Couple questions: 1) How often do these watercooled systems leak or get air into them (sign of a possible leak, definately a breach of some sort at the least) 2) what's the deal with Koolance's copper/plastic blocks? Wont the plastic warp over time, and then leak on your stuff? 3) I'm a newbie at this watercooling and plan on cooling my next system this way, but it's going to be kinda crazy.... A 64 3800+, 2 6800 gt.. etc etc. do you think i can oc everything in it 20-30% given the hardware can oc that high? should I get 2 watercooling setups, 1 for the cpu/chipset. the other for the video cards? 4) If i could cool the water better, say by chilling it somehow, will I get better oc results?
I wouldn't recommend a watercooler with a 3800+... if you want to shell out $200 on watercooling, then you want to get your money's worth on the OC, so I'd say a 3500+ or slower, maybe one of those .09 micron 939 chips like the 3000+ or 3200+. The problem is that max OC isn't *just* a factor of heat, and the 3800+ is already too close to the limit. If you got a 3000+ let's say you'd hit 2.55 GHz, and with a 3800+ you'd hit 2.6 GHz. You would have spent a lot more money to get only a tiny bit better performance. I'm not saying it would be exactly these figures, but this is the general idea. You can watercool 6800's, but the waterblock's are big and cost over $100. Just keep that in mind... Not sure about the plastic. I don't think its unsafe, but if it bothers you then just get another block. The koolance blocks aren't that great anyway.... Chilled would be better, but chilling isn't easy or cheap. One way to get a better OC is get a peltier (TEC) waterblock. But with any sub-ambient cooling, you run the risk of condensation forming around the socket. You can insulate the socket with foam, but its just one more thing to worry about. Leaking? In general I'd say not much chance. If it doesn't leak when you first turn it on, it probably won't leak on its own. Clamps hold down everything pretty tight, and nothing should really change once you have it all clamped down. I had a leak problem once when a nozzle snapped clean off the pump. But that was mostly my fault, the way I cut the tubing ended up forcing the pump to a slight angle. The pump was achored with suction cups to the bottom of the case, which put stress on the nozzle. I was fiddling around with it and it snapped off. The good news is the pump was off and everything was fine....
Right around $200 to $250 will get you a decent pump, waterblock, dual 120 mm radiator, tubing and fans.
how should i get it so i can cool the GPU and CPU with 1 system? just have the tubes off the cpu block lead right into the GPU block or stick a radiator in between? or do they make Y-tubes to come right off the pump? it wouldnt make sense not to put a radiator in between, because i have a cooling problem with my air, cuz my fan blows hot air from my CPU right over my GPU, which doesnt help it out at all. it idles at 58C and loads at 73-78C, and its only a 6600GT!
I'd assume just put them in a single loop, say chain the GPU to CPU to radiator. This will mean your CPU will be a bit hotter, but I believe perhaps on a degree or two. Going pump->GPU->Radiator->CPU->Radiator might alleviate it, but would be more expensive and require a stronger pump. I've seem pictures where they used a "Y" connector to connect the second waterblock, but seems a little iffy to me... don't have any solid proof of that, though. 6800 waterblocks aren't cheap, unfortunately, I've only seen two and they're both over $100.
For anyone thinking of making the leap, I'd just keep in mind that watercooled systems are still ultimately air-cooled. It isn't the water that cools the CPU but the air that blows over the radiator, the water just moves the heat from the heat source to the radiator. If nothing was cooling the water it would eventually simply heat up to the same temperature as the CPU. The water presumbably helps to distribute the heat more evenly throughout the radiator then in a conventional heatsink, but the big advantage of watercooling is that you can have a much larger radiator then you could a heatsink, since you can mount the radiator anywhere inside the case or even outside the case, whereas a heatsink must be physically attached to the CPU on the motherboard and cannot be too large for that reason. A larger cooling surfact will result in better and/or quieter cooling. For this reason you really don't want to bother with a system that doesn't have at least a 120 mm radiator, because otherwise you may end up with a watercooling system that is not much better (or even slightly worse) then a high-end heatsink that will likely still be significantly cheaper.
Just an adjustment to the origional post. QUOTE: Component Notes: Despite popular myth, the order of your blocks will make little to no difference on your temperatures, so it is best to just put them in the order that is most convenient for you. The best system order is as follows: Pump-->Blocks-->Radiator-->Resevior-->Pump. Having the radiator or res right after the pump, will kill your flow rate and pressure, so always put them at the end. This is incorrect,The max flow is a closed loop is determined by the most restrictive item in the loop.It does not matter where this item is located because the loop is closed. Additionally people often think running in Series ( Block to block) Makes the water warmer in the second block compaired to the first.Not so because the water is in contionos motion the temp from CPU block to GPU block if measured would be almost identicle. Flow is the single biggest factor in removing heat from the heat sources,As is Air flow across the core,It is in these 2 areas that the biggest resulat are achieved,Usually the easiest way to get lower temps is with a larger rad and or more air flow across the rad..And typically the best blocks offer a better flow rate compaired to a cheaper block...
I'll agree on point one, but not so sure about point two--it seems like the water would have to be a bit warmer if the first waterblock is removing heat effectively. I do know that adding a GPU block seems to raise CPU temperatures by only a degree or two, though, so the difference is certainly minor, if that's all you mean. I'm actually curious if there is an advantage to using two T-fitting to connect two waterblocks instead of a large, single loop. Then water would flow from the pump, then branch out to the two waterblocks, then join back together and hit the radiator.
Just noticed this review of a swiftech watercooling kit with 120 mm radiator. This is a good example of a basic mid-range kit for enthusiasts who want to overclock. It also illustrates all the components of a watercooling system for anyone still unclear on the concept. http://www.xbitlabs.com/articles/other/display/swiftech-cooling.html You can contrast it with the other two kits here, which are still quite decent but I'd venture closer to entry-level kits. http://www.xbitlabs.com/articles/other/display/swiftech-cooling_8.html Edit: Looks like Corsair is now reselling this same kit, looks to be a bit cheaper even. http://www.corsairmicro.com/corsair/COOL_water.html
You are asking good questions and I dont want to post a scientific reply but I all most have to so before i do lets try to explain.. First the water is in Motion constantly and going from cpu to gpu if you had a temp moniter between these blocks the water temp difference would amount to approx .3c a very very small temp difference ,in fact so small it is not even worth worrying about. You re asking me if setting a syaten up in a parallel circut will result in lower temps. And the answer is 99% of the time no UNLESS you specifically get products engenered to proivde better cooling in that type of set up. The most important factor in a closed loop is maintaing a High flow rate and when you introduce Y connections T connections or many loops you are creating drag and a reduction of flow. This is the easiest way for me to put this,But the more water you push through the blocks and rad the cooler the temps of the procesor and Gpu.
I'm sorry, but the part about having two water blocks not raising temperatures much just seems illogical. If the GPU is running at the exact same temperature as the CPU, then twice the amount of heat would be absorbed by the water, thus doubling the temperature change. If someone can prove me wrong (scientifically) then I would gladly like to see it.
Well I don't have a review handy, but I've seen some where they added a second block for the GPU and the CPU only seemed to get a couple degrees warmer, though this wasn't necessarily with some monster like the 6800 Ultra. I'll keep any eye out for something more compelling to link to, though... ----- Oh well, here's something: http://www.moddershq.net/reviews.asp?reviewid=54&pagenumber=1 "I would like to think I have a decent cooling setup(dual 6 x 12 heater cores), however, my CPU temperature rose 10C during my benching session with the NV68 installed compared to when I only had the CPU block in the system. I have yet to fully determine if this affects my maximum CPU overclock, however, I would assume a 10C difference in core temperature is enough to warrant a warning: make sure you have adequate means to remove the heat this block takes off the 6800." This is the description of his watercooler: "Cooling wise, I have the JPI v5(external box to house 2 heatercores), the DangerDen D4 pump, DangerDen TDX water block, and stock cooling for my graphics card." Which sounds pretty decent. So in his case, an overclocked 6800 Ultra heated up his CPU overclock by 10 degrees. Though an overclocked 6800 Ultra is surely a worst-case scenario.... I think its nearly like adding a second CPU.
Though just a thought... this does not necessarily mean the order of the blocks is important. It probably is a slow buildup to +10 degrees and reflects how much heat is being put into the loop and how much heat your radiator is capable of getting rid of, not how much heat the first block picks up on each pass to the second block. Presumbably the GPU block would drop 10 degrees if the CPU waterblock was disconnected as well. For instance, scenario A would be to have the pump, the GPU block, then the CPU block, then a dual-120 mm radiator, then back to the pump. Scenario B would it be have a "T"-connector that split the tube from the pump to seperate tubes that go to each waterblock, then to each their own single 120 mm radiator, then join back together with another "T"-connector and then to the pump. The question is, would "B" result in lower CPU temperature then "A"... I suspect it wouldn't, or at least not to any significant degree, presuming the cooling capacity of the two individual radiators would not be better then the single dual radiator.
Assuming that the two radiators removed as much heat as the dual-120mm radiator, then I suspect that B might perform slightly better, due to the fact that water loses heat not equally at different temperatures, but it is in fact a slight curve... If you know what I mean.
I'm going to expand on a few things. I have had thermodynamics at Georgia Tech. I'm an Emory University student but i took that as a summer class for my personal knowledge. Some things to remember about liquid cooling. First. most important think other than the actually cooling unit ie CPU cooler etc. is the Radiator most specifically that the radiator is cooled. This is virtually the ONLY way to increase deltaQ aka heat transfer. Some noobs often talk about getting bigger tubing or faster pumps. DO NOT BE FOOLED this is a mistake. bigger tubing or faster pumps ONLY increases efficiency when the deltaQ of the radiotor is not being pushed to the limits. You may ask why this is. I mean at first glance you woud think i'm moving more water faster of the cpu shouldn't this decrease the temperature of the CPU? The answer again is only if the radiator is not being pushed to its deltaQ limit or heat limit. Let me further explain. Imagion that you are on the daytona 500 and lets just say its 1 miles to complete a lap. now lets say on a specific .3 miles your car runs cooler. now stay with me now... If you travel 100mph on the track for one hour you will spend the same ammount of time on that particular .3miles of the track as you would if you went 50mph for 1 hour. Therefore if the heat displacement of the CPU is higher than your radiator deltaQ the only way to increase performance is to upgrade your radiotor. now you can do this 2 ways. you can either get more airflow on the radiator to increase its deltaQ or you can get a bigger radiator with the same airflow as before only you will have to make sure the airflow is sufficiently spread over the radiator. Liquid cooling is a great tool if carefully monitored and in the hands of a knowledgable user. Anyway happy cooling If anyone wants more information on this or needs better explaination just post and I will further explain myself. For those of you who are familiar with DeltaQ DeltaQknot and Q* and would like for me to really get more into the thermodynamics again just post and I will be more than happy to do so. ~ Sonny
Well, that's all fine and well, but its tough to say exactly when your pressure is already sufficient. There are also waterblocks that are meant to do better with higher pressure rather then lower (like DD RBX/TDX). Performance improvements with stronger pumps seems sufficiently proved, at least in some cases: http://www.gruntville.com/reviews/wc/Swiftech_MCP_350_vs_PolarFLO_TT/page4.php This is from a 3-block system, and the somewhat dinky Swiftech MCP350 doesn't hack it as well as the other two. Now if I was going to say that any part was the most important to cooling performance then I'd definately say the radiator, but I do not think you can neglect the others either.
Basically if you increase pressure / change pump speed / change hose size and you see an improvement that only means that the heat being generated from your components (DeltaQ) is less than the DeltaQKnot of the radiator. Basically the radiator is effectively transfering enough head to maintain near ambient temperature. Problem with changing the pump speed / pressure is that if your radiator isn't alowing enough heat dissipation to reach ambient temperature than all your doing is heating up the watter then cooling it slightly in the radiator then moving the water faster back to the cpu and continuely heating the water beyond ambient temperature. For this reason I would hesitate to increase pump speed because if you are not allowing enought time for the water to cool to ambient temperature in the radiator then your are only doing harm rather than good. This is more complicated then what I have just stated in short. when increasing speed you are also alowing for less time in the CPU block. this is where it gets sticky and the equations get rather messy. you have to deal with deltaH vaporization ( i know your water isn't reaching 100c but there is still an equilibrium that must be accounted for) oh i'm just going to stop here because I really don't want to complicate it any further... Most of the basic water cooling systems supply a radiator far superior to that of the other components therefore you could change some of the other things mentioned and increase DeltaQknot.
Well, you said yourself that in an hours time the water will spend the same time in the radiator no matter how much faster the pump goes, so I would not agree with this. You may be right about strong pumps not always or even usually being needed, but the problem once again determining when that is the case. Unfortunately, all the "science" in watercooling stays in the abstract, and there isn't some handy metric printed on all the components that tells you how strong a pump you'll need to not have low flow rate be an impediment. Its tough to draw strong conclusions even when there is a review, because a water-cooling system is made of up so many different components, you don't know how your mix along with the particular component being reviewed would change the results. I think as long as the ideal flow rate is an unknown people might as well just spring for 1/2" ID tubes and $50-$75 pumps, since as watercooling is not a cheap business to begin with its probably better not to try to save $10 or $20 bucks when you'll shell out at least $200 anyway. Though I would caution people who never expect to do more then a single block not to be too concerned with 3/8" ID tubing or a cheaper pump, especially if they have blocks known to do fine with low pressure (like Swiftech's new line). This kit's pump is fairly small yet the kit handles 250 watts fine in this review: http://www.xbitlabs.com/articles/other/display/swiftech-cooling.html
