Water block Writeup with the big dogs!!!!

[webmedic]

No he has not responded yet.

[toddalaska]

IU would like to see the tc 4 in there too, I just ordered two of them yesterday to go with the stuff Im buyiung from webmedic

[bigben2k]

Looking forward to see your latest set of tests, that you intend to run.

(PS: sorry that BillA is giving you such a hard time at OC about it!)

[webmedic]

I will proceed as is for now but I will work on getting some "other equipment for testing" and I will revisit these blocks at that time.

Yes I'm writing up the methodoligy right now. I've got half a page and am still discussing setup and whatnot. When I get done with it I will post it on my site so that I can easiley track comments on it and change the spec. After testing starts no changes will be made. If you would rather have it posted here speak up. I'm open about it. The only reason I wanted it at my place was so if things get out of hand I can ban the rebel rowsers.

Actually scratch that I just changed the setting to be moderated by the users.

[jaydee]

*Lowers and Shakes head*

I just read the OC thread and think they are waaaaay to concerned with precision accuracy. I like accurate results but come on, we are talking about water blocks for computers not missle guidence systems!!!

The results you have will only be so usefull anyway. NO ONE will be able to replicate your results because every cooling system and the systems enviroment is different!!!! One system may work better with a Maze3 and not as good with a Spir@l and another may work better with a Spir@l and not a Maze3!!! In order to get real usefull results to base upon a purchase you need to test every single possible flow rate with every single possible setup in every single enviroment there is!!!! Needless to say I don't have the Billions of dollars to make such a facility!!!

I think we all just want to see if there are any MAJOR differences in the retail blocks out there (like 3-5C differences with a few different flow rates thrown in). That does not require precision testing equipment. Hell I would be happy if you just filled up a 1 gallon jug and timed it for the flowrate!!! Would be close enough for me as long as every block tested is tested identical it should be fine.

Glad I make my own blocks.

[webmedic]

Be carefull about the jug. I'm actually working on a presicion test for this that will be beyond what bill would even call acceptable. I don't want ot say more in case I dont get it done on time but It's in the works.

[jaydee]

Originally posted by webmedic
Be carefull about the jug. I'm actually working on a presicion test for this that will be beyond what bill would even call acceptable. I don't want ot say more in case I dont get it done on time but It's in the works.

If you can do it by all means do it!!! Glad you are making the time to do this stuff.

[BillA]

perhaps more useful than shaking one's head
would be pushing a pencil



for those who have a browser that can connect to other forums,
the OC thread that has been referred to

sharpen that pencil to a very fine point,
and one of you 'testing experts' please inform us all of the minimum resolution necessary to display what’s on the above graph
- for the temperature, flow rate, and applied heat

you would send 'the innocent' on a fool's errand
get real
cut the uninformed opinions and show some numbers

is that not what you all expect of webmedic ?

on a more positive note:
I would like to know what is going on at ~0.7gpm ?
thoughts ?

[webmedic]

To make it more clear my flow test will be able to show the actuall flow a specific pump has when used witha specific waterblock. The measurement should be precise.

It will not however make it so that each block will be tested at exactly the same flow rates. This obviously would have to be controled with a flow meter, control valve, and optionaly with a valve to relieve back presure.

What will be shown is that if one block flows better then it will give better temps with the same pump than another block that does not flow as well if all other characteristics of the block are the same.

So in a normal computer a normal user would see the same simply because the increased flow whould allow the block to perform better not because the block cools better. Am I making sense. I'm not sure if my terms are right but my idea is.

At ~0.7gpm there is no temp difference between the tested blocks. They cool similarly based on your data. I don't know the math right now so I'm unsure of the minimun resulution to show this data but from temp data aluded to in the other thread it would be .01. I'm unsure though without seeing the actual temp data. You did mention that the variance was only .7c in your tim comparison. With the margin for error with current over the cunter measuring devices as .5c the differences would be indistinguishable with normal measuring devices.

I'm sure my spelling is of somewere up there and I have fat fingers so excuse the spelling.

[webmedic]

Ok I finally Have my rough draft up for the testing methods.

Please review the information and post comments. This is a rough draft and as such is up for revision as needed.

http://www.water-cool.com/modules.ph...owcontent&id=3

[webmedic]

Sorry I broke something and anonymous users can't post please post your comments here.

[FRAGN'STIEN]

I agree with jaydee. This is not a missle guidence system. billa, you shouldn't be so critical of webmedics work. You should be encouraging him not criticizing him. Are you afraid of a little competition? Let the man get on with his test. Your the only one in these forums giving him a hard time. Every one else here would like to see the results.

[BillA]

who would not like to see the results ?
the verb is "see"
but if he cannot measure it, what's to see ?

cumulative error masquerading as 'results' ?

why might it be that I am "the only one in these forums giving him a hard time" ?
(which is not correct, I've been answering his questions for several months)

could it be related to my being the only one with experience in measuring wb performance ?

I'll leave you in tranquility,
but one of the staff here I suspect will be just as difficult

[webmedic]

Yes when and if the ph man gets into this it will allot the same. Even so he recently did a review without going all out. It was a nice case to. With stuff like that available makes you wonder why people buy koolance.

[Cova]

Ok - after reading this thread, and very quickly scanning the OC thread, I've got a few suggestions...

1. As far as flow is concerned - Measure each block once in a loop of its own (eg. using #rotor's method), or perhaps a couple of times and average the result, post that flow-rate for each block you test and then forget about flow-rates for the rest of the experiment. Most of us don't know what our flow-rates are, and don't really care I think. All we know is that we have a pump rated at eg. 400GPH at 0 head - so tell us how a block performs with pumps of various flows so that when we buy a block we know if we can get by with a cheap pump or if we need to go buy a big high-flow pump.

Actually - for the flow tests, get a larger bucket for the source for #rotor's method (larger = more accurate, it should take at least a minute to drain or the variance in you stopping the watch will be too much) and drain it through just a straight piece of tubing. Then (using the same piece of tubing) test all the blocks. Then you can give as a percentage the amount ot change of flow each block showed. That percentage could not be directly applied to a system (eg. it won't kill 25% of the flow in an already much restricted system with rad) but it would make for an easy comparison of block to block which are more/less restrictive and by about how much.

2. For temp monitoring, IMHO a CPU for a heat-source and the on-board temp diode of an XP is plenty accurate. You should note though that most of the hardware readers for that diode have a precision of 0.1C, but an accuracy of +/- 1C. 1C is within the margin of error of your thermal paste and other variances in the system though, so I wouldn't worry about it. I'd like to see temps collected of the core, and coolant into and out of the block, pump, and rad - which should be 4 temp sensors from the digidoc (which again has a precision of .1C, but it's accuracy is only about +/-2C because thermistors really aren't very good). One sensor after the res (eg. pump intake), one sensor between pump and block (pump output and block input), one sensor between pump/rad (block output and rad input), one sensor after rad. Of course if you order your parts differently the sensors will tell you different things, but it should still be 4 sensors. Also of course ambient temp, and if you have extra sensors a air-temp of rad exhaust air. In my system at home I also have a sensor on top of my block (yes, Maze-2 with copper top or you would have to use a side of block or something) for block temp, but I'm not sure if it's that useful, I just had one left over - perhaps if you could attach it to the bottom of the block near the core you could get an idea of how well your TIM is working (eg. if one block reads really cool on that sensor likely the AS3 application went bad).


Overall - I'm really looking forward to this article. I think a good analogy would be to compare to a CPU review/roundup and to say that Billa's tests would be your synthetic CPU tests, while your roundup/tests would be your real-world CPU comparisons. Good luck in your reviewing.

[bigben2k]

Are you testing a Palomino or the new Thoroughbred core?

[webmedic]

In the article it lists the hardware used. I will be using an XP1600. As far as I know the new cor ehas only been released in 2200 but even still 1500 and 1600 have been discontinued. The nice thing about this 1600 is that I've had it at about 1995mhz. I could test it at the same level as most any amd cpu on the market and then some. Well within reason it will put out more whatts due to the higher voltages but you get the idea.

Thats part of what the final part of the roundup will be about. Which ones allows the most oc. All things being relative of course in my system and only if yours is like mine. The standard discaimer.

[webmedic]

Surly joe will be sending his famous bee hive block for the test. Ypu can find it here. I do believe its the one at the bottom of the page.

http://www.spodesabode.com/content/article/block

[pHaestus]

Need to stop and think about uncertainty and propagation of error. Need to think about what using a Compunurse type thermistor means to uncertainty (they are usually +/-3C). Crunch some numbers:

http://www.globalspec.com/cornerstone/ref/PartII.html

Considering even a 0.5C uncertainty in the water inlet and the "CPU temp". I am being generous in that error, BTW. Now throw a flow rate error of 10% into the mix. That's also generous. Now look at the variance in Bill's data. He (and I) aren't using sophisticated equipment because we just like spending money on testing blocks; it is required for confidence in the results!

It doesn't matter if your data points are separated when the error bars are much bigger than the spread in performance. Especially when the differences in block mounting are sometimes as large as the difference in block performance and when a 10% deviation in flow rate can skew the rankings completely.

Remember the presidential elections? That was an example of declaring a winner when the numbers were within the margin of error.

The errors of the equipment you are going to use are large, the standard deviation among successive tests is large due to random effects, and the actual differences between blocks is extremely small. Even without a technical background, you can see this is "a bad thing".

If you enjoy playing around with waterblocks, then by all means continue. I am sure a lot of people will appreciate your results.

But disparaging people who are doing quantitative testing rather than semi-quantitative roundups doesn't make anyone look good.

[pHaestus]

I should add that I am hard on other testers before they start because I personally bit off more than I could chew when I started testing waterblocks. Manufacturers wanted me to review blocks when I first made that diode reader, and I naively thought I could handle that no problem.

Many hours later, I realized my equipment was missing some components. So I got a flowmeter, then realized it wasn't precise enough and also killed flow rate. So I got a better one. The same is true with water temps and CPU diode temps. I needed pressure drop numbers too. And things continue to spiral out of control (looking at recirculating water baths now, and I still need to switch from silicone to copper tubing for my test area).

And then when my test equipment is finally adequate, I'll have hell with reproducibility due to paste application. Still, progress comes slowly.

[webmedic]

Hey you finaly made it to the party this is being discussed already and in far more depth at overclockers. I've also had personnal contact with billa about just this thing already. I'll put forward the same chalange to you. Feed me links to places I can find the stuff. I still need allot of the equipment and have already started to gather other parts like die simulator, flow meter and what not.

Unfortunately for now I still have to get this out. It will still be relevant for if a user does not have this test equipment how in the world will he get your or mine or bills results. He can't. What he will see will be fairly close with the same equipment. Obviously within a degree or two. Now I will still have most of these blocks and will continue testing as the equipment gets better. Plus I have a few things that you guys arent using. Not really helpfull in data collection but it shows interesting things.

Look here for what I'm talking about.
http://www.water-cool.com/modules.ph...orderby=titleA


Only thing for this to be usable the die simulator has to be finished first. It will however give an interesting idea of how the heat is disipated in the block whithout having to purchase 30000$ simulation software.

[webmedic]

Almost forgot. I do agree for more exact testing it will be interesting to note however who will that level of testing be the most beneficial for? The end user who cannot ever recreate those results? The companies that make the product? Both? If we assume both then why is this so?

Any way my bet is that it will the most beneficial to the companies to be able to see this kind of result. Were as some of the end users only want to see how well will pump a perform with waterblock b and rad c. Of course there will be a 2c or so margin for error at the most but for the end user is 2c close enough for them to make a determination? It will probably be closer to .5c but as billa pointed out in his test equipment he only found .7c difference between the highest and lowest. Will or can they ever be able to get the same results as you, I or bill if they are not using a high presure pump. The answer here is of course no. Some blocks will have a higher pressure drop witch will in turn cause that system with pump b to perform at a less than optimum level as compared to the same system when using a differrrent block with less restriction.

Anyway enough. Suffice it to say the testing will be having a second round when I've collected all the pieces for the next level of testing. I like you want to see these results.

[#Rotor]

but you said you are going to be taking a bunch of readings, and then average them, correct? so what is the problem...

accuracy in equipment is nothing more than a factor of tolerance ( and price, for that matter), and tolerance is a mathematical entity that can be reduced by the number of instances being averaged, with the addition of time spent, of coarse.

I guess the real quest will be to find the sweetspot, where money spent on testing equipment, will complement the time spent on doing so, without rendering the tester financially deceits.

[webmedic]

Ah yes on the page I posted showing my methodoligy I stated 5 runnes with the highest and lowest tossed out. The remaining 3 will be averaged. That is unless I get such strange varuations that I have to throw outthe whole batch and retest. If that happens I will do so.

[gone_fishin]

You will not be able to tell the difference from block A to block B with your equiptment so why keep pretending that you will? I will refrain from submitting one of my blocks to your roundup. An error in the output of your inadequate equiptment could cost someone some sales. I am surprised at the number of willing entries you have recieved.
Sorry if it sounds harsh but I can see it no other way after long carefull review.