Ultimate water block tester

[Brians256]

Originally Posted by Incoherent

Biggest problem is establishing how much energy you are actually placing on the block.

Exactly. This is why I didn't think you were serious, as I wasn't sure how well you could repeatably ensure high percentages of energy were transmitted to base of WB. Certainly an interesting idea.

[Annirak]

It's highly repeatable. The energy output of the laser is pretty constant. I've seen repeatability figures of +/-3%.

The properties of the paint are going to be pretty similar too. If calibration runs are done, I don't see any reason for this not to be highly repeatable. And the whole concept of a calibration run is a lot less intensive in this situation, compared to what we'd expect from a so-called callibration run on a heat die simulator.

[jlrii]

A very interesting idea. Eliminates the TIM completely excepting whatever value you would have to the non reflective coating on the base of the block, which would be negligible for a few 1000ths of an inch of...hmmm, what? Flat black Al anodizing? Mounting would be a breeze as well as die size, a square internally mirrored (chromed, mirrored glass?) tube, possibly square, with whatever fittings necessary to mount the laser on one end and a simple rubber gasket on the other. A vacuum applied to the tube would hold the block in place and eliminate any interference within the tube. Expensive but brilliant. Too bad the market isn't bigger.

[redleader]

Inductive coupling would be 1000x cheaper, and probably more precise. No need to worry about defraction that way, and you can measure the power injected accurately with a meter (though depending on the frequency, maybe not a cheap one).

I've seen some very nice ones used as medical devices. The medical grade stuff costs a couple hundred to make, is small, and not that complicated.

[Annirak]

Brilliant, eh? Yes, a laser based testing system would be very shiny indeed.

[Annirak]

I had wondered about inductive coupling (I'm an electronics engineering technologist) but then I realised the problem with that one. It doesn't force the heat to travel through the heatsink baseplate. It might even induce heat in the pins of a pin-based block.

No, too many variables to account for in an inductively coupled system. We're trying to eliminate variables, here.

[jlrii]

Yes, good thinking. It has always bothered me that the TIM was included, as many variables as possible should be eliminated, then attended to separately.

[Albigger]

Well, there's also laser power sensors that can be used to measure the output of the laser, of course more cost there...

[jlrii]

Plus the added advantage of pointing at those we don't like....What are the restrictions on purchasing lasers anyways?

[jaydee]

Ok I am lost. How exactly is this going to represent a CPU any better than a Die Sim? Seems to me the results would be just as inaccurate compared to real CPU's.

Water blocks are made and used on CPU's. The only real way to tell how well that block works on a CPU is to test it on a CPU. However this is still screwed because of the many different CPU's on the market. It would be the same as using a different sized die sim. The results would be shit on one CPU and excellent on another.

Just not seeing this as a solution or improvement to the problem.

[Albigger]

Originally Posted by jaydee

Ok I am lost. How exactly is this going to represent a CPU any better than a Die Sim? Seems to me the results would be just as inaccurate compared to real CPU's.

well its not. but whose goal is to reproduce variable results that (most) people get when testing on cpus - swiftechs? surely the majority are concerned with true waterblock performance?

Water blocks are made and used on CPU's. The only real way to tell how well that block works on a CPU is to test it on a CPU. However this is still screwed because of the many different CPU's on the market. It would be the same as using a different sized die sim. The results would be shit on one CPU and excellent on another.

Just not seeing this as a solution or improvement to the problem.

differen't sized die sims still an issue (there are different cpu sizes as you note) but with a laser simply changing the distance from the laser would change the focusing area, no?

for thoughts on testing on real cpus, see Joes post:
http://forums.procooling.com/vbb/sho...&postcount=470


and finally, testing with an IR laser could potentially be better than a die sim because it eliminates mounting variability.

[gone_fishin]

Originally Posted by jaydee

Ok I am lost. How exactly is this going to represent a CPU any better than a Die Sim?

It's not.

Problem I see is it could be fooled by an irregular surface on the bp (read non-flat). More surface area in a given geometry on the bottom of the bp would give skewed results. Flatter is better on a cpu but the opposite would give better results on this hypothetical bench test.

[jaydee]

Originally Posted by Albigger

well its not. but whose goal is to reproduce variable results that (most) people get when testing on cpus - swiftechs? surely the majority are concerned with true waterblock performance?

Any reviewer writing for a tech site I would hope. Otherwise why do they bother?

differen't sized die sims still an issue (there are different cpu sizes as you note) but with a laser simply changing the distance from the laser would change the focusing area, no?

Not from my experience operating a laser for 14 months. Change the distance changes everything, most importantly heat. Also changes in flatness on the base will cause issues. Lasers are very focus orientated. Doesn't take much to screw it up.

for thoughts on testing on real cpus, see Joes post:
http://forums.procooling.com/vbb/sho...&postcount=470

Seen it and already know all about it.

and finally, testing with an IR laser could potentially be better than a die sim because it eliminates mounting variability.

Sure but not by any amount worth the cost. We already have seen die sims have very good repeatability. Plenty good for water block testing. The TIM joint IMO isn't as much of an issue with Die Sims as it is with CPU's.

From what I gather from various threads the CPU TIM joint is the issue being discussed not the joint on die sims.

[Albigger]

Originally Posted by gone_fishin

It's not.

Problem I see is it could be fooled by an irregular surface on the bp (read non-flat). More surface area in a given geometry on the bottom of the bp would give skewed results. Flatter is better on a cpu but the opposite would give better results on this hypothetical bench test.

How would flatness have anything to do with how much energy is absorbed from the laser? Even if different this could easily be calibrated for with each wb.


Though there is one interesting point - the flatness of the base wouldn't matter (as I see it) but on a real cpu it would.

[gone_fishin]

Originally Posted by Albigger

How would flatness have anything to do with how much energy is absorbed from the laser? Even if different this could easily be calibrated for with each wb.


Though there is one interesting point - the flatness of the base wouldn't matter (as I see it) but on a real cpu it would.

Surface area enlarges as flatness goes down the shitter in a given box (say 14mm X 14mm for example).

[Albigger]

Originally Posted by jaydee

1)Any reviewer writing for a tech site I would hope. Otherwise why do they bother?



2)Not from my experience operating a laser for 14 months. Change the distance changes everything, most importantly heat. Also changes in flatness on the base will cause issues. Lasers are very focus orientated. Doesn't take much to screw it up.

3)Seen it and already know all about it.


4)Sure but not by any amount worth the cost. We already have seen die sims have very good repeatability. Plenty good for water block testing. The TIM joint IMO isn't as much of an issue with Die Sims as it is with CPU's.

From what I gather from various threads the CPU TIM joint is the issue being discussed not the joint on die sims.

1) well if all review sites want to do is replicate what is seen on cpus in real systems then that's fine. Why even bother with die sims, ttv's or this setup - all that's needed is review sites to stick different coolers on real cpus and users in forums to say their own results, right?

As I see it, results of die sims are not "inaccurate" as you put it, rather the other way around (results of tests done on actual cpu's are more likely to be 'inaccurate'). I guess this topic has been talked about enough around here though...

2) didn't know that - calibration with each waterblock base plate would then be necessary I think

3) ok

4) ok (though we haven't fully explored the cost yet). If it is cost prohibitive/ineffective, then we need different sized die sims? And still requires ~5mounts to ensure accuracy.

--Jay

[Albigger]

Originally Posted by gone_fishin

Surface area enlarges as flatness goes down the shitter in a given box (say 14mm X 14mm for example).

Yeah I understand that. I guess what I don't understand is how that relates to the input energy of the laser?

The laser energy is (ideally) fixed and finite. No matter if it is focused on 100sq. mm or 1000 sq. mm the same amount of energy will be absorbed (if same type of surface), right? What am I missing?

--Jay

[pauldenton]

Originally Posted by Annirak

It's highly repeatable. The energy output of the laser is pretty constant. I've seen repeatability figures of +/-3%.

hmm - isn't +/- 3% actually pretty bad for these purposes??? (or am i misunderstanding you... i assume you mean the output of the laser varies by up to 3%... )

[jaydee]

Originally Posted by Albigger

1) well if all review sites want to do is replicate what is seen on cpus in real systems then that's fine. Why even bother with die sims, ttv's or this setup - all that's needed is review sites to stick different coolers on real cpus and users in forums to say their own results, right?

Pretty much how it is done everywhere else. Problem is getting accurate temps off the CPU.

As I see it, results of die sims are not "inaccurate" as you put it, rather the other way around (results of tests done on actual cpu's are more likely to be 'inaccurate'). I guess this topic has been talked about enough around here though...

You are not following what I am saying I guess.... What are water blocks designed for? CPU's. Why on earth would anyone care how a block performs on a die sim that doesn't replicate a CPU? There is just no reason to have or know this information as it doesn't pertain to anything. This is why die sims results are not accurate because they do not tell us the needed information for which the water block was designed for. The water block was not designed to cool die sims. The data collected from die sims (I have plenty of it from my own) doesn't tell us how the block performs on the platform it was designed for.

2) didn't know that - calibration with each water block base plate would then be necessary I think

Getting the exact same focus over and over is hard to do. Makes mounting the block to a die sim seem less of a challenge in fact. Power changes with focus.

3) ok

4) ok (though we haven't fully explored the cost yet). If it is cost prohibitive/ineffective, then we need different sized die sims? And still requires ~5mounts to ensure accuracy.

Well our little 80watt laser with a 100watt tube cost $45,000 about 5 years ago. And it was setup just to cut and engrave. Nothing fancy. To get setup properly I would suspect in the $100,000+ range.

--Jay

Another thing I havn't figured out is were you intend to put the temp probe(s)? Can't put it in the area the laser is focused on yet that area is were the probe needs to be.

[Albigger]

Originally Posted by jaydee

Pretty much how it is done everywhere else. Problem is getting accurate temps off the CPU.

You are not following what I am saying I guess.... What are water blocks designed for? CPU's. Why on earth would anyone care how a block performs on a die sim that doesn't replicate a CPU? There is just no reason to have or know this information as it doesn't pertain to anything. This is why die sims results are not accurate because they do not tell us the needed information for which the water block was designed for. The water block was not designed to cool die sims. The data collected from die sims (I have plenty of it from my own) doesn't tell us how the block performs on the platform it was designed for.

Well, so far all the data I know of shows that heat dies replicate bare cpu dies just fine. There is only disagreement/ambiguity when you talk about cpus with IHS. This data you speak of that doesn't tell us how blocks perform on cpus, are you talking bare die or not?

Maybe there should be two methods of testing to replicate the two areas of usage...

Getting the exact same focus over and over is hard to do. Makes mounting the block to a die sim seem less of a challenge in fact. Power changes with focus.


Well our little 80watt laser with a 100watt tube cost $45,000 about 5 years ago. And it was setup just to cut and engrave. Nothing fancy. To get setup properly I would suspect in the $100,000+ range.

both of these seem to make this method extremely difficult and cost prohibitive.

Another thing I havn't figured out is were you intend to put the temp probe(s)? Can't put it in the area the laser is focused on yet that area is were the probe needs to be.

was addressed in the first post, point #2. But I was not really convinced this would be the best, as Annirak is proposing to put it in the laser. IF this method was used a probe inside but close to the heated surface of a copper calibration block would have to verify that the probe in the IR beam was reading temps appropriately, however I doubt it would.

Could put it just on the outside of the area of focus?

--Jay

[Annirak]

jaydee, the point of a die sim is not to replicate a CPU. It is to apply, repeatably, heat to a DUT in the same shape and approximate heat load as a CPU. What is my laser proposal but another way to apply heat, repeatably, to a DUT? There are all manner of problems with heat die simulators.

1) The temperature measurement of the base of the DUT is an approximation at best and a stab in the dark at worst. The measurement is clouded by the TIM layer, and you must be joking if you try to tell me that the temp sensor embedded near the surface of the die is equivalent to the base plate temperature.

2) The DUT must be mounted using variable TIM. Are we testing the DUT or the TIM? If we're testing the DUT, why are we using TIM. To get the heat to the DUT. Is there another way to get heat to a DUT? Hence my discussion.

3) Losses. I don't care how much you characterize a die sim, there are still thermal losses in it. Heat leaks out all over the place.

A laser based sim would have advantages in all these areas:

1) The thermal sensor can be attached strongly to the base of the waterblock, and is not in the thermal flow path, so it should have an isothermal relationship with the waterblock

2) The DUT has heat applied directly to its surface, there is no question of variability of TIM.

3) Losses are easy to characterize with a laser system. It's possible to minimize them with reflective surfaces

Operating a device does not imply an understanding of it. I've met pilots who claim that an airplane flies because of the Bernoulli effect. This is not the case. I'm not claiming that you don't understand, but I'm saying that practical experience with a device does not imply an understanding of its operation.

We're not trying to focus anything. That would be a mistake, as it would probably cut a hole straight through the DUT. The goal is to spread the beam. This spread would have to be quantified, and calibrated. That's obvious.

Jaydee, please re-read my initial post. I covered probe placement in it.

For device cost, that was covered in another post of mine. 70W laser setups can be had between $5k and $10k. I saw 100W lasers at ~$12k.

The flatness of the base will cause 0 problems. If light is going to reflect, then the only change will the be angle it reflects on. I said before that it should be painted matte black to ensure full absorbtion. The flatness doesn't matter nearly as much as the cross-sectional area exposed to the beam. Which is constant, and determined by the lensing.

AFAIK the purchase of lasers is not restricted at all. The required optics to make them cause damage at range are... difficult at best.

[Annirak]

My goal is to provide a system that characterizes how well a waterblock transfers heat to water. Once that is assessed, we can talk about how it interfaces with any given CPU.

[jaydee]

Good luck Annirak. No point for me to continue in this.

[Voodoo Rufus]

Good idea.

As far as the block is concerned the CPU is simply a heat source. So eliminating the TIM joint from the testing process will increase repeatability even more. Simulating different die sizes by altering the focus should not be difficult in principle as well. If a laser is not readily available, there are other heat sources (electric heaters on a lens?) that may fit the bill, or warrant looking into at least.

[redleader]

Originally Posted by Annirak

I had wondered about inductive coupling (I'm an electronics engineering technologist) but then I realised the problem with that one. It doesn't force the heat to travel through the heatsink baseplate.

With proper design you can put the heat where ever you like. Also, pretty much all heat will be in the baseplate. Thats why induction is often used for surface heating. It doesn't penatrate well. Plus unlike a laser, its not influenced by reflectivity or scattering.

It might even induce heat in the pins of a pin-based block.

I don't see how you could induce remotely significant heat into a pin. They'd be oriented the wrong way relative to the field, and so there'd be no real circuit. You said you'd done EE right? Think about how the eddy currents will flow when the field passes through the material.

No, too many variables to account for in an inductively coupled system. We're trying to eliminate variables, here.

Exactly what are these variables?