Why is cutting the socket out? Takes 5 minutes with a ginsu knife and you pop the top piece off so it is not any danger to motherboards.

You constantly amaze me: I want relaible temps but I won't spend any money is what I hear. NOT GONNA HAPPEN.

Jaydee: I would assume that block designers would prefer to know why a particular design is superior (or inferior). If he can run them all at say 0.5 GPM, 0.75 GPM, 1 GPM, 1.2 GPM, and finally wide open then he can get a pretty good idea of the block's performance and how it is expected to perform in other types of systems. Now you'll need water temps as well to be able to guess whether changes in performance from block to block are due to changes in the water temp from restrictive blocks or if they are due to changes in the C/W of the blocks. Can you get at this without a flow measurement? I guess so by measuring the water temperature. You have no way of adjusting flow downward with a valve reproducibly though. A flowmeter or manometer would make things a lot easier though I think.

Let's be clear here; you are doing "system testing" and so the goals are necessarily a bit different. If a DIY guy with an Eheim 1048 made a #rotor block and found that water temps were high due to restrictivenes, would he get a louder fan, use a larger radiator, buy a larger pump, or make another block with different hole placement? Any of these would solve the problem, eh?

Let's find out!

New poll

I would think that most people would be highly resistant to making that mod to their mobo.

But hey, if it's the only way, then that's what we'll have to recomend.

You make a good point pHaestus, but maybe not for the right reasons.

This test procedure is strictly meant for one user, and is not transferable, but the user may want to know how the block might perform at another flow rate.

The manometer has advantages. It could be setup to measure the pressure drop across the block, as well as the heatercore, at the same time, and still cheaply.

any progress with Sandra?

I had to replace the burner and ignitor on my grill yesterday and I am also addicted to Diablo2 at the moment.

I will try and get down to my testing facility this evening. I want to mess around with my digital pressure gauge as well so hopefully I can kill 2 birds with one stone.

Put down D2 for testing?! NEVAR!!!

I sorta doubt you even go to work anymore :)

So who are we talking about now? It seems to me there is different levels of DIY'ers.

1) Makes 1 block and thats it. Which seems to me is the majority.

2) Makes one block from scratch and if it isn't as good as he thinks it should be he will then make another or mod the one he has.

3) has nothing better to do in life and buys expensive equipment to make the best block on the planet.


#1 isn't even going to think about spending money on any testing equipment being he probably made the block himself to save money in the first place. Which from what I can tell is the majority.

#2 is where this whole deal should be pointed at being they don't really want to spend money on anything but decent tests could be helpfull if they know how to do them which the majority do not.

#3 would be more inclined to buy decent testing equipment.

Also consider that Cathar built the best wb available using a minimal amount of equipment (a bucket, a cheapo digital thermometer, and an Abit motherboard right?)

Excellent point. In fact my REV. designs have been put together the same way so far and I would bet money they will compete with the top blocks on the market. So why am I putting together a test bed with die simulator? :shrug:

So that you can get a C/W rating, at different flow rates.

Why does a DIY'er need C/W rating?

Bragging rights?

It would allow someone else to gauge how the block might perform on their system: this is where the results become "portable" to other systems.

So this doesn't do a DIY'er any real good being he can just use CPU temps as bragging rights and as a gauge how well the block works and being it is his block it shouldn't matter how it works in someone elses system. Just the CPU temps alone should be a good indication if a different DIY'er may want to use the design himself.

Good news is a pretty much knew the answers to these questions. Bad news is my perspective is changing a bit on building the die simulator.... Bah! Is it worth the $200 it is going to take to finish it? Humm......

I agree with Jaydee's points on the "braging rights" Ben. He nailed it.

Testing = accuracy

Brags = high % BS

Or at best brags by most are based on poor information. And that type individual won't care a bit if the results won't be of use to anyone else, just as Jaydee also pointed out.

Jaydee,

But you know making fishing lures won't be enough to do on it's own. Although the results would be better eating.

Any good walleye fishing in your area? If you come up with a good lure for those you could make some $$ in the area I live in.:D

LOL, I am actually desinging 2 molds. One for Bass lures and one for Walleye and pan fish. There is big resivore here Called the Columbia River. Has some good walleye fishing. Also some smaller lakes around here have some good walleye fishing to.

I usually use plastic tube baits like the gitzit for bass and just regular small tubes for walley and pan fish. Unfortuantly I go through a good 20-30 or so on a good day bass fishing and at $.15 each it starts adding up. I think I can make them for considerably less and add some features I would like to have in them. Nothing like custom made stuff. :D

I am also moving forward with the die simulator. To far into it to stop now. Not to mention it will be harder to break a simulator than a mobo. Got a nice stack of tore up mobo's and would rather no add to them. :D:D

No... the TC on the back of the CPU would give the DIY'er the ability to come up with relatively accurate temperature differential, not actual CPU temperatures. No C/W rating is possible without a testbench, and that's the only rating that is transferable to another system.

If a DIY'er gets a C/W rating from whoever's testbench, he'll know how good the design actually is. A DIY'ers "good/best design" could still be an abysmal waterblock, compared to the popular commercial units.

"temperature differential".... Ouch, now I am confused. I do know the TC would not give "actual" CPU die temp, but it would be similar to how current mobo's read except, with any luck, more accuratly? And this is how AMD tests is it not?
This will only apply to the commercial or other blocks that have been tested by the SAME testbench though. Which lies the problem. BillA and Joe C tested the same blocks but came up with different C/W's for example.

I don't know about AMD, but otherwise you're on it: temp differential is all we can really hope for, from home. We can use a HSF, tested by OC as reference, but that's it.

I'm sure that there's a reason for that, but I'm not going to get into it. The results should however, fall within a stated error range, and if the results are different from one testbench to another, and fall outside each other's margin of error, then there's a problem with one (or both) testbenches.

Yeah they should come up with similar results in a perfect world, but there are way to many variables that can change the outcome. Something as simple as mounting pressure or even disimilarites between blocks themselfs (which I belive is more serious problem than realized).

Exactly ;)

But What am I suppose to be then? I am indeed building all my blocks myself, from home, with minimal equipment to my disposal, Yet I am regarded by virtually all those that have my product on the chips, as producing product almost beyond perfection. I'm not saying this to blow my own horn, even though I'm saying it through my own horn, please understand... :).

Granted..... Most of my Customers since then has become my friends. Some of them might not be thermodynamic engineers, but still, I'm sitting with a string of happy customers. Does this make me commercial, or not?

All this, leaves me in what could be considered, a very fortunate position.... That of having a bunch of people from all over the world, not willing to go elsewhere for their blockage, heck I even have a bunch of DIY'rs standing by my product. As they discovered for themselves, the beauty of simplicity. :)

as for how I test blocks, I can only compare to my own... and testing a block is really very very simple. To know one block, to be better or not, from the next, you simply install an external probe, of which you can determine the accuracy of it readings, underneath a CPU, preferably the old stile Durons. They do not have any components on their bellies. Now, in a block swap, you have to leave the probe completely untouched, Not even moving the paste, nothing is to change. That is why having the probe beneath the CPU is so important. Then you stop looking a the ºC symbol.... As it is of absolutely no value to you now.. you can run an initial test with a std HSF, to get a point of reference. The readings you get are all relative and pertains to that rig, and that rig alone.... this will enable you to, without a doubt, get the best amongst the blocks you are comparing, provided you do not disturb anything with regards to the probe, the tubes (length or otherwise) Radiator, pump whatever..... just the block gets swapped out. This method can cost you anything from less than $5, to well over $100 , and it all depends on the probe you want to install. for a DIYer I think this is as close to heaven as one can get on that kinda budget.

If you want to go one step further, you can now obtain a tested block, tested by Your truly, the grumpy one :) of coarse ), and toss that into your collection.... you will be amazed at how easy it will be now, to know if your block can kick ass out there or not.... :
But be warned, we are running big blowers and the boost is high. MMHHAHAHAH :D

I accept the limitations of my equipment, but I've come to trust it. Not in a scientifically verifiable sense, but trust it enough that I can continue to develop.

I do a number of things that people would say isn't the right thing to do, but it seems to be working fine for me.

My budget digital thermometers do the job. I calibrate them with respect to each other and I know what to expect from them. I keep the testing environment exactly the same each time.

I've learned to work with my tools, as bodgy as they are. A 10 litre bucket flow-rate test gives number to ~5% accuracy, but this is good enough to give an idea. I never claim to have it plottable on a curve like BillA does, but just enough to give me insight on whether I'm going backwards or forwards.

The Abit's thermal probe is coated with thermal grease and pushed up hard against the base of the T'bred B, which has no components under it (much like the old Durons/T'Birds).

I've plotted the rate of change of the on-die diode readings with respect to the budget digital thermometers by changing the room temperature. They are linear in progression with respect to each other over the room temperature ranges I can impose.

My office room is nice thermally. I can get it such that the room's temperature will vary by no more than 0.1C through controlling the rooms access to the house's central heating. The test-bed is isolated from any drafts which is something that plagued my old test-bed setup.

By changing the central heating, I can slowly raise the room's temperature by about 0.1C every 10 minutes. This allows me to plot when the CPU diode flips to the next 0.5C mark. By tracking the water temperature and air temperature to 0.1C (yes - I know it's not really that accurate - but bear with me) I've seen fairly reliable and repeatable occurances of temperature differentials which is what I'm really measuring. I can generate "pseudo-accuracy" down to around 0.5C, and be fairly confident that within 0.5C, I really am seeing something significant. Is it really 0.5C though? I have no idea. It's all relative.

Nothing is absolute. Everything is relative. As long as you're aware of the limitations of the equipment, and measure how the equipment measures things realtive to each other, it is possible to develop enough confidence to know whether something is better or worse than something else.

Do I trust it enough to really know what the absolute differences are? Not on your life. That's why I sent my blocks to BillA and happily paid him to test them.

Of course I'll get on line and say that I saw a 2.0C improvement, or a 5.0C improvement, because people ask "how much better" and I have to give some sort of answer. I am always very careful to state that it only applies to my testbed, and to a certain tested CPU load.

With all due respect Cathar, no air conditioning system can maintain a .1C degree variance in any room. Your mere body heat would change it, and frankly no house or office outside governmental labs could pull it off. Any room will have pockets of air high, low, or harizontally that will be more than that (drafts or not). Lets not get carried away.
I don't doubt your final results though, thats for sure.

I understand what you're saying. I just see the temperature fluctuate around a 0.1C range at the fan intake which is at 15cm above floor level. Up at body level, yeah, the temperatures move around a fair bit.

If you doubt it, try it. Stick a digital thermometer about 15cm (6") off the floor, away from drafts, and watch it. It'll sit pretty steady despite somewhat larger fluctuations at say 120cm high in the room.

Of course it'll move by larger amounts over some period of time, but within any 10 minute period, for me at least, it's pretty steady.

The gas heating vent is not in the room, so it's like the room is suddenly getting hot air, then not. The door's closed. I'm sitting in the room. There's no heat sources than myself or the system, and the temperature basically steadies out at 15cm off the floor.

Ah, I see more clearly now. Good explination.

Although I would be hesitant to place any weight on a .1C calcutation. Of course I may be tainted. I have a 50+ year old house, and although the airconditioner/heater is new. But once again, my perceptions dont exacltly make a case for relevance.

I have seen different mounts on my tests vary the temp 2-3C in bad cases. How do you overcome this?

Like I said, I don't really trust the measurements that much to be accurate, but they are good enough for me to tell when something's better/worse. In no way do I believe a 0.1C accuracy is truly happening here for the fan intake temperature, more like +/- 0.3C at best, but in any event, it serves the job well enough.

I build up a database of mounts. I do a mount. Take measurements at equilibirum under my test load. I pull it all off, clean it up, reapply past, and remount again. I keep doing this until I establish what to expect from a "good mount". Typically this takes up to 10 mounts. I then keep remounting until I get a mount that is in line with the best mounts seen over those 10 times, and it's now here where I begin to do tests and know that I can trust the results that I'm seeing.

Again, as seen from BillA's data, the mounting variations even for someone who really knows what they're doing like Bill does, can vary by around 0.01C/W (or about 0.7-0.8C for my test load), which is about my assumed margin of measurement error, meaning that I probably couldn't pick up a variation this small.

But I try. Generally I can tell pretty quickly when I have a bad mount, and I reckon I can tell when I'm very close to having a good mount.

*slaps self*