WhatBlock Performance Review

Gallery Pics
http://www.ocforums.com/showpost.php?p=1622809&postcount=51

I’ve received messages from people asking about the performance of my WhatBlock waterblock. I was reluctant to say anything until I could perform tests that were reasonably accurate. I finally acquired the equipment necessary, so here is a very preliminary report of WhatBlock performance.

Test Environment – Test Computer
1600+ Athlon XP Palomino at stock speed
Max power output 62.8 watts
Abit KT7A motherboard
BurnK7 running at high priority
Task Manager process count = 15
Task Manager closed during testing
AMD on-die thermal diode isolated and monitored with diode reader

Test Environment – Monitoring Computer
2200+ Athlon XP Tbred running stock
Asus A7N266-VM motherboard
Monitoring diode reader through SMBus and Motherboard Monitor

Test Environment – Water Rig
Mag 2 pump submerged in plastic tub
Water supply – Continuous stream from faucet into plastic tub
- heated water was drained...no recycling of water.
Water rate: 1.0 GPM
Water temp at suction: 30.5C (0.0C variance during testing period)
Thermometer: Electro-Therm SH66A 0.1C resolution 0.2C accuracy


Test Results:
CPU temp = 40C
CPUTemp - WaterInTemp = 9.5C

Accuracy = +/- 1.2C

NOTE: Upon calibration review, CPUTemp – WaterInTemp may become 12.0C

What this means is that if the water temperature in your cooling rig is 30.5C, and the water is flowing at 1GPM, then the temperature of your 1600+ Palomino will be 40C while running BurnK7.

This should have been the final performance review. However during testing a calibration anomaly was noticed that still has to be resolved.

I was trying to make my results directly comparable to pHaestus’s results on Procooling.com. His waterblock rating is simply a temperature difference at a given flowrate. I got an accurate thermometer, and I also made a diode reader for accurate CPU temperature readings. I figured I’d do a test at 1 GPM and leave it at that for now.

In this article pHaestus described his testing process:
http://www.procooling.com/articles/html/procooling_testing_methods_-_p.php

He also describes his waterbath calibration method here:
http://www.procooling.com/articles/html/amd_thermal_diode_testing_cali.php

The first problem I have is that my processor is hotter than the one pHaestus is actually using. I thought he was using a 1600+ Palomino but he’s really using a 1700+ Tbred, which is 13.4 watts cooler than my processor. I figured I could underclock my processor, but the Palomino core is larger than the Tbred core so that throws things off as well so...I’m still working on that.

The big problem, though, is calibration. I performed a water bath calibration as he did, and got the same results. The CPU temperature readings were linear and in sync with the thermometer from 20C to 50C. I thought this was great. Then, another idea came to me. I wanted to check the temperature of the CPU while it was mounted in the motherboard by running water, at a specific temperature, through the waterblock and then checking the diode temp. I figured if the waterblock can remove heat, it could also deliver heat. So I set up the test to see if I’d get the same results as the water bath. The testing condition was exactly the same as for testing the waterblock, only the psu on the test computer was switched off so there was no power to the board. This is what I got:

Water: 19.5C, CPU: 18C
Water: 30C, CPU: 28C
Water: 50C, CPU: 47C

Not only was the CPU cooler, but the differences weren’t linear! I thought maybe EMF from the pump, but the temperature readings remained the same when the pump was suddenly switched off. I don't think EMF from the Monitoring Computer was causing the problem because it was on during the water bath test. Currently I am at a loss to understand why the “waterblock calibration” method didn’t work. I’m pretty darn sure it’s not electrical.

I do not know if pHaestus ever performed such a test. For all I know, his results could be reading low as well. Or my results might be just fine under a powered PC. I do not think pHaestus, or anyone else, has figured out how to calibrate the CPU diode while under actual test conditions (waterblock installed and with CPU powered.)

Anyways, since the difference between the thermometer and diode would seem to be around 2.5C at 40C, I listed a possible corrected temperature difference of 12.0C.

I’m open to all suggestions and comments.

EDIT: Changed GPH to GPM

aside from technical thoughts

generally, a review posted by the designer/producer is not trusted or looked upon favorably.

I'd suggest that you send the block to pH, but you already burned that bridge.

greenman100 said:

generally, a review posted by the designer/producer is not trusted or looked upon favorably.

Click to expand...

People asked. I'm just providing the best answer I can.

pHaestus has his own problems with his test setup which he has admitted to. And considering what I've discovered, at the moment I don’t view a test-run by him to be any more valid. So...no great loss.

ouch

some bad blood, maybe?

you wanted something directly comparable to pH's results

easiest way being send him the block?

no, you burned that bridge, next step is to try to emulate his testing

(much like you did his diode reading/calibration)

you've choosen the hard way many times, graystar. you will learn, given enough time

furthermore

I would expect the CPU die to be cooler

the heater is the water....the heat must travel through the block, TIM, and then hit a sensor. meanwhile, the sensor is being cooled by motherboard traces (up to 20% of total cooling, see thread at procooling)

explains the non-linearity of your measurements (remember basic formula for heat transfer, as delta T increases, heat transfer increases

so, I believe pH is calibrated properly

As so you ARE capable of referencing others work whenever you have something critical to say.

I DO have a suggestion for what to do with the block, but I doubt there's any room left with the space your head is currently using.

A more pragmatic suggestion? Consider the heat radiated by the surfaces of your wb and tubing. Those losses go up as the difference between water temp and ambient increases (either above or below ambient). Perhaps try insulating the wb and the hoses? Or pull the CPU out of the mobo, clamp it onto the bp of wb (carefully eh?), insulate the whole package, and try again?

Or you could just take a chance to take shots at someone else's test methods without any basis. Yea that looks a lot smarter. You took my "problems" pretty much completely out of context btw. And at least I have (a) methods that can let me tell when something is awry and (b) the decency to be up front about it and to explain the fix.

Also consider that the entire motherboard is behaving as a heatsink when the CPU is in the board like your second sets of tests. Perhaps measure the temperature of the backside pins or the motherboard itself? I saw all kinds of weirdness when testing with chilled water due to the motherboard cooling down and backside CPU temps responding in a nonlinear way.

The diode is linear; the properly calibrated diode/diode reader setup is linear. What it is measuring may not always be the same though! That would be my guess in this case.

greenman100 said:

ouch

some bad blood, maybe?

Click to expand...

I’m not sure what you mean. Did he not admit to having issues with his diode reader? Would you not question the validity of tests performed with faulty equipment?

you wanted something directly comparable to pH's results

Click to expand...

Yes I did. I thought that performance information would be more useful if it could be compared to existing results. I would have much preferred to be comparable to Overclockers die simulator...but that’s impossible. So I went with what I thought I could accomplish.

easiest way being send him the block?

Click to expand...

Faulty equipment...remember?

no, you burned that bridge, next step is to try to emulate his testing

(much like you did his diode reading/calibration)

you've choosen the hard way many times, graystar. you will learn, given enough time

Click to expand...

I think you have some odd obsession with flaming bridges, but besides that...I do tend to do things the hard way. But that’s because I can’t just know something...I need to understand it. pHaestus thinks my bent-diode-pins idea will solve his own diode reader issues...right? I have things to learn, but there are things to be learned from me as well.

I would expect the CPU die to be cooler...

Click to expand...

That’s exactly what I was first thought. However, a quick review of the data (along with a new bit of info) shoots that down. I experienced the problem at 19.5C. What I didn’t mention previously was that the ambient temperature was about 26-27C If the problem was simply thermal transfer, I would have experienced the reverse problem when my water temperature dropped below ambient.

Nice try though.

And it’s not a grounding issue...had that...fixed that.

...so, I believe pH is calibrated properly

Click to expand...

I do too. I believe I’m calibrated properly as well. I just think that something is mucking up my nicely calibrated devices.

pHaestus said:

As so you ARE capable of referencing others work whenever you have something critical to say.

Click to expand...

Here...this pre-dates the article. Now you can die happy.
http://www.ocforums.com/showthread.php?t=315513

And if you think you deserve credit for chopping up Socket A...

YOU ARE OUT OF YOUR FRIGGIN’ MIND!!!!

People were slicing up Socket 370 long before anyone took the knife to Socket A. Here’s someone taking one apart:
http://3feetunder.com/krick/370mod/

The note on top says that the original page is gone. That’s the case with many of the hack jobs on Socket 370 that I’ve seen. I’ve seen the diagonal cuts on sockets years and years ago. You’ve got some balls to even THINK that you were the first to do that!

I DO have a suggestion for what to do with the block, but I doubt there's any room left with the space your head is currently using.

Click to expand...

Nice. That’s about as original as your hack job.

Consider the heat radiated by the surfaces of your wb and tubing.

Click to expand...

Yeah, I did consider it. I had performed some pre-testing tests to try to find areas that might affect readings. One test was simply running water through the wb (while mounted) and measuring the temperature differences. I was thinking heat added by the pump at the time. In any case, the water out was nearly the same temp as water in...inside the accuracy of the thermometer. So I don’t know for sure what the temp difference was but it certainly wasn't in the 1.5 - 3.0C range.

Also, there’s the fact that the problem occurs very near to ambient and it also occurs, with the error in the same direction, when water temp is below ambient.

Or you could just take a chance to take shots at someone else's test methods

Click to expand...

Yeah, well...everyone else is more sophisticated and I don’t have that kind of money.

Also consider that the entire motherboard is behaving as a heatsink...

Click to expand...

Yeah I thought that also. I made a response to greenman100 on that.

The diode is linear; the properly calibrated diode/diode reader setup is linear. What it is measuring may not always be the same though! That would be my guess in this case.

Click to expand...

Well...thanks for your thoughts (some of them, anyway.) BTW have you ever tried the procedure I described of running water through the block and measuring the cpu temp?

For what its worth, you'll notice that pH's 1700 is overclocked and overvolted to hell. For someone doing this much work, im surprised that you dont even know the facts about his set up... pretty disturbing really.

Still having trouble fathoming your reasons for neglecting to mention his test methods and the fact that you're using (albeit wrongly) and not mentioning him in the final article.

Slow your roll there killer

I never claimed I invented ANYTHING; I have always pointed people to other good work that others have done before me. From the first AMD diode reader article I wrote Jan 2002:

"I want to thank Peter North again for coming up with this idea and then walking me through implementing it for VyW. All credit for this great work should be directed to him; I am just a monkey with a soldering iron and a closet full of heatsinks."

Obviously the tech docs are all out there and the techniques aren't patentable or anything. Furthermore everything I do is posted on a public forum for dissection and discussion of our readers.

But I find it funny that YOU would write an article on OCers as the authority on this stuff and not even tip a hat to me. And then you get all defensive and try to belittle what I've done as being nothing that others haven't been doing before.

Yea that's all probably technically true but no one else bothered.

You are now starting to see some of the difficulties of using the internal diode to do real testing; now consider that I've been battling the same difficulties for years now and publicly posting my findings and tips. Now consider someone just posting a rehash by and large of the same information on OCers without even mentioning that I might be a good resource for diode-related information and testing. How many DIY diode reader/diode related articles are there on OCers? 4 or 5 at least. Not a single one references PeterNorth's original howto or anything that I've done since (and my VyW article far predates them all). A tip of the hat is just courtesy IMO. Courtesy that is lacking from the many $100 prizewinners on this site.

//edits: The CPU I test with generates ~ 71W (calculated from flow rate and dT of water across wb). TBredB 1700+ at 175x13 and 1.85V in bios 1.81 in MBM

//edits: Of late, I find the actual heat output of CPUs as a function of MHz and VCore to be more interesting than continuing to fiddle with diode readers. So most of my non-wb testing time goes to doing those sorts of things for personal interest. I figure no worries as there are plenty of other thermal diode experts out there who have that all under control.

//edits: I still fail to see how your poorly-designed experiment invalidates my own testing?

//edits: Of late, I find the actual heat output of CPUs as a function of MHz and VCore to be more interesting than continuing to fiddle with diode readers.

Click to expand...

hey phaestus give me a pm on this , i have some interesting info about this.

Etacovda said:

For what its worth, you'll notice that pH's 1700 is overclocked and overvolted to hell. For someone doing this much work, im surprised that you dont even know the facts about his set up... pretty disturbing really.

Click to expand...

You’re right. I didn’t know that. If you don’t mind, please show me where that is mentioned in the following article...

ProCooling's Waterblock Test Methods and Objectives
http://www.procooling.com/articles/html/procooling_testing_methods_-_p.php

Still having trouble fathoming your reasons for neglecting to mention his test methods and the fact that you're using (albeit wrongly) and not mentioning him in the final article.

Click to expand...

You obviously are under the influence of someone else’s issues with me ‘cause it’s affecting your eyes. The only time I ever wrote anything about pHaestus’s testing methods was in this thread, and in the original message I provided the same link as the one above (since you’re so blind you couldn’t find the first one.) As for the article I wrote...there is nothing in there that originated from pHaestus. Socket chop jobs for temperature probes and for using different CPUs were at their peak in the Socket 370 era. What I did is nothing compared to what had been done back then. Those guys can do my hack job in their sleep! The names of companies with sensor readers, pin-outs, and other information came from AMD. That information was simply factual...there were no original ideas in there belonging to AMD, and as such didn’t require a reference. I think it’s obvious that I didn’t invent the names of the diode pins, and that this information came from AMD.

The only person that deserves mention is the first person that had the guts to take a saw to the socket 370. I don’t know who that is but if I ever write another article about cutting sockets I’ll mention it.

I don’t even know who the hell you are! Don't you have anything better to do with your life than hassle people you've never even talk to before?

I test using different heat loads and so I didn't include the frequency and voltage on the testing methods page. I used to paste this blurb into every wb review:

"One new development is the estimation that my TBredB 1700+ at 13x175fsb and 1.85V (BIOS) generates approximately 73W under load. I am hesitant to convert the graphs from deltaT to C/W as of yet (still playing with this sort of W estimation), but that should give you a ballpark number for CPU power. "

New calibration suggests number is closer to 71.4 W. All wbs are tested at same multiplier, fsb, and vcore so they can be compared to previous wb tests.

As I said you are technically correct that you don't "owe" me anything.

I don’t even know who the hell you are! Don't you have anything better to do with your life than hassle people you've never even talk to before?

Click to expand...

So you are saying that you never read my articles and that they didn't influence/encourage you to do what you have been doing lately? That's fine. Somewhat unbelievable, but fine.

pHaestus said:

I have always pointed people to other good work that others have done before me

Click to expand...

And so did I in the two forum posts. I still fail to see what part of my article I should have given you acknowledgement for.

You are now starting to see some of the difficulties of using the internal diode to do real testing; now consider that I've been battling the same difficulties for years now and publicly posting my findings and tips. Now consider someone just posting a rehash by and large of the same information on OCers without even mentioning that I might be a good resource for diode-related information and testing. How many DIY diode reader/diode related articles are there on OCers? 4 or 5 at least. Not a single one references PeterNorth's original howto or anything that I've done since (and my VyW article far predates them all). A tip of the hat is just courtesy IMO. Courtesy that is lacking from the many $100 prizewinners on this site.

Click to expand...

If I ever decide to write an article on diode readers, I’ll keep that in mind. However, my article was not about diode readers. I stated clearly what my article was about:

“But none of these readers will work if you can't connect to the CPU's diode pins. The following article presents a method of accessing these pins.”

My article was solely about getting to the diode pins in an electrically isolated fashion. That’s all. It had nothing to do with discussion of diode readers.

The CPU I test with generates ~ 71W (calculated from flow rate and dT of water across wb). TBredB 1700+ at 175x13 and 1.85V in bios 1.81 in MBM

Click to expand...

Thanks for that info. May I suggest adding it to the Test Methods document?

I still fail to see how your poorly-designed experiment invalidates my own testing?

Click to expand...

I never said my testing invalidated your work. YOU invalidated your testing when you admitted to sensor issues because of the motherboard pins. Your result maybe just fine, but you certainly didn't reach your goal to "emerge with your credibility unscathed."

pHaestus said:

So you are saying that you never read my articles and that they didn't influence/encourage you to do what you have been doing lately?

Click to expand...

WTF!!! Am I the only person here that's paying any attention to what he's reading??? That post was a response to Etacovda!

And you never answered my question, and you had no intention of answering my question, and you know it.

You write an article who's title ends with "How 2", and then you have this;
"I bet that you are expecting a lengthy treatise with step by step instructions. I am sorry to disappoint you, but the whole trick is to follow the specifications on the Maxim data sheet "

You want credit for that??? I think someone needs to explain to you what a HowTo is.

the mods will not be happy if this continues down its current path

speaking from experience

Please
I posted on my forums when I was having a difficulty with the diode readings, I figured out what the problem was, I corrected it, no biggie. 1 week of bad test data, 1 week of figuring out wtf was wrong, and back on track. These things happen. Is it better to remain silent or to post bad results? I am thinking not. I have lots of sanity checks and occasionally go back and retest older wbs to make sure everything's reproducible. Is that now a bad thing?

If you are doing serious testing then you should recognize that it is a learning experience.

As I said you are completely within your rights to deny any relationship or semblence to anything I have ever done. Nothing I did 15 months ago in my article about soldering wires onto CPU pins is reproduced verbatim in your OCers article. I also said on my own forums that I thought bending the CPU pins was clever. More credit to you than I got from you by far. No concession from you that I might know what the **** I am talking about? Only mention my name when you want to question 7 months of work?

I get my ethics and need to credit people from my work; when reviewing manuscripts for REPUTABLE journals there are always questions like "did the authors cite all relevant previous research?". My personal style is go post links to other related work so readers can benefit and get up to speed faster.


On to technical discussion:

My Hydrothruster shuts off at around 35C water temps and so I have not done any testing above that with my current setup. I DO see CPU diode temperatures lower than water temperature when wb is plumbed and the CPU is unpowered but never lower than the ambient temperature. The magnitude of difference between CPU and water inlet is not large below 30C water (0.2-0.5C); I assumed that it was a matter of thermal resistance between wb and the diode and other paths for heat to dissipate from wb. Consider the large difference in heat density between putting water through a wb and generating 70W in a very small CPU die. For P4s it has been shown that there can be an 8C dT from the hot spots to the diodes; that's just due to lateral thermal resistance I guess in the silicon?

I would seriously recommend pulling the CPU out, mating it to the bp of the wb, and insulating the entire apparatus. Now the calibration SHOULD yield the same results as in the water bath (right?). If not then the issue is with the calibration of your diode and reader. The lower than water temps for the chilled water run is weird; no explanation leaps to mind other than something odd with the water temp measurement (and if your diode were previously calibrated with that then why would it matter?)

//edits: Though temperatures will read lower if there is some extra resistance introduced; 0.4C/ohm or so right? The 0.125C res isn't doing you any favors either...

Graystar
the same method you have applied th pHaestus will now be applied to you:

you said that an acknowledged problem had by pHaestus led to the conclusion "I don’t view a test-run by him to be any more valid."
- this is an assertion that your present work is equal to that of pH
- while simultaneously saying that neither of you are capable of doing better

your results already reveal a HUGE anti-pH bias
enough to shade the results ?
if they serve the self-aggrandizement of Greystar ??

don't need a Ouija board to get that answer

Graystar,

Don't you mean 1 GPM (gallons per minute)? You have 1 GPH (gallons per hour) in several places and just wanted to make sure you're not acutally running 1 GPH... That's pretty slow flow. (hey, that rhymes)

pHaestus said:

Only mention my name when you want to question 7 months of work?

Click to expand...

In the post that started this thread I do not believe that I ever questioned the validity of your work. I did pose a question...as to whether or not you performed this procedure that I'm having trouble with...but I never said or implied that lack of such knowledge puts your results in question. Honestly, even if somehow you are experiencing the same issue, your results are still valid as a comparative measure simply because the non-linearity of the error is so small within the range of your results.

At this point I think I’ve exhausted my thoughts on this subject. I just have nothing more to add.

Back to the block...
I will continue to try different things in an effort to discover anything that makes repeatable, measurable changes. Here’s the latest:

All equipment starts at ambient, including a container of water. Call it 25C. Monitoring Comuputer is on and reading 25C from the test CPU. Thermometer reads water temperature as 25C. I install a waterblock, with hoses connected to it, on the CPU. Temps remain at 25C. I connect the hose to the pump (output drains to container.) Still 25C.

I do not turn on the pump. What I do next is pulse the pump...like a food processor. I watch the water coming up the hose and turn the pump off when the water is where I want it. Well, actually the water is going fast so I take educated guesses at how long to keep the pump plugged in. Anyway, here’s what I observe:

When the water fills the hose but doesn’t reach the block, CPU temp remains at 25C.
When the water just hits the block and drains back, CPU temp drops to 24C and remains there.
When the water fully runs through the block, even if it’s just for a split second, the CPU temp drops to 23C and remains there.

At this point I can leave the pump off, disconnect the pump from the hose and the temperature will remain at 23C

I can remove the block and CPU temp will immediately rise to 25C.

I can reinstall the block and CPU temp will immediately drop to 23C

If I leave the waterblock (with hose) uninstalled for an hour and reinstall, CPU temp will remain at ambient.

That’s all I got at the moment. I’m pretty sure it’s not a thermal issue, so I’m back to thinking it’s electrical.

Though temperatures will read lower if there is some extra resistance introduced; 0.4C/ohm or so right? The 0.125C res isn't doing you any favors either...

Click to expand...

Hmmm...I don’t know. I don’t think so. I mean...sure, if you can somehow create a resistive load in the diode connecting leads, that would certainly throw off the reading. But how would that happen?

I’m sorry but I don’t know what the 0.4C/ohm rating refers to, but I think you’re describing a thermistor maybe? I thought diodes are read by applying a known, constant current through the diode and reading the voltage drop across it...1.8mV per C, I think. A resistive component would change the magnitude of the current right? Don’t know what that would do the reading. I don’t know how the size of the current relates to the magnitude of the drop across the diode.

Yeah, I gotta read that increased resolution somehow. I’m trying to get a Maxim SMBus interface for free (we’ll see how that goes). If I can’t get that then I’ll have to write some software to read the sensor. Just got a MAX6692 in the mail. Time to build a new reader!