Splitting feed for a dual CPU system

orev · 05-09-2002, 09:04 PM

how can I split a feed for 2 cpus and make sure I get an equal split between them? My hose is 1/2", and I was thinking of splitting it into 2 smaller tubes, probably 3/8" (3/8" being half of 1/2" in terms of area).

Will something like this provide enough back pressure on both feeds to ensure an equal system? Anyone have other ideas?

How would you split them? There's always a "T", but are there any others? I was hoping for a "Y", but no one seems to make them.

Cyco-Dude · 05-09-2002, 09:07 PM

id use a 1/2" 'Y' fitting with your 1/2" tubing.
browse Dtek's hardware page. BeCooling also has 'Y' fittings.

Mr. Thompson · 05-09-2002, 10:18 PM

Been there, done that. I think you would be better off plumbing the blocks in series.

orev · 05-10-2002, 08:31 AM

Really? That would of course be the easiest thing to do. Assuming you have done it, do your temps vary greatly on each CPU? Is one much cooler than the other, or are they pretty much the same?

What types of problems did you run into when splitting them up first?

Thanks.

bigben2k · 05-10-2002, 01:15 PM

Ah the dual setup... I remember doing this for my aquarium!

The surefire way to do this (and most expensive) would be to put a Y splitter (as stated above) and a flow meter at the 2 ends of the Y. whichever one is flowing faster, you put a restricting valve on it, and adjust it until the flow is the same.

Now let's get back to the real world!

As long as your blocks are exactly the same, that your tubing is exactly the same length, and that the tubing is in a relatively same position, orientation AND shape, you can assume that the flow will be pretty much the same.

You can run them in series, really. As long as you've got a decent flow, you'll probably never notice any difference (scary, isn't it?).

The plus with a dual-CPU setup, is that the flow is restricted, which is something that the rad likes.

What pump are you going to run?

Are you going to be using a thermal monitoring software?

orev · 05-10-2002, 01:35 PM

Hmm, running in series is starting to sound better and better...

I'm also planning on cooling my chipset and GF4 card, so making sure that everything is the same in a split system would be tough.

I've been using temperature software to monitor things (still on air right now), and I've noticed that 1 CPU is always hotter than the other. Not sure if it's just a location thing or it actually does run hotter.

My guess would be to run the input to the cooler one first, then on to the hotter one, then to the chipset and GF4 card. Does that make sense?

bigben2k · 05-10-2002, 02:23 PM

Almost.

Feed your hot CPU first, then the other one, then the GPU, then the chipset.

This is a debate I'm having myself: why bother water-cooling the chipset? I could get a quiet little 40mm fan HS, and epoxy the thing on there. Your setup is different though. Does your chipset get hot? because if it gets barely hot, just a heatsink with no fan would do the job, even if you have no fans in your case.

As for the GPU, I'm just not willing to go that way, yet... I guess you're going for an over-clock? You might want to wait to see what you can suck out of your system, before you decide to w-cool the GPU.

Got any numbers yet?

Cyco-Dude · 05-10-2002, 03:45 PM

a typical chipset really doesnt get hot enough to warrant using a waterblock to cool it. my chipset fan crapped out, so ive been using just the bare heatsink and its working just fine (as far as i can tell).
as for the CPUs, id still run them in parrallel. this way cool water will go to both CPUs. about the temps, the temp probes could be calibrated differently, or something like that.
if you put all those blocks in series, it would probably cut your flowrate way down.

pompoen · 05-10-2002, 05:16 PM

waterblocks for a northbridge are as crazy as wb for the memory

Cyco-Dude · 05-10-2002, 05:39 PM

yet ppl do it lol. i dont need to watercool, my temps and OC are fine with a HSF. im doing it strictly for fun (and becuase its cool), anything else is a bonus.

myv65 · 05-10-2002, 07:47 PM

You should get at least as good results in series as parallel and here's why. Let's look at plumbing in parallel. Just like an electrical circuit, the resistance of two resistors in parallel is lower than the same resistors in series. For this reason, the total flow rate will be higher in a parallel arrangement. It will not be much higher, however, because the water blocks are only a portion of the total resistance. Additionally, in parallel the flow through each block will be lower than the total flow through a series arrangement. As a result, velocity in the blocks will be lower in parallel than in series. Since velocity rules in the world of convection, this is not a good thing.

In series, the total flow will be a little lower, but compensated by the higher velocity in each block.

It's also important to understand just how little water temperature changes in a cooling system. For distilled water (not changed much by 5% or less additives), delta t = watts * .227 / gph. Putting numbers to this, a typical CPU using 75 watts would cause a change of only 1°C at a flow of 23 gph. Even the wimpiest of pumps can push more like 50 gph through a couple of blocks and a radiator even using 3/8" tubing. This means even lower delta t across each block.

The delta t from CPU to block is much higher. It's also a function of velocity as much as anything else. Keep the velocity high and you'll be better off in the long run.

orev · 05-10-2002, 10:04 PM

Quote:

Originally posted by pompoen
waterblocks for a northbridge are as crazy as wb for the memory

I might normally agree, however, I'm running a dual AMD system, and the northbridge does indeed get very hot. Right now it's just got a heatsink on it, but it really needs at least a fan. But that brings me to my goal for water cooling which is...

Quiet. Right now there are at least 7 fans in this thing. I thought my old PC was loud, but now it's nothing compared to this new one. If I can get as much heat as possible out of the case using the water system, that's less fans I need and I can turn down the speed on those that I will still have.

I'm not going for OC, just an efficient cooling system. The more efficient the better.

orev · 05-10-2002, 10:38 PM

Quote:

Originally posted by bigben2k
Almost.

Feed your hot CPU first, then the other one, then the GPU, then the chipset.

This is a debate I'm having myself: why bother water-cooling the chipset? I could get a quiet little 40mm fan HS, and epoxy the thing on there. Your setup is different though. Does your chipset get hot? because if it gets barely hot, just a heatsink with no fan would do the job, even if you have no fans in your case.

As for the GPU, I'm just not willing to go that way, yet... I guess you're going for an over-clock? You might want to wait to see what you can suck out of your system, before you decide to w-cool the GPU.

Got any numbers yet?

My thinking with doing the cooler one first is that even with the added heat, the delta temp would be higher when it reaches the hotter one. Though, they are only 2C difference right now, so I doubt it really matters. I'll probably wind up doing whatever order is easiest when I run the tubing.

My numbers for the air cooled system are about 50C each... not so great, I know, but I'm using the heatsinks that came in the package, so AMD must say they're OK. Max temp is 90C.

Hopefully I'll be on H2O soon though.

pHaestus · 05-10-2002, 11:53 PM

Skip the splitters and the Y and run in series.

Data here:

http://forum.oc-forums.com/vb/showth...threadid=65078

(No surprise Mr. Thompson suggested series)

bigben2k · 05-13-2002, 09:13 AM

Orev,

With a water-cooled system, you should be able to get the temps back down to 40C.

I suspected that your chipset might be hot, because you're running a dual CPU board. So go ahead and water-cool it too. I'd still hold off on the GPU, until you've got something running and some new numbers. A graphic card usually doesn't need its cooling system replaced.

I am a proponent of water cooling for the same reason: it's quieter.

Do you have any pics of your system? It'd be nice to see a Before and After series of shots.

Nordic · 05-13-2002, 05:26 PM

My experience with northbridges is that they indeed can get pretty hot. I've got a Zalman NB32 which is a passive heat sink which replaced the puny heatsink/fan combo that originally was on my EpoX 8K7A. The Zalman gets very hot, I can just barely touch it which means it's in the neighbourhood of 50C and the heat is transferring just fine. My bus is not overclocked and the system is totally stable, in fact I can not tell if there is any difference in overclockability of the bus. Perhaps my memory is holding me back but I don't think the high temperature is a problem for system stability as far as northbridges go.

On a side note, a friend of mine also got himself a Zalman NB32 but on his KT266A board the sink hardly gets above body temperature, even when overclocked (166 MHz). This indicates that northbridges come manufactured with different line widths, my AMD 760 is prolly 0.25um while the KT266A of my friends machine might be 0.18um. This is only speculation but I don't think it's far fetched.

bigben2k · 05-14-2002, 07:40 AM

That sounds perfectly reasonable, Nordic.

I think you've just demonstrated the point that every board will be different. If you've got an 8K7A, which board does your friend have?

I've heard that some chipsets run hot, and some cold. The KT266A obviously runs cool (anyone else?), and yours, with the AMD chipset, runs hot.

Nordic · 05-14-2002, 07:58 AM

To clarify, my friend had another EpoX board, namely the 8KHA+ (rev 2.0).

Brad · 05-14-2002, 02:09 PM

the 760mpx does get really hot, i'd recommend cooling it, simply because you may run into fsb limitations with it when not cooled adequately

orev · 05-14-2002, 09:06 PM

Quote:

Originally posted by bigben2k
Orev,

With a water-cooled system, you should be able to get the temps back down to 40C.

Only 40? Really? I was hoping to get much closer to ambient. Well, I guess I'll see what happens.

Quote:

Do you have any pics of your system? It'd be nice to see a Before and After series of shots.

I've been doing my best to take pictures of every step of the way. Eventually I'll get everything up on a web site.

bigben2k · 05-15-2002, 07:25 AM

Looking forward to it.

Yes, your CPU temps will only drop to 40C, maybe a little lower, if your rig is efficient.

Remember, the purpose of water-cooling is twofold:

1-Better cooling (marginally, but still significant)

2-Be vewy, vewy quiet... I'm hunting wabbits. (E. Fudd)

Ask anyone who's running more than 6 fans, especially Delta fans...

05-09-2002, 09:04 PM	#1
orev Cooling Neophyte Join Date: Feb 2002 Posts: 35	Splitting feed for a dual CPU system how can I split a feed for 2 cpus and make sure I get an equal split between them? My hose is 1/2", and I was thinking of splitting it into 2 smaller tubes, probably 3/8" (3/8" being half of 1/2" in terms of area). Will something like this provide enough back pressure on both feeds to ensure an equal system? Anyone have other ideas? How would you split them? There's always a "T", but are there any others? I was hoping for a "Y", but no one seems to make them.

05-09-2002, 09:07 PM	#2
Cyco-Dude Cooling Savant Join Date: Feb 2002 Location: Dayton, Ohio Posts: 836	id use a 1/2" 'Y' fitting with your 1/2" tubing. browse Dtek's hardware page. BeCooling also has 'Y' fittings. Last edited by Cyco-Dude; 05-09-2002 at 09:10 PM.

05-09-2002, 10:18 PM	#3
Mr. Thompson Cooling Neophyte Join Date: Dec 2001 Location: Exeter, CA USA Posts: 56	Been there, done that. I think you would be better off plumbing the blocks in series. __________________ Knowledge is useless, it’s what you do with it that counts. If the supercooled, overclocking crowd discovers how fast duallies feel, there will be a stampede to build SMP rigs that will make the California gold rush look like a few little old ladies pushing shopping carts at the supermarket.

05-10-2002, 01:15 PM	#5
bigben2k Responsible for 2% of all the posts here. Join Date: May 2002 Location: Texas, U.S.A. Posts: 8,302	Ah the dual setup... I remember doing this for my aquarium! The surefire way to do this (and most expensive) would be to put a Y splitter (as stated above) and a flow meter at the 2 ends of the Y. whichever one is flowing faster, you put a restricting valve on it, and adjust it until the flow is the same. Now let's get back to the real world! As long as your blocks are exactly the same, that your tubing is exactly the same length, and that the tubing is in a relatively same position, orientation AND shape, you can assume that the flow will be pretty much the same. You can run them in series, really. As long as you've got a decent flow, you'll probably never notice any difference (scary, isn't it?). The plus with a dual-CPU setup, is that the flow is restricted, which is something that the rad likes. What pump are you going to run? Are you going to be using a thermal monitoring software? __________________ WBTA (Water Block Testing Alliance) Nordic Hardware WC 101

05-10-2002, 02:23 PM	#7
bigben2k Responsible for 2% of all the posts here. Join Date: May 2002 Location: Texas, U.S.A. Posts: 8,302	Almost. Feed your hot CPU first, then the other one, then the GPU, then the chipset. This is a debate I'm having myself: why bother water-cooling the chipset? I could get a quiet little 40mm fan HS, and epoxy the thing on there. Your setup is different though. Does your chipset get hot? because if it gets barely hot, just a heatsink with no fan would do the job, even if you have no fans in your case. As for the GPU, I'm just not willing to go that way, yet... I guess you're going for an over-clock? You might want to wait to see what you can suck out of your system, before you decide to w-cool the GPU. Got any numbers yet? __________________ WBTA (Water Block Testing Alliance) Nordic Hardware WC 101

05-10-2002, 08:31 AM	#4
orev Cooling Neophyte Join Date: Feb 2002 Posts: 35	Really? That would of course be the easiest thing to do. Assuming you have done it, do your temps vary greatly on each CPU? Is one much cooler than the other, or are they pretty much the same? What types of problems did you run into when splitting them up first? Thanks.

05-10-2002, 01:35 PM	#6
orev Cooling Neophyte Join Date: Feb 2002 Posts: 35	Hmm, running in series is starting to sound better and better... I'm also planning on cooling my chipset and GF4 card, so making sure that everything is the same in a split system would be tough. I've been using temperature software to monitor things (still on air right now), and I've noticed that 1 CPU is always hotter than the other. Not sure if it's just a location thing or it actually does run hotter. My guess would be to run the input to the cooler one first, then on to the hotter one, then to the chipset and GF4 card. Does that make sense?

05-10-2002, 03:45 PM	#8
Cyco-Dude Cooling Savant Join Date: Feb 2002 Location: Dayton, Ohio Posts: 836	a typical chipset really doesnt get hot enough to warrant using a waterblock to cool it. my chipset fan crapped out, so ive been using just the bare heatsink and its working just fine (as far as i can tell). as for the CPUs, id still run them in parrallel. this way cool water will go to both CPUs. about the temps, the temp probes could be calibrated differently, or something like that. if you put all those blocks in series, it would probably cut your flowrate way down.

05-10-2002, 05:16 PM	#9
pompoen Cooling Neophyte Join Date: Mar 2002 Location: belgium Posts: 41	waterblocks for a northbridge are as crazy as wb for the memory

05-10-2002, 05:39 PM	#10
Cyco-Dude Cooling Savant Join Date: Feb 2002 Location: Dayton, Ohio Posts: 836	yet ppl do it lol. i dont need to watercool, my temps and OC are fine with a HSF. im doing it strictly for fun (and becuase its cool), anything else is a bonus.

05-10-2002, 07:47 PM	#11
myv65 Cooling Savant Join Date: May 2002 Location: home Posts: 365	You should get at least as good results in series as parallel and here's why. Let's look at plumbing in parallel. Just like an electrical circuit, the resistance of two resistors in parallel is lower than the same resistors in series. For this reason, the total flow rate will be higher in a parallel arrangement. It will not be much higher, however, because the water blocks are only a portion of the total resistance. Additionally, in parallel the flow through each block will be lower than the total flow through a series arrangement. As a result, velocity in the blocks will be lower in parallel than in series. Since velocity rules in the world of convection, this is not a good thing. In series, the total flow will be a little lower, but compensated by the higher velocity in each block. It's also important to understand just how little water temperature changes in a cooling system. For distilled water (not changed much by 5% or less additives), delta t = watts * .227 / gph. Putting numbers to this, a typical CPU using 75 watts would cause a change of only 1°C at a flow of 23 gph. Even the wimpiest of pumps can push more like 50 gph through a couple of blocks and a radiator even using 3/8" tubing. This means even lower delta t across each block. The delta t from CPU to block is much higher. It's also a function of velocity as much as anything else. Keep the velocity high and you'll be better off in the long run.

05-10-2002, 11:53 PM	#14
pHaestus Big Player Making Big Money Join Date: Aug 2001 Location: irc.lostgeek.com #procooling.com Posts: 4,782	Skip the splitters and the Y and run in series. Data here: http://forum.oc-forums.com/vb/showth...threadid=65078 (No surprise Mr. Thompson suggested series)

05-13-2002, 09:13 AM	#15
bigben2k Responsible for 2% of all the posts here. Join Date: May 2002 Location: Texas, U.S.A. Posts: 8,302	Orev, With a water-cooled system, you should be able to get the temps back down to 40C. I suspected that your chipset might be hot, because you're running a dual CPU board. So go ahead and water-cool it too. I'd still hold off on the GPU, until you've got something running and some new numbers. A graphic card usually doesn't need its cooling system replaced. I am a proponent of water cooling for the same reason: it's quieter. Do you have any pics of your system? It'd be nice to see a Before and After series of shots. __________________ WBTA (Water Block Testing Alliance) Nordic Hardware WC 101

05-13-2002, 05:26 PM	#16
Nordic Cooling Neophyte Join Date: Mar 2001 Location: Umeå, Sweden Posts: 98	My experience with northbridges is that they indeed can get pretty hot. I've got a Zalman NB32 which is a passive heat sink which replaced the puny heatsink/fan combo that originally was on my EpoX 8K7A. The Zalman gets very hot, I can just barely touch it which means it's in the neighbourhood of 50C and the heat is transferring just fine. My bus is not overclocked and the system is totally stable, in fact I can not tell if there is any difference in overclockability of the bus. Perhaps my memory is holding me back but I don't think the high temperature is a problem for system stability as far as northbridges go. On a side note, a friend of mine also got himself a Zalman NB32 but on his KT266A board the sink hardly gets above body temperature, even when overclocked (166 MHz). This indicates that northbridges come manufactured with different line widths, my AMD 760 is prolly 0.25um while the KT266A of my friends machine might be 0.18um. This is only speculation but I don't think it's far fetched. __________________ Pump:Eheim 1048, Radiator:LPD radiator 100x150x20mm, CPU block:DD Maze-2, GPU block: DD GPU block, Chipset block:Zalman NB32 AMD XP 1500+, EpoX 8K7A, PoV GF3Ti200 Pic of my YY-cube

05-14-2002, 07:40 AM	#17
bigben2k Responsible for 2% of all the posts here. Join Date: May 2002 Location: Texas, U.S.A. Posts: 8,302	That sounds perfectly reasonable, Nordic. I think you've just demonstrated the point that every board will be different. If you've got an 8K7A, which board does your friend have? I've heard that some chipsets run hot, and some cold. The KT266A obviously runs cool (anyone else?), and yours, with the AMD chipset, runs hot. __________________ WBTA (Water Block Testing Alliance) Nordic Hardware WC 101

05-14-2002, 07:58 AM	#18
Nordic Cooling Neophyte Join Date: Mar 2001 Location: Umeå, Sweden Posts: 98	To clarify, my friend had another EpoX board, namely the 8KHA+ (rev 2.0). __________________ Pump:Eheim 1048, Radiator:LPD radiator 100x150x20mm, CPU block:DD Maze-2, GPU block: DD GPU block, Chipset block:Zalman NB32 AMD XP 1500+, EpoX 8K7A, PoV GF3Ti200 Pic of my YY-cube

05-14-2002, 02:09 PM	#19
Brad Thermophile Join Date: Oct 2001 Location: Nuu Zeeelin Posts: 3,175	the 760mpx does get really hot, i'd recommend cooling it, simply because you may run into fsb limitations with it when not cooled adequately __________________ 2x P3 1100's at 1400, Abit VP6, 2x Corsair 256mb PC150 sticks, 20gb 'cuda ATA-III, 2x 40gb 'cuda ATA-IV in raid 0. 20" Trinitron. No fans 2x 2400+ at 2288mhz (16.0 x 143), Iwill MPX2, 2x Kingmax PC-3200 256mb sticks, 4x 20gb 60gxp in Raid 5 on a Promise SX6000. Asus Ti4200 320/630. Cooled by Water

05-15-2002, 07:25 AM	#21
bigben2k Responsible for 2% of all the posts here. Join Date: May 2002 Location: Texas, U.S.A. Posts: 8,302	Looking forward to it. Yes, your CPU temps will only drop to 40C, maybe a little lower, if your rig is efficient. Remember, the purpose of water-cooling is twofold: 1-Better cooling (marginally, but still significant) 2-Be vewy, vewy quiet... I'm hunting wabbits. (E. Fudd) Ask anyone who's running more than 6 fans, especially Delta fans... __________________ WBTA (Water Block Testing Alliance) Nordic Hardware WC 101

Currently Active Users Viewing This Thread: 1 (0 members and 1 guests)