-
Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!
Is it time to work towards a 'Standard Heat Die' design and test procedure ??
- Thread starter BillA
- Start date
This is a great idea. Concerning the use of SolidWorks: What
are the output file capabilities? I know people who use
Rhino and other programs and then render graphics in
3DS Max for display.
Concerning the mounting capability. Is it possible to
come up with a universal design that would allow
testing of both AMD and the various Intel CPUs?
In a real CPU a certain about of heat is dissapated out
the back of the die. In what way can we build a
simulator so this "secondary path" becomes part of
the die simulator so ad hoc "adjustments" need not
be made?
What can be said about the requirements for the
power supply running the simulator?
are the output file capabilities? I know people who use
Rhino and other programs and then render graphics in
3DS Max for display.
Concerning the mounting capability. Is it possible to
come up with a universal design that would allow
testing of both AMD and the various Intel CPUs?
In a real CPU a certain about of heat is dissapated out
the back of the die. In what way can we build a
simulator so this "secondary path" becomes part of
the die simulator so ad hoc "adjustments" need not
be made?
What can be said about the requirements for the
power supply running the simulator?
- Joined
- Jan 6, 2002
I just so happen to have SolidWorks2001 and can help there.
Talk about opening a 55 Gallon barrel of worms. When it comes to standardized testing, there are nearly as many opinions as there are people. If you've read my thread about why the temperatures that are published on sites and message boards are nearly useless, you can see why some type of standardization would be nice.
Problem is, no matter how rigorous the standardization is, there will always be certain variables that will not be in control and thus a certain degree of inaccuracy in comparisons. I spent a rather large chunk of my life while writing my thesis wrestling with these inaccuracies and I just about went bald from it.
Anyhow, I'd like to hear a few more opinions from others out there before going much further on my comments and suggestions. I think this could be a great thread if people are willing to sit down and discuss it, and not worry about bruised egos in the process.
Talk about opening a 55 Gallon barrel of worms. When it comes to standardized testing, there are nearly as many opinions as there are people. If you've read my thread about why the temperatures that are published on sites and message boards are nearly useless, you can see why some type of standardization would be nice.
Problem is, no matter how rigorous the standardization is, there will always be certain variables that will not be in control and thus a certain degree of inaccuracy in comparisons. I spent a rather large chunk of my life while writing my thesis wrestling with these inaccuracies and I just about went bald from it.
Anyhow, I'd like to hear a few more opinions from others out there before going much further on my comments and suggestions. I think this could be a great thread if people are willing to sit down and discuss it, and not worry about bruised egos in the process.
- Joined
- Jan 29, 2002
- Location
- Calgary, Alberta
Just get Highlander944 in on this discussion and we'll have another thread where only four people in this forum will understand what is going on. LOL
Seriously though, I'll leave hows and whys up to the experts. But if we could get a way to compare different blocks, a clear indication of what we are actually comparing and some results that the average person is able to reference and make comparisons on different blocks, I think this would help all of us.
I will now leave this thread so the expert can hashout how to do this. I look forward the results.
My 2 cent (too bad its only about 1.2 Canadian)
Seriously though, I'll leave hows and whys up to the experts. But if we could get a way to compare different blocks, a clear indication of what we are actually comparing and some results that the average person is able to reference and make comparisons on different blocks, I think this would help all of us.
I will now leave this thread so the expert can hashout how to do this. I look forward the results.
My 2 cent (too bad its only about 1.2 Canadian)
Last edited by a moderator:
I just wanted to chip in that I'm not interested in the effects of the mounting mechanism. I would much rather see an apparatus that applied a reproducible block pressure on the core. I can work around bad mounting mechanisms, but I can't work around bad block design.
Thanks!
Thanks!
- Joined
- Jan 6, 2002
If we are going to arrive at a solution that is as accurate and as reproducible as possible, we have to be able to isolate the test subject as much as possible. As absolutely many variables as can be, must be eliminated. To me this says that if you want to test the water block and its design, the hold down system and the radiator (or whatever) must be eliminated from the equation. So must the processor and it's associated energy output. Unless of course a water block is designed for a specific processor. However, this would also mean having different testers (or top heating plates) to simulate the different processors.
Personally, I think it would be best to pick one platform and develop it before muddying the waters with more variables. Stick with a configuration that either specfically tests the die size of the Athlon XP, or come up with a generic plate that is a compromise between the Athlon and the Throroughbred. I'm afraid the PIV and it's heat spreader would have to be treated as a seperate case.
If this is to work, everyone would have to have the exact same simulator, using the same heater, placing all thermocouples in the same place, etc. etc. To this I think we can all agree. Where there is most likely going to be more variation, is the method of applying the voltage and supplying the fluid to the block. Bill has a pretty nice setup, but I highly doubt many will be able to or have the resources to build something similar. However, only goal that really needs to be met is that the voltage supplied is constant and accurately known, as well as the fluid volume flow rate and temperature.
Flow rate, now this is going to be a mess. I don't believe measurements need to be taken at exact flow rates, but whatever the flow rate used, it must be measured accurately. If enough data points are taken at various flow rates, an accurate enough curve should be able to be produced. Simple yet accurate flow meters could be used I suppose. Problem is, accurate flow meters are not exceptionally cheap. I do feel strongly that the 'fill the bucket and time it' methods are not acceptable and should not be used for accurate measurements, especially at lower flow rates where there is substantial variance.
Bill, just throwing out your best guess, how much do you think it would cost to build a simulator like yours in low volumes? I'm guessing there might be enough interest for maybe 5-6 or so to begin with. Also, what exactly to you need to be done with SolidWorks? On a side note for Tecumseh, almost any 3D modeling software worth its beans will output IGES files which is what CNC machines need to do their work.
If we start with isolating just the block, then later move on to the effects of the hold down, different radiators, etc. I think we will be in good shape. Establishing one tightly controlled method for as few variables as possible first, then moving on to more would be our best path.
For the record, I'm spending most of my spare time working on numerical models to help better understand the effects of different designs, flow rates, induced turbulence, etc. I've not had any time to work on any empirical testing of my own just yet, and may never be in the best position to do so anyhow. Give me numbers and I'll crunch them, if I can weasel in any extra time I'd love to work on some real world testing.
Personally, I think it would be best to pick one platform and develop it before muddying the waters with more variables. Stick with a configuration that either specfically tests the die size of the Athlon XP, or come up with a generic plate that is a compromise between the Athlon and the Throroughbred. I'm afraid the PIV and it's heat spreader would have to be treated as a seperate case.
If this is to work, everyone would have to have the exact same simulator, using the same heater, placing all thermocouples in the same place, etc. etc. To this I think we can all agree. Where there is most likely going to be more variation, is the method of applying the voltage and supplying the fluid to the block. Bill has a pretty nice setup, but I highly doubt many will be able to or have the resources to build something similar. However, only goal that really needs to be met is that the voltage supplied is constant and accurately known, as well as the fluid volume flow rate and temperature.
Flow rate, now this is going to be a mess. I don't believe measurements need to be taken at exact flow rates, but whatever the flow rate used, it must be measured accurately. If enough data points are taken at various flow rates, an accurate enough curve should be able to be produced. Simple yet accurate flow meters could be used I suppose. Problem is, accurate flow meters are not exceptionally cheap. I do feel strongly that the 'fill the bucket and time it' methods are not acceptable and should not be used for accurate measurements, especially at lower flow rates where there is substantial variance.
Bill, just throwing out your best guess, how much do you think it would cost to build a simulator like yours in low volumes? I'm guessing there might be enough interest for maybe 5-6 or so to begin with. Also, what exactly to you need to be done with SolidWorks? On a side note for Tecumseh, almost any 3D modeling software worth its beans will output IGES files which is what CNC machines need to do their work.
If we start with isolating just the block, then later move on to the effects of the hold down, different radiators, etc. I think we will be in good shape. Establishing one tightly controlled method for as few variables as possible first, then moving on to more would be our best path.
For the record, I'm spending most of my spare time working on numerical models to help better understand the effects of different designs, flow rates, induced turbulence, etc. I've not had any time to work on any empirical testing of my own just yet, and may never be in the best position to do so anyhow. Give me numbers and I'll crunch them, if I can weasel in any extra time I'd love to work on some real world testing.
Aesik said:
Flow rate measurment will be a mess. I have looked into various
flow sensors and guages available for the rates we will need
and none of them seem accurate enough. But, before I forget,
are we only going for WB testing? What about HSFs?
Is it universal that CNC machine software can take an IGES
file and convert it to its native format? Since IGES is a loose
group of standards or poorly enforced, how reliable is this
capability?
I'm glad you are working on your numerical codes, so if you are
like me you'd love to have loads of high-quality empirical data
to compare against.
- Joined
- Jan 6, 2002
Tecumseh said:
Flow rate measurment will be a mess. I have looked into various
flow sensors and guages available for the rates we will need
and none of them seem accurate enough. But, before I forget,
are we only going for WB testing? What about HSFs?
Is it universal that CNC machine software can take an IGES
file and convert it to its native format? Since IGES is a loose
group of standards or poorly enforced, how reliable is this
capability?
I'm glad you are working on your numerical codes, so if you are
like me you'd love to have loads of high-quality empirical data
to compare against.
I'd say stick with one thing first, meaning work out the details for testing water blocks then later worry about HSFs.
Most newer CNC machines do quite well with IGES files. I've seen a few problems with them, but most knowlegable machinists can overcome them easily.
And yes, I'd love lots of empirical and trustworthy data to validate my models!
- Joined
- Dec 20, 2001
- Location
- New Port Richey
Hey to all
imho..you need a controllable heat source(pelt)with thermistor or diode feedback to the voltage control IC of the power supply. (to the sense input) To me, an absolute heat source is basic for the measurement of heatsink efficiency.
just my 2 penneys worth
imho..you need a controllable heat source(pelt)with thermistor or diode feedback to the voltage control IC of the power supply. (to the sense input) To me, an absolute heat source is basic for the measurement of heatsink efficiency.
just my 2 penneys worth
- Joined
- Sep 9, 2001
- Location
- Pocatello, ID
I'm going to jump in just a bit over my head here and make a suggestion for mounting.
I think you're right that you need to be able to isolate the wb from the mounting assembly. However, most end users will not modify the mounting assembly of a wb. So if a great block has a lousy mounting assembly, for the end user it's only a mediocre wb.
Fortunately I don't think you have to choose between isolating the assembly and not in designing the simulator.
I notice that Bill's simulator has four posts. Simply have a separate board with an appropriate mechanism for mounting the block. Then mount that board to the simulator via the 4 posts. The mounting of the board could be done either by using calibrated springs, thus isolating the effect of the wb mounting assembly, or it could be mounted at a reference height so that pressure would vary according to the wb mounting assembly. In this way you could test either the isolated block or the block including mounting assembly.
One other thing, how accurate does the flow rate measurement need to be? Also, how long will the test likely run. I'm guessing that in order to ensure constant input temps, the water won't be recycled.
nihili
*Edit* Ok, having read BillA's list of equipment, that last paragraph of mine is gibberish. I'll leave it in, but it's probably best ignored.
I think you're right that you need to be able to isolate the wb from the mounting assembly. However, most end users will not modify the mounting assembly of a wb. So if a great block has a lousy mounting assembly, for the end user it's only a mediocre wb.
Fortunately I don't think you have to choose between isolating the assembly and not in designing the simulator.
I notice that Bill's simulator has four posts. Simply have a separate board with an appropriate mechanism for mounting the block. Then mount that board to the simulator via the 4 posts. The mounting of the board could be done either by using calibrated springs, thus isolating the effect of the wb mounting assembly, or it could be mounted at a reference height so that pressure would vary according to the wb mounting assembly. In this way you could test either the isolated block or the block including mounting assembly.
One other thing, how accurate does the flow rate measurement need to be? Also, how long will the test likely run. I'm guessing that in order to ensure constant input temps, the water won't be recycled.
nihili
*Edit* Ok, having read BillA's list of equipment, that last paragraph of mine is gibberish. I'll leave it in, but it's probably best ignored.
Last edited:
- Joined
- Nov 18, 2001
- Location
- Troutman, NC
Actually, the CNC machine needs G-Code (also called ISO or EIA programs). What the CAD/CAM software usually outputs is called APT (Automatically Programmed Tool). APT is a "universal" cutter location language (in ASCII Text) that is then run through a Post Processor for the particular CNC machine in question. There are some CAM programs that put out their own G-Code, but these are usually pretty limited.
IGES (Initial Graphics Exchange Specification) is a "common" file format that many CAD/CAM packages can read/write to, kind of like an .rtf can be read by both Word and WordPerfect. IGES works well for 3D wireframes, but it has its limitations as far as actual solids are concerned. It's the CAD/CAM software that uses the IGES file. There are some newer "open control" CNC machines that have their own built in CAM software that would probably take an IGES file directly (usually pretty limited as well), it all depends on the machine.
Sorry for being long winded, but I finally get to post about something that I actually know a little about. I've been a CNC programmer/Mfg Engineer for 5 years. I know CATIA, APT, AutoCAD and a few others. With the manual for the particular CNC machine, I could probably write a straight up G-Code program for anything that y'all would need. I can also operate just about any CNC machine, so if someone would like to donate one...
Anyway, I'd love to help out if y'all need it.
- Joined
- Jan 6, 2002
stymee had to all go and get technical on us! (slight inside joke there) The machinist I work with takes the IGES files I give him and puts them right into the CNC machine. I guess the software on the machine itself is doing any of the converting that it needs to.
I have both Freehand and Solidworks so if necessary I could work on them, but like Bill said, perhaps someone else has more time that would like to contribute that way.
Bill, I'll take any data you have (spread sheet format would be nice) for now as anything is good for me to use as my numerical model evolves.
BTW, there is a chance I could get some really cheap machine time assuming I could find a good source for the metal stock.
I have both Freehand and Solidworks so if necessary I could work on them, but like Bill said, perhaps someone else has more time that would like to contribute that way.
Bill, I'll take any data you have (spread sheet format would be nice) for now as anything is good for me to use as my numerical model evolves.
BTW, there is a chance I could get some really cheap machine time assuming I could find a good source for the metal stock.
- Joined
- Nov 25, 2001
- Location
- Niagara Falls, Ontario, Canada
Howabout to emulate secondary heat paths you do something like bolt the die simulator to a 20x20cm double sided (copper clad both sides) chunk of fiberglass PCB material through a 1/4 inch chunk of PVC to emulate the socket, with bolts that are brass and calculated to approximate the area of the pins of a typical CPU threaded into a specified thickness backplate and a close fit in the holes through the board with brass nuts and washers clamping to the back of the PCB and providing some thermal contact. Then stick your heater bits on top of that.
Oh yeah, and mount that on a plate of steel with standard motherboard mounts.
However, you could say this benefits air cooling over water, but then so does a motherboard.
Results should state air pressure and humidity as well as room temperature to be thorough. If you really want to go to town, the local value of g too but I don't really see that mking too much difference apart from for purely convective cooling.
Um, you might want to specify something about light sourcing, since if you had a 100W desk lamp only a foot or so away (maybe to take piccies while it's running) that might well screw up the results, as would doing it on a bench under a window in sunlight.
By the way, anyone suggest you just use a standard 100W (or 150W) filament lamp bulb in a standard heinz baked bean can, bolted to a wood block ANYONE can get hold of those parts, that would be a real basic and accessible test. Cost less than $5. Temps to be taken from a standard "sandwich" self tapped to the top of the can with 4 screws, which could be a specified thermistor from a major manufacturer and taken and quoted as the resistance of this with a standard multimeter. This should be made from aluminum, easier to get hold of, be the dimensions specified, for top plate bottom plate and U shape middle plate cutout for thermistor, which should be glooped in with fresh standard zinc oxide thermal compound. That seems to me a real simple and accessible solution for everyone here to use, not just those that want to spend $$$ and have the skills and want to review professionally.
Road Warrior
Oh yeah, and mount that on a plate of steel with standard motherboard mounts.
However, you could say this benefits air cooling over water, but then so does a motherboard.
Results should state air pressure and humidity as well as room temperature to be thorough. If you really want to go to town, the local value of g too but I don't really see that mking too much difference apart from for purely convective cooling.
Um, you might want to specify something about light sourcing, since if you had a 100W desk lamp only a foot or so away (maybe to take piccies while it's running) that might well screw up the results, as would doing it on a bench under a window in sunlight.
By the way, anyone suggest you just use a standard 100W (or 150W) filament lamp bulb in a standard heinz baked bean can, bolted to a wood block ANYONE can get hold of those parts, that would be a real basic and accessible test. Cost less than $5. Temps to be taken from a standard "sandwich" self tapped to the top of the can with 4 screws, which could be a specified thermistor from a major manufacturer and taken and quoted as the resistance of this with a standard multimeter. This should be made from aluminum, easier to get hold of, be the dimensions specified, for top plate bottom plate and U shape middle plate cutout for thermistor, which should be glooped in with fresh standard zinc oxide thermal compound. That seems to me a real simple and accessible solution for everyone here to use, not just those that want to spend $$$ and have the skills and want to review professionally.
Road Warrior
- Joined
- Nov 25, 2001
- Location
- Niagara Falls, Ontario, Canada
What I meant really in my last paragraph was that it's all very nice figuring out a $200+ 50 manhour test setup, but it would be also be real good just to figure out a $10 5 manhour standard test that ordinary stiffs could use and compare with.
- Joined
- Jan 29, 2002
- Location
- Calgary, Alberta
ROTFLMAO, Now that I just fell of the bar stool
and picked my **s up. Just wanted to say, Road Warrior that you have successfully brought this space shuttle mission down to earth. I was waiting for a comment like this. I've being trying to keep up with the BillA and Aesik threads ever since Hoots round up came out. These guys are killing me and I am a graduate engineer(structural).Thanks for the laugh,
BTW keep these threads up, I am really learning alot. I've even dug up the old Thermo and Fluids books out of the milk crate.
