Add another Quad Rad

Elfie · Mar 11, 2010

I'm slightly puzzled with a test I've just done. I bought a used Black Ice Extreme Quad rad (no damage that I can see) for one of my other PCs but I thought I'd add it to the loop in my sig to see what effect it would have. I still used the single Phobya pump which is supposed to deliver 800lph which I calculate to be just under 3gpm and a max head of 4m. It made absolutley sod all difference. I appreciate that the new rad requires fans with high cfm and I'd fitted 1200 rpm Noctua's with a claimed 59cfm (which is more realistically about 45-50cfm) but still, I expected a few degrees drop in temperature.

What bit of the mechanics of watercooling am I not understanding?

Elfie · Mar 11, 2010

Sorry, meant to post in the Watercooling forum and missed.

Conumdrum · Mar 11, 2010

You already have a 480 and a 240 in the loop? Then added the other rad?

If so before you added the rad your DT was very low to start with. Adding another radwill bring the DT down probably less than 1C, and that really won't make any diff in the CPU or GPU temps.

Your already overraded, adding another is n't needed. In fact, there is no reason to have the stealth 240 in the loop. it's a known restictive (flow wise) rad. That pump is a bit weak, you add mora tubing etc, you could be inpacting your flow rate.

muddocktor · Mar 11, 2010

Yeah, I'll go with what Conumdrum posted too. You are the point of diminishing returns on your loop presently and adding another rad isn't going to show noticeable gains.

hokiealumnus · Mar 11, 2010

Agreed. The best thing you can do to help when you already have that much rad is a water block, and you already have one of the best (assuming it's either the Cu or LT model). Vapor says the EK HF he's testing kicks the tar out of anything else right now. Hasn't published numbers, but I'm anxious to see them.

Not sure if it would make a huge difference, but a pump with some more head pressure possibly? The MCP355 is rated at 6.1m, yours is 4.0m. Not even positive it would change anything, but that would be another option for you to try if you wanted to experiment.

Elfie · Mar 11, 2010

Thanks for the heads up. I wasn't planning on keeping the rad in the loop as it's intended for another system. I did wonder if I'd reached the limit of how much heat could be transferred through the block (and it is the copper version). At the moment I'm getting temps in the high 60's with the fans on slow and low 60's with them on full. I know the i7's are OK to run hot but as I do a lot of video editing I was just hoping I could get them down into the low 60's with the fans on slow so that when summer comes the temps would stay under 70 and it would be a very quiet system.

P.S. Thanks for moving the thread.

Almost forgot to add that adding the BIS 240 did nothing for the temps either way although I do have a HD4870 in the loop now.

Elfie · Mar 12, 2010

Still puzzled. I looked up the stats for the MCP355, MCP655 & the Phobya 400

MCP355 Head 4.7M, Flow 600l/h
MCP655 Head 3.7M, Flow 1500l/h
Phobya Head 4.0M, Flow 800l/h

If my pump is weak what is the key attribute, flow or head pressure? My (limited) understanding was that as long as the flow rate is around 1.5g/m then that was sufficient.

If I'm over-radded why are 2 x 320 rads often recommended for overclocked i7s with a GFX card in the loop. Admittedly my HD4870 doesn't seem to strain my temperatures at all. I've removed the BIS 240 from the loop now.

Sorry to be a pain but I thought I was beginning to understand this and obviously still haven't grasped some concepts yet.

muddocktor · Mar 12, 2010

I never said anything about your pump as I'm not familiar with it at all. From the specs, it looks to be a good choice as the specs say it has both good head pressure and good flow rate.

And there is nothing wrong with having too much rad. That's like saying you have too much sex.

Too much is still real good.

I'm over radded myself on my main rig as OCCT/Linpack option only gets my temps to the very low 50's on the cpu and Furmark can't get my gpu to 40 C. But if I added another rad to my loop besides my PA120.3 and my Bonnie heater core, my temps wouldn't be any better than they are presently. You can only cool down close to ambient temps unless you go to active cooling via tec or phase or heat exchanger to the cold air outside or what have you. Your present loop was efficient enough at cooling your heatload that adding the other radiator didn't help matters a bit because you were already getting optimum heat exchange.

RGE · Mar 12, 2010

I can tell you my numbers for illustration, it helps if you can measure your ambient and water temps.

mine with 1 x 360 rad, mcp355 and i7 950 with vcore 1.34 OCed to 4.4 with HK block in single loop. After an hour of full load, ambient 24C, water temp 30C, core temp avg 68C.

So basically delta ambient to water is 6C, and adding more rads will only decrease that 6C, nothing else. The biggest problem is delta from core to water which is 38C (68C-water temp 30C), changing tim/block may help a little with that delta, but already using best tim and one best waterblocks. Changing the rads wont help with that. So look at it this way, ambient is 24C, water to core temp (basically gradient from core to IHS) is 38C, so no matter what I cant get below 62C load, even with 50 rads and powerful fans.

So now I add a second 360 rad, at best I can get another 3C, cut that 6C in half. However I am using relatively quiet fans 1200-1500 rpms. Instead of getting a second rad, I could have just gotten more powerful fans, but I dont want noise.

First I added 295gtx to single loop, my water temp only increased 1C, since at idle it only puts about about 80W or so (at gpu load, yes delta is way up there 15C or so). But at cpu load only, had 7C delta with 1200 rpm fans, then got second rad with quiet fans 1200 rpms, my delta is now ~3.5-4C from water to ambient at cpu only load.

So after 2x360 rads (without gtx 295) my delta water to air dropped from 6C to 3C, so temps 3C better with second rad. With added GTX at idle which adds 1C to water temp, I only got 2C total better temps with adding a second rad with adding gtx. However with both cpu and gpu fully loaded, that second rad becomes very noticable, since cuts my water delta down from 15C to about half that.

Elfie · Mar 12, 2010

Thanks again for the help.

rge - your step by step examples has given me a greater understanding than anything else I've read so far.

I suspect a friend of mine is being 'economical with the actualité' (to quote one of our pretentious politicians) when he claims a Core i7 at 4.4GHz runs at 50ºC load with a Black Ice 360 rad and a pair of 295s at no more than 42ºC with another triple rad on a separate loop.

Neuromancer · Mar 12, 2010

rge said:
I can tell you my numbers for illustration, it helps if you can measure your ambient and water temps.

mine with 1 x 360 rad, mcp355 and i7 950 with vcore 1.34 OCed to 4.4 with HK block in single loop. After an hour of full load, ambient 24C, water temp 30C, core temp avg 68C.

So basically delta ambient to water is 6C, and adding more rads will only decrease that 6C, nothing else. The biggest problem is delta from core to water which is 38C (68C-water temp 30C), changing tim/block may help a little with that delta, but already using best tim and one best waterblocks. Changing the rads wont help with that. So look at it this way, ambient is 24C, water to core temp (basically gradient from core to IHS) is 38C, so no matter what I cant get below 62C load, even with 50 rads and powerful fans.

So now I add a second 360 rad, at best I can get another 3C, cut that 6C in half. However I am using relatively quiet fans 1200-1500 rpms. Instead of getting a second rad, I could have just gotten more powerful fans, but I dont want noise.

First I added 295gtx to single loop, my water temp only increased 1C, since at idle it only puts about about 80W or so (at gpu load, yes delta is way up there 15C or so). But at cpu load only, had 7C delta with 1200 rpm fans, then got second rad with quiet fans 1200 rpms, my delta is now ~3.5-4C from water to ambient at cpu only load.

So after 2x360 rads (without gtx 295) my delta water to air dropped from 6C to 3C, so temps 3C better with second rad. With added GTX at idle which adds 1C to water temp, I only got 2C total better temps with adding a second rad with adding gtx. However with both cpu and gpu fully loaded, that second rad becomes very noticable, since cuts my water delta down from 15C to about half that.

Hey, great post

Was curious about stepping (c0 I guesS?) and block you are using. Also wondering if you were using HW temp sensors (I suspect so since you also have measured water temps) 68C seems a little high to me, unless you have a bad mount or a dirty block. My own experience is skewed as I do not have a HW temp sensor (yet.. its still in shipping), and I had my Rad in the window, :attn:

but the temp difference still seems pretty high.

thx

Conumdrum · Mar 12, 2010

Good discussion! A hot overclocked i7 under load is about 250 watts from what I have read. You really push it I don't see why 300 watts could be created. A GPU comes in SOOO many flavors of wattage.

It's not really about the chip on the CPU or GPU, it's just the heat created. Once you got an idea you then work the rads depending of your case and your noise situation.

I can see two 295's at idle on a 120x3 rad. No matter the rad I don't think you would see 42C under load, thats just wrong. No maybe while gaming, but put them in SLI and run Furmark for an hour. Unless they are at really low ambients and the room is silly silly cold.

Same for the CPU temps. Whats the load proggy, measuring proggy and load time?

My i7 965 at stock idles at around 32C, my two GTX 280s idle at 36C or so. Very quiet fans. Loads at 51 or so CPU, about the same on GPU. Setup is in my sig, lots like your friend.

hokiealumnus · Mar 12, 2010

Wasn't you that mentioned pressure mudd, twas me. I'm sorry to say that I got what I thought was a spec that was greater than 4.7M (6.1M to be precise), but it appears to be an erroneous number on Sidewinder's web page. Please disregard that particular comment, read rge's brilliant post and be enlightened like I have been.

joshhua5 · Mar 13, 2010

if he was to add another pump would taht get tempetures down?

RGE · Mar 13, 2010

Neuromancer said:
Hey, great post

Was curious about stepping (c0 I guesS?) and block you are using. Also wondering if you were using HW temp sensors (I suspect so since you also have measured water temps) 68C seems a little high to me, unless you have a bad mount or a dirty block. My own experience is skewed as I do not have a HW temp sensor (yet.. its still in shipping), and I had my Rad in the window, but the temp difference still seems pretty high.

thx

Thanks...it is D0 stepping. I have flukes with calibrated thermocouples which I use most of the time, and a crystalfonz with dallas temp probes which I dont have set up right now, wires are messy...flukes are easier to just use as needed.

Currently using HK3.0 wb, but also have Swiftech XT, GTZ and now EK supreme HF, but mounted 16x straight times in testing XT vs HK, and all about same delta..somewhere posted in this forum.

If look on xtremes thread 4.5+ghz stress with linx, a ~35C delta water to core is ~avg for 4.4 to 4.5ghz and 1.35-1.4 vcore on i7 chips, those running 4.6+ at much higher vcore some of them have 40-50+ deltas. At 4ghz, I load in high 40's/low 50's. Scaling in wattage and heat goes to crap once you get above 4.2.

Here is one at 4.6ghz, 1.376vcore in linx thread , loading at 75C with 24C ambients with 3x360rads just cpu loop (probably ambient to water only 2C so near 50C water to core), and his is one of the cooler ones at that mhz
http://www.xtremesystems.org/forums/showpost.php?p=4143844&postcount=64 You can find some in that thread using crazy vcore 1.6+ where wattage on cpu alone according to everest is over 300W (on mine anyways at those settings), system over 700W and delta >60C.

My C0 stepping i940 actually ran 8C cooler if both run at exact same vcore/mhz (my i950 used more watts at same setting according to everest). But i950 D0 clocks much higher. And rad in window during winter definitely helps, I can drop my load temps 15+C that way...it is how I ran my 4.6 linx 20 run in that thread

.

Edit: Elfie if your friend with i7 at 4.4 has his rad in outside ambient temps or using A/C vent to keep ambient near 10C...

Elfie · Mar 13, 2010

I've just an unscientific test over the last few hours. I put a thermometer in the res and left it at idle for an hour than ran Prime95 small fft for 2 hours with the following results:

Ambient 18ºC
Water idle 22ºC
Water load 27ºC
CPU load 65ºC

I can see now that an extra rad would only make a tiny bit of difference if any.

I should mention that to run my RAM at 2000MHz I have to use a QPI/VTT voltage of 1.6v which increase the temps by roughly 4ºC compared to running at 1600MHz with 1.4v.

Conumdrum · Mar 13, 2010

I'm surprised it didn't drop it more. Wondered what your DT would be with the rad back in?

Methinks your CPU temp is more limited by the CPU block, just not able to extract more heat. But your at 4.2 so yea thats more a reason why I guess. It's not your CPU block.

Still your got a DT of 9C, I though it would be lower. Ohh! Can you run it again with the fans on max? Curious about how much diff the fans make.

Elfie · Mar 14, 2010

Unfortunately that was with the fans on max. I know that with the fans on low that the CPU temp increases by 4-5ºC.

I'm going to remount the CPU block anyway and use a different retaining mechanism just on the off-chance it might make a difference.

Elfie · Mar 14, 2010

I've just done another water temp test. This time I reduced the overclock to 3.8GHz with a vcore of 1.18v rather than 1.33v for 4.2GHz. HWMonitor was showing the CPU power consumption as 112W rather than 152W.

Ambient 18ºC
Water idle 22ºC
Water load 25ºC
CPU load 53ºC

So with a 40W reduction in heat being dissipated the Delta dropped 2ºC but the CPU temp dropped 12ºC. Quite a big reduction for 40W less. I don't quite know what all that means though.

I'll probably stick to that overclock now.

Conumdrum · Mar 15, 2010

Aha, yep, I'm not a Thermo engineer either. I do know the copper block, the size of the interface between the CPU and block, effieciency of the paste, bla, bla, bla can only do so much to pull the heat out off the chip.

Just can't absorb the heat above a certain heatload. If the water was MUCH cooler it would help a lot. Not ambient, below ambient and zero.

Thats why chilled water, phase, or Ln2 is the next step. The cooling stuff is so much colder than the watercooling it can absorb more heat.

You kinda demonstrated what watercooling can handle and it still can only do so much.

Think your move to back off to 3.8 for long term is a smart one. Not a ton of difference in performance.

Add another Quad Rad

Member

Member

Member

Retired

Water Cooled Moderator

Member

Member

Retired

Senior Member

Member

Member

Member

Water Cooled Moderator

Member

Senior Member

Member

Member

Member

Member

Member

Similar threads