• Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!

Corsair H60 vs. Thermalright Ultima 90 vs. Corsair H100

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.

Tech Tweaker

Member
Joined
Dec 14, 2010
I have been doing a performance comparison between the Corsair H60 and the Thermalright Ultima 90.

For those who don't know, the Ultima 90 is a smaller version of the Ultra 120 Extreme (TRUE) and performance is roughly the same between the two.

Testing was done on the 775 system in my signature (Core 2 Quad Q6600 at stock 2.4GHz with stock voltage (1.23V)).

Temperatures in degrees Celsius with Thermalright Ultima 90:
Core| 0 | 1 | 2 | 3 | Ambient temperature
at load: | 44| 45 | 42 | 38 | 72°F/22°C
at idle: |30 | 33 | 33 | 27 |72°F/22°C

Temperatures in degrees Celsius with Corsair H60:
Core| 0 | 1 | 2 | 3 | Ambient temperature
at load: | 43 | 44 | 43 | 41 | 72°F/22°C
at idle: | 26 | 30 | 29 | 25 | 72°F/22°C

Temperatures in degrees Celsius with Corsair H100 (open case, side panels and top panel removed):
Core| 0 | 1 | 2 | 3 | Ambient temperature
at load (push fans on low, pull fans on full): |42 | 43 | 42 | 38 | 72°F/22°C
at load (push and pull fans on full): | 38 | 39 | 38 | 35| 72°F/22°C
at idle (push fans on low, pull fans on full): | 25 | 28 | 28 | 24 | 72°F/22°C
at idle (push and pull fans on full): | 25 | 28 | 28 | 24 | 72°F/22°C

Temperatures in degrees Celsius with Corsair H100 (closed case, all intake and exhaust fans on full):
Core| 0 | 1 | 2 | 3 | Ambient temperature
at load (push fans on low, pull fans on full): | 43 | 44 | 43 | 40 | 72°F/22°C
at load (push and pull fans on full): | 42 | 43 | 42 | 38 |72°F/22°C
at idle (push fans on low, pull fans on full): | 26 | 31 | 30 | 25 | 72°F/22°C
at idle (push and pull fans on full): | 26 | 30 | 30 | 25 | 72°F/22°C


So, the H60 seems to be just barely better than an Ultima 90 or TRUE. It's also a little bit quieter, just going by my ears since I don't have a sound meter on hand. The hottest core went down by just two degrees, but for some reason the coolest-running core actually went up in temperature when testing at the highest ambient temperatures and I'm still a bit confused by that.

Unfortunately I wasn't able to keep temperatures constant since my computer room isn't exactly well regulated temperature-wise, and I kept having to open the window to get cooler air into the room as it was getting a bit hot at times, which is why testing happened a multiple temperatures.

I will probably try to test on my socket 939 system as well, since I was also wondering if the H60 would be a good upgrade for that system or not, but since I am running a similar air cooler on that computer as well I don't expect any great gains there either unfortunately.

The H100 puts up some impressive numbers as far as temperature drop going from the Ultima 90/TRUE's little brother. I'm seeing around a 10° drop at load with fans on full (or 6-7° with push fans on low), and a 5° drop at idle.
 
Last edited:
Nicely done! :thup:

I would much rather see a test like your than most of the long winded reviews. You have taken care to keep track of what your ambient temp is doing for each test. Very good!

When I play around doing similar testing to yours I used to use room ambient. I now put a temp probe 30mm in front of cooler intake. I think this gives me a more accurate reading of what the air temperature going into cooler is at all times. I use a cheap indoor/outdoor thermometer with a wired remote. Terrarium thermometer will work too.
I got a new Thermalright AXP-100 downflow HTPC cooler with new TY-100 fan. Great little cooler but when I was playing around and reversed the fan to pull instead of push into cooler I got 5-8c cooler readings. Obviously the fan isn't pulling that much more air than it can push so I started hunting for the reason. I was testing on an open worktop so wasn't case induced. But downflow exhaust was curling back up from RAM, GPU, I/O blocks, heatsinks etc and being sucked back in. After I put the probe 30mm above intake I also put a TY-140 fans 20cm away and 15cm above worktop blowing over the cooler... and it was still 5-8c hotter!
 
Interesting. So these were tested in completely different environments (closed case, open case etc)? EDIT: Ok I see now. What where the air coolers tested in/on???

Can you make the data more legible though? A table perhaps? What voltage and clocks were the Q6600 run at?

You can also 'normalize' the temperatures (and should). If you have one test that was 22C and another 25C, you raise the 22C temps up 3C across the board. That way, its an apples to apples comparison using the same ambient temperatures. Just pick an ambient value and add or subtract around it to make it accurate.
 
Last edited:
Interesting. So these were tested in completely different environments (closed case, open case etc)? EDIT: Ok I see now. What where the air coolers tested in/on???

Can you make the data more legible though? A table perhaps? What voltage and clocks were the Q6600 run at?

You can also 'normalize' the temperatures (and should). If you have one test that was 22C and another 25C, you raise the 22C temps up 3C across the board. That way, its an apples to apples comparison using the same ambient temperatures. Just pick an ambient value and add or subtract around it to make it accurate.

Unless noted otherwise all coolers were tested in a closed case environment with all fans on full.

Unfortunately I have no idea how to make a table on here. :shrug: The Q6600 was at stock clocks and stock voltage (1.23V as reported by Hardware Monitor).

Well, unfortunately it's not quite a direct relationship between the drop in ambient temperature and the drop in the CPU temp, so changing the temperatures by mathematical addition or subtraction might not give entirely accurate results. On the Ultima 90 where I dropped 2°C on the ambient temp the CPU temp average only dropped by 1.5°C.

Although on that same note when I dropped the ambient temperature from 23.5°C to 21°C during the H60 load tests the average core temp also dropped by 2.5°C.

I was actually shocked when I managed to maintain the exact same ambient temperature during the H100 testing.
 
Well, unfortunately it's not quite a direct relationship between the drop in ambient temperature and the drop in the CPU temp, so changing the temperatures by mathematical addition or subtraction might not give entirely accurate results.
Yes, it does. This is what skinny has done and martin when testing (most respected water cooler testing people around). Aside from them, its a scientific fact. These need to be normalized to be more accurate. The difference you see is likely in your temperature monitoring device.
 
Can you make the data more legible though? A table perhaps? What voltage and clocks were the Q6600 run at?

You can also 'normalize' the temperatures (and should). If you have one test that was 22C and another 25C, you raise the 22C temps up 3C across the board. That way, its an apples to apples comparison using the same ambient temperatures. Just pick an ambient value and add or subtract around it to make it accurate.
Per your request I formed some tables. Took me forever to figure out how that worked.

These numbers can't be right, after "normalizing" the temperatures from the Ultima 90 and H60 tests it looks like either the H100 is a terrible or mediocre cooler or the Ultima 90 is an awesome cooler for reasons unknown.
 
Its possible you had better mounts on one and not on the other? I assume you mounted these multiple times to rule out bad mounts or TIM applications? Another variable is the testing environment.. closed case, open case, different cases, those werent the same as I understand it so that is another area for a potential cause for concern/variability in your results. Tests in case A vs case B will be different due to different airflow patterns.

Basically, if you want this to be accurate you have to minimize the variables between the testing (remounting a couple of times and dropping highest lowest then average, same EXACT environment for all items being tested, etc), which wasnt done.
 
Last edited:
Its possible you had better mounts on one and not on the other? I assume you mounted these multiple times to rule out bad mounts or TIM applications? Another variable is the testing environment.. closed case, open case, different cases, those werent the same as I understand it so that is another area for a potential cause for concern/variability in your results. Tests in case A vs case B will be different due to different airflow patterns.

Basically, if you want this to be accurate you have to minimize the variables between the testing (remounting a couple of times and dropping highest lowest then average, same EXACT environment for all items being tested, etc), which wasnt done.

Yep, mounted multiple times.

They were all in the same case.

Only difference was one round of the H100 testing was done with the case open, the other round of H100 testing was done with the case closed, exactly as the other tests were done.
 
Ok, cool! That eliminates that... but then you have mounts and TIM which are the biggest variable of them all. ;)
 
I don't see enough of a diff in temps to say one is better than the other. 3C at load (using core #3) is within ambient air and testing variabilities. If it was a 3970k overclocked then the difference would be more due to the fact that each cooler has a % effiency. Say the H60 is 12% less capable than the H100 per watt made vs dissipated.

It would show a lot more difference with a more typical OCF user with that. Silicon Die size, CPU heat spreader size is very diff on modern chips.

And having to normalize temps is a necessity for good compairisons. As is 3+ applications of the TIM with the same paste on each test.

You got the testing bug and you'll get better in time if you continue your efforts. Thanks for your input and efforts.
 
I don't see enough of a diff in temps to say one is better than the other. 3C at load (using core #3) is within ambient air and testing variabilities. If it was a 3970k overclocked then the difference would be more due to the fact that each cooler has a % effiency. Say the H60 is 12% less capable than the H100 per watt made vs dissipated.

It would show a lot more difference with a more typical OCF user with that. Silicon Die size, CPU heat spreader size is very diff on modern chips.

And having to normalize temps is a necessity for good compairisons. As is 3+ applications of the TIM with the same paste on each test.

You got the testing bug and you'll get better in time if you continue your efforts. Thanks for your input and efforts.

The fact that the difference is minimal on a stock CPU is important. Most reviews show the difference between cooler at high load and readers fall into the belief that their little cooler is no good because DDB cooler is 12c cooler on XB3864 CPU that's pulling 180watts to runt 4.9GHz on 8 cores. How many people even have software that can utilize that many cores or that kind of speed??? This is part of the hype many young and/or inexperienced buyers fall for. Most of us.. myself included have little need for the huge cooling ability of the big coolers I use them because I like quiet. And more cooling means less airspeed means less fan and air noise.

Most systems would run 3-8c cooler if their owners payed more attention to getting hot exhaust CPU and GPU cooler air out of the case before it mixes with case air. It doesn't take much 30-35c exhaust air to raise the case temp from 23c to 28c.. and that's 5c more CPU heat.. and the air out of cooler goes up to 38-40c... case goes up to 32c.. until 15-20 minutes into test things reach a balance.. or test is shutdown. ;)

It really isn't how many fans are blowing into case or how many are blowing out or how many extra fans in case to blow at GPU or NB or.... It's all about cool air to coolers/heatsinks and hot air out of case.

*climbs off of hovercraft*
 
I don't see enough of a diff in temps to say one is better than the other. 3C at load (using core #3) is within ambient air and testing variabilities. If it was a 3970k overclocked then the difference would be more due to the fact that each cooler has a % effiency. Say the H60 is 12% less capable than the H100 per watt made vs dissipated.

It would show a lot more difference with a more typical OCF user with that. Silicon Die size, CPU heat spreader size is very diff on modern chips.

And having to normalize temps is a necessity for good compairisons. As is 3+ applications of the TIM with the same paste on each test.

You got the testing bug and you'll get better in time if you continue your efforts. Thanks for your input and efforts.

maybe we can ask the OP to OC his chip first, to something ~4GHz @ ~1.5v
before we ask him to buy a new chip for test :D


Testing was done on the 775 system in my signature (Core 2 Quad Q6600 at stock 2.4GHz with stock voltage (1.23V)).
 
Back