that smart looking owl with glasses answered that already...Quote:
Originally Posted by NaeKuh
3
:ROTF::ROTF::ROTF: Okay, okay...I'll go back to testing CPU blocks now. :sofa:
Printable View
that smart looking owl with glasses answered that already...Quote:
Originally Posted by NaeKuh
3
:ROTF::ROTF::ROTF: Okay, okay...I'll go back to testing CPU blocks now. :sofa:
Quote:
Originally Posted by NaeKuh
Why does that matter for the case of the argument though? The water level doesn't change while the system is running...
My vote goes toward it being a by product of the pressure the loop is under while running.
Flexible tubing.Quote:
Originally Posted by NaeKuh
When under flow the system as a whole is under more pressure than when it's off. Assuming under flow with a D5 at say 1.5GPM, the pump will be producing close to 4.5psi difference(depending on the top). Now you need to add the static pressure of the water in height above the loop. If a loop runs up 2 feet and back down, the maximum static pressure is 2ft H2O or .9 PSI. If the reservoir is allowed to vent, the maximum pressure point would be right behind the pump at about 5.4 PSI, and the average system pressure is then around 2.7psi depending on where the restrictions are. When the system is off, you only have static pressure working on the system which is the .9 psi or .5 psi on average
So there is a difference of about 2.2 psi on average acting upon the tubing throughout the loop. Not a bunch of pressure, but with our low durometer tubing, it's enough for a little bit of flex.
This flexing of the tubing, allows the tubing to expand and hold more volume. This volume has to come from somewhere and the only place to give is the reservoir. Even if the reservoir is sealed, the liquid will overcome the air. Air is compressible and expandable and even when sealed will expand rather easily. Water is not compressible so the level will drop when running and the air (if sealed) will expand down and create negative pressure. This is also why you'll get that air sucking in thing when you crack the reservoir when running. The volume of the loop is larger when it's running, so it's just trying to equalize the air pressure in the reservoir with the atmospheric pressure.
This is also why there is some importance in keeping reservoir closer to the pump inlet. If it's allowed to vent, it prevents that inlet side from going negative pressure, it will always have static pressure on the inlet tube. If it's sealed off too much and the reservoir has no air space, that inlet tube can possibly get enough negative pressure that it will collapse.
Anyhow, that's just the tubing flexing and should be exactly the same no matter what loop order is. It may change a little bit if the reservoir is allowed to vent or is sealed.
I have a theory where pressure and pump placement can make a difference. I blame Naekuh for bring up the fluid level thing, it steered me this way. I shouldn't be so quick to dismiss the idea, if you guys are measuring something there could very well be a good reason for it. This is what I suspect is the reason:
Pressure Induced Bowing of the Base
With some blocks these days getting fairly thin in the base, and some that are still fairly flat, I think this is possible.
We've been able to measure bowing of the base being up to 2C worth of gain on some processors, so I suspect there could be a similar gain if pressure was enough to bow the base of a thin flat block enough to start providing some expansion and ballooning of the base surface. There IS a difference in pressure around the loop and it's at it's maximum directly behind the pump. So depending on the pump, you might see a difference of 5PSI or more depending on the location in the loop.
This would be very block specific because it depends on the structure of the base. Microchannels provide greater rigidity than pins. The top may also account for some of the benefit if reversed bowing is occuring. The biggest criteria though I think would be how flat the base is. If it's already bowed alot, it wouldn't see any gains, but if it's really flat, then there would be potential.
This would probably also depend on the processor type, and the mounting pressure applied.
Anyhow, one way to test this on a suspect block that appears to be gaining by pump location, is to simply test the block base out under load by the pump. Use a straightedge and a backlight to get a really good look at the base flatness with the pump off, then run a pump right before the block and check the base flatness again (maybe even cut off flow rate behind the block for maximum pressure). There may also be a very slight increase in system flow rate if the block is extremely restrictive, and that may in itself provide some benefit, but it would likely be fairly small.
This is the only thing I could think of that could explain where pressure in itself would provide a benefit, but it's not a benefit from the standpoint of fluids and fluid thermal efficiency. It would be a base bowing benefit, and it would only apply to certain blocks, more specifically blocks that are thin and flat based.
Anyhow, if this pump before the block thing does work, I suspect it's more of a structural/mechanical deal with the base bowing and contact. I just can't for the life of me come up with a fluid based reason except for the minor system flow rate increase thing.:shrug:
Copper foil block anyone? Interesting theory Martin :up:
thanks to the this interesting read, I have a much better understanding of how this flow/head pressure/velocity thing works.
One question I have though not exactly related to the OP's original question, is how does this relate to running two identical rads in series vs. parallel? I've always heard the explanation that two rads in series gives less flow and more pressure drop than 2 rads in parallel. The only test results I've seen in the past seemed to indicate that there really is not a difference running identical rads in series or parallel.
With parallel you end up cutting the velocity in half of what it would be in a single rad or two rads in series
Yeah, if they are identical, you would see roughly half the flow rate running through each radiator. It's not exactly half though, because you've reduced pressure drop, your system flow rate increase.
You may see 2.0GPM system flow rate in a split parallel, but each radiator would see only 1.0GPM because they are run in split parallel loops.
If you ran both in series, your overall system flow rate would be reduce a bit by the higher pressure drop. That might turn out to be 1.8GPM, but each radiator would get the full 1.8GPM.
If you compare those flow rates to thermal efficiency however, the difference is pretty small.
http://www.swiftech.com/assets/image...0_HD_vs_FR.gif
At a 10C water/air delta the difference with the fans swiftech tested with would be roughly 400 watts parallel vs. 440watts series, or a 10% difference.
So at a 10C delta you could see maybe 1 degree difference or so, not huge, but I'd rather run series myself.
I havn't seen too much testing on this, actually I thought swiftech was going to publish something on their stacking of rads a while back. What ever happened to that swiftech stacked testing??
Thanks for putting numbers on it Martin. I understood about halving the velocity through the rads, but the heat dissipated chart in conjunction with the flow rates being roughly halved makes sense.
Done with GTZ tests....still waiting on the KL-350, then it'll be ~4-7 days after that for results from the KL-350.
Anyway, here's what I have so far:
http://vapor.skinneelabs.com/Other/L...ZLoopOrder.PNG
4120MHz, 1.46v 920. Test procedure the same as my normal. i7 has 1:1 ambient:CPU temp scaling though, so there's no correction factor used (only difference I suppose).
Quote:
Originally Posted by Hondacity
oops i quoted myself :DQuote:
Originally Posted by Vapor
all hail fluid dynamics and the laws of physics :bows:
Awesome work!!:clap::clap: Thanks for testing the flow rate too, that's also confirms the pump and flow rate is not affected either.Quote:
Originally Posted by Vapor
So far it appears the suspected "Pressure increase= performance increase" and/or "Pump before the Block" Myth is getting busted!!:up: