whats your favorite WC

[Cathar]

Well if you guys with P4's want to try to see just how far out your P4 diode is, then try this little trick.

If you have a heatsink, turn off the fan. If you have water-cooling, turn off the pump.

P4's have a user-hidden internal diode that is calibrated to shut the CPU down when it hits 135C. You can't actually see the reading of this diode, but that's what its job is. When the CPU hits 135C, the CPU trips a signal that automatically powers off the system.

Run the CPU at stock speeds. Turn off any thermal throttling. Turn off your pump or your heatsink fan, and run Sandra Burn-in while monitoring your CPU temperature with Motherboard Monitor, or whatever.

If your P4 is like mine, the CPU will trigger a critical thermal shutdown when the user visible diode is reading around the 70C mark. Another P4 CPU of mine will do it when reading 55C. A friends will do it when reading 90C.

In all cases, the P4 diode is actually at 135C by the factory calibrated internal thermal critical shut-down diode. This diode is calibrated by Intel properly because it prevents CPU death from overheating, so they have to get it set right. For the user-visible diode, well that's not really calibrated in any fashion, and it is open to being tampered with by the BIOS on the motherboard.

It's just a little experiment that you guys might like to run to open the eyes on just how accurate you think your CPU's diode reading really is.

Once you try it, it'll really change your opinion on arguing anything at all about whether or not reported temperatures are "high", "low" or "correct". They're all a total joke.

 

[kronchev]

my fav is i'm-not-too-lazy-to-pick-out-parts-and-cut-a-hole, by DIY

 

[kronchev]

Well if you guys with P4's want to try to see just how far out your P4 diode is, then try this little trick.

If you have a heatsink, turn off the fan. If you have water-cooling, turn off the pump.

P4's have a user-hidden internal diode that is calibrated to shut the CPU down when it hits 135C. You can't actually see the reading of this diode, but that's what its job is. When the CPU hits 135C, the CPU trips a signal that automatically powers off the system.

Run the CPU at stock speeds. Turn off any thermal throttling. Turn off your pump or your heatsink fan, and run Sandra Burn-in while monitoring your CPU temperature with Motherboard Monitor, or whatever.

If your P4 is like mine, the CPU will trigger a critical thermal shutdown when the user visible diode is reading around the 70C mark. Another P4 CPU of mine will do it when reading 55C. A friends will do it when reading 90C.

In all cases, the P4 diode is actually at 135C by the factory calibrated internal thermal critical shut-down diode. This diode is calibrated by Intel properly because it prevents CPU death from overheating, so they have to get it set right. For the user-visible diode, well that's not really calibrated in any fashion, and it is open to being tampered with by the BIOS on the motherboard.

It's just a little experiment that you guys might like to run to open the eyes on just how accurate you think your CPU's diode reading really is.

Once you try it, it'll really change your opinion on arguing anything at all about whether or not reported temperatures are "high", "low" or "correct". They're all a total joke.

um no. theres no way that itll shut down at 135C because by then youd already have disaster on your hands. 100C, maybe. but 135C is waaay too high. also "Turn off any thermal throttling" is the worst advice you can ever give someone. plus turning off a WC system and then letting the temps get about boiling is probabily the dumbest piece of advice ive ever heard.

/edit: internet says its 95-101 depending on your proc. however it does have a 135C absolute failure trigger, but you wont hit it.

 

[[RIP]Zeus]

my fav is i'm-not-too-lazy-to-pick-out-parts-and-cut-a-hole, by DIY

AGREED!!!!!
if your to lazy to do water cooling right don't do it at all...KITS SUCK

 

[Cathar]

um no. theres no way that itll shut down at 135C because by then youd already have disaster on your hands. 100C, maybe. but

135C is waaay too high. also "Turn off any thermal throttling" is the worst advice you can ever give someone. plus turning off a WC system

and then letting the temps get about boiling is probabily the dumbest piece of advice ive ever heard.

/edit: internet says its 95-101 depending on your proc. however it does have a 135C absolute failure trigger, but you wont hit it.

Let me re-explain. There are two thermal diodes on the P4.

One, which the user cannot see, is calibrated correctly and shuts the CPU down when the CPU hits 135C.

The other, which is the one the user can see, is not calibrated, and it is the one that users use to report their CPU temperatures.

So when people are saying on the Internet "The CPU gets to 95-101 before it shuts down", what's really going on is that the user visible diode is reading 95-101, while the calibrated internal diode which the user cannot read the value of, has hit 135C, and shut the CPU down.

i.e. there is a ~40C discrepency in this case between the user visible diode, and the actual temperature of the CPU die according to the calibrated critical thermal shutdown diode.

The point being that people arguing whether some board is too low or not is a bit silly - every board grossly understated the actual CPU die temperature - this is what has lead to the perception held by many that Intel CPU's "run cool".

 

[ErrOnReq]

my fav is i'm-not-too-lazy-to-pick-out-parts-and-cut-a-hole, by DIY

AGREED!!!!!
if your to lazy to do water cooling right don't do it at all...KITS SUCK

What does a kit have to do with cutting a hole? You can get kits with huge radiators that require that a hole be cut for proper mounting...

You are so adamant that DIY is the way to go and anything slightly resembling a kit sucks because you are not "doing it yourself." What about the blocks themselves? You are buying a premade block, you didn't make the block yourself. Radiators too... make your own! PREMADE SUCKS!!!

You know what sucks? Your logic... which sucks just as much as your useless comments in this thread.

 

[ProOC]

I dont really see anything wrong with a kit. I have a swiftech kit and it owns air. I was able to drop my temps in half and my setup looks great. I am glad to see some people mentioning swiftech because I don't see it much around here with all the Danger den buzz (I must say that does have me wanting to go custom and get a dd block soon though) only because my server needs some water and this swiftech kit is cool. Any water will be fun. Just have fun with it whatever you do.

 

[MrHappyGoLucky]

The point being that people arguing whether some board is too low or not is a bit silly - every board grossly understated the actual CPU die temperature - this is what has lead to the perception held by many that Intel CPU's "run cool".

And hense the reason we should not look at the temp from MBM as how good a cooling setup performs.

 

[Hadez026]

just a guess here, but i think by a kit kronchev and [RIP]Zeus were refering to something like this. but i doubt they have a problem with a kit like this. They're just trying to make the point that while kits arent always the cheapest and highest preforming. you can usually do a lot better for the same by buying parts individually, but it does require a lot of research. remember, google is ur best (and only) friend.

 

[Mcseiam]

Now what kind of [H] watercooling solution is that? Not removing the mobo.... bah!

Swiftech kits are nice, but they typically underperform compaired to Dangerden or Dtek because swifty uses 3/8'' ID tubing rather than 1/2'', leading to lower flowrates and pressure through the system. Then add the price of a basic swifty kit and its small radiator to the equation, and dangerden and dtek come out on top in terms of price and performance. Just my .02cents...

Swiftech offers 1/2" fittings on all their blocks

I would recommend Swiftech without a second thought - their products are well made, work very well and are easy to come by. I have owned and used 4 Swiftech CPU blocks (2 peltier models) and have never had an issue with them.

I am currently using Aqua-computer parts from Germany and would happily recommend them as well. Even with 3/8" tubing and an OC'd P4 and an OC'd 9700Pro in my system the temps are quite low. Not to mention that they are some of the best looking parts around

check them out here

and OneMadPoptart, your behavior is unacceptable and inadvisable - keep it up and you may lose your righ to post. You are welcome to your opinions and are free to express them, but you will do so with respect

 

[ErrOnReq]

just a guess here, but i think by a kit kronchev and [RIP]Zeus were refering to something like this. but i doubt they have a problem with a kit like this. They're just trying to make the point that while kits arent always the cheapest and highest preforming. you can usually do a lot better for the same by buying parts individually, but it does require a lot of research. remember, google is ur best (and only) friend.

Regardless of what they were trying to say, they have said it in a bit more constructive manner... But yeah, those kits you refer to are exactly what I was talking about earlier.

 

[[RIP]Zeus]

custom kits from dangerden and swiftech are still CUSTOM kits
what me and kronchev have a prob with is kits by koolence and other companys who make simlar types...they are no where near a DD or a SWIFTY kits....these kits you can customize no them pos koolence and shit like that....

 

[kronchev]

Let me re-explain. There are two thermal diodes on the P4.

One, which the user cannot see, is calibrated correctly and shuts the CPU down when the CPU hits 135C.

The other, which is the one the user can see, is not calibrated, and it is the one that users use to report their CPU temperatures.

So when people are saying on the Internet "The CPU gets to 95-101 before it shuts down", what's really going on is that the user visible diode is reading 95-101, while the calibrated internal diode which the user cannot read the value of, has hit 135C, and shut the CPU down.

i.e. there is a ~40C discrepency in this case between the user visible diode, and the actual temperature of the CPU die according to the calibrated critical thermal shutdown diode.

The point being that people arguing whether some board is too low or not is a bit silly - every board grossly understated the actual CPU die temperature - this is what has lead to the perception held by many that Intel CPU's "run cool".

ok, i get it, but that seems like a HUGE dT, almost a dangerous one. i should start putting thermal probes next to my die so i can see how bad AMD is compared to that. thats just...wow.

 

[kronchev]

What does a kit have to do with cutting a hole? You can get kits with huge radiators that require that a hole be cut for proper mounting...

You are so adamant that DIY is the way to go and anything slightly resembling a kit sucks because you are not "doing it yourself." What about the blocks themselves? You are buying a premade block, you didn't make the block yourself. Radiators too... make your own! PREMADE SUCKS!!!

You know what sucks? Your logic... which sucks just as much as your useless comments in this thread.

im going to make my own block and i would make my own rad if i had a torch. so yes you fail.


as for not DIY, i say that the DD kits suck because you shouldnt get all your stuff from DD. I did that for my first kit and I grossly overpaid. Built an equivalent one with a better pump and rad and it was $50 cheaper ($100 vs $150). this is with replacing EVERYTHING but a few clamps and the bottle of water wetter I had left

so to make it clear, prebuilt kits are useless, but DIY kits are overly expensive.

 

[Cathar]

ok, i get it, but that seems like a HUGE dT, almost a dangerous one.

Well you have to take into account that CPU's can run at up to around a 155-160C critical gate temperature before things start to get really dangerous for the CPU, and that's what the non-visible diode is measuring.

Intel makes reference to Tcase and Tgate. Tgate is what the non-visible diode measures, and it is the highest temperature within a CPU, being what's going on at the transistor switching level. Tcase, which is what the user visible diode measures, as I understand it, sits at the top corner of the die and mostly just measures the temperature of the interface between the CPU die and the heat-spreader (heat-spreader being the "case" in question), or if you've pulled off the heat-spreader, then the temperature of the interface between the CPU and your heatsink, and this explains why there's such a large discrepancy.

I do not believe that it is a fixed 40C thing though. These things are relative, much like sticking a thermal probe on the outside of an oven door. The disparity between the readings gets larger when the numbers get bigger. i.e. while it may be 40C difference at 95C visible, it will probably be more like a 20C difference at a 60C visible.

Remember though, some user-visible probes are seriously out of whack. I've had one of my CPUs thermally shut-down when it was reading just 55C, which is an 80C discrepency.

 

[kronchev]

Well you have to take into account that CPU's can run at up to around a 155-160C critical gate temperature before things start to get really dangerous for the CPU, and that's what the non-visible diode is measuring.

Intel makes reference to Tcase and Tgate. Tgate is what the non-visible diode measures, and it is the highest temperature within a CPU, being what's going on at the transistor switching level. Tcase, which is what the user visible diode measures, as I understand it, sits at the top corner of the die and mostly just measures the temperature of the interface between the CPU die and the heat-spreader (heat-spreader being the "case" in question), or if you've pulled off the heat-spreader, then the temperature of the interface between the CPU and your heatsink, and this explains why there's such a large discrepancy.

I do not believe that it is a fixed 40C thing though. These things are relative, much like sticking a thermal probe on the outside of an over door. The disparity between the readings gets larger when the numbers get bigger. i.e. while it may be 40C difference at 95C visible, it will probably be more like a 20C difference at a 60C visible.

Remember though, some user-visible probes are seriously out of whack. I've had one of my CPUs thermally shut-down when it was reading just 55C, which is an 80C discrepency.

ill definitly keep that all in mind. now im glad I only use AMD stuff, although that has me worried about what A64's are doing. any info on those?

also in your opinion, since youre well versed in this, what is the best way to measure temps?

 

[HeThatKnows]

Tcase, which is what the user visible diode measures, as I understand it, sits at the top corner of the die and mostly just measures the temperature of the interface between the CPU die and the heat-spreader (heat-spreader being the "case" in question),

Tcase is an industry standard term for the temperature of an IC's packaging, in this case the chip-carrier-PCB the die is soldered to. The diode is on the underside of the die along with the rest of the circuitry; though it would be cool to see ICs with circuit on both sides of the silicon wafer. In any event, it's still a long way from hottest area.

 

[ErrOnReq]

custom kits from dangerden and swiftech are still CUSTOM kits
what me and kronchev have a prob with is kits by koolence and other companys who make simlar types...they are no where near a DD or a SWIFTY kits....these kits you can customize no them pos koolence and shit like that....

im going to make my own block and i would make my own rad if i had a torch. so yes you fail.


as for not DIY, i say that the DD kits suck because you shouldnt get all your stuff from DD. I did that for my first kit and I grossly overpaid. Built an equivalent one with a better pump and rad and it was $50 cheaper ($100 vs $150). this is with replacing EVERYTHING but a few clamps and the bottle of water wetter I had left

so to make it clear, prebuilt kits are useless, but DIY kits are overly expensive.

Glad you could finally explain your opinion and possibly help the thread starter understand why. That was much better than your original comments:

my fav is i'm-not-too-lazy-to-pick-out-parts-and-cut-a-hole, by DIY

AGREED!!!!!
if your to lazy to do water cooling right don't do it at all...KITS SUCK

 

[cgrant26]

I'd like to suggest not just looking at heatercores for radiators. While the performance may not be quite as good, oil coolers have a much better variety of sizes, including thicker depths. Many of them also have threaded inlets which makes installing barbs a breeze. (no brazing needed)
I'm running a slightly unusual combo:

D-tec Whitewater
Maze4 GPU block
Maxijet 1200
B&M Supercooler (3x10x2 measure) with threaded outlets
Custom built 1 gallon plexi reservior

I have the B&M cooler mounted inside a window AC unit. I chose this small size cooler because I originally was going to mount it in the AC's outlet, I ended up mounting it internally instead and there is room for a cooler 5 times the size. Even with such a small cooler, a relatively low flow pump (295 GPH / 5'4 head) and 2 waterblocks, I still manage reported temps of:

Idle Temps:
CPU: 26* C
Board: 27*C

Load temps: (Running COD and Prime95 for 1 hour simultaniously)
CPU: 32* C
Board 28*C

Ambient 76*F +_ 3*F

The system:
AthlonXP 2500+ Barton
MSI KT3 Ultra 2
Radeon 9800 Pro
512mb GEIL PC3500 Ultra Platinum

Now these temps on their own don't mean a whole lot, but compared to what this same system reported on air cooling, it is definately a major improvement. Another benefit to having a window AC in my computer room, is the nearly constant temps in this room.

Swiftech MCX462-A W/80mm Vantec Tornado
2 Vantec Tornado case fans
1 120mm window fan blowing on Vid

CPU Idle: 36*C
Board Idle: 27*C

CPU Load: 42*C
Board Load: 29*C

Ambient 76*F +_ 3*F

I guess if there is a point to this, it's that reported temps are strictly relative. If you post temps, they mean nothing unless you have some sort of control temp from the same system as well as ambient temps. (or at least confirm them being the same in all temp cases) While I still have no clue what my CPU is REALLY running at, I know it is relatively MUCH cooler than it was.