Re: product testing today – who is being served ?
 

Does the Swissflow needs special attachments? Otherwise yes, I'll find attachments for the Sponsler.

Swissflow can be had for ~$50 US, got mine in a group buy for $32 (+ shipping from the UK).

I know of no other reasonably priced flowmeters that is commonly available, and economical.

The Sponsler was a lucky find on eBay at $42.

I'll advise once I have both the Swissflow and the fittings for the Sponsler.

Re: product testing today – who is being served ?
 

The Swissflows I got have BSP threads. They sent me some BSP to NPT adapters. They used to sell a barbed version but I don't know if they still do. I like my Swissflow other than it is restrictive and they recommend a 20 micron filter. I would be very interested in the results.

I pretty much have all this stuff mentioned in pH post and already used it in previous testing. Another bonus of the CF633 is it will read and log the Swissflow flow meter output aswell.

Only thing I disliked about the CF633 is the speed of the temp changes. That is why I ended up not using it for temps. I use my ColeParmer Unit with the YSI probe to help mount blocks. It scans so fast you can see the temps changing when adjusting the mounting pressure. Helps point out to much or to little pressure on one side.

The biggest problem I have with CPU testing is getting the CPU temp. Not to many can just get the IHS milled.

Re: product testing today – who is being served ?
 

Darn, mine's not going to have any fittings. On to McMaster...

I hear you about milling the IHS. It's not an easy job, simply because you have to have a solid mill, and a beefy motor, to run a 1mm endmill, even though it's only a 0.6mm depth. I only know that Fettig is so equipped.

I suggested contracting someone to groove all of these IHS for us...

Re: product testing today – who is being served ?
 

I could do it on mine but I am not going to simply because of liability and I don't have the time.

This brings me to the next question of the cost of the probe and the meter that it will take for this measurement?

Re: product testing today – who is being served ?
 

I have a CF635 coming in the mail that I will get reviewed and posted in the next couple weeks (focusing on getting site work done still until then)

Re: product testing today – who is being served ?
 

reality ck guys
look at those DOW sensor packages, good for air and water (in a sheath)
not gonna work in Incoherent's die sim or inlet into the top of an IHS
- there may be a thermistor soln, there is a known TC soln

jd
I buy Fluke 2190As all the time off eBay, avg ~$50 inc shipping plus $75 or so to cal
compare specs, best deal on the planet
you are presumably a more serious tester, but if you are unwilling to groove your test CPU; then with a guesstimated C and a guesstimated W what are your efforts worth ?
the whole point of this is to have a known temp related to the CPU temp on a known platform

an approach from the other side ?, what is a 'reasonable' cost for a CPU based thermal test setup ?

and if grooving the CPU is a deal-breaker, we should probably let this be and consider it a higher-end testing activity
note that the 775 has no pins, less fragile, and 'no known' internal TIM joint problems
(a CPU heat source, not your product endorsement jd)

Re: product testing today – who is being served ?
 

"and if grooving the CPU is a deal-breaker, we should probably let this be and consider it a higher-end testing activity"

Hmm I thought avoiding higher end testing activity was the whole point?

For AMD you can tap into the diode like I do. Works ok I guess; I see nothing wrong with it anyway (lol?). My plan is to get an AMD64 "grooved" on top as per your specs and then publish offsets between die and IHS temps and watch them over time while I'm testing on this platform. Cost of that setup is $100 plus being able to solder and not being skeered to desolder pins from your 939 chip and resolder wires on it.

//edit: HONESTLY my old socketA setup was a pretty good example of a readily attainable test platform for everyone. Nothing too expensive in that package; it worked well enough for general use. It required a lot of fiddling of course and an eye for whether the instrumentation was still working properly I guess.

Re: product testing today – who is being served ?
 

the 'fiddling part' is what we need to avoid
there must be a staight-forward means to verify a baseline, and very stable operation thereafter
the troubleshooting ability of 'reviewers' should be considered zip, yours is a high-end system by that definition pH

my comment about dropping this was intended for jd, if he won't consider grooving his IHS then he can make a heat die
(that will make a groove seem pretty attractive)

Re: product testing today – who is being served ?
 

I think you need to see what people use now and get a feel for what they will do to improve it.

I can pretty much guarantee they will not:

1) Spend more than what is in their pocket
2) Send anything out for calibration or calibrate what they have
3) Mill their IHS
4) Solder pins
5) Use any methodology that takes more than an hour to test a block

The sites/people that are actually whiling to do anything to improve their test bench already do so. The rest just don't care. That is the way I see it anyway. I think the problem isn't so much in the test setup but the willingness to build it and learn how to troubleshoot it. pH has already laid the ground work for what can be done with his Socket A bench and it has been out for a year or 2 now? How many have adopted his work or have done similar? Can count the one's I know on one hand.

Let's face it most reviewers have no clue how to get anything resembling consistent results and many simply claim they need not bother with better methods.

Anyway we are already deep into the pockets for this setup not counting the computer itself. Already to much money.

CF633 + probes $150
Used Fluke calibrated $150
Usable pump $100
Flow meter $50
Radiator/Fan $25 (heater core style)
Misc. valves/plumbing $25

guess we are throwing out dP all together?

You are already at $500 conservatively.

Then add in a dedicated computer....

Re: product testing today – who is being served ?
 

I am not saying I won't (although I won't because I don't need to), I am saying most don't have the ability to get it done. I don't think it is an option.

Re: product testing today – who is being served ?
 

Not being willing to take more than an hour to test a block to me seems rather lame... to test a HSF I would assume it would take 3 or 4 hours of testing... to test a waterblock I would assume the same to do repeated tests to make sure your moutning is not fubar.

Re: product testing today – who is being served ?
 

Exactly my point. However most tests I see done (that actually state times) are done in less than an hour. 30 mins usually. I don't even start taking temps for a hour or 2 on each mount. Have to get equilibrium established.

This brings me to my next question on who exactly is this thread intended for?

Re: product testing today – who is being served ?
 

This thread is intended to help bring an acceptable testing solution so that reviewer can get comparible data across websites, without having the massive expense and days of time it takes to test something.

3 or 4 hours are nothing compared to what a block would take to test at R+D levels. Hell it would take 3 hours to setup a test on high end gear.

Re: product testing today – who is being served ?
 

Ok so who are we talking to? Shouldn't the reviewers interested in such a bench be in the conversation? Otherwise we are talking to ourselfs with our own biases.

We need outside reviewers with their perspective on what they will and will not do and then work with that info. What seems simple to us is not so for them. That not only will give us a better idea what to look for but may also educate the reviewer on better methods to get the desired results. We can design this bench but out of 10 people using it you will get 10 different results if the testing methods/conditions are not the same.

Re: product testing today – who is being served ?
 

last line, post #1

but before we try to sell something we need to know if what we would propose:
works
is repeatable
at a tolerable cost

THEN address the notions of the reviewers

Re: product testing today – who is being served ?
 

$500 is already more than most will pay IMO but carry on. We can give e-bay lessons I guess.

What happens if the CPU and socket gets discontinued?

Re: product testing today – who is being served ?
 

Well what of an open source testing rig similar to the setup incoherent made? We provide CNC drwgs and circuitboard schematics/parts lists/etc? We provide test results for a few reference blocks? Like the innovatek system but with better parts? In some ways this is actually easier than testing with a real CPU. And with a single platform then we have the ability to write detailed directions and usage notes too.

The problem I see there is what happens when a measurement is out of calibration? It's very hard to make this into a black box that'd run for years.

And if you send people to ebay as a start for surplus/used parts with no experience then I think they're doomed anyway.

Re: product testing today – who is being served ?
 

Hum... I see a digression in this thread, from one statement:
Maybe you can expand on that JD?

You already have a heat die, we're just proposing to cap it off with a salvaged IHS. It has the advantage of reducing/eliminating the maintenance on the top surface of the CPU simulator. Plus, the temp measurement is done through an inexpensive thermocouple (measuring instrument aside).

If someone (another reviewer) wants to run tests with a real CPU, he can have it grooved, and wire into the CPU temp diode, just as pHaestus did.

I don't understand the objection here.

Bill suggests:
Picking up the 2190A is easy. Having it calibrated is certainly up to the owner of the equipment, but it's relatively inexpensive, and validates the measurements. Heck, as stingy as I can be, even I would do it! Running a calibration procedure oneself is possible, but I've never done it, so I can't say with any kind of certainty that I would be able to do it successfully.

What I do know, is that we all need a starting point, e.g. a reference temperature measurement; ONE calibrated instrument does that. If you want to run your calibration procedure yourself after that, knock yourself out. I'm (barely) starting out with a 0.1 C graduated mercury thermometer, ranged from 19 to 27 deg C. $30, and I haven't taken a single temp measurement yet...

Re: product testing today – who is being served ?
 

I thought we were trying to put together a CPU based test bench all reviewers could afford and operate.

Re: product testing today – who is being served ?
 

Yeah agreed, but then reviewers have to spend time testing the ref. blocks (even if they have done so previously) and then spend time pulling their hair out when their setup gives diff. results?

Are you guys going to provide troubleshooting/support for them too?

Re: product testing today – who is being served ?
 

I don't think that we excluded a heat die just yet. In fact, I believe that a heat die is still preferable, but true CPU as a heat die would allow us to figure out what offset to apply to our results, to give out data that a user can relate to, i.e. :

I have an (AMD/Intel) proccy model XXX and I'm considering ABC cooling solution. Our testing results show that one would be able to get temperature X, but apply correction Y to figure out what one would actually read.

I'm not suggesting that we do both, but rather that we get together; one does real CPU, another a die sim. This requires that we measure temps the same way, so that we can relate. Grooved IHS measurement can be done by every tester.

Al, I'd be more than happy to assist anyone that has testing issues.

Re: product testing today – who is being served ?
 

no Ben, no heat die is any part of this discussion
no pH, ditto

you are not going to get anyone at all to make a convoluted heat die (even me probably)

if a TC is put on the IHS it will have to be read with a thermometer
anyone with a good price/performance suggestion (not off eBay) ?

refocus
if the goal is a CPU based testing system, then it could be considered in 2 parts:
1) grooved CPU and TC thermometer, CFxxx and DOW sensors for air (water optional) - suitable for sink or kit testing, components by substitution
2) as above plus flow and pressure measurement capability - to create parametric performance curves for components

lets focus on #1 only for now

Re: product testing today – who is being served ?
 

Be noted the CF633 and sensors need serious calibration. In my experience the probes are far from consistent from each other. I got 8 probes with mine and only 3 would read somewhat close together. Others were off .5 to 5C. Also had one that would not scale liner to the others. For ever 10C on the others it would raise 9C for example. I found it to be a pain in the ass but it worked out ok after cherry picking the probes. Kinds sucks to have to buy a bunch of probes and hand pick though. Also the probes are far from water tight. They need good water proofing.

I still don't see people grooving their IHS's though.

Re: product testing today – who is being served ?
 

I'd like to see this also, but with both setups as professional as can be, so as to correlate/interpret the data. i.e. not what we're after in this thread.


Not in house, for sure. May be easier if we source a machinist or two, even better if we get mulitple reviewers to commit and send many procs to be done at the same time. I fear this may be unlikely, depends on cost though...

--Jay

Re: product testing today – who is being served ?
 

re TCs
http://cgi.ebay.com/Fluke-2190A-Digi...QQcmdZViewItem
these things are not difficult to find

Re: product testing today – who is being served ?
 

Bill whats that wrong with the one I currently have (or if you think it will work for a test - it would seem that the resolution is similar to the fluke)

http://www.procooling.com/index.php?...icles&disp=100

Re: product testing today – who is being served ?
 

I was pondering last time when I was testing with TC's to pick up a calibrator such as : http://www.extech.com/instrument/pro...433201data.pdf

seems to be a "cheap" way to make sure your gear is still returning valid data.

Re: product testing today – who is being served ?
 

same resolution, the accuracy/uncertanty ~1/2 that of the Fluke (±0.5 vs. 0.3 or so, as I recall)
yes, a calibrator will work as a ck (eBay ?)

Re: product testing today – who is being served ?
 

Then could it be an option to provide custom made IHS, and already grooved, according to specs?
People would then cap off their AMD 64 cpu's, substitute their original IHS for these, and the fear of damaging the processor if one would attempt to groove the IHS would disappear.

Re: product testing today – who is being served ?
 

so a DIY TIM joint is automagically better than a factory one ?
in your dreams, going to take some data to convince me

my effort is to 'qualify' an existing CPU as a heat source,
not to make a heat die with an IHS as I do not think that such is within anyone's budget, talk this one to death

IF a CPU is to be selected for use as a heat source, why on earth pick a CPU with known TIM joint reliability issues ?

it seems the 'problem' with this idea is that using an Intel CPU is out of favor with the OCing community, though how Intel CPU heat differs from AMD heat is unclear

lets revisit die temps
what are the silicon temp 'limits' ? (not the 'case' (IHS) temp given in the thermal design gides eh)
I happen to know a number of these values, short term, long term, etc. but this info is not public
why the pursuit of die temps when the limits are unknown ?