New Air 740 and H115i, mediocre temps LLF advice

I had decided to replace my 800D with a new Air 740. I bought the 800D back in 2011 when I had the ambition to do my own watercooling, but as the years went by and the tedium involved with a custom water set up kept me at bay, I stayed with the AIOs. When I upgraded my I5 to a I7 I bought the H80 and set up the fans in a push/pull configuration along with 2 other 120mm fans on the top sucking fresh air into the case, single 140mm exhaust. Temps were typically in the low 40s to 50s and it handled everything I threw at it without moving past 56C so while idle wasn't what I expected, the top end seemed good.

I loved the case, it was very expandible and adapted to quite a few hardware changes over the years, but it was a chore to move and I wasn't at all using it to its liquid cooling potential as originally intended. In addition, my temps were reading in the high 40s-53C while gaming, mobo was warmer than I liked too and I chalked it up to limited airflow that the case wasn't designed to provide.

So I decided to go for a midtower instead. Since cubes had hit the market I took quite a shine to them and I like the Air 740 (despite some glaring flaws in its design like no magnetic filter on top, why not retain the verticle 5.25" bays on the other side? And the HDD cage gets too warm with 3 drives installed unless a fan is mounted in the back to increase air circulation). But...despite these flaws, still like it. However I'm disappointed with the temps which are sitting at exactly the same 45-48C in a 25.5C room (78 degrees).

In contrast, my other box (a Lian Li K 62-love that case too) is running with a H60 and it's running 34-37C at usual standard activity, peaking at 44C running most games in my library. For twice the radiator size and running 2x140mm fans, I expected at least the same temp range from the H115i vs the H60, and TBH, a 2-3C lower at idle.

After doing a component transplant from the 800D to the Air 740, I was disappointed by the lackluster temps on the CPU with the H80, I tried several different mounting positions: front, top and back with no difference in temps, leaving me to think that the H80's position in one case was as good as the other. Finally I broke down and decided I'd just move up the purchase of the H115i a few months ahead of schedule.

Before my H115i arrived, I had put the H80 up front with 2 120s below pulling air in, along with 2x140 pulling air in on top. They were pulling air in through a aftermarket filter and set to "quiet mode" but regardless if "quiet" or "aggressive" the mobo ran 48-50C. I tested the airflow with a incense stick and it seemed to push a decent amount of air past the mesh, and I got the same numbers for all areas from 3 different temp programs. I was a little concerned about air turbulence though, as the front top was pushing air in the path of the H80 as it push/pulled air in from the outside towards the back, but without a fog machine to test I couldn't say.

Currently my cooling config is set up like this:

• H115i front mounted, sucking air in push through radiator, no pull fans
  
• 2x120mm bottom as exhaust
  
• 1x 140mm back exhaust
  
• 1x 140mm top exhaust
  

Now I'm considering yanking out one of the bottom fans as I likely have too much negative pressure in the case and given that I live in the desert, don't think that's a good idea.

Anyway, sorry for all the babble trying to explain my rig's set up. My question is, are there any suggestions for why a H115 is giving temps in the 44-50C range and not lower? Maybe add another pair on the other side of it? TIA.

It might be interesting to transplant the H60 into the Air 740 machine to determine if it's the water cooler or the rest of the system that's to blame.

It seems reasonable to assume that the H115i, with much more cooling surface on the radiator, would work better. Of course, that assumes that heat is getting to the radiator to begin with.

If the H60 cools better when replacing the H115i in the same system, it could indicate a flaw with the H115i. Perhaps with the new "v2" water block design?

My question is, are there any suggestions for why a H115 is giving temps in the 44-50C range and not lower? Maybe add another pair on the other side of it? TIA.

Load CPU temps in 44-50 range? I am not sure it can be lower without exotic cooling elements. You haven't mentioned anything about frequency, voltages, load type, or any of the other things that put your CPU temps into context. Now that you have a H115i, you should be able to get a coolant temperature reading (H115i Temp in Link) and that should give you a rough idea of how much heat is in the system and how much is unavoidable at the point of creation. You need a starting temp/load temp number for this. Most of these things will have a more significant impact on your CPU temperatures compared to case airflow, particularly on a CPU only load.

Since you took the trouble to describe the layouts, we might as well talk about them anyway. It seems you went from overwhelming heavy intake bias with the H80 (front, top, bottom) to a heavy exhaust bias with the H115i. Neither one is particularly beneficial to your system. Despite all the garbage floating around out there about how to set up your case, the general principle is fairly simple. The air in your case warms up from radiant heat coming from your internal components. You want to remove that air from the case and replace it with cooler air from outside the case. Just like moving people or anything else, if you have more entrances than exits or vice versa, you are going to get back up. Pressure in this environment will not speed up the airflow. Put your hand at the back mesh in that all intake set-up and you will not feel a stream of air coming through. "Positive venting pressure" is a poor substitute for an actual fan moving air in the same way people jumping the fence to get out is a poor substitute for having proper exits.

Your GPU is likely the biggest source of heat in the case and that is certainly use dependent. Without more information about hardware and how you use it, it's hard to specifically recommend an exact set-up. However, you likely would be better off moving the H115i to the top as exhaust. This will dump the waste heat directly out of the case. The front fans are then unobstructed and you can use the front fan filter now in good conscience. The one on the 740 is offset from the fans and is not nearly as crippling as the 540 before it. Rear 140/120mm to exhaust. If you do not have something to cool in the bottom of the case, don't use the fans at all. Take them out and put down a dust filter if that is the prime concern. It is difficult to filter that area when fans are installed. Otherwise, I would use them as a gentle intake to allow lower front fan speeds when under load conditions.

Hello C-Attack;

Thanks for your reply, the comment about airflow was particularly educational and I don't think I've ever run across a more illustrative description before.:biggrin:

Load CPU temps in 44-50 range? I am not sure it can be lower without exotic cooling elements. You haven't mentioned anything about frequency, voltages, load type, or any of the other things that put your CPU temps into context.

Good point, sorry about that. The joy of doing 2 things at once, both are half assed :o: The load tops out around 55C, so I'm not unhappy/worried there and makes me quite pleased with the H115i, idle however is where I'm left scratching my head. This I-3700 isn't overclocked any, its running at stock speeds.

Now that you have a H115i, you should be able to get a coolant temperature reading (H115i Temp in Link) and that should give you a rough idea of how much heat is in the system and how much is unavoidable at the point of creation. You need a starting temp/load temp number for this.

Right, everything I've been able to dig up on the Corsair Link software suggested much the same which is where I'm left scratching my head. According to the Link software, the H116i's temp yesterday was pretty steady @33C , but this morning its holding steady at 27C from starting up , fan is at 1020 rpm and the pump at 2820-2850 rpm. I'm pretty sure that will go up over the course of the day.

My assumption was the cpu therefore should be around the same temp or perhaps elevated a few. However when I fired up the system this morning the same differences in temp were shown, radiator being on average 10C cooler than the CPU.

Your GPU is likely the biggest source of heat in the case and that is certainly use dependent. Without more information about hardware and how you use it, it's hard to specifically recommend an exact set-up.

The GPU sits at around 32C during idle times and rarely gets past 40C even during long gaming sessions. Most games I play don't stress the GPU much which is why I haven't been that moved to upgrade the hardware for so long. The most demanding I play (when time allows) being Skyrim modded heavily till I get gorgeous pictures running a good 50-60 FPS with multiple foes on the screen, that's the only time I've seen the GPU hit in the mid/high 40s.

However, you likely would be better off moving the H115i to the top as exhaust. This will dump the waste heat directly out of the case. The front fans are then unobstructed and you can use the front fan filter now in good conscience. The one on the 740 is offset from the fans and is not nearly as crippling as the 540 before it. Rear 140/120mm to exhaust. If you do not have something to cool in the bottom of the case, don't use the fans at all. Take them out and put down a dust filter if that is the prime concern. It is difficult to filter that area when fans are installed. Otherwise, I would use them as a gentle intake to allow lower front fan speeds when under load conditions.

Thanks to your great illustrative description earlier, that recommendation sounds solid. I'll switch fan/rad placement based on that tomorrow morning. The lower half of the case only carries the GPU and a USB pci card for the time being. I've been debating plugging in a SBX in again as I picked up a great Logitech 7.1 surround system yesterday for a trade of a spare 540GTX I had laying around. :headbang:

While I'm at it, if you wouldn't mind some more advice. Currently I have a Artic Freeze F8 mounted on the PSU side of the case. The 80mm doesn't quite line up unfortunately, but a 92mm seems unlikely to fit without some modification. I had to mount the little guy there as I run 3 HDDs in the HD cage and they were heating up to 55C after the system had been running for ~10 hours, I do a fair amount of work on this rig so those HDDs need to stay in the green temp zone.

Unfortunately that fan, despite being labeled silent, is the single source of noise I got from that case. I've been astonished at how quiet the system runs otherwise, its damn near silent with the current fan set up (there'll be less with your suggested placement) except for the little 80mm buzzing away and is audible from 15' away. Any suggestions on what to do for that side?

So what we are really trying to figure out is why the "CPU temperature" sits around 40C at idle? I skipped over the Ivy/Sandy Bridge series, so I am going to generalize a little and perhaps someone with specific knowledge of that line can add to this.

1) Idle temps are affected by other variables. They are both room/case temp dependent, but also power state dependent as well. I am going to assume you are running the default Windows "Balanced" power plan. If you are running the Performance plan, it will keep the cores up all the time and has a noticeable affect on idle core temperatures. In addition to this, C-states and EIST/speedstep do play a role. Obviously if the voltage steps down to 0.6 and the frequency to 1000 MHz or whatever, that will produce less heat than a processor constantly at 3.4-3.9 GHz.

2) What is "CPU temperature"? Does this information come from the motherboard sensors, Intel sensors, etc. This is something very chip and board specific. Rather than try and answer that, the easier thing to do is differentiate between that value and the individual core temperatures. I suspect your actual core temps are lower than the "CPU temp". The problem is people do not always specify which value they are referring to and 'CPU temp' may not be equivalent between models. I almost always talk about core temperatures. Those are the ones I worry about and are most affected by operation and voltage. CPU Temp on my X99 is just weird. Right now as I am typing, it reads a solid +5C over my warmest core. I tend to think of CPU temp as the cumulative heat produced by all 6 cores measured at the heat spreader. It matches my coolant temp, so that fits as well. However, the one I cannot explain is package temperature. It reads +15C over my cores and another +8C over the CPU temp, when at idle with a paltry 0.757 vcore. When under constant 100% load, the cpu cores pass the CPU temp by an average of +10C while the package temperature hovers just above the warmest core in the 60's or 70's at my settings. The only temperature values that consistently make sense are the core values. There may be explanations for how and why the other values behave as they do for each CPU model, but nevertheless it becomes difficult to use them as universal measuring standards. What monitoring software are you using to obtain the "CPU" temp data? I believe Link uses the package temp, but perhaps that is different depending on motherboard and what data is available. Either way, you have to make sure you are comparing apples to apples and while a CPU temp of 39C at idle might bother me, looking down at my core temps right now between 22-25C lets me ignore that package temp value.

For the general day to day stuff, yes, the coolant temp will be lower in the morning and warmer at night after running all day. Two parts to that: 1) coolant temp basis is still room temperature, even when the system is powered down. Colder mornings bring colder coolant temps (and colder CPU temps). As the day wears on the room temp usually goes up a few degrees, or even a lot if you have a sunny side to the house. 2) Even in a perfectly planned system, there will always be some heat absorbed by the materials in the case. The longer you run, the more heat the case will retain. It is usually small amounts, but by the end of the day it will be a few degrees warmer in the case, even if the room temp did not change. It's a very small factor, but in combination with normal room temp changes usual creates one you notice. You have plenty of fan and pump speed at idle. You could probably get by with a few hundred RPM less at idle at the cost of a whole 1C difference. Even with my heavy watt draw on this set-up, 1200 rpm fan speed is about all I need for a 100% load. The next 400-500 rpm don't matter for shorter runs of 1 hour or less.

I do have 2x140 as intake at the bottom of my 740. I thought it would help with m.2 temps, but actually it makes no difference at all (for a 960). However, it does let me keep the front 3x120 under 1000 when gaming and that makes volume levels quite reasonable. If the dust concerns are not critical, by all means put 2 in at the bottom. The sound card might like it, even though we normally don't get temp readings on those.

Oh, the HDD tray. I am assuming you have three stuffed in that drive slot in the front end of the 'PSU/cold side'. I have been all SSD for a few years now, so I am afraid I cannot offer first hand advice. Someone else asked me about this a few months ago and my recommendation was to find one of those enclosures with the fan built in. I am not sure any combination of 80 or 92mm fans on the back grill will do anything for drives in the front. It's too far away. Rather than 'cool the drive', you are really trying to blow the hot air off it. The fan has to be close for HDDs. Since there is space in the back, it may also be possible to mount a new HDD enclosure against the back grill, allowing hot air to be moved out off the drives or simply as a way to split the three apart. I suspect the latter is the most temperature efficient solution. As for buzzy 80mm fans, they all are like that. You can only slow it down. I had a bunch of 92's in my last build to cover some funny corners and drives. Drove me crazy. I even used the 92mm version of yours for a while. I thought the 92's were buzzy. I can only imagine the 80's. I need to find that other thread.

A brief update: I took your advice regarding radiator and fan placement, though I retained a single 140mm fan on the bottom rear as exhaust set to a low rpm to assist exhaust. The mobo's temp zone for that was running 3C higher than the one on the other side of the GPU card. With the reconfiguration the mobo's temps went from sitting solid at 50C down to 42C, a temp I am far more comfortable with seeing long term.

As for the CPU, your assumptions were correct about power profile. As for where I'm getting my readings I have several sources. One is the Link, another was Specy, and the one I trust the most is the utility that came with the motherboard specifically included for being able to overclock the CPU and RAM. They each however give me the same numbers, or upon occasion when there is a variance, its literally no more than 1C.

You're also correct that I've been referring to the packet value, since it seemed to be a average between the 4 cores I have. All of my utilities show the cores separately, as well as a load percentage. I'm not sure though if any of those 3 are letting me see the core temps, up till now I've thought that's what I've been looking at but maybe no?

Something I've noticed over the past few days using this new radiator and the Link is that the temp in the radiator is 13C lower than what the CPU package. What relevance that amounts to I have no idea sadly :o:

The HDD cage is holding 3 drives, some tweaks to the fan connections let me control the little guy from being like a angry bee in my ear to being fairly quiet. Its keep the 3 HDDs running in the high 30s low 40s when they're active.

I guess what the takeaway from all this is about expectation management. While I've picked up some great info and understanding from the exchange, it seems like perhaps my expecting average temps being a lot lower was misplaced. I was looking at the performance I got out of a H60 on my other rig's I5 (4 cores) being so low I figured with a double radiator I'd see at least a match of those temps and definitely a improvement vs the H80 that I had before. The rig is a lot quieter though, so there was that gain at least, I just expected a more vertical improvement instead of a lateral move.

At any rate C-Attack, your replies have been very educational and you have my sincere thanks for having taken the time to make them. :P:

. As for where I'm getting my readings I have several sources. One is the Link, another was Specy, and the one I trust the most is the utility that came with the motherboard specifically included for being able to overclock the CPU and RAM. They each however give me the same numbers, or upon occasion when there is a variance, its literally no more than 1C.

Yes, I wasn't suggesting the readings were inaccurate, but more like the two observations below and comparing them to what others post. Is the package temp really the one to care about on Ivy bridge? I don't know. That it is +13C over the coolant temp makes me wonder where that sensor is and how it operates. It is occasionally refereed to as "socket temperature", but that may not be accurate either. Mine is the same way, +15C over the core temps tonight. I would never refer to my package temp as my CPU temp, whether that is correct or not. I suspect most other people do not either. There likely is a maximum recommended package temp, but other than exceeding it, I don't think you should use it as the daily assessment of fitness.

.You're also correct that I've been referring to the packet value, since it seemed to be a average between the 4 cores I have. All of my utilities show the cores separately, as well as a load percentage. I'm not sure though if any of those 3 are letting me see the core temps, up till now I've thought that's what I've been looking at but maybe no?

 

Something I've noticed over the past few days using this new radiator and the Link is that the temp in the radiator is 13C lower than what the CPU package. What relevance that amounts to I have no idea sadly :o:

I guess what the takeaway from all this is about expectation management. While I've picked up some great info and understanding from the exchange, it seems like perhaps my expecting average temps being a lot lower was misplaced. I was looking at the performance I got out of a H60 on my other rig's I5 (4 cores) being so low I figured with a double radiator I'd see at least a match of those temps and definitely a improvement vs the H80 that I had before. The rig is a lot quieter though, so there was that gain at least, I just expected a more vertical improvement instead of a lateral move.

If you dig up any multi-cooler review where they list a whole range of different models and sizes, you'll see the "idle temp" has a range of about 2C from first to last. Often times, a small cooler like a H60 might even do better than a larger double radiator on this specific aspect. The reasoning is pretty straightforward. With you C-states on, frequency and voltage stepped down in idle, your CPU is only outputting a pittance of wattage. 25 watts? 13? It really depends on low deep you let the c-states go. Regardless, it is a tiny amount of heat to get rid of. Therefore, the normal parts of a cooler than help disperse heat, like radiator surface area and fan speed, have very little bearing on the resulting temperatures. What might make that 1 or 2 C difference is the pump flow rate. A smaller cooler like a H60 usually has a higher flow rate (pump speed) than a larger cooler. It has to in order to overcome the surface area deficit it faces when under load. The trade-off is that is is louder in this state. I have a Asetek non-Corsair 120mm cooler, much like a H60 on my Titan (P). It needs to handle more wattage than my 280mm radiator on the 5930K CPU. It is able to do so and peak GPU temps are about 50C. Great, except that it is loud both at idle and load. At load because I need higher fan speeds. At idle because of the pump speed. The real gain in moving to a larger cooler is you can let the fans run at low speeds and you do not necessarily need higher pump speeds to offset other limitations. This allows you run quieter at idle and load, usually with better load temperatures. Aside from the mechanical reasons, idle temperature of anything is influenced by room/case temperature more than anything else. The afternoon sun will have more influence on your idle temps that most other settings, simply because the room temperature will go up. Your load temps should be better with lower fan speeds on the H115i.

What is in the Lian Li? Is it the same CPU, different serial number? Or was the current set-up previously in that case?

There likely is a maximum recommended package temp, but other than exceeding it, I don't think you should use it as the daily assessment of fitness.

 

If you dig up any multi-cooler review where they list a whole range of different models and sizes, you'll see the "idle temp" has a range of about 2C from first to last. Often times, a small cooler like a H60 might even do better than a larger double radiator on this specific aspect. The reasoning is pretty straightforward. With you C-states on, frequency and voltage stepped down in idle, your CPU is only outputting a pittance of wattage. 25 watts? 13? It really depends on low deep you let the c-states go. Regardless, it is a tiny amount of heat to get rid of. Therefore, the normal parts of a cooler than help disperse heat, like radiator surface area and fan speed, have very little bearing on the resulting temperatures. What might make that 1 or 2 C difference is the pump flow rate. A smaller cooler like a H60 usually has a higher flow rate (pump speed) than a larger cooler. It has to in order to overcome the surface area deficit it faces when under load. The trade-off is that is is louder in this state. I have a Asetek non-Corsair 120mm cooler, much like a H60 on my Titan (P). It needs to handle more wattage than my 280mm radiator on the 5930K CPU. It is able to do so and peak GPU temps are about 50C. Great, except that it is loud both at idle and load. At load because I need higher fan speeds. At idle because of the pump speed. The real gain in moving to a larger cooler is you can let the fans run at low speeds and you do not necessarily need higher pump speeds to offset other limitations. This allows you run quieter at idle and load, usually with better load temperatures. Aside from the mechanical reasons, idle temperature of anything is influenced by room/case temperature more than anything else. The afternoon sun will have more influence on your idle temps that most other settings, simply because the room temperature will go up. Your load temps should be better with lower fan speeds on the H115i.

 

What is in the Lian Li? Is it the same CPU, different serial number? Or was the current set-up previously in that case?

As a result of this exchange I've been digging into the topic of packet temp, core, and additional temp monitoring utilities to compare readings with. Actually learned quite a bit thanks to you and these various articles, a bit humbling considering I'm a bit of a chip-head and have been building systems for 20+ years now.

The temp ranges for the Ivy line is much better than I had thought, the danger zone is when it hits 105C which I had previously mistakenly thought was 70C, turns out that's still considered plenty safe if that's where your most demanding temps range. My previous anxiety regarding temps was likely memory fade from what I once knew to be safe temp ranges being downgraded to 'max temp' and knowing that when your edging into your overhead something needing some looking into. Refreshing one's knowledge not a bad thing after all :o:

I installed CPUID HWMonitor yesterday and ran some tests, using Prime95 and then later running Deadspace, then Xcom2. I didn't do a long test though, time constraints kept me locked into a total window of 6 hours. What I found both interesting and comforting was that running 4 different temp monitoring utilities, they all pretty much said the same thing. The only area they tended to differ on was the CPU Packet values, which were usually within +/- 2C of each other.

Since HWMonitor came highly recommended specifically for monitoring core temps, I compared it to Corsair's Link the most and the two utilities were in agreement 98% of the time, the only place they ever disagreed on was the packet temps. The result of all that being I now trust the Link's readings and don't feel the need for the other utilities anymore.

Another thing that I noticed is with those tests the H80 didn't fare as well as the 115I did. With the 115 the temps didn't go past 55C, with the H80 they reached 63C, and running the 115 the temps would settle down faster than with the H80 and was far quieter in comparison. I hadn't noticed by how much the H115 was quieter till I did some active comparison. My take away there being that the H115 is overall a better performer once it has some work to do. I should've figured as much going into this, again referring to my previous comment about expectations going in.

To answer the Q about the Lian Li's set up. The mobo is the same model, and is sporting a I5 3570 with a 570GTX GPU. Your comment about how the H60 does it's thing with pump speeds, idle rating, etc was a nice little aha for me. I took a little time this morning to run prime95 on it for a couple of hours before I had to start using it assist with some decoding and the utility showed it's top temps running at 60C, though interestingly it cooled off faster than the H80 did by around 20 minutes or so. More affirmation about your comment regarding smaller pumps and their speed :biggrin: And it was definitely the loudest of the 3 when pushed to that level.