The new build ...

[Rob Williams]

This is a view of the new system. The old system meticulously cleaned, all fans removed & cleaned with a tooth brush ... if your going to do it, do it well ... what a PITA.

The fans must've been fun to do with a toothbrush! It's circumstances like this when I tend to lose patience and just throw out the old equipment. Kudos for not taking the same lazy route I tend to.

My brain did turn off for a bit as I was working out where to put the springs. In my opinion, now, that I have it done. I would make 1st priority to put the springs on the hot side of the system .... in between the water block & the radiator in other words.

Alright, I find this interesting, because I never gave it much thought before. You're saying that springs should be used on the hotter parts of the tubing, as it's more likely to collapse? Not that I've dealt with the stuff before, but I wouldn't have installed it with that mentality. It makes perfect sense though. Perhaps one goal would be to make sure the tubing would never get that hot.

The rig's looking good though man!

 

[Psi*]

I got tired of spending $$ & especially waiting on UPS ... and these fans are quiet.

Hot? What is hot ... well being the geek of multiple disciplines (or is it undisciplined) that I am, I have a pyrometer (infrared thermometer) used for checking my tire temperatures when racing. Basically it allows reading temperature w/o contact. Sooooo, probing around all over the system I get about a 1 degree temperature change and that likely exceeds the accuracy of the method. Even the inlet & outlet of the radiator only has about a degree difference if I use my imagination some. I am not saying that this is a good method, but it is what I have.

When OC-ed 4.4GHz (or so) & running OCCT, it reports core #1 temp to reach 71 deg C. The pyrometer reports 27 to 29 deg C with not a lot of rhyme or reason ... again speaks to the measurement method, but it is what I have. Core #1 idle temperature is 38 deg C, BTW. Stock idle for Core #1 was around 27 deg C

But, what I think it indicates is that water flow is ample. And, that was a priority of the new build & I used as justification to get rid of the 1/4" tubing. I never thought to probe that system, but there was a definite temperature differential across the old radiator.

I have read someone's explanation of what you want in a water cooled system. He tried to use some logic about a low flow thru the radiator so that you would see a more pronounced delta. I cannot agree with that. My own analysis suggests maximum flow equals maximum thermal exchange. And it says that if I do want lower temperatures that can accomplished with increasing the air flow by adding external fans to the radiator. A bigger radiator also would do the job, but there is a dis-proportionately larger PITA factor involved with that.

To your last point or question, this *was* guesstimate & opinion at that time. Lots of materials "flow" ... even glass ... over time. So putting the springs on the tubing is a precaution for the future and I can't offer science or numbers, but just gut feel. The 1/4" tubing developed a tendency to collapse at the grommets on the back of the case. They did not start out that way.

And rethinking where the hotter section of the tubing is ... flow is so high that the temperature can be considered uniform ... without a better way to measure.

 

[Tharic-Nar]

I've never dealt with water cooling, so take whatever i say with a pinch of salt...

When measuring with your IR-Thermo, all you're picking up is the external temperature of what your measuring, not the water temperature. The tubing itself is plastic, thus an insulator, so it would take a reading of its external - thus cooler - surface. You might want to check the water directly with a thermocouple or something in the reservoir for a more accurate reading.

There is some logic to the reduced water pressure through the radiator for extra cooling. Less pressure, more time for heat transfer, but my guess is this is negligible since the aluminum and water has very good thermal conductivity and you would need a very high water pressure to negate their effects. If you were dealing with an iron radiator or some such with a lower thermal conductivity, then yes, low pressure would help (but kind of pointless since the CPU would overheat before you could gain any benefit). Since water to aluminum heat transfer is fast and aluminum to air transfer is slow, you need to increase the air flow to compensate. Bigger radiator, more air transfer, more cooling, that or add a second radiator to the loop.

 

[Psi*]

Thanks for the thoughts, and to add to your information.

The radiator is copper. So was my 1st radiator although I did not think so because it is so light. The Swiftech description for the new one says it is and I did scratch a bit of paint ... WHO KNEW?!

Yes, measuring on the tubing or even at the plexi-glass reservoir gives much lower numbers & they are quite variable at the same location. I have also been measuring on the radiator & the top of the water block ... there are 2 sort of head chambers on each sides of the radiator that have the fittings that the tubing attaches to. They appear to be 2 separate chambers (just occurred to me attach a pic!).

As if it makes a difference, the bottom tube with the yellow Tie Wrap is the gazinta meaning that the other end of the tube is on the water block and is the out flow. So the top with the blue Tie Wrap is the "cold" side or the gazouta of the radiator ... out flow.

Per some vague class in thermodynamics long long ago and far far away or I dreamt it (easily), says that the temperature on both sides (inside & outside) of those little cans is the same (aka a very small delta) because the metal is thin. This is the basis of my assumption and is where I am measuring hot side & cold side temperatures ... the 1 degree.

I am oriented to doing measurements with thermocouples. Notice the "s". But, this is a production machine for number crunching that i needed desperately to get running. I am planning a nearly identical machine in the next several weeks. I'll lose sleep about whether to "thermocouple" or not.

In a nutshell, I am just very surprised that there is not more easily discernible temperature difference between the cans on the radiator.

 

[Tharic-Nar]

Well, copper has better thermal conduction than aluminum, it's not a huge amount, but there is a difference none the less, so the same logic can still be applied, you would need to have a very high pressure to negate the heat transfer of water --> copper. I admit, thermal conduction is a little more complicated than this, since there is the amount of energy a material needs to heat up as well as how quickly it can dissipate that energy... or something like that, since i have little comprehension of material science, thermo and hydrodynamics .

There are thermocouples you can get that you can stick within a water loop to get temperatures, people complain about them saying they reduce pressure, but for testing purposes, that doesn't matter so much. So checking the temps at different points - as you suggested with thermocouples, would help. Before and after the water block as well as before and after the radiator.

But as you said you need the comp up and running now and i don't think the how's and efficiencies come into play when just getting it working is more important.

 

[Psi*]

completed

OC-ed to 4.34 GHz. In the bottom pic next to the Swiftech Apogee XT, you can see an 80 mm fan clipped onto the top of the memory heatsinks.

I have seen some people pee-ing all over Swiftech and glowing over the Watercool HeatKiller. Reason given is the XT is cast & made in China versus the HK is built in Germany. I have read that the HK is built in Germany & not sure about assertions about the XT. Casting? Why could that be a problem with copper? And, I don't believe that it could be cast. Although I think it comes down purely to thermal performance.

 

[Rob Williams]

When OC-ed 4.4GHz (or so) & running OCCT, it reports core #1 temp to reach 71 deg C. The pyrometer reports 27 to 29 deg C with not a lot of rhyme or reason ... again speaks to the measurement method, but it is what I have. Core #1 idle temperature is 38 deg C, BTW. Stock idle for Core #1 was around 27 deg C

That's rather incredible to me... both the idle and load. I could only dream about 71°C with such a major overclock.

Sorry it took me so long to respond back in here. How's the system been running since your last post? The last couple of pictures look fantastic... I'm a little jealous when I look at your setup, both from a performance and cleanliness aspect (I shudder to think of all the dust in this machine).

 

[Psi*]

Currently running BLCK = 210 & CPU Ratio = 21 for 4410 MHz, QPI = 5.03 GT/s. Temps get to low 80s now & OCCT linpack eventually crashes over night. I do have the RAM pushed to 6 7 7 18 & voltages don't seem to make much difference although I can't tell you what they are at the moment. They must make some difference but fiddling with them is kind of like making love with boxing gloves on ... your getting vague response but your not really sure. Not that I would know about that ...the boxing gloves that is.

BUT, that doesn't matter so much to me as my number crunching software doesn't load the CPU quite as much as OCCT linpack. 8 threads from the number cruncher & temps gets to the mid 70s ... I am a very happy camper with this.

A customer (in a mid sized company) of mine just bought a Dell e/w 2X W5580s e/w 12 GB DDR3. They spent $6500 I spent <$2000. I am anxious to see how we bench with the same number crunchers. Will know that later next week.

Rob, you talked me into this i7 920! We will reconcile in a good way, some day.

 

[Kougar]

Very impressive results Psi!

I agree working with the memory settings is a bit frustration, other than the main settings and back-2-back CAS delay I don't tighten anything else and leave them on auto. Outside of synthetic memory testing it's just not going to show results... and it's very easy to make a stable system go ever so slightly unstable.

 

[Rob Williams]

They must make some difference but fiddling with them is kind of like making love with boxing gloves on ... your getting vague response but your not really sure.

Haha! Nice analogy there. Though, as I'm under the weather, it hurt to laugh :-(

BUT, that doesn't matter so much to me as my number crunching software doesn't load the CPU quite as much as OCCT linpack. 8 threads from the number cruncher & temps gets to the mid 70s ... I am a very happy camper with this.

That's just it. Linpack is like the worst of the worst case scenario, so for your uses, it could likely be considered completely stable.

Rob, you talked me into this i7 920! We will reconcile in a good way, some day.

Just keep showing love to Techgage, and I'll be a happy camper Glad you are enjoying the CPU so much man.