How to test a homemade block?

[bigben2k]

Originally posted by 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).

Just a reminder...

The only reasonable goal here, is to allow a user to state that his waterblock gives his CPU temps an X degree(s) improvement, over his HSF.

No cross comparison is possible. No actual temps are possible either.

[MMZ_TimeLord]

I have read your articles in the past pHaestus, and I love them.

A quick question on that last one... if I use the MAX6655EVSYS to measure temps on a P4 as opposed to a AMD... is the probe in the CPU similar? :shrug:

The setup you described in the article is only a challenge in the wire soldering area... and I've done military spec micro soldering, so no biggie.

The MAX6655EVSYS is only $90 ... so, no big deal. I have a laptop with '98 on it and could be used for the monitor.

Anything else? Does the calibration you did HAVE to be done? I presume it does. This is to verify linearity and baseline for temp measurement... correct?

Again, thank you pHaestus for all your hard work!

One day this n00b may understand...

[jaydee]

Originally posted by bigben2k



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

I will try to get back to you tonight on that. Have way to much work to do "at work" today.

As for the newbs not welcome, well I would say yes. But I don't classify a newb as a guy that just signed up. I didn't think Zymrgy was a newb by any standard even after his 1st post. it is all in the the person if he is a newb or not. We have some newbs here with almost 1,000 posts!

[MMZ_TimeLord]

pHaestus...

One other quick question.

Did you sever the circuit to those two pins for the CPU temp diode on the motherboard? If not... could this not skew your readings?

Just curious...

[jaydee]

Originally posted by bigben2k



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

This isn't the exact ones I was thinking of but I will throw it out there. It has to be better than a CompuNurse. If nothing else is has other fetures that might be usefull. It is probably garbage (can't seem to find jack on google about it) but being thats the results we will get with the ideas in this thread then it should be fine.

http://cgi.ebay.com/ws/eBayISAPI.dll...&category=4678

I found a site a while ago that had a bunch of thermometers and the specs with them and there was a few that where $20-$30 that had pretty good specs. With any luck I booked marked it at home.

[edit]I just bid on it. So if I get it I will see whats it is about. Maybe it has specs with it. Need a new multi meter anyway.....[/edit]

[pHaestus]

One diode one reader. You would have to get a P4 board that you could cut traces for. And even then Intel says that there can be big differences in core temp across the die. I avoid messing with them altogether (P4s)

I have the AMD chip on an 8K7A+ which doesnt have any internal diode reading circuitry.

[Zymrgy]

Ok, going thorugh this post I would like to give my input reguarding the original topic. Even I am not a fool enough to believe that I have any means to accuratly measure my actual CPU temps. However, I do think I can measure temperature difference with a bit more accuracy. That is....I would run whatever "burn in" test that struck my fancy while running my HSF setup. I would measure my room temp with a regular thermometer. Record these values. Swap the HSF for my block & repeat.

Now, I know that the sensor in my mobo is not 100% accurate at measuring temp...that is 50 deg C might be +/- 5deg. However, I think that it would stay in the same tolerance range...that is if 50 was really 48 then it wouls stand to reason that 40 would be close to 38.

So maybe rather than stating your temp, state the difference between the air & watercooled. Whenever I actually get my block done this is exactally how I plan to test it.

Flow rate...the 5 gallon bucket sounds like a good idea to me. Not 100% acctuare, but close enough to get the job done.

[jaydee]

Originally posted by Zymrgy
Ok, going thorugh this post I would like to give my input reguarding the original topic. Even I am not a fool enough to believe that I have any means to accuratly measure my actual CPU temps. However, I do think I can measure temperature difference with a bit more accuracy. That is....I would run whatever "burn in" test that struck my fancy while running my HSF setup. I would measure my room temp with a regular thermometer. Record these values. Swap the HSF for my block & repeat.

Now, I know that the sensor in my mobo is not 100% accurate at measuring temp...that is 50 deg C might be +/- 5deg. However, I think that it would stay in the same tolerance range...that is if 50 was really 48 then it wouls stand to reason that 40 would be close to 38.

So maybe rather than stating your temp, state the difference between the air & watercooled. Whenever I actually get my block done this is exactally how I plan to test it.

Flow rate...the 5 gallon bucket sounds like a good idea to me. Not 100% acctuare, but close enough to get the job done.

Well your a total newb when it comes to testing, but you sure make up for it in others areas.

[jaydee]

Originally posted by jaydee116


http://cgi.ebay.com/ws/eBayISAPI.dll...&category=4678


[edit]I just bid on it. So if I get it I will see whats it is about. Maybe it has specs with it. Need a new multi meter anyway.....[/edit]

I won the auction and finally found the manufactures home page and a .pdf of the specs. Picture below of the Temp accuracy. Doesn't look great.... Will be picking up a Fluke in the next few months and I will test it againt it.

[bigben2k]

All right, let's address part#1: the flow test.

So far, pHaestus despises the bucket test (and rightfully so), and proposes a manometer instead, which is very easy (heck, I've done it, in the kitchen!). I think that there are various opinions on this otherwise.

Now I don't want to start a poll for everything here... and I tend to lean towards what pHaestus proposes, because it's simple, cheap, and relatively accurate (with the right pump).

Can we drop the bucket test, as an acceptable flow test?

The way I see it, if a typical system has a flow of ~50 gph (source: OC article), then we're looking to see a 5 gallon bucket fill up in 6 minutes, and assuming that the tester is slow on the trigger, he might measure the time at plus or minus 10 seconds.

Is this 3% error unacceptable? Is the manometer going to be that much more accurate?

[jaydee]

Originally posted by bigben2k
All right, let's address part#1: the flow test.

So far, pHaestus despises the bucket test (and rightfully so), and proposes a manometer instead, which is very easy (heck, I've done it, in the kitchen!). I think that there are various opinions on this otherwise.

Now I don't want to start a poll for everything here... and I tend to lean towards what pHaestus proposes, because it's simple, cheap, and relatively accurate (with the right pump).

Can we drop the bucket test, as an acceptable flow test?

The way I see it, if a typical system has a flow of ~50 gph (source: OC article), then we're looking to see a 5 gallon bucket fill up in 6 minutes, and assuming that the tester is slow on the trigger, he might measure the time at plus or minus 10 seconds.

Is this 3% error unacceptable? Is the manometer going to be that much more accurate?

How about you tell us. Do both a few times each and see what the results are?

I am still a little confused on this whole GPH deal though. What gph are we trying to find? The pumps gph? the block's gph? the pump with the block hooked up to it gph?

Seems to me you would want the pumps GPH, the pump with the block gph, and then the complete systems gph.

Then we would know what that the pump can do, how restrictive the block itself it, and then the gph of the system while testing. All in which seems pretty necassary to me.

[bigben2k]

Oh yes.

In this case, we would want the block and pump only. Using the pump's curve, we can then extrapolate the pressure drop.

[Cathar]

Near as I can tell, as long as at least a 10l bucket is used, and it's accurately marked at 10l, the margin of error with the bucket approach is around +/- 5%. Not quite good enough for pressure drop prediction, as the quadratic nature of the pressure drop equation can make a 5% difference at say 4lpm be quite grossly mispredicted at 10lpm (by up to 30% inaccurate). If you're going to use the bucket approach, you can't really extrapolate the pressure results with any degree of confidence.

So is it good enough? For a single test at a single flow rate, yeah, it's "okay" as a rough indication for the home user.

[Zymrgy]

Ok...now that I see how specific you all want to be on this...

I have made several of these things for test the flow rate in water in a nuclear reactor.

(damn I do nice work )

anyways, from what I gather they are highly accurate, like +/- .25%. Of couse this is a larger model, but my thinking is is that it can be scaled down. It is called a venturi flow meter. The liquid (or gas) travels through a series of very specifid tapered diameters ( tolerance of +/- .0005 in) ( on the inside ) & pass by a couple of very specific ports that are a certian length. The fluid passing by these ports creates some sort of vaccum & there ya go. However, at this point we have gotten way past the point of being a hobbiest.

Doing a bit of searching on the web I did find this little unit that looks like it might be ideal for testing flow....does not look expensive & the accuracy looks to be acceptable.

Pretty simple...water flows against a piston that is mounted on the spring. The spring compresses & you read where the plunger stopped on a scale on the outside of the clear tube. There is a option to "build it yourself" on the page, so I submitted a quote for one with 1/2in fittings,brass fittings, with a brass ball valve with a max flow rate of 3gallons per minute. Be interesting to see the price.
http://www.flowmeters.com/catalog_li...tion_set_id=73

[jaydee]

Originally posted by bigben2k
Oh yes.

In this case, we would want the block and pump only. Using the pump's curve, we can then extrapolate the pressure drop.

I think you should explain what we need this pressure drop measurement for. In fact maybe explain each measurement that "should" be taken and why they should be taken. I don't think to many of us DIY'ers really understand why these measurments are needed. Hell, I am not even sure anymore (maybe I never was).

Once we get that straiteded out then maybe move on to procedure and tools.

[bigben2k]

Thanks Zymrgy, but those have a high pressure drop, and in a cooling loop, with our poor little centrifugal pumps, we need all the pressure we can get!

I have a rotameter type flowmeter, similar to the one you displayed, and it's not too bad, with a claimed 5% error margin.

It's been suggested to wait for a magnetic one to appear on Ebay, but all in all, we're passed giving a practical, and available solution to the DIY block maker.

[Zymrgy]

would the pressure drop really matter that much just for testing the flow rate outside of the system? I mean have the thing piped from a swimming pool or other handy large container, then to you pump, rad, & waterblock & have the other end squirting wherever?

[pHaestus]

Well what is the DIY block builder interested in? What are the goals here?

What can I do to improve my waterblock's performance, and how does it compare to a typical commercial block?

Is it agreed that this is the question?

Now what are the limitations?

I don't want to spend any money on testing equipment of any kind. Maybe $10-20, but that's pushing it.

Ok so this means we are pretty limited in terms of test equipment even on ebay. The only real solution is to use thermistors (cheap cheap cheap) and calibrate them versus a more accurate temperature probe. I keep pimping this site, but it is a really elegant idea and solution and it is all open source and free:

http://www.benchtest.com/gp_Temp.html

He uses a sound card's gameport to get 4 thermistor readings pulled right into MBM. He also has all the software needed and an explanation of how to calibrate everything. You should be able to get a waterblock inlet water temperature, a radiator air in temp and a radiator air out temp and a reading for the underside of the CPU from these 4 probes. You will probably have a bit of hassle tracking down the right sized thermistors but otherwise no worries.

Regarding flow rates, I would suggest using the manometer instead of any sort of direct flow rate measurement. Just choose a pump that has a published P-Q curve and take it as truth. If you can measure the P from your manometer which is plumbed on either side of the pump, then you can just read the flow rate off that curve. Put a ball valve in the loop between the outlet of the wb and your res (and connect the pump to res with large ID hose). You should then be able to adjust flow rates 0.5-1.5 or so assuming you are using a 1/2" system and a relatively low resistance radiator. You will probably have to spend a bit of time fussing with the manometer to get the readings reproducibly.

Is this perfect? No. Taking temps at a variety of flow rates as well as wide open should give you a pretty good idea of where the changes in performance are coming from though. I assumed that is the goal here.

[pHaestus]

[MMZ_TimeLord]

CPUBurn gets my vote!

[jaydee]

Originally posted by MMZ_TimeLord
CPUBurn gets my vote!

I have used them all aswell and CPU Burn usually gets it the warmest from my experience. But as long as you use the same thing everytime it probably isn't to critical.

[Joe]

The util I used in my reviews was also stable and solid as CPU Burn. I noticed oscilations when I used other tests. http://www.procooling.com/files/burntest.exe and use the "CPU warming" setting to get the stable heating results that mimic the one pH just posted, would be interested in seeing if there is any difference with this app and the one pH uses.

[leejsmith]

i just built one of the game port probes today.

the actual build is very easy but the calibration is proving to be a differant story.

I have a small multimeter with a thermocouple that i am using to calibrate the thermisters and so far 2 out of the 4 are giving the same temps as the thermocouple i thought i had the other 2 correct but one is now giving +1C and the other -1C.

will try again with the other 2 in the morning.

the only problem i have with the software are the temps it reports as there is no decimal place shown. This must be a software issue but i dont know delphi so i cant make any changes to the code.

it works very well mith MBM.

[Gooserider]

Interesting thread, but I'm not seeing a lot of comment from us lowly newb types... Thought I'd pitch in my two cents worth

I fall in the category of the target audience pretty well. I'm making my blocks, and would like to see how they do. I'm not so much interested in comparing to an air block (I might not even get an air block!) as in finding out if I have 'acceptable' performance.

OTOH, I don't want to spend a lot of money just on testing gear, in part because I don't have it.

I am planning to get a Digi-doc (or equivalent, but I haven't seen any comparable alternatives mentioned)

I have a cheapo DVM with a temp probe.

One other constraint that I have, and I suspect others might share, is that I am NOT willing to do warranty voiding, and potentially destructive mods like cutting up my CPU sockets, or other non-reversible changes. (My GF is paying 1/2 on the computer, and wouldn't agree even if I wanted to...)

I'm somewhat more inclined towards spending on stuff that can be used on a regular basis (i.e. for system health monitoring) than I am for things that are only for testing.

I don't see absolute numbers as essential, however I do think it is important that they be off in a predictable way, and that they be repeatable. If a temp is off 5* in one test, I would like to see it off the same 5* in the rest. Secondly, If I run a test today, I should be able to run the same test tomorrow and get the same results.

I also have lots of buckets. However I know some folks have problems with that test, and it does seem a bit suspect to me. I'm not sure I get what Phaestus is talking about with is 'manometer test' however - could someone put up a plumbing diagram, or point me towards where the test is described in more detail? (Everyone else acts like they know what he's talking about, but I'm confused I guess that's why they say I'm a newbie...)

Thanks,

Gooserider

[bigben2k]

Here's a quick diagram (below):

In short, you put tees at the inlet and outlet of what you're measuring (in this case, a heatercore). The coolant comes in one side, and out the other. The pressure drop will appear in the form of a different height of water, in the vertical tubes.

FYI, it's a lot (!) easier if you mount the vertical tubes on a board, and put some horizontal lines, for measurements.