New Method For Testing Water Block Performance

[GlassMan]

With the release of the Apogee water block from Swiftech, the has been a lot of discussion about the effectiveness of waterblock testing techiques in making valid predictions of water block performance in the real world, both bare die, and IHS CPU's.
I think we can agree that under identical conditions, the block that delivers the lowest cpu temperature is the higher performing block. The typical approach is to establish identical conditions and measure temperatures. The problem is that these temperature measurements take place at various distances from the core, and/or taken from imitation cores. The correlation to conditions in the cpu core are
questionable.

Things are futher complicated by these three recent results. Three different test methodologies by
competent testers has given us this mess to figure out.

TTV test by Swiftech http://www.swiftnets.com/products/Apogee.asp#


Thermal die test by Robotech at SystemCooling http://www.systemcooling.com/swiftech_apogee-09.html


Thermal Probe testing by Robotech posted at ProForums

http://forums.procooling.com/vbb/sho...2&postcount=93

The first 2 contradict each other, the third shows that the last few years of waterblock developement

would have been better spent blowing on our CPU's.

This is not exceptable,so I propose a new method to help us move forward. the key,
http://www.intel.com/technology/maga...ng/it04021.pdf


the Intel Thermal Control Circuit (TCC) controlled by the Integrated Thermal Sensor (ITS) which is totally seperate from the thermal diode we all know and ignore. Located in the hottest part of the P4 core.

I propose changing conditions in a controlled way to reach a known cpu temperature, the TCC trip point. My belief is that conditions can be set that will just trip the curcuit in a pattern related to the load program used. Changing out the water block with no other changes will result in the trip point not being reached, or being reached more often. The difference of clock speed to match the previous pattern will provide a measurement of performance.

Both voltage and clock speed can be used to contol cpu wattage. Ambiant temperature can be controlled, and heat rejection of the waterloop can be controlled. All directly

One program to detect throttling, http://www.panopsys.com/Downloads.html Throttle watch 2.01. It offers real time detection and logging.

If it works there are many questions we can get insight into. P4 w/epoxied IHS can be compared to P4's w/TIM IHS, which can be compared to naked P4's w/o TIM and IHS. Data from the same P4 with and without IHS will be useful. These actual cpu measurements may corelate to a current test method. A posible method of calculating cpu wattage.
For A64 users, any cpu data is better than thermal probe and diode testing.

Although anyone with a P4 can do this testing and get results, the best results will still come from dedicated Water Block testors, with there superior equipement and skill.

I'm sure there are many problem that I haven't thought of, but I think it is worth trying out. Coming to any conclusions about IHS's and WB performance will need the sacrifice of quite a few Northwood IHS's.

EDIT: forgot to post this link , s775 http://download.intel.com/design/Pen...s/30255304.pdf section 4

[BillA]

P4 is dead silicon, and has been for 2+ years
P4s have known/reported Tim1 variation (?)
move forward in time/technology

something in a 775 package ?

[GlassMan]

Same thing will work with all current Intel CPU's. With dual cores as well. The p4's are important because the IHS can be removed and the same cpu tested. This may give clues into how the IHS changes how heat is absorbed by the waterblock.

[BillA]

pointless, Tim1 is different
but do as you wish

my concern with Intel has long been throttling, preferable to measure pwr in or out - correct for losses, etc.
be most interesting to correlate 'your' temp with that of the IHS (Tim1 C/W)

you going to do it ?

[GlassMan]

I don't have a p4 I'm an AMD user, and all different pentiums #'s confuse me
Actually I think the main benifit will be in measuring Water Block performance on an actual cpu. Can opening the p4's is secondary if interesting (to me).

[pHaestus]

For A64 users, any cpu data is better than thermal probe and diode testing.

You don't know what you are talking about. You think what you proposed is superior to my modded 939?

[GlassMan]

I don't know, I was refering to the Robotech results. I didn't mean to imply you couldn't do better than that. When you do the testing and can resolve differences between water blocks we will know if your system is useful.
There are compression issues and resolution issues. (It's been a while since I read your method, I'll read it again). Also as you mention losing pins is always a threat. (Perhaps use 2 hold down plates , the lower with widened holes so you can slowly twist the block till TIM lets loose, the upper to positively locate the block). (I lost my best s754, tear).
I'm all for A64 testing. If the results of northwood IHS on, off corralate with yours, it increases the reliance IHS users can put in your IHS off results.

I think my system has potential*, as we are working directly with the hottest part of the cpu, and the wattage levels generated by FSB adjustments are very small and calcula(ta)ble (% anyway) resulting in a finely graduated scale. Even if the scale doesn't workout, it should answer if block a>b>c, at least on the relevent intels.

I suggest it as a way of verifying the results gotten by other methods more than a replacement. That is possible, but we don't even know if it works. If results can't be corralated, they are all questionable.


* definately with the same cpu, and as I believe the calibration procedure intel refers to is to a set temperature, not to specific cpu's as has been speculated. Trying to get exact heat flux of each cpu and setting the TCC to that is let's say counter to mass production)

[BillA]

GM
unintelligble
can you distinguish between AMD (bare die) and Intel (775 w/IHS) testing ?
they are not now the same, nor proposed to be so

[GlassMan]

Bill, I don't understand what you mean by distinquish. There is no relation. If the system works, it will be excellent for comparing waterblocks mounted on 775's. Since AMD's aren't throttled it doesn't apply to AMD's at all.
I would also be excellent on 478's.
Any differences noted in 478 IHS vs 478 bare will most likely apply to A64 IHS vs A64 bare. In other words if 478 IHS vs 478 bare fall right in line with each other it would imply that IHS testing is comperable to bare testing, and vice versa.
IF that was the case, it would imply that bare die A64 test results were applicable to A64's with a heatspreader.
pHaestus may be able to distinguish waterblocks on IHS A64's. Robotech's results show it is not easy.

[BillA]

I do not think you will find someone (else ?) to bother testing a P4
pointless to test an AMD CPU w/IHS
cannot test a 775 w/o IHS
I have no idea what your proposed matrix is, or the actual methodology

[GlassMan]

Why not, they aren't water cooled? See graph 1 and 2, one or both are wrong as to applying to a cpu. I have suggested an independent way test the performance on a cpu, the P4 or PD
So I said
So I said
I think you have read far more into it than I have claimed.

[Ice Czar]

what I see are two sets of very similar performance curves with shifted definative values

attributing the shift primarilly to the sensors as opposed to other variables seems premature

say for instance
http://www.magnetrol.com/us/html/vie...sp?pdf=mii\552
flow profile (page 19)
since each block has a different flow circuit, its reasonable to assume the starting turbulence level could play a role in the subsequent heat transfer in the block (or simply at the measurement point)
the 2 test setups not being identical, likely not even identical from block to block with a given reviewer (though of course theyd be far closer)

change the circuit length slightly, tubing curvature, pump hieght a little bit and you introduce a variable, one which might favor one block over another
a hypothesis that would bear investigation and can easily be proven or disproven

that is just one variable that immediately comes to mind
I think we are likely seeing compounded variables here that need to be attributed and quantified
there would be literally dozens of small differences between the two basic test systems (graphed)
the call to make incremental changes to identify or ruleout these makes alot of sense to me

actually thats one reason Im working on the environmental test chamber
inorder to control and replicate ambient temperature, pressure and humidity from test to test

[BillA]

"the environmental test chamber in order to control and replicate ambient temperature, pressure and humidity from test to test"
??
done a trial calc to ck the expected difference due to reasonable barometric changes ?
same with humidity ?
and I would be kinda interested in what you see the cost being to control each ?
you seem to be edging towards mass flow control on both sides, in realtime ? (if so, why ?)

- I only note them, and simply stop testing when the equipment is below the dew point (Houston)
on a related note, I'll start a blog on building an instrumented wind tunnel this week

[bigben2k]

Originally Posted by BillA

...on a related note, I'll start a blog on building an instrumented wind tunnel this week

Looking forward to that one, as I have been in such preparations myself. Cheers.

[BillA]

joint effort ? (dual blog ?)

ok, e-mail coming in a bit

[bigben2k]

Am interested, yes.

[Ice Czar]

Originally Posted by BillA

done a trial calc to ck the expected difference due to reasonable barometric changes ?
same with humidity ?
and I would be kinda interested in what you see the cost being to control each ?
you seem to be edging towards mass flow control on both sides, in realtime ? (if so, why ?)

primary reason pressure: the Lab is at 5236 feet, and Colorado has the highest mean altitude of any of the 50 states, pressurizing the chamber is a goal but one that is the lowest of the three.

primary reason temperature: Chamber will be employed in stress testing more than just cooling solutions, want to do decent MTBF runs on PSUs, has the added advantage of eventually determining how a decreased temperature differential effects secondary heatsources in a case and how that can impact the primary cooling solutions. (Im going to be reviewing alot of turnkey solutions, cases ect)

primary reason humidity actually that is how Im going to control the environmental test chamber employing Dalton's Law of Partial Pressures, it will be controlled at a targeted dew point temperature and then reheated to a specific dry bulb temperature
How it will be done.pdf
its easier to build a chipchiller for a swampcooler and add a heater, controling water temperature Id think would be easier than a limited amount of air mass. A conclusion I arrived at when attempting to get a stable ambient in my computer room with just air cooling. (I ended up with a nasty cycle, its a small room)

cost is considerably less than the PSU test equipment

as far as expectations, I dont know, I suspect that as the temperature differential decreases in a case it all goes to hell in a handbasket a little quicker than is currently apprciated in the enthusiasts community especially when you consider the heat from VRMs and additional drives, overwhelming the remaining primary systems. I had a corollary experience of that this summer when I overwhelmed the AC for the computer room, just a different scale. The ambient temperature plays a much larger role in a wholistic view than is credited or illustrated to the n00bs, I mean to address that.

[bigben2k]

Originally Posted by BillA

joint effort ? (dual blog ?)

ok, e-mail coming in a bit

Did I miss your email?

[BillA]

long bit, I'm on it now

IC
I would opine:
pressure can be addressed by calc, same as barometric but a constant
temp is essential, but you will never tie up the chamber doing MTBF stuff (many tests running to failure)
I have speced conditioned air for fiberglassing underground pipeline joints, would never consider such for a control system. Should be interesting, I think the tail will wag the dog.

last paragraph not clear
your sink is the air, in a chamber the air temp is a real temp assuming the chamber can dump the heat being input; you may have to piggyback another AC system to bump capacity (thats what I did with a Forma Scientific incubator, to ~600W capacity)
what are you using for a chamber ?