H75 High Temps @ i7 920

Hello guys! :biggrin:

 

I have recently acquired a Corsair H75 water cooling system for my i7 920 CPU considering going for a light OC from 2.66Ghz to 3.20 or 3.40Ghz.

 

Situation is that my registered temperatures using the H75 are, on a bare minimum, of 35-37C (IDLE) even when running at stock speed of 2.66Ghz. :wtfeh: Should it not be on the region of 25C instead?

 

According to this website: http://www.hardwarecanucks.com/forum/hardware-canucks-reviews/65552-corsair-hydro-h75-h105-cpu-coolers-review-6.html my temperatures at stock speed are basically the ones these guys get when their CPU is @ 3.80Ghz... :wtfman: so what's going on here??

 

My doubt is basically if the cooling unit is working properly, because the temps are very similar to the ones registered with the huge air cooling unit I previously had installed. Only just slightly cooler by 3-4 degrees...

 

How can I find that out what is happening? How can I test this better?

 

INFO: Fans were placed to push/pull air to the inside of case as per Corsair manual. Also tried inverting the flow direction to get air out, but no big difference (it actually increased 1-2 degrees).

 

I have been using Open Hardware Monitor to check temps and Prime95 to stress test the CPU.

 

All the available help with this is highly appreciated. :feedback:

 

I really wanted to get more out of the CPU, but would rather not burn it in doing so... :nono:

 

Thank you!

No, you're too warm. There is an install guide in the stickies at the top of the cooling section. Most common problem by far is not aligning the two little notches on the back plate. Additionally, as the owner of two different X58/1366 boards, I always found the mounting a little more difficult than later generations of motherboards. I don't know whether that generation was just a little bit thinner, but I did need to end up using washers on one of them. I ran my 930@4.2 for years and even 4.4 on occasion with plenty of headroom. You will be able to easily run your overclock once we get this straightened out.

 

If it's not the backplate and mounting feels secure, post back and we can look at other things.

Hello :hihi: c-attack

 

Thanks for the quick reply.

 

I really do not think the back plate is the issue as it does not have any notches.

It is this plastic model that seemingly has no major hurdles in putting it into place... (or has it?). :thinking:

 

It is the one on the attached picture...

 

I will (anyway) take the system out and re-install it to check for any possible tweaks on placement. :winking:

@CorsairDustin

 

OP posted Stress temps attachment. It shows to be 69*C on hottest core, and 67*C on coolest core under load. That is at 3.2GHz.

 

PS: don't worry about idle temps. My CPU under custom water loop sits at 40*C, and max under load is 43*C :D i5 4670T.

Hi again guys

 

Thank you all for replying.

 

So, just re-seated the whole thing.

 

Back plate is rightly placed: vertically without touching the back of the mobo (as it should);

 

Fans blowing air in as per Corsair's installation manual booklet.

 

Still I get the same temps basically once more: 67-69C on load @ 3.20Ghz.

 

Please check attached file.

 

What else can I check?

Sorry, I forgot the H75 has a different style backplate. Just make sure the thicker sides of it run vertically. You absolutely do not have to run the H75 to bring air into your case. I know it says it in the instructions, but for most people running it as exhaust offers better overall system health. As you have already seen, it makes little difference in CPU testing. It can make quite a difference when GPU and other long duration loads are factored in. If you describe your case layout, we can make a recommendation. However, that is not responsible for your higher than expected temperatures.

 

In your BIOS or fan controller, check to make sure the pump is receiving the full 12V - maximum possible speed. It should be pretty close to 1500 rpm, but some variation is fine.

 

What kind of fan speeds did you use on the H75 when you ran the stress test?

Hello c-attack

 

Thanks for the input. Much appreciated! ;):

Please check the pic I took from BIOS Hardware Monitor just moments ago.

Please share your thoughts as I am completely lost in these matters.

 

I will probably revert the airflow again at some point...

Always made more sense to me to do so :biggrin: as I initially installed it that way: exhaust.

 

Weirdly enough, although Corsair booklet mentions installing the H75 pulling air in, their own tutorial video shows the exact opposite: [ame=

] [/ame] :confused:

 

Now regarding case airflow. Case is an Antec 300.

2 front fans pulling air in. Side panel fan also pulling air in.

 

Upper case fan exhausts alone for now.

And (for now) the H75 is also pulling air in..

 

Ps: These fans are a bit noisy and even more when on exhaust seemingly. They are bit more silent now :D:

OK, pump speed looks good at near 1500. The fans aren't going to run full blast when you are in the BIOS. Do you have a way to control the fan speed from the desktop or to program a curve in the BIOS?

 

The test data in reviews will be measured with both fans at 2000 rpm. You do not need to run that kind of fan speed for anything other than a benchmark test. It's too loud. But if you are trying to get an accurate comparison, a quick 5 minute test with the fans on full would give you a more accurate comparison. It's possible you could reach the temperatures you did if the fans were restricted to a low speed.

 

If you have only CPU concerns, don't run a discrete graphics card, or in some other way need to prioritize CPU temperatures, installing the cooler as intake may make more sense. The downside is that exhaust air coming out of it will likely be at least 10C warmer than the air going in. In short, you boost the temperature of everything in your case by 10C, unless you have some very strong exhaust. In your situation, with three fans already as intake, it would make sense to use the H75 as exhaust in that rear port and the top as well. That's pretty much the standard install for a 120x120 cooler in a normal tower.

 

It's possible using it as exhaust does sound louder. The way the outer most fan interacts with the case mesh can change the noise. With a 25mm rad and two fans, you can pretty much stick anything you want on there and it will be effective. But that is something for another time.

I just noticed your voltage in the BIOS screen shot is ringing in at 1.24V. Are you on Auto-voltage? It's been a while, but my sister 930 chip would probably run about 1.25V at 4.0GHz. The HWC review has some graphs for OC's 920's, but unfortunately they don't seem to list the voltage they used. Your temps are in line with the higher overclocked results, which might use similar voltage. **A quick scan around indicated 1.25V may be enough for 3.6-3.8GHz. This would account for the temperatures and also gives you an idea of how high you an go.

Hello again!

 

I have no fan controlling programs... any good ones to suggest? :D:

I will flip the fans round next time I open the case :laughing: should be rather soon actually..

 

I never touched the CPU voltage as I was unsure about the H75 temps..

If voltage is turned up I guess the CPU will get more and more hot and if the cooler is not getting safer temps I guess I should not touch that, or should I? :confused::confused::confused:

 

;): Check out this link: http://www.mefeedia.com/entry/asus-p...cing/12157000

 

Simple OC steps get you there seemingly. The thing is if the temps rise as they do, I'll fry the processor in no time... :mad:

Speedfan is probably the most popular aftermarket software based fan control. I haven't used it in a long time, so I am afraid I can't offer much guidance there. It looks like you have some fan control in the BIOS (Q-Fan control), but I don't know how much flexibility the program had back then. You may be able to control the fans when hooked into CPU fan header, enabling Q-Fan control, and running the pump header from a different CHA_FAN header set to full speed. Some of the other CHA_FAN headers may also be controllable.

 

As for the voltage, I can't be sure whether the voltage is set to AUTO from the one BIOS screen, but it seems way out of specification. It's completely normal for the BIOS to have pre-programmed voltage tables that are on the high side. This is to make sure even the worst CPU will run appropriately. But from everything I have seen, 1.25-1.35 volts is the range for people running 3.6-3.8 GHz. You are running 2.6. It shouldn't be anywhere near that number. It's possible it was changed during a prior overclock attempt or even inadvertently. Right now, just enter the BIOS and and go the AI tweaker (2nd tab from left). The first line should read AUTO. Go down to the lower paragraph. CPU voltage should also read AUTO. Most BIOS used a drop down menu when you hit enter at that time. If necessary, type the word AUTO into the field. The other option would be to find the reset to optimal defaults entry in the BIOS. This would change all features back to stock, including anything else you have done (like XMP memory presets). This isn't necessary, but it a thorough and certain way to return to the stock settings.

 

Your temperatures are safe now, just higher than expected. The correct voltage should lower your temperatures significantly, if it is in fact at 1.24V now. That's the part we need to confirm. Your HW monitor data charts also indicated this is true. My hunch is you are stuck on a manual fixed voltage of 1.24-1.25.

 

**Well, I take that back. Apparently, 1.25V is the average stock Vcore, which is just shocking since it is also enough to put on a 1.0 GHz OC. Still, double check the BIOS settings. If the mount is good and voltage is nominal, then we are running out of other possibilities. But perhaps we have been looking at this wrong the whole time. If you set up a 3.2GHz overclock with 1.25V, you can't compare it to the another test's CPU frequency --- you need to compare it to another tester's voltage. The HW Cannuks didn't list their voltages, but the 3.8 is likely between 1.25 and 1.35V and the 3.4 between 1.15 and 1.25V. This puts you right in the middle of the two higher test results of 64 and 70 peak temperature. If my estimation of the test voltages is close, then your H75 is right in line with what's expected.

Hiya c-attack! :hihi:

 

Hope you are ok.

Once again, thank you for replying. ;):

 

Fans are now reverted back to blowing air out the back of the case. :sunglasse

 

Q-Fan control seems to be a feature mainly designed to make fans go faster or slower regarding CPU needs, but mainly to reduce noise..

 

When turned on, CPU gets hotter as the pump does not get the ~1500 RPM needed as you mentioned. Unsure if this is best turned on or off TBH... :thinking:

 

Now, regarding the BIOS settings: voltage is on AUTO.

 

I do have an X.M.P. profile activated as my Pc struggles to detect my full 12Gb RAM memory. Triple channel silliness... :(:

 

I have therefore activated X.M.P. Profile #1 with a 9-9-9-24-2N 1.50V so that it detects all 6x 2Gb's 1600Mhz I have there.

 

Last time I turned the profile off my Pc only detected 8Gb. :mad:

So, I had to take all the RAM out, place all dimms back again one by one(!) and restarting every time so that it "finds" the 12Gb's again... :eek:

 

Do you think the voltage is incorrect because of this and this is what causes the overheating?

 

Status Quo: CPU @ 3.20Ghz ; DRAM @ 1604Mhz

H75 Fans are connected to the CPU Fan header

H75 Pump is connected to Chassis Fan #1 header

 

Ps: I believe that the fan names on Open Hardware Monitor are all swapped around... :mad:

 

Cheers!

First, we need to make sure the pump is spinning close to 1500 rpm (+-50). It needs a 12V signal, so no speed limitations. Asus have their own vocabulary, and it changes, so the terms are not always clear. If you are unable to set the pump to "full speed" or 100% in the BIOS for the motherboard header where the pump is connected, then you may need to get a molex to 3 pin fan cable and likely some other way to to connect the molex cable to the power supply (like a SATA to molex adapter). Some boards from that generation did have molex adapters already installed, so that would make it easier. However, your original BIOS picture did show it at 1500, so I am hopeful there is a way to do it from the motherboard. The Q-fan control may be appropriate for controlling the speed of the two fans on the H75.

 

OK, the XMP information was the missing part of the puzzle. You've done nothing wrong and you do not have to lower your memory speed (remove the XMP preset) to lower your temperatures. On that 900 generation of chips, there were few memory multipliers available. When you overclocked the memory with the XMP set, it also increased your CPU core frequency and the voltage to compensate (AUTO). Everyone seems to agree the stock voltage table is very heavy. Most people running 3.6 seem to be just under the voltage you are running now. So you can do one of two things. 1) Leave it alone. Your cooler is working fine and your temps are appropriate for that voltage. You are not in a danger zone for CPU temperature. 2) Lower the voltage by entering a manual number into that field. Every chip is different and may require a different voltage to be stable. The 920 seems to be more volatile than most with a wide variety of results. I have seen everything from 1.08 to 1.175 for 3.2GHz. It's a popular speed because of the common 1600 MHz memory speed and overclock. You should be able to read up quite a bit. I would suggest starting with a value of 1.18 or 1.20 in the voltage field. This should be more than you need. Play, stress test, or do nothing --- your temperature should take a noticeable drop. You will certainly be stable at that point and cooler. From there, if wish to fine tune it more, you can start dropping the voltage in increments of 0.02. This is not necessary, but would allow you to reduce temperatures further. However, at some point you may wish to start going to other way and raise your CPU clock speed. Most people are able to run 3.6 at or below the 1.25V you are at now. Your temperatures would be the same as you have now.

your temps are fine my i930@3.5 idles around 35-40c(depending on ambient-right now CPU is @ 35c, ambient is 24.5c - so about10~c warmer then room temp.), 56-60c heavy load,

I have the h100i so I might get slightly better cooling performance.

 

what are your ambient temps?, temps can also be effected by a bunch of stuff like video card throwing heat off and hot air gets sucked through rad...etc, don't forget to double check the voltages in your bios, setting them to auto or normal can be disastrous if you OC

Hello!

 

Ok CPU voltage set to 1.18 as suggested ;): and system seems cooler and stable. Please check BIOS documenting pics attached.

 

On the "BIOS CPU voltage info" file it reads max to be 2.1... :eek: so it has been really high up until now no?? :eek: Blimey.

 

Anyway... DRAM Bus voltage is set on the minimum of 1.5. Should it be higher?

BIOS HM attachment for further checking... :D:

 

And last - but not least - stress tested with the above specs and took a print screen with cores load on 100% for roughly 1 hour with Prime95.

 

:feedback:

@ noir

 

Hello ;):

 

Thanks for the reply.

I have no exact idea on how the temperature is in my room as have no way of measuring it. But I'd say it's way cooler than outside. :D: Say about 20-25C inside now?

The HDD is at 35C inside case...

 

"double check the voltages in your bios, setting them to auto or normal can be disastrous if you OC" can you explain further?

 

Cheers!

OK, the pump appears to be running full speed, so nothing to worry about there. The test results in the HW Cannuks article are done with both fans running at just under 2000 rpm. Putting yours at that speed would likely bring your numbers down under their temperatures, however 2000 rpm is loud and really for benchmarks or extreme loads only. You don't need that kind of fan speed for normal work, including gaming. It would be nice to get control over them so you learn the effect fan speed has on temperature and if this is mounted in the rear exhaust slot, these fans can help remove internal GPU and case heat as well.

 

All of the BIOS setting have a maximum possible range, as determined by the motherboard manufacturer. MOST of them have a high input value that is exceptionally larger than any one could every use under normal circumstances. For your chip, 1.25 Vcore is about the top end for casual users. Advanced users may push up to 1.35. Anything above that is out of bounds. I don't see any reason to change any other BIOS settings at this point, unless you decide to try and lower your Vcore a little bit more. Eventually, you may get a x124 BSOD. Don't worry you haven't damaged the computer. It means the voltage was too low. Raise it back up to the previous level and you're done.

 

Do not change your DRAM voltage. It's part of the XMP preset. 1.65V is the more common voltage at 1600. You have a better set of memory that can run 1.50V at the same speed and thus operates a lower temperatures. There is no reason to change your memory settings. It's a high skill, low reward type of tuning for most users.

 

Don't worry about room temperature and the internal case heat affecting the CPU. Those are factors, but the GPU's don't factor into standard stress testing and we know the higher 1.25V is responsible for the unexpected test results. Room and case temperature do play a factor. Just be aware you will have slightly higher numbers in Summer compared to other seasons. I assume weather control is out of your hands. You are well under the danger point so there is no chance of overheating the CPU on a hot day. The "disastrous" remark was hyperbole, but Noir is correct -- your problem was the AUTO voltage table was using a much higher voltage than we needed for your CPU frequency and so we got higher CPU temperatures. No damage done. Setting a very high BCLK (like 200 for 4.0GHz) and then telling the BIOS to auto-volt would likely give you very higher voltages past the 1.35V safe zone. I believe this is what Noir was referencing.

Cool c-attack :cool:

 

The whole thing does make a bit more sense to me right now. :o:

Thanks a lot for your time and expertise in helping me out.

Much appreciated! ;):

 

So... if I want to up the speed for the CPU say to 3.80Ghz or even 4Ghz :p: will this voltage value hold or do I need to put that higher again and thus getting higher temps? :confused:

 

The whole idea behind getting the H75 was to OC the CPU up as much as possible to a stable speed and temperature. :sunglasse

 

I merely stopped at 3.20Ghz as it was the last checkpoint before heat becoming too high.

 

Thanks a bunch!

well I can't say for sure what voltages are safe since every pc is different, even the same CPU can be different (D0, C0 revisions ..etc) some also leak more heat then others do.

 

If you have a D0 revision 920 CPU you might be-able to reach 4ghz with less then 1.3 volts on the vcore, the C0 revision is known to need more voltage (you can check the cpu revision in a program like CPU-Z)

 

Voltages when set to auto or normal in the bios will automatically adjust when you OC for, example I first OC my CPU to 3.5 ghz and I noticed a little orange LED light up under my northbridge, (looking this up I found out this is a overvolt led, off normal, green safe, orange warning, and red is extreme), so basically leaving it to auto caused it to overvolt my northbrige to semi-dangerous levels. (also causing more heat), I was getting up to 70C on my x58 chipset-after manually adjusting it's down to 45c-55c

(btw since watercooling removes the cpu fan, you might want to consider a little fan for the northbridge since no cpu fan= not as much airflow around that part, see how it is first)

 

so when you adjust the BCLK and multiplier the system is automatically adjusting the voltage for you, this is why you should set everything to stock, go back into the bios then write down all the voltage values under the voltage menu, then start your OC and manually set the voltages to the original stock voltage and slowly adjust them until everything is stable (you can even undervolt if everything is already stable- to get even better temps-though this is usually unlikely ;)), with luck you will only have to slightly adjust the vcore and IOH to reach 4 Ghz :)

 

but as a general rule RAM is ok up to 1.6-1.65 volts (you shouldn't go any higher then that unless it's specifically rated for it.)

I think you will likely be able to run 3.6GHz @ the 1.25V you were at before. You would need to manually enter the voltage value. Do not leave it on AUTO --- that value will change when you increase the BCLK. This will bring your temperatures back to what you got in the original test, but with a 400Mhz boost. However, and this is a big one, there are no guarantees on this stuff and it can take quite a bit of time and patience to sort it out. In looking for information on your chip, it seems there is a very large amount of variability on the 920 chips, particularly compared to the 930 I used. It's always true every chip is different. But with 920's every chip can be very different. You'll need to do your research on this and read up. Remember the voltages determine heat. Anything at 1.25V, no matter what CPU speed, will give you temperatures like your first test run in the upper 60's. I would suggest you not add more than 1.25V until you get comfortable working with the BIOS and have complete control over your cooling (fans).

Well... turns out the Pc shut down when playing BF4 :mad:

 

Low voltage I guess :(:

Placed back on AUTO for now.

You can also start working the other way with the voltage. 1.20 --> 1.22 --> 1.24. If it was a BSOD, there should have been a error code. It can be helpful in trying to figure out the issue.

Hello guys! :hihi:

 

So the Pc restarted with no warning what so ever yesterday when playing Battlefield 4.

Not a BSOD but a forced restart... :mad:

 

Looking at what you both wrote I went in BIOS to put the values in as described.

Turns out that when punching in "2.25" the BIOS reverts back to what it finds to be the max: 1.70?? :thinking:

(check attached)

 

Did not see this one coming... :eek::eek::eek::eek:

Puzzling to say the least. :confused::confused:

 

:mad: What now??

 

Cheers!

STOP!

 

Remember when I suggested you not go higher than 1.25V you were already at? That is a very real safety line. Experienced users may go up as high 1.35, but you are well past the danger zone. Immediately put it back. Immediately.