L33T
06-14-2007, 05:27 PM
I wrote this because I wanted to explain just how difficult, not easy, it is to build a DIY TEC system, to serve not as a deterrent so much as a admonition to all who would follow my path. I delineate the past few months to those who were unaware, and for those who were with me, as a humorous account, but a caution nonetheless.
Despite the censure of some people -- particularly those not-so-nice people from XFN and other lesser known forums such as [H]|F, SharkyForums, AnandTech, &c (I'm so getting the ban hammer if they read this; but I present only the facts: XS is definitively better than them all combined), that TEC'g is the path of lackluster success at best, I took that path nonetheless, being humorously abetted by several of my online buddies here -- LittleOwl, NoL, SerialKiller, &c...
TEC seemed liked a good plan: it appeared to be quiet; it seemed neat, and it seemed compact and cheap. It also seemed easy -- at least to the extent that it was quite doable within a reasonable amount of time, whereas the phase change solution required that first I pass several exams. Of course, it isn't very easy to gauge these metrics anymore, since TEC is a dying art: it is truly life -- the slightest bit -- on life support, ready to fade into oblivion in a day or a minute or a second; blink, and you might not notice as it disappears completely.
Zenjirou was the half the inspiration with his race car radiator [1]; my ex-roommate Rick "Slick Rick" was the other. Zenjirou simply provided empirical evidence that a silent system was achievable. Unfortunately, his system involved slightly older hardware; mine involved not only newer hardware, but also TEC units that would be be outputting an enormous amount of energy every second, which would quickly heat up such an excellent KOYO radiator. Using a room fan on the lowest setting, while helping a significantly, increased the noise level of the room. Simultaneously, the extra power supplies -- particularly the 24V ASTEC power supply was extremely loud with its industrial fans. Truly, using this computer was most enjoyable when wearing "ears" (ear protection for shooting).
You won't find too many examples of TEC systems even in XS. Building one, I lived on an island, in an anachronism, shared by but a few people. NOL is quite obsessive about TEC'g, as is LO. Even so, why couldn't I be simply a microcosm of this? Alas, things phase out for a reason [2]. Phase change truly is the way to go. Even NOL has stopped TEC except for the NB and possibly GFX cards. On the other hand, the 8800GTX isn't quite TEC-friendly. I run Vista with standard drivers, but even without a digital temperature reading, I can tell what the temperature is simply by fingering the back of the GFX card. The temperature starts at ambient temperature and then starts to decline; at this point it feels rather cool to the touch. However, it irrevocably begins its ascent some time thereafter.
A radiator can be small enough, but a small radiator requires high-powered fans to keep the temperature cool; on the other hand, a big radiator simply is big, by definition of the word big (http://dictionary.reference.com/browse/big) -- and a KOYO radiator is not big; it is huge. To its credit, it does fit neatly behind objects, such as desks and what not, and can be hidden. On the other hand, its performance suffers when hidden, since airflow is diminished typically.
Browse around online and you'll find the highest performing TEC, the venerable 437W monster being sold on FrozenCPU.com for just $49.99 plus shipping. The pumps you already have. The water blocks you sort of already have, but you'll find out that those don't really work as TEC water blocks. Copper plates seem cheap enough at various sources. Alas, TEC is not as cheap as one would suppose due to its various hidden costs, like the muggers hidden around every dark corner within the dark alley. In my situation, I used an MCW5002 [3] -- it was what I already had, so my plan was to modify the base so as to convert it into a TEC water block (i.e., the MCW5002T). This involved milling the bottom of the base flat, which my friend was more than happy to do for me. To me, this was a cost, since it did cost my friend time and energy; that he was doing me a favor. Directly to me, it cost shipping to and back. However, that was not the end, there were also the cold plates and hot plates, and various types of sandpaper. There were so many surfaces to lap that in the end I simply did not both to sandpaper some faces at all.
Yet, of these past few days I thought I finished the TEC project when in fact I had barely begun. Quite clearly, my temperatures were anything but good. To paraphrase SerialKiller, my temperatures were tantamount to a water-cooled system [4]. My consumption of electricity however was on par with that of an auto cascade system. The former proved quite true simply by searching the various forums. The latter I knew in part already: I ran two TEC units: a 437W TEC for the CPU and a 320W TEC for the GFX. Both I ran at full power; in all, I drew quite close to a kilowatt of power -- an obscene amount, if I may say so, and I can.
The TEC system also seemed easy to build. I realized that I had a good knowledge of electronics. While I am a mechE, I am not an HVAC technician, and concordantly, I realized that phase change would require a high "leather shoe cost", as my brother would say [5]. It follows that it would be irrational to purchase torches and what not to do something that I have little relative experience in doing. It would be better for me to work harder to pay someone to do the job on my behalf. However, I reasoned that since I could build TEC PSU's easily enough, I would be able to build a TEC system relatively easily. I did pursue the subject of phase change "on the side", since I was after all, very interested in the topic.
Of course, the TEC system was not easy to build. It required much machining and much designing of my 24V power supply, which took an incredible amount of time to build. The cost was in any case, only slightly less than that of a retail power supply manufactured by Meanwell, for example.
I decided a few days ago that there was really no need for me to build my own phase change or chilling unit. I could simply commission one of the many trusted members of XS to do that task. To that extent, NOL has offered to build me a chilling unit. His pricing cost just under $500 shipped, which considering the cost of a water cooled TEC system, seems quite acceptable.
Looking back at the past few months I realized that I didn't make any mistake per-se. The power supply was worth building for the sake of learning, as electrical engineering is closely tied to mechanical engineering in most concentrations of the latter. Building the TEC system was frustrating to say the least, and yielded quite a bit of frustration at the end, since the performance was sub-par. Why was it sub-par, when other review sites would laud the 437W TEC? That was a mistake: I simply forgot an item on the to-do list -- a detail in the scope of the project, and quickly lost amongst a myriad of objectives. I forgot to machine the base of the cold plate. As you may know, a TEC block contains at least the water block, the TEC unit, and the cold plate; sometimes, it may contain an extra hot plate as well. Now, imagine an LGA motherboard, and think about the surrounding area of the CPU. There are quite a few capacitors around the CPU. In fact, those capacitors are taller than the height of the mounted CPU. Therefore, it is common that cold plates have extrusions in the general shape of a single partition of a Mayan pyramid. In my situation, I simply forgot about this task. I initially planned to machine the base, as I may prove by referring to the various posts I made regarding that topic, but when the time came months later, I simply did not remember. Thus, when I mounted the TEC water block, it failed to provide any but the least bit of contact along the edge of the processor. Hence, the temperature was so high. That would at least partially explain the enormous discrepancy between the high and low temperatures.
Even so, the temperature of the 8800GTX was still not that great: when I fingered the back of the gfx card, it was mildly warm; it could only follow that the 320W TEC was insufficient to cool the 8800GTX to temperatures that I would perceive as cool (i.e., temperatures that are below the dermal temperature, or roughly a few degrees below the average body temperature; below, because my body temperature is lower; I tend to live in igloos that have windows).
Concordantly, looking back, had I foreseen this outcome, I would certainly have pursued the path of water chilling or phase change instead of that of TEC. I would have simply used the money to purchase a system that would be immediately usable, instead of waiting months for a system that at this point, still technically does not work. Yes, it did work -- but it is terribly inefficient, and unbearably loud. It was also quite a mess, even in my basement den. Reviewing the various metrics, I would conclude that the majority are in favor of a pre-built phase change system (whether direct die or chilled). In terms of noise, the TEC system had the combined noise output of the stand-up fan used to cool the radiator as well as the industrial fan insdie the TEC power supply. In terms of cost, the purchase costs are roughly the same, but the hidden cost is that of electricity (which while I am at school is free, so that should not be considered; however, the consumption per-se should still be considered, since rooms typically have a certain allotment, and exceeding that threshold will cause the circuit breaker to trip). In terms of aesthetics, a TEC system allows the liquids in the system to be visibly UV-reactive, which I do enjoy. Externally however, I have two power supplies next to my system; I have additionally, a giant race car radiator leaning against the wall, and a stand up fan that is lying down and pointing at the radiator. In that context, a simple box for a chilling system is exponentially more appealing than the system I have right now.
___________________
1. http://www.xtremesystems.org/forums/showthread.php?t=82851
2. Pardon the pun please.
3. The MCW5002 and MCW5002T are not considered efficient enough to cool a Xeon 3060 with a 437W TEC unit well.
4. While idling, the temperature would drop to 4-5 degrees C; however, the temperature would quickly rise to roughly 40 degrees at stock voltages while running WCG at 100% load; at higher voltages, the temperature would increase sharply. However, there is an explanation for this which will be described later.
5. My brother is in economics. His quote metaphorically explains the reasoning behind purchasing pre-made goods instead of DIY in cases where the person has imperfect ability to produce the good.
Despite the censure of some people -- particularly those not-so-nice people from XFN and other lesser known forums such as [H]|F, SharkyForums, AnandTech, &c (I'm so getting the ban hammer if they read this; but I present only the facts: XS is definitively better than them all combined), that TEC'g is the path of lackluster success at best, I took that path nonetheless, being humorously abetted by several of my online buddies here -- LittleOwl, NoL, SerialKiller, &c...
TEC seemed liked a good plan: it appeared to be quiet; it seemed neat, and it seemed compact and cheap. It also seemed easy -- at least to the extent that it was quite doable within a reasonable amount of time, whereas the phase change solution required that first I pass several exams. Of course, it isn't very easy to gauge these metrics anymore, since TEC is a dying art: it is truly life -- the slightest bit -- on life support, ready to fade into oblivion in a day or a minute or a second; blink, and you might not notice as it disappears completely.
Zenjirou was the half the inspiration with his race car radiator [1]; my ex-roommate Rick "Slick Rick" was the other. Zenjirou simply provided empirical evidence that a silent system was achievable. Unfortunately, his system involved slightly older hardware; mine involved not only newer hardware, but also TEC units that would be be outputting an enormous amount of energy every second, which would quickly heat up such an excellent KOYO radiator. Using a room fan on the lowest setting, while helping a significantly, increased the noise level of the room. Simultaneously, the extra power supplies -- particularly the 24V ASTEC power supply was extremely loud with its industrial fans. Truly, using this computer was most enjoyable when wearing "ears" (ear protection for shooting).
You won't find too many examples of TEC systems even in XS. Building one, I lived on an island, in an anachronism, shared by but a few people. NOL is quite obsessive about TEC'g, as is LO. Even so, why couldn't I be simply a microcosm of this? Alas, things phase out for a reason [2]. Phase change truly is the way to go. Even NOL has stopped TEC except for the NB and possibly GFX cards. On the other hand, the 8800GTX isn't quite TEC-friendly. I run Vista with standard drivers, but even without a digital temperature reading, I can tell what the temperature is simply by fingering the back of the GFX card. The temperature starts at ambient temperature and then starts to decline; at this point it feels rather cool to the touch. However, it irrevocably begins its ascent some time thereafter.
A radiator can be small enough, but a small radiator requires high-powered fans to keep the temperature cool; on the other hand, a big radiator simply is big, by definition of the word big (http://dictionary.reference.com/browse/big) -- and a KOYO radiator is not big; it is huge. To its credit, it does fit neatly behind objects, such as desks and what not, and can be hidden. On the other hand, its performance suffers when hidden, since airflow is diminished typically.
Browse around online and you'll find the highest performing TEC, the venerable 437W monster being sold on FrozenCPU.com for just $49.99 plus shipping. The pumps you already have. The water blocks you sort of already have, but you'll find out that those don't really work as TEC water blocks. Copper plates seem cheap enough at various sources. Alas, TEC is not as cheap as one would suppose due to its various hidden costs, like the muggers hidden around every dark corner within the dark alley. In my situation, I used an MCW5002 [3] -- it was what I already had, so my plan was to modify the base so as to convert it into a TEC water block (i.e., the MCW5002T). This involved milling the bottom of the base flat, which my friend was more than happy to do for me. To me, this was a cost, since it did cost my friend time and energy; that he was doing me a favor. Directly to me, it cost shipping to and back. However, that was not the end, there were also the cold plates and hot plates, and various types of sandpaper. There were so many surfaces to lap that in the end I simply did not both to sandpaper some faces at all.
Yet, of these past few days I thought I finished the TEC project when in fact I had barely begun. Quite clearly, my temperatures were anything but good. To paraphrase SerialKiller, my temperatures were tantamount to a water-cooled system [4]. My consumption of electricity however was on par with that of an auto cascade system. The former proved quite true simply by searching the various forums. The latter I knew in part already: I ran two TEC units: a 437W TEC for the CPU and a 320W TEC for the GFX. Both I ran at full power; in all, I drew quite close to a kilowatt of power -- an obscene amount, if I may say so, and I can.
The TEC system also seemed easy to build. I realized that I had a good knowledge of electronics. While I am a mechE, I am not an HVAC technician, and concordantly, I realized that phase change would require a high "leather shoe cost", as my brother would say [5]. It follows that it would be irrational to purchase torches and what not to do something that I have little relative experience in doing. It would be better for me to work harder to pay someone to do the job on my behalf. However, I reasoned that since I could build TEC PSU's easily enough, I would be able to build a TEC system relatively easily. I did pursue the subject of phase change "on the side", since I was after all, very interested in the topic.
Of course, the TEC system was not easy to build. It required much machining and much designing of my 24V power supply, which took an incredible amount of time to build. The cost was in any case, only slightly less than that of a retail power supply manufactured by Meanwell, for example.
I decided a few days ago that there was really no need for me to build my own phase change or chilling unit. I could simply commission one of the many trusted members of XS to do that task. To that extent, NOL has offered to build me a chilling unit. His pricing cost just under $500 shipped, which considering the cost of a water cooled TEC system, seems quite acceptable.
Looking back at the past few months I realized that I didn't make any mistake per-se. The power supply was worth building for the sake of learning, as electrical engineering is closely tied to mechanical engineering in most concentrations of the latter. Building the TEC system was frustrating to say the least, and yielded quite a bit of frustration at the end, since the performance was sub-par. Why was it sub-par, when other review sites would laud the 437W TEC? That was a mistake: I simply forgot an item on the to-do list -- a detail in the scope of the project, and quickly lost amongst a myriad of objectives. I forgot to machine the base of the cold plate. As you may know, a TEC block contains at least the water block, the TEC unit, and the cold plate; sometimes, it may contain an extra hot plate as well. Now, imagine an LGA motherboard, and think about the surrounding area of the CPU. There are quite a few capacitors around the CPU. In fact, those capacitors are taller than the height of the mounted CPU. Therefore, it is common that cold plates have extrusions in the general shape of a single partition of a Mayan pyramid. In my situation, I simply forgot about this task. I initially planned to machine the base, as I may prove by referring to the various posts I made regarding that topic, but when the time came months later, I simply did not remember. Thus, when I mounted the TEC water block, it failed to provide any but the least bit of contact along the edge of the processor. Hence, the temperature was so high. That would at least partially explain the enormous discrepancy between the high and low temperatures.
Even so, the temperature of the 8800GTX was still not that great: when I fingered the back of the gfx card, it was mildly warm; it could only follow that the 320W TEC was insufficient to cool the 8800GTX to temperatures that I would perceive as cool (i.e., temperatures that are below the dermal temperature, or roughly a few degrees below the average body temperature; below, because my body temperature is lower; I tend to live in igloos that have windows).
Concordantly, looking back, had I foreseen this outcome, I would certainly have pursued the path of water chilling or phase change instead of that of TEC. I would have simply used the money to purchase a system that would be immediately usable, instead of waiting months for a system that at this point, still technically does not work. Yes, it did work -- but it is terribly inefficient, and unbearably loud. It was also quite a mess, even in my basement den. Reviewing the various metrics, I would conclude that the majority are in favor of a pre-built phase change system (whether direct die or chilled). In terms of noise, the TEC system had the combined noise output of the stand-up fan used to cool the radiator as well as the industrial fan insdie the TEC power supply. In terms of cost, the purchase costs are roughly the same, but the hidden cost is that of electricity (which while I am at school is free, so that should not be considered; however, the consumption per-se should still be considered, since rooms typically have a certain allotment, and exceeding that threshold will cause the circuit breaker to trip). In terms of aesthetics, a TEC system allows the liquids in the system to be visibly UV-reactive, which I do enjoy. Externally however, I have two power supplies next to my system; I have additionally, a giant race car radiator leaning against the wall, and a stand up fan that is lying down and pointing at the radiator. In that context, a simple box for a chilling system is exponentially more appealing than the system I have right now.
___________________
1. http://www.xtremesystems.org/forums/showthread.php?t=82851
2. Pardon the pun please.
3. The MCW5002 and MCW5002T are not considered efficient enough to cool a Xeon 3060 with a 437W TEC unit well.
4. While idling, the temperature would drop to 4-5 degrees C; however, the temperature would quickly rise to roughly 40 degrees at stock voltages while running WCG at 100% load; at higher voltages, the temperature would increase sharply. However, there is an explanation for this which will be described later.
5. My brother is in economics. His quote metaphorically explains the reasoning behind purchasing pre-made goods instead of DIY in cases where the person has imperfect ability to produce the good.