPU-core and water, I think the conclusion would be the same. 10 Degress is very little, i would guess most CPUs can handle at least 40 degrees (celsius). (20+40=60 with water at 20).

As a side note, I think i have the weakes pump known to man (120 l/h)

. And it keeps my CPU at a cool 27 degrees.

AtW

pdf27 · Post by **pdf27** » Mon Jun 20, 2005 10:26 pm

ATWindsor wrote:Yeah, my calculations i sbased on 5 l/m but i wrote 5 l/h, sorry about that.
You are right in your assumption that i was calculating the diffrencese between water in and out. But if the temperature is the diffrence between CPU-core and water, I think the conclusion would be the same. 10 Degress is very little, i would guess most CPUs can handle at least 40 degrees (celsius). (20+40=60 with water at 20).

As a side note, I think i have the weakes pump known to man (120 l/h) . And it keeps my CPU at a cool 27 degrees.

AtW

Most can. Remember that those figures were for a particular chip (Athlon XP?). If you're running a hotter chip the dTs will be proportionally greater. And while they can handle higher temperatures running at stock, if you want to overclock at all, or run a tiny radiator (you will have another dT between water and air as you have already mentioned - this will depend on the heat load, air temperature and radiator airflow - most radiators dump the maximum possible heat to air anyway, with airflow being the limiting factor) you may need the extra headroom of a better block.

Plus, there's the geek factor

Spooky Neo · Post by **Spooky Neo** » Tue Jun 21, 2005 11:44 am

Hey chylld,

what's the current watercooling setup in the computer you shown? The one with the SLK3700B...As I can see you're cooling a GPU and a CPU? A BIP and a Swiftech CPU Waterblock unit, but what else?

chylld · Post by **chylld** » Tue Jun 21, 2005 3:03 pm

the setup in that picture was:

Pump: Eheim 1048
Radiator: Black Ice Extreme
Fan: Stock bqe rear case fan
CPU waterblock: swiftech mcw5002
GPU waterblock: swiftech mcw50

after not too long, i got sick of the high temps and also the cpu waterblock was leaking, so i moved everything (except the waterblocks) to an external box, upgraded the radiator to a BA, upgraded the cpu waterblock to a silverprop cyclone se and upgraded the fans to 2x papst 4412fgl. and my temps dropped by more than 20 degrees

Spooky Neo · Post by **Spooky Neo** » Tue Jun 21, 2005 4:45 pm

chylld wrote:the setup in that picture was:

Pump: Eheim 1048
Radiator: Black Ice Extreme
Fan: Stock bqe rear case fan
CPU waterblock: swiftech mcw5002
GPU waterblock: swiftech mcw50

after not too long, i got sick of the high temps and also the cpu waterblock was leaking, so i moved everything (except the waterblocks) to an external box, upgraded the radiator to a BA, upgraded the cpu waterblock to a silverprop cyclone se and upgraded the fans to 2x papst 4412fgl. and my temps dropped by more than 20 degrees

hum I want my watercooling to be internal so I might gain some heating but anyway...

So you're telling me to put the radiator in front of the 3000B, at the place of the fan cage. And what about the air? Which schematic is the best?

(Airflow is reprensted by an arrow)

Inside computer<-------Fan----Radiator----Case----Fresh air

(so the fan suck the air from the radiator and push it into the case. By doing this, the fan also force the air to passe trought the radiator and this one is fresh air)

Inside computer<-------Radiator----Fan----Case

Which one?

chylld · Post by **chylld** » Tue Jun 21, 2005 5:36 pm

the first one. pull air through the radiator with the fan, don't try to push air through it

another option you might want to consider is the use of 2 radiators, i've seen people put one BIX at the front bottom and another at the rear exhaust, like your original idea and my suggestion combined. it'll add a lot of weight and clutter to the case, though.

Spooky Neo · Post by **Spooky Neo** » Tue Jun 21, 2005 5:41 pm

chylld wrote:the first one. pull air through the radiator with the fan, don't try to push air through it

another option you might want to consider is the use of 2 radiators, i've seen people put one BIX at the front bottom and another at the rear exhaust, like your original idea and my suggestion combined. it'll add a lot of weight and clutter to the case, though.

...but that way, I'll have enough cooling capacity for my CPU and GPUs...sounds really cool! More expensive but if I want to keep my watercooling inside...

hum what would be the schematics for the back?

chylld · Post by **chylld** » Tue Jun 21, 2005 5:51 pm

hmm it's your decision, normally i would say "just try the basic setup and then expand later" but with watercooling you really want to get it right first time. taking the circuit out of your system and changing parts and cleaning up the spill is really not a fun thing to do!

(especially when you're using uv green coolant hehe)

just thought i'd get that out of the way. with the rear, there are 2 options: intake or exhaust. both setups would involve a fan mounted to a shroud mounted to the radiator, with the case wall somewhere in between.

with that fan as intake, you'll need to add exhaust fans at the top of the case to get rid of all of the hot air (or your psu fan might ramp up) but both radiators will be getting fresh air, so your temps would probably be lower.

with that fan as exhaust, your temps will be higher since the rear radiator will be getting warmed case air, but you won't need to make any further mods.

another issue you should really consider is the restriction of another radiator. in a basic wc setup the flow rate isn't usually a problem, but radiators are quite restrictive in design and additional ones might slow the flow rate enough to hinder the effectiveness of your waterblocks. overall though i think it'll be fine

Spooky Neo · Post by **Spooky Neo** » Tue Jun 21, 2005 6:02 pm

chylld wrote:hmm it's your decision, normally i would say "just try the basic setup and then expand later" but with watercooling you really want to get it right first time. taking the circuit out of your system and changing parts and cleaning up the spill is really not a fun thing to do! (especially when you're using uv green coolant hehe)

just thought i'd get that out of the way. with the rear, there are 2 options: intake or exhaust. both setups would involve a fan mounted to a shroud mounted to the radiator, with the case wall somewhere in between.

with that fan as intake, you'll need to add exhaust fans at the top of the case to get rid of all of the hot air (or your psu fan might ramp up) but both radiators will be getting fresh air, so your temps would probably be lower.

with that fan as exhaust, your temps will be higher since the rear radiator will be getting warmed case air, but you won't need to make any further mods.

another issue you should really consider is the restriction of another radiator. in a basic wc setup the flow rate isn't usually a problem, but radiators are quite restrictive in design and additional ones might slow the flow rate enough to hinder the effectiveness of your waterblocks. overall though i think it'll be fine

thinking about it, I'm wondering if I'm ok with the idea of taking out the hard drive cage...'cos a radiator + a 120mm fan will probably stuck on the hard drive cage and I need hard drives. Already has 2, + 1 floppy. Doh.

As for the restriction, I just have to get a powerful pump and it should do it no? Something like the C-Systems one which is about 120GPH...

chylld · Post by **chylld** » Tue Jun 21, 2005 6:25 pm

yeh a more powerful pump, or 2 weaker pumps in series. that should give you enough flow.

i didn't take out my hard drive cage either for the same reason - unfortunately that left me with the rear exhaust hole as the only option.

not fair, is it?

Spooky Neo · Post by **Spooky Neo** » Tue Jun 21, 2005 6:33 pm

chylld wrote:yeh a more powerful pump, or 2 weaker pumps in series. that should give you enough flow.

i didn't take out my hard drive cage either for the same reason - unfortunately that left me with the rear exhaust hole as the only option.

not fair, is it?

maybe my reading is bad tonight, but you didn't have to take out your drive cage with a radiator and a fan? Oh wait! you're using the 3700 right? The one with the hard drive cage inverted to 90 degrees? If so, you get more space that way compared to the 3000B

chylld · Post by **chylld** » Tue Jun 21, 2005 6:59 pm

a little bit more space, but not enough useable space. also in the bqe the drive cage is riveted to the case

(not hard to remove, but still, a hassle)

Spooky Neo · Post by **Spooky Neo** » Tue Jun 21, 2005 7:50 pm

chylld wrote:a little bit more space, but not enough useable space. also in the bqe the drive cage is riveted to the case (not hard to remove, but still, a hassle)

geez that's not really good for me...

is the BQE isn't enough to put the radiator+fan on the front, what will be my 3000B lol

chylld · Post by **chylld** » Tue Jun 21, 2005 8:00 pm

by not enough space, i mean that with the drive cage installed, there isn't enough space. with the cage removed i'm pretty sure there would be.

another option is to mount the radiator at the very top of the case, it requires cutting a big hole in the top but it works.. here are some pics of such a case:

i can't remember what fan orientation he's using, or if he had a fan box sitting on top of the computer... but you get the idea

the radiator he used was a BA heatercore. BA stands for "big-arse" btw, in case you didn't know, it's a designation for anything the size of the black ice extreme 2 but with more surface area (denser fins) and for about half the price

i don't know how well that setup would work with a BIX.