H100i GTX Pump speed Perf. difference

Good day,

I need some advice. I recently bought a 2nd hand Core i7 4770k and now I am dialing in the overclock. Currently I have it sitting at 4.2Ghz at 1.25V on the core and the CPU is cooled by a H100i GTX with 2 ML120 fans. I did a 30min stress test using Intel Extreme Tuning Utility. The results were are follows:

Test 1 (30min)

CPU package temp: 83ºC

Cooler liquid temp: 44.1ºC

Pump speed: 1920RPM (Quiet mode)

Fan speed: 1440RPM (Custom curve)

Test 2 (30min)

CPU package temp: 79ºC

Cooler liquid temp: 42ºC

Pump speed: 3150RPM (Performance mode)

Fan speed: 1260RPM (Custom curve)

The peak CPU package temp may only be 4ºC lower but the average temps across all cores dropped by about 8-10ºC when running the pump in performance mode. I knew there would be a difference, but I would like to know if this is in line with what other users also experienced?

Note: I bought the cooler in April of 2015, so it's almost 3years old.

The difference between your tests from a cooler performance standpoint is 2C. That is the difference between the two coolant temperatures. I'll give it 1C for the pump speed difference and 1C for the fan speed difference. Those are in line with what most people see.

If you have a genuine 8-10C core average temp difference, it is mostly likely something else happened or intervened during the test. Background activity, anti-virus software picking a poor time to run, Windows update, etc. There are a lot of different potential reasons, but none of them is the cooler directly. It would have to go up +8-10C on coolant difference as well. How much cool down time did you allow between tests? You would probably need at least 30 minutes to match the test duration and I would then still take stock of the case temperatures to see if the first set were lower on a cold PC. If you were going to run this kind of test again, set your fan speeds to a fixed percentage or rpm before the test starts and let it run for 2 minutes before initiating the load. That will take fan speed out of the equation and also allow the coolant temp to stabilize at the higher fan speed before the test begins. This is being picky, but when the expected differences are small you somewhat need to be.

Niggling aside, there are two things that standout for me. First, you coolant temp on XTU is hitting the 42-44C range. I would only expect this if you were in a very warm environment (case or room). Normal coolant delta for XTU on a 95TDP CPU should be no more than 6-8C. What was your original baseline H100i GTX temp? That is the critical assessment of whether things are as they should be.

The other thing I noticed is your package temp line is a bit erratic. XTU normally produces a very smooth sine wave like temp graph. It is possible the TIM material is ready for a clean and change, but I would hold off until the coolant temp question is dealt with. Also, you before running XTU you can go into the settings wrench on the graph and enable cores 1-4 (0-3) to show along with package temp. This can provide more information as to general behavior, and also if you have one hot core (common) or if the TIM is not even.

Tell us the baseline coolant temp before testing (room temp too). You don't have to run 30 min. 10-15 min is plenty to establish cooler functionality. Once you get past that, other things come into play like case air flow that may muddy the analysis. If there is a substantial difference between 15 and 30 min, that tells us something as well.

The cool down time I used between the test was about 15min. I am running the tests again with a longer cool down time and I set the fan speed to 50% The ambient room temp is about 28ºC and the coolent temp at said room temp is 34.8ºC at 50% fan speed. I will report back as soon as they are done. Attached screenshot is of the system in idle state just before I started running the test again.

As you can see there are two thermal probe readings on the screen. CP1 which sits at the front of the case measuring intake air temps and then CP2 which sits at the top of the case measuring hot exhaust air. These probes are connected to a Corsair Commander Pro. The case fan setup is as follows. 3 x ML120 front mounted intake fans and 3 x ML120 top mounted exhaust fans. The H100i GTX is also top mounted

Test 3 (15min)

CPU package temp: 84ºC

Cooler liquid temp: 45.1ºC

Pump speed: 1920RPM (Quiet mode)

Fan speed: 1260RPM 50%

15min cool down time..

Test 4 (15min)

CPU package temp: 78ºC

Cooler liquid temp: 41ºC

Pump speed: 3120RPM (Performance mode)

Fan speed: 1260RPM 50%

What do you make of these results?

Your coolant temps are a goodly bit higher than I would expect to see.

What is your ambient temperature?

Also, in looking at your Idle state images, you appear to have relatively high motherboard temperature sensors. This would tend to indicate that your ambient temperature is also high, which will skew ALL of your temperatures. Now, I don't fully trust the readings from Link for motherboard sensors so it's be good to get those temperatures confirmed as accurate, preferably with a motherboard vendor-supplied tool.

Can you detail how your case fans are configured and what your ambient/room temperature is?

The ambient room temp is about 28ºC and the coolent temp at said room temp is 34.8ºC at 50% fan speed.

OK then. Typical coolant temp delta over ambient room temperature is 4-7C, with both processor type and case configuration being the prime factors. You are at the outer edge, but its either Summer or you are tropical and that tends to be how things go. Nobody sets benchmarks or brags about their performance when it is hot in the room.

Your coolant temps appear to start dropping off as soon as the load stops. That's is good. I misread the 3/4 test coolant temps on first write and that initial analysis of normal function may have been premature. I don't like the 4C difference between them on a short duration test with fixed fan speeds. While not conclusive, it is starting to look like there is a developing flow restriction of some type. This seemingly is the most mentioned failing on the GTX/v2 series. The higher pressure 3000 rpm pump speed may be able to force liquid through the loop faster than the 2000 rpm setting, something that is normally irrelevant. I can run these types of tests with my a 350W GPU and the single component loop at pump speeds of 2000 and 5000 rpm. It is still only 1C difference and this echoes hardware specific testing of your cooler by others. Presumably it still works sufficiently for non-maximal loads, but you might want to ponder your next move. It still should be under warranty, but that may or may not be the course of action you wish to take.

I cannot explain the why the CPU package temp graph is so erratic. The heatpaste I used was applied yesterday and I use Coolermaster Mastergel Nano

Don't worry about this too much. It was a general perception and I really need to see core line graphs as well to be more certain. Nothing wrong with the TIM brand and that is never a real issue. Also, if you just put it on, it may still be curing. Some TIMs have specific cure times and it may settle down. I don't see any reason to wipe it off and do it again. Your BIOS setting are capable of creating this effect on some boards as well.

To that end, the one thing I cannot see is your actual max Vcore. I see what you set it to in the info above, but BIOS set vcore and actual synthetic load vcore are often different. This is not a major issue on Z87/97, but is is always something to check on. A +40C core temp over coolant temp delta is not unreasonable, although I thought it might be slightly less at your voltage. This is why I ask, although at this point there is another concern.

OK then. Typical coolant temp delta over ambient room temperature is 4-7C, with both processor type and case configuration being the prime factors. You are at the outer edge, but its either Summer or you are tropical and that tends to be how things go. Nobody sets benchmarks or brags about their performance when it is hot in the room.

 

Also, your coolant temps appear to start dropping off as soon as the load stops. You are within normal ranges. I don't see any major problems. Is your cooler slipping a degree with age? Maybe. But 1C is not a good reason to change coolers unless you have another reason.

 

 

 

Don't worry about this too much. It was a general perception and I really need to see core line graphs as well to be more certain. Nothing wrong with the TIM brand and that is never a real issue. Also, if you just put it on, it may still be curing. Some TIMs have specific cure times and it may settle down. I don't see any reason to wipe it off and do it again. Your BIOS setting are capable of creating this effect on some boards as well.

 

To that end, the one thing I cannot see is your actual max Vcore. I see what you set it to in the info above, but BIOS set vcore and actual synthetic load vcore are often different. This is not a major issue on Z87/97, but is is always something to check on. A +40C core temp over coolant temp delta is not unreasonable, although I thought it might be slightly less at your voltage. This is why I ask.

I live in South Africa and right now we are in the last month or two of a hot as hell summer. Outside of room temps range from 35ºC - 40ºC so the aircon is struggling to keep the room at a decent temperature. Where can I see what the Vcore max is? In the BIOS of using something like CPUZ?

Your coolant temps appear to start dropping off as soon as the load stops. That's is good. I misread the 3/4 test coolant temps on first write and that initial analysis of normal function may have been premature. I don't like the 4C difference between them on a short duration test with fixed fan speeds. While not conclusive, it is starting to look like there is a developing flow restriction of some type. This seemingly is the most mentioned failing on the GTX/v2 series. The higher pressure 3000 rpm pump speed may be able to force liquid through the loop faster than the 2000 rpm setting, something that is normally irrelevant. I can run these types of tests with my a 350W GPU and the single component loop at pump speeds of 2000 and 5000 rpm. It is still only 1C difference and this echoes hardware specific testing of your cooler by others. Presumably it still works sufficiently for non-maximal loads, but you might want to ponder your next move. It still should be under warranty, but that may or may not be the course of action you wish to take.

IMO the cooler is losing a degree or two in performance as it ages. Like stated above its almost 3years old and I think it has a 5year warranty. So what I think I'll do is monitor the temps periodically to see if anything changes and then if I do notice sudden drops in performance I will start the RMA process before the 5 year warranty expires. Would you recommend this or do you have another idea?

Your coolant temps are a goodly bit higher than I would expect to see.

 

What is your ambient temperature?

 

Also, in looking at your Idle state images, you appear to have relatively high motherboard temperature sensors. This would tend to indicate that your ambient temperature is also high, which will skew ALL of your temperatures. Now, I don't fully trust the readings from Link for motherboard sensors so it's be good to get those temperatures confirmed as accurate, preferably with a motherboard vendor-supplied tool.

 

Can you detail how your case fans are configured and what your ambient/room temperature is?

The case fan config is listed in my 2nd reply to the OP. Ambient room temps sit at around 28ºC

OK ... you have a CoPro and 2 more sensors that don't appear to be used. One of those should be used to measure the radiator intake air. That's going to be warmer than the intake air at the front of the case.

That's a key variable that you're missing in this scenario.

Another question ... check for airflow around the system vents - such as in the back over the motherboard or the empty expansion slots. When this system is at full load, do you have air being sucked in to the vents or being blown out?

How are the fan curves for your intake and your exhaust being controlled?

Where can I see what the Vcore max is? In the BIOS of using something like CPUZ?

This should register in Link as well, but you may need to configure the graphing panel to show the peak. That might explain some slightly higher package or core temps, but it would not explain the coolant temps. When you pour on voltage, the cpu cores heat up much more than any total heat into the cooling system.

This should register in Link as well, but you may need to configure the graphing panel to show the peak. That might explain some slightly higher package or core temps, but it would not explain the coolant temps. When you pour on voltage, the cpu cores heat up much more than any total heat into the cooling system.

I added it to the CL graph and as soon as the XTU stress test starts it goes up to 1.27V and stays there for the rest of the test. So its a offset of about 0.02V. Do you think I should adjust the Vcore in the BIOS down a bit to compensate for this?

OK ... you have a CoPro and 2 more sensors that don't appear to be used. One of those should be used to measure the radiator intake air. That's going to be warmer than the intake air at the front of the case.

 

That's a key variable that you're missing in this scenario.

 

Another question ... check for airflow around the system vents - such as in the back over the motherboard or the empty expansion slots. When this system is at full load, do you have air being sucked in to the vents or being blown out?

 

How are the fan curves for your intake and your exhaust being controlled?

There is a image called Airflow Diagram attached below.

You will see the blue arrows indicate 3 x 120mm intake fans and the red arrows indicate 3 x 120mm exhaust fans (2 are fitted to the 100i) The extra 120mm fan (CP Fan #4 in CL) and all 3 intake fans are grouped to respond to the GPU core temp. The rear 120mm fan mount is blank. Nothing is mounted there.

There is a image called Airflow Diagram attached below.

 

You will see the blue arrows indicate 3 x 120mm intake fans and the red arrows indicate 3 x 120mm exhaust fans (2 are fitted to the 100i) The extra 120mm fan (CP Fan #4 in CL) and all 3 intake fans are grouped to respond to the GPU core temp. The rear 120mm fan mount is blank. Nothing is mounted there.

OK ... what about the rest? Your voltage is just fine; 1.27V shouldn't increase your temps too much.

I suspect that you have airflow issues. The questions that I asked were trying to get a feel for that. You especially need to take a look at your radiator intake temperature - that's key to understanding why your radiator is getting so warm. The motherboard temp readings in Link seem to indicate this as well but, like I said, I don't fully trust them. Using the additional sensors from the Commander Pro would help us really understand what's going on with your thermals.

And, based on your diagram and the fan speeds indicated, it seems likely that you have a negative pressure scenario.

I added it to the CL graph and as soon as the XTU stress test starts it goes up to 1.27V and stays there for the rest of the test. So its a offset of about 0.02V. Do you think I should adjust the Vcore in the BIOS down a bit to compensate for this?

No, that is completely normal and what is expected. That closes this line of inquiry off.

Obviously the general problem is things are hot and this is a fact of life in these seasonal conditions. However, the 3C discrepancy between the two pump speeds is the telling mark that differentiates it from a host of other problems. If you had inadequate airflow, crazy fan configurations, a pile of dust blocking the radiator, etc., the test results would still be the same for both pump speeds. It would be hot, but it should be equally so in both tests. If we accept the principle that the pump speed should not make any difference for a short term, moderate to high load, (we say this all the time) then a 3C difference is meaningful and only attributable to the internal workings of the cooling system.

As for short term relief, if you have some consistent GPU loads you probably want an exhaust fan at the rear to get more waste heat out the back and not through the top radiator. Even with an additional fan at the rear, that should maintain positive pressure since the front 120x3 will out pace the other 3x120 by a good measure. The top radiator restricted fans can only move a fraction (maybe half) of their free air volume specification.

OK ... what about the rest? Your voltage is just fine; 1.27V shouldn't increase your temps too much.

 

I suspect that you have airflow issues. The questions that I asked were trying to get a feel for that. You especially need to take a look at your radiator intake temperature - that's key to understanding why your radiator is getting so warm. The motherboard temp readings in Link seem to indicate this as well but, like I said, I don't fully trust them. Using the additional sensors from the Commander Pro would help us really understand what's going on with your thermals.

 

And, based on your diagram and the fan speeds indicated, it seems likely that you have a negative pressure scenario.

I will take a look at radiator intake temps as soon as I can. I am working away from home until the end of the month. I have my PC with me, but the box with the extra thermal probes are at home. Maybe I will temporarily move the probe at the front intake to the intake on the radiator to see what readings I get there.

I might also later on add a extra probe closer to the GPU to see how much heat its dumping into the system which needs to be extracted.

I will take a look at radiator intake temps as soon as I can. I am working away from home until the end of the month. I have my PC with me, but the box with the extra thermal probes are at home. Maybe I will temporarily move the probe at the front intake to the intake on the radiator to see what readings I get there.

 

I might also later on add a extra probe closer to the GPU to see how much heat its dumping into the system which needs to be extracted.

Yeah, that's the other one that's worth looking at. Honestly, given a choice between radiator exhaust and radiator intake, I'd watch intake. It'll be more telling about what we're seeing here.