ASUS P8Z77-M pro + SpeedFan (renamed: Question about cooling sandy bridge)

[velis]

Please see post #17 for current state of this thread.

i7-2600K to be exact

I'm using water cooling. Please let's stick with this in this thread. A debate water vs air is not what I'm after.

This CPU stays @room temps all the time, but when I put any load on just one core, it immediately jumps to >60°C. This makes it extremely hard to control fans on the radiators. I have no idea how much wattage the computer is actually consuming and am forced to run the fans at 100% (since 60°C is hot).

But actually, this is not the case: if only one core is loaded, the fans spin up immediately despite the fact that this is only for a few seconds at a time (app compilation, for example).

I have been watching other sensors and none really convey the true state of my computer at any given moment. Perhaps chasis temp (some sensor on the MB I suppose) is the most accurate, but that only moves a few degrees no matter what happens. As such it cannot be used for this purpouse.

What I'd really need is a sensor that would see water temps, but that doesn't seem to really be available.
The only other alternative I see are those automatic fan controllers that come with their own sensors. I could just stick those sensors into the radiator and hopefully set all the cooling / pump voltage based on that?

What do you guys use?

 

[Topweasel]

Why do you say 60°C is hot? At full load on my H80 I am roughly at 58°C. As far as I know Sandy bridge is good almost up to 80°C. I haven't seen to many coolers water or air keep a decent SB under 55°C and it would be even harder with a IVB or HW.

Looks like the thermal limit on the 2600k is 72°C http://ark.intel.com/products/52214.

If your really worried make sure you aren't set at max performance on the power settings. Anytime the CPU is in use it'll be running at full clock speed and power usage.

Nothing about your setup is really worrying. You have a large delta between where you are at and it's thermal limit.

 

[BonzaiDuck]

i7-2600K to be exact

I'm using water cooling. Please let's stick with this in this thread. A debate water vs air is not what I'm after.

This CPU stays @room temps all the time, but when I put any load on just one core, it immediately jumps to >60°C. This makes it extremely hard to control fans on the radiators. I have no idea how much wattage the computer is actually consuming and am forced to run the fans at 100% (since 60°C is hot).

But actually, this is not the case: if only one core is loaded, the fans spin up immediately despite the fact that this is only for a few seconds at a time (app compilation, for example).

I have been watching other sensors and none really convey the true state of my computer at any given moment. Perhaps chasis temp (some sensor on the MB I suppose) is the most accurate, but that only moves a few degrees no matter what happens. As such it cannot be used for this purpouse.

What I'd really need is a sensor that would see water temps, but that doesn't seem to really be available.
The only other alternative I see are those automatic fan controllers that come with their own sensors. I could just stick those sensors into the radiator and hopefully set all the cooling / pump voltage based on that?

What do you guys use?

This isn't about the air and water war, but my first water project ever is scheduled for next year. I use high-end heatpipes on my SB-K/2600K/2700K chips.

Frankly, I've never noticed anything like what the OP reports. I'm "evaluating" the 2700K right now and waiting for a replacement board. At 4.6 GHz and under the usual extreme stress tests, the average-of-core-maximums after an hour is about 68C. the lead core may show a very occasional peak at 73C. And I'm talking about ALL cores on my systems. The variability in core temperatures under any load or idle condition stays within the expected range of error for sensors: a range that is less than 10C at any given time -- 10C at the widest delta between the high core and the typically low core.

But under modest loading conditions, neither of the processors barely crack 50. In maybe one game, I'd seen perhaps 55C.

So the question remains: What causes a single core of the OP's system to show these temperature spikes? What he describes is a "spike" -- an increase that just as quickly recedes. If one core simply pegs at 100% usage, you still wouldn't get the temperature spike, but the 100% usage could be from DPCs and interrupts coming from a driver conflict. [And solved that, too!] When I saw this latter phenomenon, all the core temperatures were pretty much in the same range, despite the Core-0 loading.

What about the software that monitors these temperatures? What is the polling interval? The polling interval can be changed.

60C is not "that hot," but these temperature spikes seem unusual. How is the BIOS set up? Does it increase "turbo" "By All Cores?"

I just don't have an answer for what OP is describing, but I sure have some questions. How about this: Is there more than one sensor-monitoring program running simultaneously on OP's system?

Somebody has to have an answer for this.

Maybe Intel's Indium solder-gun missed a patch on his processor?

 

[Cerb]

OK...

1. Why does the CPU temp spiking make it hard to control the fans on the radiators?

2. Do the fans need to spin up to 100% because of this, or not?

3. What are you using now to control the fans? Specifically, if "not" to #2, Speedfan allows the sum of results to control a fan header. So, you could, for example, set up a curve where some random sensor's value sets the base % (minimum idle speed), then 0 to 1/4 of max for each core, over normal ranges, with a quick rise to 100% at abnormally high values.

4. Are you sure you have enough water flow? A temperature spike is not going to be affected much by the radiator's cooling. Just like with big heatpipe coolers, but more so, there will be a lag time, as the heat is physically moved with the water. The CPU block should be getting cool water all of the time, and that should be what handles the spike, as the fans speed up to then handle what's in the radiator.

 

[Cerb]

Looks like the thermal limit on the 2600k is 72°C http://ark.intel.com/products/52214.

That's the elusive Tcase-max. It is not the limit for the cores, but the limit for the center of the IHS. The cores being 15-20C hotter is fine, with pretty much any enthusiast cooling, and preventing cores from occasionally spiking up past that value is basically impossible, without major undervolting. It looks like 98C is the tjmax for all desktop SBs (you want to stay well under it, but core temps of 72C for a bit at a time are not going to be a problem).

 

[BonzaiDuck]

That's the elusive Tcase-max. It is not the limit for the cores, but the limit for the center of the IHS. The cores being 15-20C hotter is fine, with pretty much any enthusiast cooling, and preventing cores from occasionally spiking up past that value is basically impossible, without major undervolting. It looks like 98C is the tjmax for all desktop SBs (you want to stay well under it, but core temps of 72C for a bit at a time are not going to be a problem).

No disagreement there. I'd only say that if your cooling solution brings it up to Avg.of-Maximums at 85C and you can find any way to lower it 10C, you're in better shape if you can keep the same clock and voltage.

The only other problem: the BIOS thermal monitoring feature or possible something like OCCT will enforce a tolerance limit on the system. I found my OCCT terminates when one, single sensor passes 87C. But that can all be changed, I think, in the "settings" menu -- possibly in the BIOS. I would only guess that the lead thermal sensor -- highest registered value -- would correspond to an average somewhere around 82c. In that case, you only have to find 7C-worth in "grains of rice."

You see what other people achieved in their overclocks. For Sandy, you should be able to get within 200Mhz of "superb" clock speeds, even with air-cooling.

I'm posting results I've had this week for a cheap ASUS mobo with only 4+1 phase-power and an EVGA ACX cooler with an AP-30 Gentle Typhoon exhaust fan. [Nothing gentle about it . . ] Affinitized LinX with this cooler gives me an avg.-of-Max's at 68.5C with 4600 clock and ~1.35V. the droop is a little greater for Prime95 sFFT, showing about 1.33V, and the temperatures are ~64C. Room ambient 77F. This with an i7-2700K.

I can't say what better phase-power design will offer until a week from now. I can get this to 4700 on this board, but I don't think that speed or 4800 is such a great idea.

And after all -- there's a chip lottery. I think they marketed the 2700K for noobs and mainstreamers who really don't want to bother. It's spec'd to run just 100Mhz faster than the 2600K at stock, and I don't think they cared about being more discriminating about picking the processors that only offer up 1GHz stable overclocks above that.

I DID have it at 4700 with excessive LLC and an offset of over 50mV. I think the average-of-maximum temperatures didn't much exceed 69 or 70C, but I'll have to try it again with better tweaks. If you ask me, I could more likely kill this motherboard than the processor. So . . . . next week.

Does anybody know if you can fit a Swiftech H2O 240X in a HAF 922? I don't even think I found anybody selling them. For the price, I'll likely just go straight to custom water.

 

[velis]

Seems the temp difference raised some eyebrows, I didn't think it'd be so important, so I didn't list my OC details and such in OP. I just ran a few prime tests to report more accurate numbers. These are the immediate package temps reported for # of threads started:
1 - 48
2 - 55
4 - 65
8 - 70

These are temps reached in seconds from starting the test. They will rise later as water heats up as well.
The processor is overclocked to 4800 MHz. The rise in temps is definitely not as pronounced at stock speeds, naturally.
Readings are from CPU-Z Hardware monitor utility. Core temps ALWAYS lower than package.
I also cool the graphics card with the same loop.

To answer Cerb's questions:
1. This makes it hard to control fan speeds because a single few seconds compilation will spin up the fans when thresholds are set to 50°C or so. If I set thresholds to 60°C then I have an issue where the gfx card will heat up the water too much without the processor actually breaking the 60°C threshold.
These are of course two different usage scenarios: the first is development, the second is gaming. I don't really want to adjust ASUS Fan Xpert+ settings every time I want to play games (but I have to right now).

2. If I want to cover the gaming scenario, the fans will spin up also in dev scenario, yes.

3. I have not noticed that with SpeedFan. I think it too can't see GFX card temps which could be a good second reading for such a combined index. Additionally, I've had issues just making it work. I don't think I succeded at all on this motherboard (ASUS P8Z77-M pro). Will try again. So currently I'm using ASUS Fan xpert+ as mentioned in #1. It only uses CPU package temp for doing it's magic.

4. I'm using a single pump, but I believe the water flow is sufficient, yes. It's an EK DCP2.2, rated for 400l/h. I do not feel any warmup in tube going directly out of the CPU even though admittedly that's not a really good test. Still, even if the pump only manages 0.05 l/s, that should be enough IMO. I used to have two pumps, but haven't noticed any change in performance going back to one.

 

[BonzaiDuck]

Seems the temp difference raised some eyebrows, I didn't think it'd be so important, so I didn't list my OC details and such in OP. I just ran a few prime tests to report more accurate numbers. These are the immediate package temps reported for # of threads started:
1 - 48
2 - 55
4 - 65
8 - 70

These are temps reached in seconds from starting the test. They will rise later as water heats up as well.
The processor is overclocked to 4800 MHz. The rise in temps is definitely not as pronounced at stock speeds, naturally.
Readings are from CPU-Z Hardware monitor utility. Core temps ALWAYS lower than package.
I also cool the graphics card with the same loop.

To answer Cerb's questions:
1. This makes it hard to control fan speeds because a single few seconds compilation will spin up the fans when thresholds are set to 50°C or so. If I set thresholds to 60°C then I have an issue where the gfx card will heat up the water too much without the processor actually breaking the 60°C threshold.
These are of course two different usage scenarios: the first is development, the second is gaming. I don't really want to adjust ASUS Fan Xpert+ settings every time I want to play games (but I have to right now).

2. If I want to cover the gaming scenario, the fans will spin up also in dev scenario, yes.

3. I have not noticed that with SpeedFan. I think it too can't see GFX card temps which could be a good second reading for such a combined index. Additionally, I've had issues just making it work. I don't think I succeded at all on this motherboard (ASUS P8Z77-M pro). Will try again. So currently I'm using ASUS Fan xpert+ as mentioned in #1. It only uses CPU package temp for doing it's magic.

4. I'm using a single pump, but I believe the water flow is sufficient, yes. It's an EK DCP2.2, rated for 400l/h. I do not feel any warmup in tube going directly out of the CPU even though admittedly that's not a really good test. Still, even if the pump only manages 0.05 l/s, that should be enough IMO. I used to have two pumps, but haven't noticed any change in performance going back to one.

I'd want dual pumps for two reasons: better head pressure and a backup for safety. I think those are reasons cited in this forum by the Chief Water Guy.

That choice is a lot less dire with a case, a few fans and an airflow strategy. So I'll only build it with dual pumps.

 

[velis]

Installed SpeedFan. Immediate observations:
It can now see GPU temp. It cannot see CPU Package temp. It cannot control fans, at least not with ASUS utilities still installed and active.
It is only the CPU package temp that rises so quickly, not "the other" CPU temp reading. That one goes up gradually, but very slowly - presumably because of WC.
ASUS Fan Xpert uses CPU package temp, not CPU temp.

@BonzaiDuck: That's exactly what I had two pumps for before. However, this time I chose a minimal case for my setup (GD05) and only installed one pump due to space contraints - though I could install two if I wanted.

 

[BonzaiDuck]

Installed SpeedFan. Immediate observations:
It can now see GPU temp. It cannot see CPU Package temp. It cannot control fans, at least not with ASUS utilities still installed and active.
It is only the CPU package temp that rises so quickly, not "the other" CPU temp reading. That one goes up gradually, but very slowly - presumably because of WC.
ASUS Fan Xpert uses CPU package temp, not CPU temp.

@BonzaiDuck: That's exactly what I had two pumps for before. However, this time I chose a minimal case for my setup (GD05) and only installed one pump due to space contraints - though I could install two if I wanted.

Somehow I still think that's the challenge with a water system. You either build a monster case with an LED-lit aquarium in some Volenti cooler rig, or you start trading off disk cages for water parts. To do it well should take some planning if the case cannot be a monster case.

 

[velis]

It is done well: this setup easily handles a 2600K + Radeon 5870. Now I bought a GTX 970 and I have no reason to believe it won't be able to handle that. Even if there is a problem, I still have room for 2x120 rad in addition to two 3x120 rads I have right now.

My only issue is with WHEN the fans should spin up.

Update: It seems I cannot make SpeedFan control fan speeds even if I do disable the ASUS utility (I haven't uninstalled it yet).

 

[BonzaiDuck]

It is done well: this setup easily handles a 2600K + Radeon 5870. Now I bought a GTX 970 and I have no reason to believe it won't be able to handle that. Even if there is a problem, I still have room for 2x120 rad in addition to two 3x120 rads I have right now.

My only issue is with WHEN the fans should spin up.

Update: It seems I cannot make SpeedFan control fan speeds even if I do disable the ASUS utility (I haven't uninstalled it yet).

Which board and chipset is it? Is it the Z68 or Z77 chipset -- essentially?

 

[velis]

ASUS P8Z77-M pro

 

[BonzaiDuck]

ASUS P8Z77-M pro

That mATX board should work well for you. If you hadn't already done so, you should explore the BIOS fan settings before configuring with Fan Xpert.

There are already "profiles" in BIOS that work by themselves, and to use Fan Xpert, you need to select "User" in the BIOS menu. The BIOS and the software have to work together if you use the software at all.

Truth is, we seem to be working on the same project, more or less. But I'm using a glitzy-bling heatpipe cooler, which seems to be outperforming all the others despite the glitzy-blingy way it is marketed. [And I don't really care if the bundled fan has "Red-LED." Geeesh!!]

Also, if you still have trouble with the onboard fan controller, there's a $10 NZXT splitter/controller device. You might want to look into it, but onboard fan control is the best and simplest way to go.

Now I've refreshed myself about the contents of this thread. I wouldn't know for sure, but the "simple" serial inclusion of the VGA card in the loop may actually be complicating things for you. I couldn't say for sure. Let's say that what you've described hasn't been in the inventory of my firsthand experience.

 

[Cerb]

1. This makes it hard to control fan speeds because a single few seconds compilation will spin up the fans when thresholds are set to 50°C or so. If I set thresholds to 60°C then I have an issue where the gfx card will heat up the water too much without the processor actually breaking the 60°C threshold.
These are of course two different usage scenarios: the first is development, the second is gaming. I don't really want to adjust ASUS Fan Xpert+ settings every time I want to play games (but I have to right now).

Does Fan Xpert allow you to set the fan curves based on multiple temperatures?

3. I have not noticed that with SpeedFan. I think it too can't see GFX card temps which could be a good second reading for such a combined index. Additionally, I've had issues just making it work. I don't think I succeded at all on this motherboard (ASUS P8Z77-M pro). Will try again. So currently I'm using ASUS Fan xpert+ as mentioned in #1. It only uses CPU package temp for doing it's magic.

Using the video card temp is one of the reasons I use it. It works for Geforces, at least. You will have to set your BIOS to manual control, and may need to also set that in Speedfan for the chip, before it takes over. It's not entirely intuitive, I'll grant that.

Once Speedfan is actually controlling the fans, you can set up fan curves based on multiple temperatures, either the maximum of a set of values, or their sum. With a shared loop, sum would make the most sense. I've personally found CPU core 0 alone to be a good enough way of handling CPU load. Between Turbo, and the memory subsystem chewing up power under load, the added cores tend to be a fairly small addition to total power, compared to the first one loading up.

 

[BonzaiDuck]

Does Fan Xpert allow you to set the fan curves based on multiple temperatures?

Using the video card temp is one of the reasons I use it. It works for Geforces, at least. You will have to set your BIOS to manual control, and may need to also set that in Speedfan for the chip, before it takes over. It's not entirely intuitive, I'll grant that.

Once Speedfan is actually controlling the fans, you can set up fan curves based on multiple temperatures, either the maximum of a set of values, or their sum. With a shared loop, sum would make the most sense. I've personally found CPU core 0 alone to be a good enough way of handling CPU load. Between Turbo, and the memory subsystem chewing up power under load, the added cores tend to be a fairly small addition to total power, compared to the first one loading up.

A broken line approximates a curve. There are essentially three threshold temperatures in Fan Xpert from AI Suite II: A minimum, below which % duty cycle is a constant; a midpoint, that defines a linear ascent to that point in % Duty cycle as temperature increases; and a high-point, defining a line with any upward slope that deviates or follows the slope of the first line. So of course, speed increases for "many different temperatures." There can be two rates of increase between the minimum point and the high-point. The high-point defines a maximum % duty-cycle.

That kind of fan profile would be all you would need for thermal control of each group of fans. for that version of AI Suite, you have two groups: CPU, and Chassis. You could run other fans from the CPU_FAN PWM signal and they wouldn't necessarily need to be "CPU fans" but they'd have to be PWM fans. With the Chassis group, I could tie in some Sunon Mag-Lev 40mm fans that cool VRM and other mobo components such as the Northbridge and its heatsink. The only tricky part is to choose fans so that you can mix them for each duty cycle profile and keep the noise at the desired level while getting the optimum airflow you want or need from those fans.

With three groups of fans, or more, you could probably do almost anything a little better. But it is possible to define a minimum-fan strategy that does everything that is needed. Cheaper motherboards have fewer groups of fans or fewer fan ports generally. Even with those limitations, you can buy a cheap NZXT device that would allow you to control both 3-pin and PWM fans simultaneously -- probably from a PWM signal. It might even work both ways. I'll look that up in case there is a "want to know."

Here. There are two of these GRID items, one being a controller:

http://www.newegg.com/Product/Produc...-014-_-Product

A study of the gallery pictures and "features" description tells me there is a USB connection to the motherboard. So with this device, you have "extra software bloat." You always have another possibility of wiring low-amperage fans in parallel and connecting them to a three-pin mobo CHA-Fan port.

If you make sure you use PWM resources to the fullest, you can have PWM control of both CPU and CHA fan ports on various motherboards. On some ASUS mid-range boards, there is at least one PWM CHA_FAN port, and two CPU_FAN PWM ports. So for the price of $10 each, you could use two swiftech 8W-SPL-PWM_SAT splitters for two groups of fans. You might then have a spare CHA fan port and a PWR fan port to play with in at least a couple different ways. IF the software allows separate control of all fan ports, you have that many groups to play with. The GRID device allows more. And the point of it: once you've defined fan profiles with something like Fan Xpert, you don't need to use anything more than BIOS features monitored by either board software or shareware. So adding the NZXT software for its own controller utilizing USB shouldn't really present a problem.

Ultimately, I could see spending $10 to $20 for PWM splitters (2) and the $30 on something like the NZXT controller. But I don't want to add the NZXT controller if I don't need it. And that goes for front-panel fan controllers of all kinds. Aquaero 5 is nice, but if you don't need it, well, you don't need it.

 

[velis]

Exactly. I don't want to buy something I don't need.

So far I have reached a conclusion that my optimum reading would be CPUTIN instead of Package temperature. CPUTIN is read nicely by the CPUFan utility while the ASUS Fan Xpert only reads package temp.

That makes fixing SpeedFan the next goal. Is there anyone with this mobo that has SpeedFan controlling the fans?