How to test a homemade block?

[bigben2k]

Following this thread, I propose that we, together, come up with a way to test a home made waterblock.

First, I have to stress that this is strictly for a single user, making a single block, with extremely limited resources.

Here are a few references, for the purpose:

Bill Adam's DIY Waterblock roundup

Bill Adam's testbench

Bench vs system testing: pros and cons (Bill Adams)



#1: First, I'd like to check that everyone agrees that a flow test is important, if not at least for the increase in understanding about pumps and pressure, at least to gain an understanding about the block's performance.

How?: The old, innacurate bucket test. Take two buckets, one full, one empty, and run the water from one bucket to the other, and time it. The longer the test, the more accurate it will be. This should be done with the pump and block alone. We can then have a rough idea of what the block is capable of, using the pump's curve.

Error: the added tubing, the height difference (if any) between the two vessels, and variations in the pump's performance, compared to the curve.


Now, let's see what we can do about the actual performance of the block, shall we?

What *load* program should we recommend?

Can we assume that most people have a HSF, for reference purposes?

Can we trust a manufacturer's program (i.e. AsusProbe), or should we stick with the popular MBM, or just the BIOS (for *idle*)?

Can we assume that MBM reads the CPU diode accurately for any mobo?

I think we can all agree that although knowing exactly which CPU was used, there is no point in claiming an actual power output from it. So if we can't measure Watts, are we limited to temperatures only?

Using a reference (i.e. the HSF), can one state temperature differentials, and trust those readings?

Is measuring the water temp irrelevant?

Can we establish, for certain, the amount of time that the system needs to "sit still", in order for temperatures to stabilize?

About the TIM joint:
Would it be fair to say that mounting it three times (let's forget about ten, for now) would give us a fair assesment of the performance?

Should it be left to settle for 48 hours, each time, or can it be "burned in", and for how long?

Should we just skip the paste, and use a thermal pad, just for testing?


About the room temp:
I think it's necessary (!), but I don't see the point in stating the mobo's temp. Can we state, once and for all, that the room temperature has to be measured, using a good old fashion thermometer?

Is the humidity relevant?



Many Q's... IMO, all a home user can do, is state a temp difference, as an inprovement over a HSF, or a previous version of his block, at Idle, and at load.

Let's clear all this up, once and for all, OK? We can make an article out of it, and reference it, instead of flaming the new faces that show up in here!

[MMZ_TimeLord]

This sounds like a great idea for us n00bs out there...

Without even a sensor I could tell you that going from that 900Mhz (OCed to 1.1Ghz) Celeron to a P4 2.66 Ghz (OCed to 3.00Ghz) has REALLY raised my water temps.

The water itself feels warm to me. Not hot... but warm.


My suggestions...

As for flow through the system. I think that just about anyone could get a couple of 5 gallon (minimum) paint buckets and do the flow with that. It should be fairly accurate.

The height of the tubing to get over the top of the bucket should cancel each other out if the middle both buckets are the same height from their connection to the system.

You could also "approximate" the physical layout of your system on the bench for the flow test by simply putting boxes under each item so it's the same height from the pump as it would be in your system.


After the system is installed...

I think just about anyone (if they have it) can use a room thermometer and their old HSF setup for reference and MBM5 for the CPU measurements with something like SETI or Folding@Home running to push the CPU to 100%. (my F@H runs 24/7) I think that as long as the same "load" application used before and after, then it should be okay.

As for time to stabilize the temps... I would say let the load application run for about an hour.

Not super accurate... but it would be adequate for most people.

[Tuff]

Thx BB

Tuff

[leejsmith]

i use sandra burnin to test because it records the results while doing the test.

i tried toast but the temps didnt get as high as sandra.

combine this with a water probe , room temp probe and cpu probe under the cpu it should provide all the informaton needed.

mbm can do the same but it's limited to a thousand records.

i look for (cpu probe - room temp) and (cpu probe - water temp) for my results.

i agree with the bucket test very easy to do.

normaly i let the system sit at idle for 1 hour and then run sandra burnin for 1 hour. taking temp readings at the begining of idle ,begining load and end of load.

i test againts a maze 3 and will do 3 tests on each alternating the blocks each time.

at the end you can copy and past the sandra results into note pad and name the file based on block and room temp.

example

SiSoftware Sandra

Run 1 executing...
Board Temperature : 24.0°C (Min 24.0°C; Avg 24.0°C; Max 24.0°C)
CPU Temperature : 36.5°C (Min 36.5°C; Avg 36.5°C; Max 36.5°C)
CPU Fan Speed : 2948rpm (Min 2948rpm; Avg 2948rpm; Max 2948rpm)
CPU Core Voltage : 1.62V (Min 1.62V; Avg 1.62V; Max 1.62V)
CPU Core Power : 87W (Min 87W; Avg 87W; Max 87W)
CPU Cooling System Thermal Resistance : 0.14°C/W (Min 0.14°C/W; Avg 0.14°C/W; Max 0.14°C/W)
CPU Arithmetic Benchmark : Started on 10 June 2003 at 21:47:13...
CPU Arithmetic Benchmark : Finished on 10 June 2003 at 21:47:55.

Run 2 executing...
Board Temperature : 24.0°C (Min 24.0°C; Avg 24.0°C; Max 24.0°C)
CPU Temperature : 39.5°C (Min 36.5°C; Avg 38.0°C; Max 39.5°C)
CPU Fan Speed : 2948rpm (Min 2948rpm; Avg 2948rpm; Max 2948rpm)
CPU Core Voltage : 1.60V (Min 1.60V; Avg 1.61V; Max 1.62V)
CPU Core Power : 85W (Min 85W; Avg 86W; Max 87W)
CPU Cooling System Thermal Resistance : 0.18°C/W (Min 0.14°C/W; Avg 0.16°C/W; Max 0.18°C/W)
CPU Arithmetic Benchmark : Started on 10 June 2003 at 21:47:55...
CPU Arithmetic Benchmark : Finished on 10 June 2003 at 21:48:37.

Run 3 executing...
Board Temperature : 25.0°C (Min 24.0°C; Avg 24.3°C; Max 25.0°C)
CPU Temperature : 38.5°C (Min 36.5°C; Avg 38.2°C; Max 39.5°C)
CPU Fan Speed : 2948rpm (Min 2948rpm; Avg 2948rpm; Max 2948rpm)
CPU Core Voltage : 1.55V (Min 1.55V; Avg 1.59V; Max 1.62V)
CPU Core Power : 80W (Min 80W; Avg 84W; Max 87W)
CPU Cooling System Thermal Resistance : 0.17°C/W (Min 0.14°C/W; Avg 0.17°C/W; Max 0.18°C/W)
CPU Arithmetic Benchmark : Started on 10 June 2003 at 21:48:37...
CPU Arithmetic Benchmark : Finished on 10 June 2003 at 21:49:18.

Run 4 executing...
Board Temperature : 25.0°C (Min 24.0°C; Avg 24.5°C; Max 25.0°C)
CPU Temperature : 41.0°C (Min 36.5°C; Avg 38.9°C; Max 41.0°C)
CPU Fan Speed : 2909rpm (Min 2909rpm; Avg 2938rpm; Max 2948rpm)
CPU Core Voltage : 1.54V (Min 1.54V; Avg 1.58V; Max 1.62V)
CPU Core Power : 78W (Min 78W; Avg 82W; Max 87W)
CPU Cooling System Thermal Resistance : 0.20°C/W (Min 0.14°C/W; Avg 0.18°C/W; Max 0.20°C/W)
CPU Arithmetic Benchmark : Started on 10 June 2003 at 21:49:18...
CPU Arithmetic Benchmark : Finished on 10 June 2003 at 21:50:00.

Run 5 executing...
Board Temperature : 25.0°C (Min 24.0°C; Avg 24.6°C; Max 25.0°C)
CPU Temperature : 41.0°C (Min 36.5°C; Avg 39.3°C; Max 41.0°C)
CPU Fan Speed : 2948rpm (Min 2909rpm; Avg 2940rpm; Max 2948rpm)
CPU Core Voltage : 1.57V (Min 1.54V; Avg 1.57V; Max 1.62V)
CPU Core Power : 81W (Min 78W; Avg 82W; Max 87W)
CPU Cooling System Thermal Resistance : 0.20°C/W (Min 0.14°C/W; Avg 0.18°C/W; Max 0.20°C/W)
CPU Arithmetic Benchmark : Started on 10 June 2003 at 21:50:00...
CPU Arithmetic Benchmark : Finished on 10 June 2003 at 21:50:41.

Burn-in Wizard
Finished Successfully : Yes

[jaydee]

Originally posted by bigben2k
Following this thread, I propose that we, together, come up with a way to test a home made waterblock.

First, I have to stress that this is strictly for a single user, making a single block, with extremely limited resources.

Fail to understand why it is so important that a single user's block really needs tested. If the user built it, and he/she is happy with it, then whats the point? Bragging rights? Is this what we are going to do, make a bragging right test method?

Here are a few references, for the purpose:

Bill Adam's DIY Waterblock roundup

Bill Adam's testbench

Bench vs system testing: pros and cons (Bill Adams)

We should all book mark these.

#1: First, I'd like to check that everyone agrees that a flow test is important, if not at least for the increase in understanding about pumps and pressure, at least to gain an understanding about the block's performance.

Well I guess. Not sure what it really matters if it is just a DIY block though. But if the block is going to be modified then it can be of some use.

How?: The old, innacurate bucket test. Take two buckets, one full, one empty, and run the water from one bucket to the other, and time it. The longer the test, the more accurate it will be. This should be done with the pump and block alone. We can then have a rough idea of what the block is capable of, using the pump's curve.

Error: the added tubing, the height difference (if any) between the two vessels, and variations in the pump's performance, compared to the curve.

Probably be better to have a bucket marked at a certain point then the source of the water be larger than that marked pont. Your not going to be to sucessfull in getting the last half gallon out of the second bucket in other words. I used to just fill the bath tub up and then pump in to a 5 gal bucket that was market at 5 gallons and then timed it. Do it 5 times and pull an average.

Now, let's see what we can do about the actual performance of the block, shall we?

What *load* program should we recommend?

Toast, CPU Burn, something of the sort. Need to come up with one standard one that everyone would be using.

Can we assume that most people have a HSF, for reference purposes?

I would think so.

Can we trust a manufacturer's program (i.e. AsusProbe), or should we stick with the popular MBM, or just the BIOS (for *idle*)?

I always liked MBprobe myself. http://mbprobe.livewiredev.com/ Seems to read the same as MBM but a hell of a lot simpler to use.

Can we assume that MBM reads the CPU diode accurately for any mobo?

Not really. We also need to know which probe it is reading off of, internal or external.

I think we can all agree that although knowing exactly which CPU was used, there is no point in claiming an actual power output from it. So if we can't measure Watts, are we limited to temperatures only?

I would think so. The watt calculators are worthless. Each board will run a CPU a little different than another even with identical settings so power usage in the equasion really isn't a option.

Using a reference (i.e. the HSF), can one state temperature differentials, and trust those readings?

Probably not.

Is measuring the water temp irrelevant?

Stumped on this one.

Can we establish, for certain, the amount of time that the system needs to "sit still", in order for temperatures to stabilize?

No. It takes how ever long it takes. Going to vary depending on the system. But it should be easlily figured out by the tester when equelibrium has set in.

About the TIM joint:
Would it be fair to say that mounting it three times (let's forget about ten, for now) would give us a fair assesment of the performance?

Not sure on this one. I would be pretty sure of myself at 3 times, but what about a newb that hasn't installed blocks 100's of times?

Should it be left to settle for 48 hours, each time, or can it be "burned in", and for how long?

Don't know.

Should we just skip the paste, and use a thermal pad, just for testing?

Don't know.

About the room temp:
I think it's necessary (!), but I don't see the point in stating the mobo's temp. Can we state, once and for all, that the room temperature has to be measured, using a good old fashion thermometer?

Would think so.

Is the humidity relevant?

Could be, does moist air cool better than dry air?

Many Q's... IMO, all a home user can do, is state a temp difference, as an inprovement over a HSF, or a previous version of his block, at Idle, and at load.

Still don't understand what idle temps have to do with anything.

Let's clear all this up, once and for all, OK? We can make an article out of it, and reference it, instead of flaming the new faces that show up in here!

Don't think it will be all to easy. And telling someone their testing method is BS is not a flame.

[bigben2k]

Originally posted by jaydee116
Fail to understand why it is so important that a single user's block really needs tested. If the user built it, and he/she is happy with it, then whats the point?

Ok, since I wasn't clear, let me detail it:

You're at home, and you get the crave to build a waterblock, after going over whatever Forum.

You come up with an idea, and decide to give it a shot, so you build a prototype.

Then you show it in the forums, and pick up a few more tips, to improve it, which you want to try.


At this point, you really want to have some idea of how much better it's going to be, maybe because you might try a different design alltogether, later.

Using your logic, one would throw their hands up, and call it quits. Is that what you really want to see happen, Jaydee?

[pHaestus]

Look at what your starting parameters are:

A bucket to measure flow rates

An assumption that every motherboard CPU temperatures are comparable (we HAVE gone over this before, haven't we?) This is patently untrue.

There is no difference in your proposed "testing" and the worthless "Post your CPU temps" threads at every overclocking forum throughout the web.

What can you reliably measure?

Water temperature and air temperature: Sure if you buy a decent device to calibrate everything. There's a reason the $10 Compunurses don't come with tracable certs...

Flow rates? Well maybe if you settle upon a low enough rate and everyone buys a ball valve and is equally adept at "bucket filling". I would kinda point to this as an "unlikely" measurement. Why not make a manometer like Owen Stevens and then use pump's P-Q curve? Would probably be more accurate than your bucket...

CPU power? Another big ****ing unknown. Changing RAM timings, changing FSB, using different motherboards (they all give difft voltage eh?) and using different CPUs to start with. Gonna just blindly assume Radiate is right? It's not.

What you will end up with is super low C/W values because you will overestimate W (radiate is crazy high) and underestimate delta T (by a fair amount most likely).

It sounds to me like you want to do "feel good" crap rather than real testing. That's fine but I wouldn't expect to get it posted as an article on THIS site.

[N8]

Originally posted by pHaestus


It sounds to me like you want to do "feel good" crap rather than real testing. That's fine but I wouldn't expect to get it posted as an article on THIS site.

Heheh, taking over for BillA? I like this kind of response.......

[bigben2k]

I didn't expect that from you, pHaestus.

This is about how the average user can make some measurements, of his home made waterblock. We all know the problems, and the limitations, and it's very clear that a C/W rating is absolutely out of the question, without a proper testbench, and I don't believe that anyone suggested that it was possible to do that at home, with kitchen items!

But...

What can a user do, to measure, or at least confirm, that a modification to his home made waterblock is an actual performance improvement?

I like the manometer idea, and I agree that it would be more accurate than the bucket test.

For those who don't know, a manometer, in this case, would consist of two tubes, connected at the pump's inlet and outlet via a tee. What you do is fill it up with water, about half way, then turn the pump on. You then measure the difference in level of the water, between your two tubes. That result is your pressure difference. Since a pump will provide different flows at different pressures, it's then possible to look up the pressure on the pump's curve, to determine the flow rate.

Of course this assumes that the block has a reasonable flow restriction. For reference, an Eheim 1250 has a max head (water column) of 2 meters (~ 6 feet).


I think a few of us need our egos deflated, and come down for a "walk in the trenches".

[bigben2k]

Originally posted by N8
Heheh, taking over for BillA? I like this kind of response.......

Now wait a minute...

Is this forum for entertainment, or is it actually meant to be useful?:shrug:

[pHaestus]

Yea perhaps a bit harsh but jeez.

We've been through this a few times before for sure. If really interested in some practical testing for those with more sense than to try and do it with error bars perhaps some logic could be applied.

1) Everyone will need the same CPU and IMO it should be a ceramic based AMD (TBird or MP). Take that CPU and thermal epoxy a thermocouple or thermistor under the core's center. Then cut away a bit of the socket to run the wire out. Then you can use AMD's tech doc to extrapolate the CPU die temp from the reading of CPU under. You will want to do a voltage modification on your board so you can tune the voltage under load to the desired amount. You'll want to do a mod to the socket so you can record Vcore as well rather than relying on the motherboard's circuitry. Use a CPU loading program that is insensitive to fsb and ram timings (I suggest CPU Burn). That's a decent start towards cross-person comparability.

Is any of this really "bench testing"? No because the power of the CPU cannot be quantified.

[bigben2k]

Originally posted by pHaestus
Yea perhaps a bit harsh but jeez.

We've been through this a few times before for sure. If really interested in some practical testing for those with more sense than to try and do it with error bars perhaps some logic could be applied.

1) Everyone will need the same CPU and IMO it should be a ceramic based AMD (TBird or MP). Take that CPU and thermal epoxy a thermocouple or thermistor under the core's center. Then cut away a bit of the socket to run the wire out. Then you can use AMD's tech doc to extrapolate the CPU die temp from the reading of CPU under. You will want to do a voltage modification on your board so you can tune the voltage under load to the desired amount. You'll want to do a mod to the socket so you can record Vcore as well rather than relying on the motherboard's circuitry. Use a CPU loading program that is insensitive to fsb and ram timings (I suggest CPU Burn). That's a decent start towards cross-person comparability.

Is any of this really "bench testing"? No because the power of the CPU cannot be quantified.

That's outside the scope, and topic of this thread.

Let me repeat: You're at home, and you get the crave to build a waterblock, after going over whatever Forum.

You come up with an idea, and decide to give it a shot, so you build a prototype.

Then you show it in the forums, and pick up a few more tips, to improve it, which you want to try.


At this point, you really want to have some idea of how much better it's going to be, maybe because you might try a different design alltogether, later.

Using your logic, one would throw their hands up, and call it quits. Is that what you really want to see happen, pHaestus?

All you have, is a mobo, a HSF (possibly benchmarked at OC), and whatever you'd expect to have around the house, more or less.

[jaydee]

Originally posted by bigben2k
Ok, since I wasn't clear, let me detail it:

You're at home, and you get the crave to build a waterblock, after going over whatever Forum.

You come up with an idea, and decide to give it a shot, so you build a prototype.

Then you show it in the forums, and pick up a few more tips, to improve it, which you want to try.


At this point, you really want to have some idea of how much better it's going to be, maybe because you might try a different design alltogether, later.

Using your logic, one would throw their hands up, and call it quits. Is that what you really want to see happen, Jaydee?

WTF? Its one thing to show your block off which I am certainly in favor of, but it is another to say it beats blocks you don't even have (especially the best on the market). If you built the block and put it in your system you should have a pretty good idea how it works compared to whatever you had before. If you built another block or modified the current one and put it back in the system you should have a pretty good idea what the differences are from the original. We shouldn't need this spacific testing method for that. Would think it would be common sence, but I guess not eh? :shrug: This doesn't give you the right to try and compare other blocks to yours and expect people to buy into it. if he just gave results based on his block compared to HIS Maze 3 and showed a decent testing prcedure then all may have been ok. Non of which happened.

I don't understand your reasoning here Ben. You say this is provoked from the thread you linked and that we were flaming Tuff. Well let me make it perfectly clear why Tuff is getting his ass handed to him.

And this can be used as an example by the new guys as to what NOT to do.

originally posted by Tuff
Have had the block up and running..so far 3 deg less than the 1st Gen MicroCup block...which would put in on par with a WW as the 2nd gen block gets 6 deg lower.

I am basing this assumption on the fact that Cathar has said that his WW was on average 5 deg less than the Maze 3. "If memory serves me right"

The first gen block was tested against a Maze 3 and scored on average 3 deg less.


I will hopefully be sending this block off to www.benchtest.com for some critique. But before that it will be put into another system to help varify results that I have found.

Tuff

Ok, so now Tuff is using results from Cathars Maze 3 vrs. the WW testing and somehow magically thinking that his Maze 3 can somehow be used to compare results from Cathars?

Cathar pointed out his Maze 3 was properly lapped and shouldn't be used as a comparision with his own results, but Tuff seemed not to understand and simply thought that lapping his Maze 3 would somehow give more accurate results even though he doesn't have the SAME equipment Cathar used. I would have though it was made blatently obvious to Tuff that his whole concept there was not even remotely usefull or right. But I was dead wrong.

And let's move on:

originally posted by Tuff
I have been told many times...remount the block 10 times...same day...same room temp...ect...Just get tired of hearing it.




Tuff

Speaks for itself. Doesn't want to listen to anyone.

next:

originally posted by Tuff
Sensor 1
6/9/2002;4:40:04 PM;2075 MHz;22°
6/9/2002;4:39:55 PM;2075 MHz;30°
6/9/2002;4:39:46 PM;2075 MHz;25°
6/9/2002;4:39:37 PM;2075 MHz;23°
6/9/2002;4:39:28 PM;2075 MHz;23°
6/9/2002;4:39:19 PM;2075 MHz;24°
6/9/2002;4:39:10 PM;2075 MHz;28°
6/9/2002;4:39:01 PM;2075 MHz;26°
6/9/2002;4:38:52 PM;2075 MHz;22°
6/9/2002;4:38:43 PM;2075 MHz;25°
6/9/2002;4:38:34 PM;2075 MHz;26°
6/9/2002;4:38:25 PM;2075 MHz;25°
6/9/2002;4:38:16 PM;2075 MHz;26°
6/9/2002;4:38:07 PM;2075 MHz;30°
6/9/2002;4:37:58 PM;2075 MHz;23°
6/9/2002;4:37:49 PM;2075 MHz;23°
6/9/2002;4:37:40 PM;2075 MHz;18°
6/9/2002;4:37:31 PM;2075 MHz;26°
6/9/2002;4:37:22 PM;2075 MHz;24°
6/9/2002;4:37:13 PM;2075 MHz;26°

average is 23-24 deg

Readings from MBM.

Tuff

Posted this and somehow thinks this is acceptable? Is there no common sence?

And then this:

originally posted by Tuff
Those are the temps that MBM reports....and gives an average at the end.


Its off the Diode which is able to respond very quickly to temp changes...if anyone out there thinks that a Cpu temp does not change and stays around the same number...you are fooling yourself.

Like I said the MBM is reporting an average temp...the sum of all divided by the amount of readings.

I could have easily posted the other temps off the monitors..that just sit idle and move back and forth by 1 deg...but I am assuming that is not accurate.

The only way to get an accurate temp is to get the original temp off the water...and find out the difference when it leaves the block.

Would it have been better for me to post that Asus Probe says 26 deg and just sits there???

Tuff

Still under the impression that his testing is ok????

And next:

originally posted by Tuff
Thats not accurate at all...The waterblock is there to MOVE Heat to the water..not control the temp of the cpu. The amount of heat that is removed is proportional to the Temp of the cpu.

Think of a dimmer switch on a light. We can limit the amount of light that is being used...therefore we have a lower temp. Once we turn the light on full...we get a higher temp.

Works the same with a Cpu...although the voltage is at a contstant...if we do not demand more usage from the cpu it will have a lower temp We increase the usage we get a higher temp.

The waterblock has NO IDEA how much heat is generated until it has to absorb it. But the DIODE does. The diode gives a more accurate reading than any other measuring device..that is supplied onboard. It is able to tell what is happening at an instant...while a normal thermometer will slowly raise and lower.

For this reason..the diode is used by the vendors for monitoring the heat. If the temp raised to high..the comp shuts down...if it was regulated by a thermometer...everyone would have a cooked cpu by now.

Tuff

This is pretty much wrong and I pointed it out on the thread just to be told this:

originally posted by Tuff
JayDee before you post...think.

thats all I have to say.

Tuff

Instead of trying to tell me why he though I was wrong?

Yeah Ben, maybe I don't understand what he is trying to say. :shrug:

He doesn't seem to want to learn about testing or he could have just asked instead of telling people how wrong he thinks they are with nothing to back it up?

Testing is one thing but what Tuff is trying to pull is another.

[bigben2k]

So what do you want, a retraction? an apology? An admission of guilt? What?

[GTA]

So let me get this straight....

1. I have a system, its watercooled.
2. I build a waterblock.
3. I get some temps for my waterblock under load.
4. I remove my waterblock
5. I stick a commercial block in there
6. I get some temps for the commercial block
7. I compare the 2 temps

Now, at the end of this, I could fairly legitimatly say that one block was better than the other, in my system.

Okay, my results are subjective, and couldn't really be compared with anyone elses. For example, if I stick the commerical block in there, eg. Maze3, and get temps of 50 degrees, and then someone says, "Well, my block gets me temps of 32 degrees, so therefore my design beats a maze 3 by 18 degrees, and as the WW only beats a maze 3 by 5 degrees, my block is 13 degrees better than a WW"

Thats not logical, its daft in fact.

But for my purposes, my test is fine, I can say, "okay, my block beats a maze3 by 2 degrees, nice one, now how can i make it better? " I make some changes, and now I'm 3 degrees better.

I see no problem with that whatsoever. Okay, I could mount and remount to get a more accurate picture of the temp difference between my block and the commercial one. And indeed I should mount and remount lots of times. But 10? I'm not that bothered, maybe try each block twice, if one block wins both times, then I'm perfectly happy in myself to say that that block is "better"

And I'd post to that effect. " My block, in my system, beats a maze3 in my system by 2 degrees"

I get the feeling that if I posted that, I'd get flamed to death. And I really don't see the problem.

[jaydee]

Originally posted by GTA
So let me get this straight....

1. I have a system, its watercooled.
2. I build a waterblock.
3. I get some temps for my waterblock under load.
4. I remove my waterblock
5. I stick a commercial block in there
6. I get some temps for the commercial block
7. I compare the 2 temps

Now, at the end of this, I could fairly legitimatly say that one block was better than the other, in my system.

Okay, my results are subjective, and couldn't really be compared with anyone elses. For example, if I stick the commerical block in there, eg. Maze3, and get temps of 50 degrees, and then someone says, "Well, my block gets me temps of 32 degrees, so therefore my design beats a maze 3 by 18 degrees, and as the WW only beats a maze 3 by 5 degrees, my block is 13 degrees better than a WW"

Thats not logical, its daft in fact.

But for my purposes, my test is fine, I can say, "okay, my block beats a maze3 by 2 degrees, nice one, now how can i make it better? " I make some changes, and now I'm 3 degrees better.

I see no problem with that whatsoever. Okay, I could mount and remount to get a more accurate picture of the temp difference between my block and the commercial one. And indeed I should mount and remount lots of times. But 10? I'm not that bothered, maybe try each block twice, if one block wins both times, then I'm perfectly happy in myself to say that that block is "better"

And I'd post to that effect. " My block, in my system, beats a maze3 in my system by 2 degrees"

I get the feeling that if I posted that, I'd get flamed to death. And I really don't see the problem.

Well said actually. Thats what I was getting at. I don't see why you would get flamed for that. You might get knocked a little bit on testing procedure, but I doubt a flame would come out. If you man enough to acknowlege your testing is a bit flawed then nothing should come of it. Didn't in any of my REV. Block threads.

Example of a flame:

"You are a total dumb ass that couldn't measure your dick with a ruler". Thats a flame. This is not: "Your testing methods are not right" which some seem to think is?

origianlly posted by bigben2k
So what do you want, a retraction? an apology? An admission of guilt? What?

Does this really need answering? :shrug:

[Cathar]

I agree with what GTA said.

About the best you can do is say "my block is better than commercial block X (that I actually own and have used) by Y degree for my personal system/testbed".

If you're really serious, you can maybe try it out on 3 different systems and see if the differences stand up.

Saying my block is better than block X by Y degrees on my system, therefore it must be better than block A by B degrees based on a test result somewhere else, is a totally flawed assumption to make.

I'm reminded of a reseller closely associated with a certain commerical block maker saying that because they outperform a Silverprop Cyclone 5 by 2 degrees on some test bed, so based on the VR-Zone review of the White Water vs the Cyclone 5 taken on a singular P4 setup (which I suspect has a faulty thermal diode), that their Maze 2 look-alike design is the equal of the White Water, which they had never tested, and then using these claims in marketing material.

"hop-step-jump" comparisons between different systems cannot be made, and really this is the basis of what has sparked off the issue that created this thread.

Everyone has the right to say that the block they made measured on their system is doing wonders. They also have the right to expect to be viewed with a certain amount of disbelief when they start using "hop step jump" comparisons to say they have just invented the best block ever.

[pHaestus]

I would suggest that a good place to start is the HardOCP waterblock roundup. Same system, same everything just different blocks. This is the scenario that you envision for your testing correct?

http://www.hardocp.com/images/articl...RZGkjI_4_3.gif

The swiftech MCW462-UH with the 1/2" barbs is right on top. Now let's look at Bill's testing:

http://thermal-management-testing.co...st_results.htm

The Swiftech MCW-462UH is the absolute worst performing block due to the decreased velocity at the die from the "neutering" to add 1/2" NPT barbs.

What was actually measured in the system testing of H|OCP then? How is that relevant to people trying to make a better block? You are making everything 100% system dependent and removing any chance of actually learning something about WHY blocks work with this "swap one for another and it should all be internally comparable" nonsense.

What of the relative size of the blocks? As you cover up more of the socket area with big copper hunks, this affects the temperature of the PCB surface. This will also affect the CPU temp (whether from your in socket probes or from the die).

I would agree that the needless flaming of newcomers and "Post your CPU temps here" people is not a good thing. But I fear you are headed pretty far into the "placate rather than educate" side of things now which is an even worse thing.

[Cathar]

Disregard this post. Nothing to see here.

[pHaestus]

awww I missed it

[bigben2k]

Originally posted by pHaestus
I would suggest that a good place to start is the HardOCP waterblock roundup. Same system, same everything just different blocks. This is the scenario that you envision for your testing correct?

...
I would agree that the needless flaming of newcomers and "Post your CPU temps here" people is not a good thing. But I fear you are headed pretty far into the "placate rather than educate" side of things now which is an even worse thing.

Ok, I'll take that as "some progress".

So is there any hope for a DIY block maker to measure anything?

I don't know how [H] conducted their test, which would probably have been more accurate, if they had repeated it three times (at least), no? They should have at least been able to see that there's a problem with their testing, because of the variance, no?

Ok, so what about using a thermal pad: wouldn't it give more consistent results?


Otherwise, I can see what's emerging here, in what GTA said. Since the power applied, which is an unknown to a home user, will result in various CPU temps, system to system comparisons aren't valid, and we have to make that clear, and understandable to our new guys.

Example: The White Water waterblock was tested (uinder a specific set of circumstances) to have a C/W rating of 0.19. So if the source power is 50 Watts, the CPU temp will be 0.19 * 50 = 9.5 degrees above ambient.

On someone else's PC, if the power is 60 Watts, either because it's overclocked, or because it's a different CPU alltogether, whatever, then the temp difference would be 0.19 * 60 = 11.4 degrees above ambient.

Add to that, the different set of circumstances, which will include the flow rate (most important) and the different cooling solutions, and room temp.

[pHaestus]

I don't know how [H] conducted their test, which would probably have been more accurate, if they had repeated it three times (at least), no? They should have at least been able to see that there's a problem with their testing, because of the variance, no?

Missing the point completely. My point was that they set out to test waterblocks, and only proved in the end that different waterblocks have different flow resistance. Let's ignore the larger issues with the review (did room and water temperatures change? Sure they did) and just stick with that. What that would tell a would be wb maker is that the best performance comes from using the least flow restriction. Why is this the case though? Could it be because higher flow rates put the radiator in a better position to produce lower water temps? How does that relate back to the would be wb maker? Wrong impressions.

If you want to test your watercooling SYSTEM then you really should just learn how to calibrate thermistors and then plumb in 4 water (wb inlet and outlet and rad inlet and outlet) and 2 air (radiator intake side and exhaust) sensors. Monitor how changing the layout, fan speed, waterblock, etc affects temperatures. Then you can start to make some inferences in where improvements can be made.

[pHaestus]

Yes GTA and to some extent Ben are both talking about SYSTEM performance. Nothing wrong with that, but I don't see how that is a helpful thing for a waterblock maker to know necessarily other than identifying target pressure drop for most users.

[jaydee]

Originally posted by bigben2k
Ok, I'll take that as "some progress".

Your more optomistic than I am.

So is there any hope for a DIY block maker to measure anything?

I guess it would really depend on how "into it" they are. If they are just making one or they plan to make several prototypes to get a final acceptable version they are happy with.

If they are going to dump the kind of money it takes into making a high performance block then dropping some cash on acceptable testing equipment (or at very least sorting mobo's untill they find a decent one) shouldn't be a big issue.

If they just want to make one water block then why bother. Be happy if the computer runs cooler and/or quieter than whatever they used before eh?

I don't know how [H] conducted their test, which would probably have been more accurate, if they had repeated it three times (at least), no? They should have at least been able to see that there's a problem with their testing, because of the variance, no?

What if it wasn't in the mounting? What if the board they used is the issue? CPU? To many variables....

Ok, so what about using a thermal pad: wouldn't it give more consistent results?

Well you can make about 50 mounts with a small tube of silver compound, or you can make what 5 with the same amount of cash in thermal pads? That will get expensive fast, talking 100's of dollars for a series of prototypes as opposed to 2 tubes of silver compound which can be had for $10. Not to mention pads are harder to clean up then paste. At least the pads I have used in the past. I will use silver compund simply because I know how to apply it. I guess you can try and dumb it down and have them use pads but is that really the right thing to do? Are pads consistant? You would have to use the same brand, make, and model every time.

Otherwise, I can see what's emerging here, in what GTA said. Since the power applied, which is an unknown to a home user, will result in various CPU temps, system to system comparisons aren't valid, and we have to make that clear, and understandable to our new guys.

Ummm, isn't that what we have tried to do and just end up gettng called a flamer? Interesting indeed....

Example: The White Water waterblock was tested (under a specific set of circumstances) to have a C/W rating of 0.19. So if the source power is 50 Watts, the CPU temp will be 0.19 * 50 = 9.5 degrees above ambient.

On someone else's PC, if the power is 60 Watts, either because it's overclocked, or because it's a different CPU alltogether, whatever, then the temp difference would be 0.19 * 60 = 11.4 degrees above ambient.

Unfortunatly we don't know the voltages.

Add to that, the different set of circumstances, which will include the flow rate (most important) and the different cooling solutions, and room temp.

Which will all have to be measured somewhat decently.

The only reason the temps of DIY'er blocks should even be posted is so we have an idea how their design works. But if the temps are way off then whats the use? Doesn't do anyone here any good eh?

Promote proper testing and maybe people will be more inclined properly test and their results maybe more usefull for anyone thinking about making a block.

We should be concentrating more on an advanced DIY test setup that is relatively inexpensive and we can all agree the results will be loosly acceptable IMO. Being this forums is not geared towards the newbs to begine with why should we be catering to them?

What would be the "acceptable" is the question. I am sure we all have different thoughts on this.:shrug: I have been banging my head off the wall for 8 months now trying to find a way I would be happy with by testing my blocks and sharing the results. I am getting close to getting it all together, but still a few months out and it certainly isn't inexpensive. I already have several hundred $'s in it and I have yet to buy the expensive parts. Yes, all this for my own DIY blocks... So like I said, maybe it just depends on how into they are...

[Khledar]

It does depend on how into it you are and what your intents are, I recall reading Cathar's statement (see below) regarding his extesnsive initial testing, including statements about lapping the Maze 3 and running 10 tests, etc.

I think a standard setup would be difficult, but if someone truly feels the need to discuss their results what I said in the Tuff thread I still believe applies:

Originally Posted by Khledar
I mean this has all been discussed before - the more thorough a set of results (testing methods, etc.) - the more acurately they can be interpreted by the readers... Maybe wanna indulge us with your setup Tuff?

If you give a full description of how you tested it, maybe a few pictures and a short paragraph - I doubt anyone will challenge you unless there is something serious that needs to be improved. I think it's generally understood that we're not talking about professional testing, but a gauge of one block to another. If your methods are well thought out and lack any serious problems they will be taken for what they are. But necessary explanation of what has gone on during testing is required.

I mean when BillA was testing blocks he didn't just post the nice C/W vs flow graphs etc and say glad you tuned in. There was a huge explanation of his setup, how he tested, what he did to eliminate error. I'm not saying we all need to do this, but in a short few paragraphs (getting back to the beginning of this post) Cathar was able to give some sort of credibility to his testing with respect to the Maze 3.

[EDIT: Below]

In case you're not tuning into the Tuff thread:

Originally Posted by Cathar
My Maze 3 was lapped properly by me after I received it. The base wasn't that flat to begin with and I picked up an easy 2C from that lapping of it. Since I'm comparing against a well known product, I want that product to be operating in the best possible way that it can be made to, and to remove any possible inherent performance variances that may exist as it comes from the factory. That way I'm comparing against a well defined "high point" for a design, rather than a possible "middle or low point" in which performing "better" can suddenly be absorbed by merely picking/receiving a better example of the competing block.

All tests were run to equilibrium and the environment was very carefully controlled. Each block was mounted at least 10 times to minimise mounting variations which could, on occasion amount to 3C differences, and the various blocks were alternated between mounts to ensure that results were repeatable and comparable with earlier results and to provide additional control and detection for test setup and environment variances.

Since I'm developing a block, I feel that I absolutely owe it to myself to do these things, since it costs me money to make blocks and I don't want to be fooling myself as to what direction I'm heading at any point in time, or worse, that should anyone buy a block from me and perceive that I have fooled them as well.

I don't know how any of that applies to what is going on here. I just thought I'd share what I do to determine how well a block that I make performs relative to any other block. If I'm going to claim it's better than another block, it must be better than the best possible representation of that other block.