How to test a homemade block?

[MMZ_TimeLord]

Okay,

I have sat and read everything since my post... and I'm throughly disgusted.

This thread started out as a simple attempt to post some BASIC and "AVERAGE JOE" level testing proceedures to allow a home user to COMPARE HIS OWN BLOCKS IN HIS OWN SYSTEM!!!

It has degenerated into a "your going to make people think they can compare temps of blocks on different systems...". Just stop feeding the troll? :shrug:

That WILL happen whether you make a testing proceedure or not.

If you can NOT contribute to some kind of testing methodology... dont' post please. I am really looking forward to some kind of refined proceedure that us "SHMUCKS" can use to get some idea of whether or not our system is truely being improved or not. Nothing elaborate or highly accurate... just general... like "hey my CPU temp dropped 3 degrees Celcius with just that pump change" just by running tests before and after.

This is not rocket science or nuclear physics... it's simple... and usefull for that casual hobbiest to utilize so they can ENJOY the hobby.

I for one am NOT looking forward to trying to compare my block/system to WW or any other block out there. I am ONLY concerned with MY system and MY changes and how well they do or don't work.

Can we get this back on topic please?

[pHaestus]

Well as long as everyone feels good that's what counts right?

[Khledar]

Lol pH

Don't be so harsh, remember there is no objective truth.


.....

[utabintarbo]

Originally posted by pHaestus
Well as long as everyone feels good that's what counts right?

No, that's not it at all. What Ben is after here is a HOW-TO for us regular people to run a basic comparison with enough credibility that we don't get our ass handed back to us for posting a temp.

I think GTA had it spot-on. If blockA is X degrees better than blockB over several mountings in our system, then we should be able to claim, with some measure of confidence, that blockB is an improvement. Wildly swinging MBM readings should give us pause as to the accuracy of the system, but I believe this behavior to be an exception, rather than the rule(?).

Somehow cathar managed to be taken seriously without all of the BillA test equipment. What is it about his set-up can we replicate so that others can be taken seriously enough not to be dismissed out-of-hand?

Bob

[Tuff]

Is it safe to say...that a MB will be consistant with itself? ie..it will have the same readings day after day...what I mean is if I test a block and have the same conditions room temp...ambient ect...will I get the same results each time.

-------------

If this is true.

We or the DIY guy starts with a single block...gets his average temp...with mounts...with burn programs ect.

He or she can then use this temp as a BASE to start.

Then if the person chooses they can make another block and then look for differences in the temp. Keeping in mind that what we are looking for is the difference...not the actual temp. As I have been told...some MB might read 35 deg...but could actually be higher or lower.


All the above using the exact same procedures for mounting and testing.

------------

If it is not true.

Then there would be NO way to tell what is really going on, unless you have an outside way of finding out.

ie...water in water out.

Tuff

[MadDogMe]

Jesus!, I expected better of some of the people on this forum ...

Is this site still about watercooling or is it just a place to strut and stroke your 'superior' knowledge?. Several years of 'experience' puts you above helping people nowadays?. Let one of the do gooders deal with that crap eh?. All you have to do is sit back and pick holes. If you get tired of repeating something, put it somewhere in an artical (but not THIS site eh?, it'd probly get lost amongst the rest!!) where it can be seen!...


Stop looking at the 'rarified' end of the spectrum(most of it does'nt even bloody apply anyway, it left the point of 'diminishing returns' behind an age ago!) and realise some people need help at the beginning, try to help or don't poke your nose in with rubbish when other people DO try. Some people don't know about basic stuff and if you have that learning curve behind you then you're in the best possible spot to help them. Sitting back and ripping holes without explaing why in a way the person understands is the lowest form of entertainment, but I see people indulging in it here quite frequently nowadays ...


ProCooling!, Where the intelectualy enlightened come to pick holes in lesser mortals once they've outstayed their welcome at overclockers.com?!! :shrug: but as long as they feel good that's all that matters right?...

[bigben2k]

Thank you MadDog, my point exactly. Has ProCooling lost its edge? Is it's reason for existing solely for the "serious" tester, where noobs are not welcome?

Or do we want to use that knowledge and do something with it?


Tuff: I think that the mobo is reliable within itself, but the TIM joint is still a variable, as Bill demonstrated several times.


Let me try to recap what we have so far:

Part #1: perform a flow test.

mount the block and the pump together, nothing else.

Option 1, do the bucket test
Option #2 (more accurate) setup a manometer

Then compare the results to the pump's curve, to determine the pressure or flow rate.


Part #2: measure the performance

Run three tests with a standard HSF (preferably something that OC reviewed ?), re-mounting it each time. Use the same thermal paste than with the waterblock. (should the TIM be "cooked" for a couple of hours?)

[edit]Repeat the following three times, re-mounting the block each time.

Use SiSoft Sandra Burnin

Measure the room temperature (forget about the mobo temperature: it's irrelevant). Use a regular alcohol thermometer. What about humidity?

Measure the water temperature. Use an alcohol thermometer.

State results as a temp difference with the HSF, or as a difference with a previous version of the block. Never state an actual temp (keep it to yourself, or post it at OC )

Understand that the results cannot be related to anyone else's system, because:
a) temps may be reported innacurately, either in your system, or someone else's.

b) The power (watts) that your CPU put in the cooling loop, may be different than someone else, either because of overclocking (frequency and/or voltage), or flow rate, or the TIM joint. That will make a very significant difference.


Have we covered it all? Do the "senior members" approve?

[jaydee]

MadDogMe: How many posts do you have here and how much have you learned? By your posts not much. Maybe sit back, shut up, and learn something useful and post it instead of the flame bait you consistently spew with the do gooder attitude.

MMZ_TimeLord: Did you read the thread? It has all been covered dude. There is no "How To" on this subject and there shouldn't be. Is it really that hard to look at what the thermal software is telling you? That's all you can to do with a mobo. Common sense man. That's what GTA explained.

utabintarbo: Do people really need to be told how to read MBM? That's all you can do with a mobo. I guess you could make a "How To" but it would consist of one short paragraph;

"Load up MBM, Load up CPU Burn, Let run until equilibrium sets in (when the water temp stops rising), Check room temps (at least once an hour while testing), start testing for 12hrs, stop testing, gather results, do it over 2-5 more times. If the results are inconsistent then get another mobo".

I am sure common sense can fill in the minor other things that should be done there aswell.....yeah right eh?

It just baffles me this needs to be said on this forum, especially to people that have almost a 1,000 posts here. I expected BETTER from some of the people here aswell MadDogMe including yourself and especially Ben. It seems we need to pamper the one's that don't have any common sense. Don't tell them they are doing anything wrong or your a flammer! Don't try and keep this site above the others and dumb it down so all the kiddies can do things half assed or the do gooders will come out and start flaming you for knowing something. Heaven for bit there should be anything PRO about this site........ We should change it to AmatureHourCooling.Com to suit some of you much better! Maybe we can get Joe to start up AmatureHourCooling.com so all you that want to spread half assed BS to newbs have a place to do it!

Sorry I cannot be as polite as pHaestus and throw a simple one liner out to cover the same things and think COMMON sense would under stand it.

Maybe some of you should consider starting your own site and cater to the newbs as you wish. Lets keep this place a little more professional eh?

[bigben2k]

Originally posted by jaydee116
...There is no "How To" on this subject and there shouldn't be.

Sounds to me like you don't want to contribute anything: why don't you just stay out of this thread then?

[pHaestus]

What we have here is a theoretical person throwing out what a lot of people should try, and a person with some personal experience saying "that's not gonna work". Then everyone gets mad at the guy who's tried it before? Swell. And just because I have a bit of experience doesnt mean I have infinite patience. Are you asking me if mobos reliably report CPU temperature? **** no they don't! I have been saying that for years. Now you want me to go back and say "ok that's good enough". The compunurses that people use are also pretty awful for temperature measurement.

Did you not read my suggestion above? Get a thermocouple reader that takes two probes and epoxy 1 type T under a ceramic core chip. That is the best way that the layman can get a temperature reading that will have SOME relation to actual CPU measurement. Cost? Prolly $50 on ebay plus the cost of making type T thermocouples. I am NOT just strutting around; all the equipment needed to do what you want was suggested earlier and summarily ignored.

The onboard temperature probes sometimes can't even rank coolers in the correct order. And when they do they typically make them appear to perform closer together than they actually do.

The way it works with "reak world" experiences is that they are only useful if you trust the bearer and if they are done CAREFULLY and with a bit of THOUGHT. That's why Cathar at OCAU and Hoot over at O/cers have been able to do some interesting testing with their systems and not get totally flamed; people trusted their word and they took the time to do the best that they could with what they had.

I don't know exactly what is going on here. Are you guys implying that the majority of DIY waterblock designers dont know how to tell if one block is better from another one? I haven't heard that complaint before now. #rotor's been using the TC under the core tip for years btw. Are you looking for a way that results people post can somehow become valid when using the same crappy equipment? Not likely to happen. You should try to do an uncertainty estimate for a C/W using temp probes that are off by 2-3C each and a power error of 20%. Put those error bars on your result, and conclude that all points with error bars which overlap are statistically the same. Wat I am being the theoretical one then; let me just tell you. Statistically all blocks would be the same then. So why bother testing or tweaking designs or anything?

I stay out of the ridicuous temperature postings threads and certainly don't flame DIYers about their test methods. But I don't care to have that "good enough" methodology bred and encouraged as a valid testing method either. It's a very short leap from there to people basing buying decisions on flawed testing from the forum.

Oh well back to strutting about.

[pHaestus]

Ben have you not read my latest diode article? Was there NOTHING of interest for the current discussion?

Tuff: I think that the mobo is reliable within itself, but the TIM joint is still a variable, as Bill demonstrated several times.

The mobo is not reliable within itself because there is temperature compression that makes the results useless. There are articles on overclockers and void your warranty and even this site (the first diode reader article) that show this clearly. That is why I have little patience; I have written article after article and post after post on this subject and you just keep missing the point of them. Hey I know let's use a mobo. GUN

A bucket? Come on. Get two Ts and plumb them in on either side of the pump and make a manometer. Will be easier and faster and more accurate.

Sandra? Use CPUBurn (aka K7Burn) in high priority. Again I JUST wrote an article...

The insinuation that I am not helpful to the community at large is insulting. I think if you will look at the articles I have written at this site you will find all sorts of useful info for your project. But it seems that evey single comment and suggestion from articles and from the forum is summarily ignored by you, Ben. I have TRIED the methods you are proposing. 2 years ago I started out trying them. I have a big folder of absolute crap data in fact that is sitting on the file server. I have upgraded and revised and then upgraded again to get somewhere close to the test system that I really need. You just sit here and throw out plans and ideas that I have already found unacceptable and expect me to nod my head and give you a gold star for effort?

Let me ask you a question: Would you be satisfied if I wrote articles for Procooling with a bucket and a mobo and Sandra burnin?

[bigben2k]

Sorry pHaestus, but if you took that to include yourself, you were wrong.

Your input is very valuable, especially after all the time you spent with it.

In this case though, given that the results only have value to the tester, aren't we aiming for repeatability? I mean, I thought it was pretty clear that a C/W rating is absolutely out of the question, for a DIY working in his garage/kitchen, and I'd like to think that everyone agrees that an actual temp is also out of the question, which leaves a relative temp, which is also only of value to the tester.

For those interested, here's pHaestus' article.

CPUBurn and toast seem to run very tight code that keeps temperature swings under load to a minimum.

The results also point to the use of a program like CPUBurn to load the CPU rather than one that is dependent upon other system components.

What's your experience with Sandra?

[Joe]

This whole thread is rather interesting. We have 2 factions here as pH mentioned... on theoretical and one experimental who have done this already.

I fully agree with pH's angst towards the whole "good enough" testing setup. Using a motherboard temp monitoring and Sandra is about the biggest waste of time I have heard of. I mean I agree that for home testing of your blocks, it should be inexpensive and also should be consistent.

Why is everyone bucking the thought that you may need to invest like 60$ in a thermometer that may give you some semi accurate temps? or putting some scientific/engineering thought into how you test? I mean what’s the point here? You want to learn how your block performs, or just sorta how it performs in a round about way?

Put some time and money into testing if you want it to give you anything in return. If all you want is shit numbers, use shit hardware, and a shitty testing philosophy.

I personally think that investing in a good thermometer, flow meter, and Manometer are essential to any testing. I think a CPU can still be used as a heat source fairly well... since the costs of going to a simulated Die setup could be over 1000$ alone.

People have mentioned in this thread that they 'expected' more from some people in here... I think this thead rocks, cause there are some down right stupid theories being voiced as fact. I am happy to see Cathar, pH, and the others in here giving some REAL experience based input to try and combat the whole "good enough" disease that has consumed 99% of all websites that think they can "test" anything. Just look at the 3DGheyMan reviews... or hardocp... or system cooling for perfect examples of how degenerate reviews can become if you go by "well that’s good enough".

The main reason I haven’t reviewed jack in the last year is simple... if I cant do it right I am not going to do it, and I haven’t had the time/money to spend on building a test rig that I trust for results. pH has though, and I trust his testing will be quite impressive, the man knows what he’s talking about

[utabintarbo]

Originally posted by pHaestus
What we have here is a theoretical person throwing out what a lot of people should try, and a person with some personal experience saying "that's not gonna work". Then everyone gets mad at the guy who's tried it before? Swell. And just because I have a bit of experience doesnt mean I have infinite patience.

Well, I suppose a little suggestion in the right direction would be appreciated, rather than a simple "that's not gonna work". Not too much trouble, eh?

Originally posted by pHaestus

Are you asking me if mobos reliably report CPU temperature? **** no they don't! I have been saying that for years. Now you want me to go back and say "ok that's good enough". The compunurses that people use are also pretty awful for temperature measurement.

Stipulated.

Originally posted by pHaestus

Did you not read my suggestion above? Get a thermocouple reader that takes two probes and epoxy 1 type T under a ceramic core chip. That is the best way that the layman can get a temperature reading that will have SOME relation to actual CPU measurement. Cost? Prolly $50 on ebay plus the cost of making type T thermocouples. I am NOT just strutting around; all the equipment needed to do what you want was suggested earlier and summarily ignored.

The onboard temperature probes sometimes can't even rank coolers in the correct order. And when they do they typically make them appear to perform closer together than they actually do.

This is helpful, though it probably is ruled out for a fair-sized portion of DIY-ers.

Originally posted by pHaestus

The way it works with "reak world" experiences is that they are only useful if you trust the bearer and if they are done CAREFULLY and with a bit of THOUGHT. That's why Cathar at OCAU and Hoot over at O/cers have been able to do some interesting testing with their systems and not get totally flamed; people trusted their word and they took the time to do the best that they could with what they had.

I believe this trust is the aim of this HOW-TO. Not to the level of those mentioned, perhaps, but useful for comparison purposes.

Originally posted by pHaestus

I don't know exactly what is going on here. Are you guys implying that the majority of DIY waterblock designers dont know how to tell if one block is better from another one?
...

Apparently not, by any possible objective measurement. That is, barring the investment in the type of device you mentioned.

Given the fact that many here are young and not financially "established", and even the more mature can be heard complaining about their finances, I see the market for the TC device relatively small. Should those not wishing to make the investment be left voiceless?

If a consistent methodology can be formulated, we might at least have a fair shot at making a reasonable judgement, without having to send our $7 drilled w/b to BillA for $30(?) worth of testing.

Originally posted by pHaestus

I stay out of the ridicuous temperature postings threads and certainly don't flame DIYers about their test methods. But I don't care to have that "good enough" methodology bred and encouraged as a valid testing method either. It's a very short leap from there to people basing buying decisions on flawed testing from the forum.

...

Caveat Emptor!

Bob

[bigben2k]

Fair enough, let's find out how much a DIY builder would be willing to spend, on measurement tools.

Goto this thread and post your opinion.

[GTA]

Something does strike me about this talk of "good enough"

BillA's tests were fantastic because they were repeatable, even testing one block in the summer, one in the winter, whatever really, made no difference, because of the way his system was set up.

HOWEVER, slight problem I realised today.

Take for example Bills test of the OcPC Atlantis block, he gave it a c/w rating of somewhere between 0.33 and 0.23, depending on flowrate. For this example, CPU wattage doesn't really matter much, so assume 100W for ease of calculation.

so, between 23 and 33 degrees.

Same method with the WW test data gets us 17-22 as a range.

WW wins by between 5 degrees at high flow rates, and 11 degrees at lower flow rates.

Despite the fact that this set of temps was produced using top level test equpiment, you still can't compare it to your own data.

Eg.

I test an OcPC atlantis in my system. I get temps of 50 degrees. Using my limited testing method, above, I build a theoretical block that gets me 42 degrees. Nice one, I've beaten the atlantis by 8 degrees.

Say I'm using a very powerfull pump, can I go onto assume that my block beats the WW by 3 degrees?

No, despite the fact that Bill Adams testing procedure is very accurate indeed.

Okay, so I improve my testing procedure. I know my flowrate accuratly, I know my CPU wattage more accuratly, I measure temperature more accuratly.

My block in my setup now beats the OcPC block in my setup by 7.1 degrees, on average, over many many test runs.

Can I now say my block beats a WW?

Nope, still can't.

This poses a problem. No matter how good my testing is, its still only really going to benefit me. None of you will be able to compare your results to mine, no matter how accurate I am.

And you still won't be able to predict what temps you're going to get if you put a block I've tested in your system.

You tell me your flowrate and CPU wattage, and I can give you the c/w result that I got at that flowrate, multiply that by your CPU wattage, and I'd get a temp.

You won't get that temp in your system.


So, is accurate testing useless? Well, no, it'd help me, because in making changes to my own blocks, I'd be more likely to be able to tell that a physical change to the furniture has made a statistically relevant difference to the performance of the block.

Would my accurate tests of commercial blocks help you compare your DIY efforts to commercial blocks?

No.

My accurate test is useless to you, unless :

1. You send me your DIY block, and I test it.
2. You're shopping for a commercial block ( AKA : Leave now and don't leave ass prints on my door )


Something to think about anyway

[bigben2k]

Good one GTA.

The application here really is limited to the maker, and doesn't have any value outside of that.


Here's another thought:

What if there was a service available to test a block? Would a DIY blockmaker be willing to send it out, each time he made a modification to it, while shelling out X$ each time?

I think not.

So what's he left with?

Could he gauge an improvement based solely on the water temp (and room temp)? I think that the temp difference would fall below what one can measure.

Would a Compunurse have any value here, not for accuracy, but for repeatability, if one probe measured the air temp coming into the rad, and another probe was thermal epoxied to a short section of copper pipe, stuck in the loop? Non-linearity aside, could it indicate an improvement? (in a simpler "Yes or No" statement).

[jaydee]

Originally posted by bigben2k
Sorry pHaestus, but if you took that to include yourself, you were wrong.

No he is NOT wrong. Everything I have been saying in this thread is DIRECTLY related to him. pHaestus is the very reason I think the way I do about testing. All the effort he has put into his testing methods and ideas that he has posted in articles on THIS site to try and raise the level of testing is being severly insulted and ignored by this thread. It is a shame!

You say I don't want to contribute anything which is furthest from the truth Ben. I am trying to contribute the idea that this half assed testing shouldn't be promoted on THIS site. I am trying to contribute the idea that we should step up on THIS site and do things better than everywhere else. I am trying to contribute that the knowlege we have should be improved upon instead of going backwards upon.

Yet I get flamed and bitched at for trying to promote a higher level of learning and execution and get called an arrogant, egotistical, ass hole, and flamer in the process through multiple threads.

Would you be satified Ben if your Radius was designed to be good enough? Did you spend all the time and energy and soon to be money on it just to be good enough? Would you be satisfied with the results from this half assed testing method your promoting here on your Radius? Wouldn't you want to test it decently after spending the time and money making it? Would you settle for half assed results that are not conclusive or accurate by any respectable degree?

Sure the DIY'er may not want to spend money on testing. Then he/she can go somewhere else and post his "leet" temps eh?

There is plenty of sites that cater to newbs and half assed procedure. This should not be one of them and I think Joe has made it clear several times he isn't interested in a dumbed down site catered to newbs. Correct me if I am wrong Joe.

[jaydee]

Originally posted by bigben2k


Would a Compunurse have any value here, not for accuracy, but for repeatability, if one probe measured the air temp coming into the rad, and another probe was thermal epoxied to a short section of copper pipe, stuck in the loop? Non-linearity aside, could it indicate an improvement? (in a simpler "Yes or No" statement).

No, and you can get much better probes for $20-$30.

[bigben2k]

Originally posted by jaydee116
I am trying to contribute the idea that this half assed testing shouldn't be promoted on THIS site. I am trying to contribute the idea that we should step up on THIS site and do things better than everywhere else. I am trying to contribute that the knowlege we have should be improved upon instead of going backwards upon.

So newbs aren't welcome?

I'm hoping to straighten up this half-assed testing, into something that would have a reasonable level of acceptance, with a minimal cost, as I believe that that's what the newbs are after.

I'm not aiming for perfection here, just something that's actually useful. If you want to fork out $2'000 for a test bench, go ahead, but you can't expect each and every home/DIY block maker to do the same now can you?

Can you put yourself in their shoes for a minute, with your knowledge and experience, and come up with something that they can use?

I think we all agree that there is virtually no value in home testing, the way we've often seen done.

So is there a way for a home/DIY maker to run any kind of reliable test, or is it completely hopeless?

[pHaestus]

Why not go with a finger? Just touch the block.

Are you asking if you will get the same reading every time you use a thermistor? Nope (hence the services which calibrate thermal devices!) Thermistors are typically more linear than thermocouples though but with much worse accuracy for the type you are proposing. Every time you move the thermal sensors around and power down/on the readers you will probably get some issues. The worst is cutting the wires. Compunurses are terrible because every time the probe gets bent then it can affect the reading also. Best to calibrate often and well.

Many DMMs can accept a thermocouple; doesnt everyone have a multimeter around or can at least find a use for one? I use my DMM all the time for electronics projects. www.Allelectronics.com used to sell them for $20. Get a DMM and rig up a design where thermistors can be inserted into the loop reversibly (I have a suggestion on this actually that I will post tonight if I get a chance to run by home depot). Calibrate your thermistors vs the thermocouple every month or so (or more often if you change things around). Then use the DMM for measuring the temperature under the CPU core and the thermistors for waterblock in and out and radiator in and out and air in and out of the radiator. A digidoc would accept enough thermistors actually; don't know much about the quality of their electronics.

Use the manometer for flow rates assuming that your pump has a reliable P-Q curve (the Via Aqua is IMO pretty far off). Total cost of that is about $5 for 2 T fittings and a bunch of vinyl tubing. And there you have it.

The big problems to deal with are the issues with motherboard readings for CPU and the need for regular recalibration of the cheapo thermistors.

[pHaestus]

The whole impetus for this thread and the bulk of the complaining about temps people post is the shittiness of motherboard measurements. Given that some people see motherboard CPU temps only 3-4C over the room temperature under load, there is no way those people can compare the performance of different blocks on their systems. Much less compare from person to person (I would just give up on that).

[pHaestus]

The calibration is a PITA with a digidoc though because you will need to label the probes and calibrate them all and save all the curves in Excel. Once that is done then you can do the testing and record the numbers and then go back into Excel and throw the numbers into the linear regression equation from your calibration data to get a real number.

I will test Sandra CPU burn in tonight; I will have to dig up a heatsink; have a waterblock on that test setup atm.

[bigben2k]

Good input, I like the sound of it!

So the Compunurse is ruled out, period. (it'll have a remaining usage as a quick visual check of the system, in case something gets exceptionally hot).

I think that the $50+ Digidoc is going to be out, but we'll see (pending survey).


Jaydee: what probe setup did you have in mind, for $20-30?



All: I'm personally keeping in mind that whatever anyone says, should be confirmed by at least two other sources (which makes anything difficult to do around here, BTW). So anyone please, if you have an opinion, voice it!

[bigben2k]

Originally posted by pHaestus
I will test Sandra CPU burn in tonight; I will have to dig up a heatsink; have a waterblock on that test setup atm.

Thank you! I was hoping you'd look into it.