how hot can all in one liquid cooling get?

[HectiKCookiE]

so today I went and brought an nzxt kraken g10 for my amd r9 290x for cooling i am using a used antec kuhler h20 620, the thing is that when i ran the valley bench mark i was getting 60fps with my gpu clocked at 1000mhz but the max temp on my gpu was 75c im not sure how hot all in one water coolers should get. is this temp ok or should i throttle it a bit?

 

[Hecknologist]

It's a little high but nothing to worry about, the 290x is a beast of a power consumer so heat output will always be high.

[Tripnator]

 

so today I went and brought an nzxt kraken g10 for my amd r9 290x for cooling i am using a used antec kuhler h20 620, the thing is that when i ran the valley bench mark i was getting 60fps with my gpu clocked at 1000mhz but the max temp on my gpu was 75c im not sure how hot all in one water coolers should get. is this temp ok or should i throttle it a bit?

 

you have put coolers on your VRM right? the kracken on a 290x will fry those chips unless you put coolers on them? 

[Faceman]

you have put coolers on your VRM right? the kracken on a 290x will fry those chips unless you put coolers on them? 

Not necessary, purely precautionary.

 

The Kraken G10 gets a lot of bad publicity because people claim that it has insufficient Voltage Regulation Module(VRM) and Video Card RAM(VRAM) Cooling.  For starters, the VRAM never gets hot enough to be worried about, so lets focus on the VRMs.  Does the GPU need additional heat sinks for the VRAM and VRMs?  In short, no.

 

   The whole "insufficient VRAM/VRM cooling" began with a review of the G10 by Puget Systems.  This review is probably the most critical review of the G10 out there.  While it makes an interesting point, I have to explain its critical flaw to a certain extend.  To begin, their Unigine temperature test, which is most closely related to video game performance, their measured temperatures are perfectly fine, they are even lower with the G10 than with the  stock heat sink for both the GTX Titan and Reference R9 290X(Notorious for high heat).  The only time when we see the temperatures higher with the G10 than the stock heat sink are when they run Kombustor.

 

     Kombustor/Furmark is a program that should NEVER be run.  It puts unnecessary load on the GPU that is completely unrealistic.  It is a card killer.  It is uncommon, but happens enough that it needs to be mentioned.  Kombustor can and will brick your card.  It happened to me earlier this year, and it happens to many others.  I like Tiny Tom Logan's analogy of this stress test.  "It is like the Navy Seals, it is incredibly rigorous and has a low pass rate."

     I like to give my components 2 weeks of break in time before overclocking to make sure everything is working properly.  This is what I did with my brand new, MSi GTX 770 Lightning.  I had used the card for 2 weeks, no problems, everything going fine.  As you know with a Lightning, this is a card built specifically for overclocking.  Without knowing what I know now, I ran Kombustor at stock settings to get a baseline for my overclocks.  It took all of 3 minutes running Kombustor on a brand new, factory settings card for it to die.  I am not the only one this has happened to, so I always tell my story whenever I get a chance because Kombustor is a dangerous program, and should not be used.  If you want to run some benchmarks for your GPU, use Unigine, 3DMark, and In-game benchmark utilities.  Not Furmark/Kombustor.

 

     Back to the Puget review.  They ran Kombustor during their test, which as a reviewer, they should do in order to review the G10 in all scenarios to give the reader an idea of what to expect.  My problem with the review and the negative perception it has created is that it is an unrealistic scenario, one which I would not recommend anyone do at home, and you are unlikely to encounter that type of load during regular use.  I'm not aware of any real world applications where your GPU would be put under that much stress.  So lets look at the part of the review that pertains to real world scenarios, their Unigine portion of the test.  The VRM temperatures are perfectly fine, even lower with the G10 than the stock heat sink.  This is with "naked" cards.  A lot of non-reference cards which a lot of people own come with mid-plates that passively cool the VRAM and VRMs. 

 

     To give you an idea of what VRM temperatures are like for those of us at home, we asked members of the OCN G10 Owner's Club, who have cards that can monitor the VRM temperatures to run some tests.  Simple before and after tests. There are two reviews that I will point to. The Asus DCUII GTX 780Ti, Post #1147 and a reference R9 290 Post #1486 & Post #1500

   You will see that the guy with the Asus GTX 780Ti was getting DCUII temperatures of 86C on the Core and 104C on the VRMs, his card was overclocked.  You can look at the link to see specifics of his OC, nothing major, very average.  Once he did the G10 modification without extra heat sinks, he was able to further overclock his card, and temperatures dropped to 49C on the Core, and 81C on the VRMs.  This tells us two things.  #1, the VRMs can withstand a lot of heat, #2 the G10's Fan does an excellent job of cooling the VRMs on its own. 

     There was another test done with an R9 290.  They went as far as to do a before and after test with heat sinks on their G10. that they went from something like 90C on the core with the reference cooler all the way down to 55C with G10 + H55. VRMs went from 110C down to 75C.  Most importantly, when they added heat sinks the VRM temperatures dropped from 75C to 60C.  So adding heat sinks does help out if you are concerned about VRM temperatures. This was with a reference R9 290, so it was a naked card without mid-plate to passively help cool VRAM and VRMs.  These are just two examples, if you read through that thread, there are many more.

      In summation,VRM temperatures are not a concern as long as you don't run a dangerous program like Kombustor/Furmark.  For awhile now people have been thinking that its a hardware problem, when in reality, its a software problem.  You will be perfectly fine buying just a Kraken G10 + AIO of your choice without extra heat sinks.  Any extras you decide to buy are optional.  Here is a list of my optional extras you might want to buy to go along with your G10:

 

1 Pack of Cosmos mini Aluminum VGA Heatsinks(20pc) - $6 You can place these on your VRMs for peace of mind if you like.  It is what I did, even with a card that has a mid-plate.  My VRMs have never exceeded 54C according to my NZXT Sentry 3 Fan Controller.

 

1 Pack of Sekisui Double-Sided Adhesive Thermal Heatsink Tape(2pc) - $3 The heat sinks come with thermal tape pre-applied, but it is not very good quality and people report them falling off.  Buy this high end thermal tape to get better heat transfer, and to make sure your heat sinks stay on.  You will have to remove the old pre-applied tape before using the good tape.

 

Aftermarket Thermal Compound, it comes with the AIO Cooler, but if you want to buy some performance stuff to shave off an additional 1-4C buy your own.  Can also be used on your CPU.  It is also nice to have just in case you need to remount anything.  The Cooler Master Seidon 240M actually comes with a tube that you can apply yourself, so you will have some extra.  This TIM is good, but definitely not "performance" stuff.  I recommend Gelid GC-Extreme - $11

 

 VGA to PWM Fan Adapter - $5  This Fan Adapter is very nice because it allows you to plug the G10's 92mm fan directly into the GPU instead of into the valuable motherboard fan headers.  Once plugged into the video card, you can control the fan speed in MSI Afterburner.  Set it to a constant speed, not a fan curve.  The reason being is that the temperature is derived from the core.  The fan is for the VRMs.  Set it to a constant value of close to 100%.

 

Cooler Master Blade Master 120mm PWM Fan - $12  A 2nd 120mm fan for your H55 will go a long way in cooling performance, and noise.  Having two fans allows you to run them at lower speeds, keeping the noise down, while maintaining the same awesomely low temperatures as a single fan at high speed, or crank them both up for even better performance.

 

Anti-Kink Coils - $4 per tube.  One tube per AIO is necessary.  This is a pure aesthetics purchase, but it does make your tubing stand out and look really cool.  If you do buy the AIOs that I recommended, you will need 1/2 OD Anti-Kink Coils.  You have to get one size up OD in order to make it fit properly.  Adds a very cool effect to your case.

[asquirrel]

 

so today I went and brought an nzxt kraken g10 for my amd r9 290x for cooling i am using a used antec kuhler h20 620, the thing is that when i ran the valley bench mark i was getting 60fps with my gpu clocked at 1000mhz but the max temp on my gpu was 75c im not sure how hot all in one water coolers should get. is this temp ok or should i throttle it a bit?

 

The AIO is only 120x25mm. This would be considered a 'slim' 120mm radiator. Expecting this to dissipate more than 120W of heat is...a bit silly. Considering the 290x has a 290W TDP, you are seeing such high temps (though, in fairness, 75C is cooler than most air cooled GPUs run...) because your radiator is saturated. It simply cannot dissipate heat any faster than it already is, you need more surface area.

 

Said another way, you have a radiator which can be reasonably expected to get rid of 120W, maybe 150W, of heat. You are asking it to get rid of 290W of heat. Do you see the problem here?

[Faceman]

The AIO is only 120x25mm. This would be considered a 'slim' 120mm radiator. Expecting this to dissipate more than 120W of heat is...a bit silly. Considering the 290x has a 290W TDP, you are seeing such high temps (though, in fairness, 75C is cooler than most air cooled GPUs run...) because your radiator is saturated. It simply cannot dissipate heat any faster than it already is, you need more surface area.

 

Said another way, you have a radiator which can be reasonably expected to get rid of 120W, maybe 150W, of heat. You are asking it to get rid of 290W of heat. Do you see the problem here?

There are a slew of people on the OCN G10 Owner's Club with H55s on their overclocked R9 290Xs getting 55-60C.  My GTX 780 is a 250W TDP GPU, and my temperatures are in the 40s, never exceeding 49C.  The problem is not the radiator size, but likely the mounting.  It is not making proper contact with the GPU die would be my guess.

[asquirrel]

There are a slew of people on the OCN G10 Owner's Club with H55s on their overclocked R9 290Xs getting 55-60C.  My GTX 780 is a 250W TDP GPU, and my temperatures are in the 40s, never exceeding 49C.  The problem is not the radiator size, but likely the mounting.  It is not making proper contact with the GPU die would be my guess.

I...highly doubt this. 40-60 during idle, sure. Under full load? No...I highly doubt that. Unless they are in an ice cold room / their house temp is around 50F. Physics says "unlikely".

[Faceman]

I...highly doubt this. 40-60 during idle, sure. Under full load? No...I highly doubt that. Unless they are in an ice cold room / their house temp is around 50F. Physics says "unlikely".

Go to the OCN G10 Owner's Club and see for yourself. 50-60C is the average temp for 780s and R9 290s under gaming load. With the 780 skewing lower and the R9 290 higher.  Idle temperatures are in the mid 20s to mid 30s.

 

Here is an excerpt from my How To Water Cool your CPU and GPU for Under $200 USD Guide.  These are just two examples, but I have them on hand without having to search.

 

  "To give you an idea of what VRM temperatures are like for those of us at home, we asked members of the OCN G10 Owner's Club, who have cards that can monitor the VRM temperatures to run some tests.  Simple before and after tests. There are two reviews that I will point to. The Asus DCUII GTX 780Ti, Post #1147 and a reference R9 290 Post #1486 & Post #1500

   You will see that the guy with the Asus GTX 780Ti was getting DCUII temperatures of 86C on the Core and 104C on the VRMs, his card was overclocked.  You can look at the link to see specifics of his OC, nothing major, very average.  Once he did the G10 modification without extra heat sinks, he was able to further overclock his card, and temperatures dropped to 49C on the Core, and 81C on the VRMs.  This tells us two things.  #1, the VRMs can withstand a lot of heat, #2 the G10's Fan does an excellent job of cooling the VRMs on its own. 

     There was another test done with an R9 290.  They went as far as to do a before and after test with heat sinks on their G10. that they went from 90C on the core with the reference cooler all the way down to 55C with G10 + H55. VRMs went from 110C down to 75C.  Most importantly, when they added heat sinks the VRM temperatures dropped from 75C to 60C.  So adding heat sinks does help out if you are concerned about VRM temperatures. This was with a reference R9 290, so it was a naked card without mid-plate to passively help cool VRAM and VRMs.  These are just two examples, if you read through that thread, there are many more.

      In summation,VRM temperatures are not a concern as long as you don't run a dangerous program like Kombustor/Furmark.  For awhile now people have been thinking that its a hardware problem, when in reality, its a software problem."

 

I am an avid member on the OCN G10 Owner's Club, and I can assure you that the average temperature for R9 290(X) users is 55-60C under load for people all over the world.  This also falls in line with the results that were achieved during the most critical review of the G10.

[asquirrel]

-snip-

You are debating physics with heresay and forum content. The G10 isn't a cooler, it's a bracket. If the people in that club were using 240mm radiator coolers or larger, or even the 120mmx80mm H80 from Corsair, I'd buy it. But expecting a 290x to be cooled to 55C with a 120mm radiator is just plain silly without a room *far* colder than I would consider normal (~70F). Here, have some fizzicks to ruin your Faux News style 'fact' spouting:

 

http://skinneelabs.com/hwlabs-gtx-360-radiator-review-data-supplement/3/

http://skinneelabs.com/swiftech-mcr320-radiator-review/3/

 

Two *very* good 360mm radiators paired with decent fans, using designs that were considered leading edge 2 years ago. The peak result is 518W of cooling. Do the math, that's 1.438 (repeating) Watts of heat per linear mm {1.21 Watts per linear mm for the 34mm thick rad}. And that was with a radiator that was 54mm thick, almost twice the thickness of the OP's radiator. And using fans more powerful than OPs, and using a pump more powerful than OP's.

 

There is no f***ing way that a radiator that small could remove 300W of heat. Your G10 owner's club members are either full of lies, not running their cards at peak output when they quote those numbers, or have not run the test long enough (10m or so) to fully heat all components of the loop.

 

The third is the one I find most likely. It takes a few minutes to fully heat soak a water cooling loop. But, sir, Physics says UNLIKELY!

[Faceman]

You are debating physics with heresay and forum content. The G10 isn't a cooler, it's a bracket. If the people in that club were using 240mm radiator coolers or larger, or even the 120mmx80mm H80 from Corsair, I'd buy it. But expecting a 290x to be cooled to 55C with a 120mm radiator is just plain silly without a room *far* colder than I would consider normal (~70F). Here, have some fizzicks to ruin your Faux News style 'fact' spouting:

 

http://skinneelabs.com/hwlabs-gtx-360-radiator-review-data-supplement/3/

http://skinneelabs.com/swiftech-mcr320-radiator-review/3/

 

Two *very* good 360mm radiators paired with decent fans, using designs that were considered leading edge 2 years ago. The peak result is 518W of cooling. Do the math, that's 1.438 (repeating) Watts of heat per linear mm {1.21 Watts per linear mm for the 34mm thick rad}. And that was with a radiator that was 54mm thick, almost twice the thickness of the OP's radiator. And using fans more powerful than OPs, and using a pump more powerful than OP's.

 

There is no f***ing way that a radiator that small could remove 300W of heat. Your G10 owner's club members are either full of lies, not running their cards at peak output when they quote those numbers, or have not run the test long enough (10m or so) to fully heat all components of the loop.

 

The third is the one I find most likely. It takes a few minutes to fully heat soak a water cooling loop. But, sir, Physics says UNLIKELY!

Then how in the world am I only getting 52C max on my 250W GTX 780 SC with a 120mm, 30mm thick radiator with two fans in push pull at a moderate 1500rpm?  This is the result I am getting in a 23.8C room.

 

The H80i isn't compatible with the G10, and its not 80mm thick.

 

Physics are laws, but something must be unaccounted for because this is the result that everyone is getting who is doing the G10 modification with 120mm radiators.  The H55 is the most commonly used AIO in the club and whether it be a 280, 770, 290 or 780, their temperatures fall between 50 and 60C from all over the world.  Those who do use 240mm coolers are only getting marginal better temperatures, 6-8C better than those of us with 120mms.

 

Since you wont take my, or 100 other users' words for it, how about some credible reviewers?

http://www.anandtech.com/show/7719/nzxt-kraken-g10-review-liquid-cooling-for-your-gpu

http://www.pugetsystems.com/labs/articles/NZXT-Kraken-G10-Review-527/

http://www.kitguru.net/components/graphic-cards/henry-butt/nzxt-kraken-g10-liquid-gpu-cooling-bracket-review/

http://www.tweaktown.com/reviews/6318/nzxt-kraken-g10-liquid-cooling-gpu-mounting-kit-review/index6.html

http://www.overclock3d.net/reviews/cases_cooling/nzxt_kraken_g10_gpu_heatsink_review/1

http://www.legitreviews.com/nzxt-kraken-g10-gpu-water-cooler-review-on-an-amd-radeon-r9-290x_130344/4

http://www.overclockersclub.com/reviews/kraken_g10/4.htm

http://www.brightsideofnews.com/2014/09/16/nzxt-kraken-g10-gpu-bracket-provides-great-cooling-solution/

http://techgage.com/article/nzxt-kraken-g10-gpu-bracket-review/2/

http://www.hitechlegion.com/reviews/cooling/liquid/38158-nzxt-kraken-g10-liquid-cooling-gpu-adapter?showall=&start=2

 

Clearly, everyone has just found a way to defy the laws of physics.

 

I would also link you Linus' review and unboxing, but he inexplicably doesn't show temperature results, just how much FPS you can gain on a 290 by overclocking it further.  Really a ridiculous video.

[asquirrel]

Then how in the world am I only getting 52C max on my 250W GTX 780 SC with a 120mm, 30mm thick radiator with two fans in push pull at a moderate 1500rpm?  This is the result I am getting in a 23.8C room.

 

The H80i isn't compatible with the G10, and its not 80mm thick.

 

Physics are laws, but something must be unaccounted for because this is the result that everyone is getting who is doing the G10 modification with 120mm radiators.  The H55 is the most commonly used AIO in the club and whether it be a 280, 770, 290 or 780, their temperatures fall between 50 and 60C from all over the world.  Those who do use 240mm coolers are only getting marginal better temperatures, 6-8C better than those of us with 120mms.

 

Since you wont take my, or 100 other users' words for it, how about some credible reviewers?

http://www.anandtech.com/show/7719/nzxt-kraken-g10-review-liquid-cooling-for-your-gpu

http://www.pugetsystems.com/labs/articles/NZXT-Kraken-G10-Review-527/

http://www.kitguru.net/components/graphic-cards/henry-butt/nzxt-kraken-g10-liquid-gpu-cooling-bracket-review/

http://www.tweaktown.com/reviews/6318/nzxt-kraken-g10-liquid-cooling-gpu-mounting-kit-review/index6.html

http://www.overclock3d.net/reviews/cases_cooling/nzxt_kraken_g10_gpu_heatsink_review/1

http://www.legitreviews.com/nzxt-kraken-g10-gpu-water-cooler-review-on-an-amd-radeon-r9-290x_130344/4

http://www.overclockersclub.com/reviews/kraken_g10/4.htm

http://www.brightsideofnews.com/2014/09/16/nzxt-kraken-g10-gpu-bracket-provides-great-cooling-solution/

http://techgage.com/article/nzxt-kraken-g10-gpu-bracket-review/2/

http://www.hitechlegion.com/reviews/cooling/liquid/38158-nzxt-kraken-g10-liquid-cooling-gpu-adapter?showall=&start=2

 

Clearly, everyone has just found a way to defy the laws of physics.

 

I would also link you Linus' review and unboxing, but he inexplicably doesn't show temperature results, just how much FPS you can gain on a 290 by overclocking it further.  Really a ridiculous video.

Wow...you really can't be bothered to think can you?

 

http://www.anandtech.com/show/7719/nzxt-kraken-g10-review-liquid-cooling-for-your-gpu

Manages 50C on a GTX 770, which has a TDP of 230W according to http://www.anandtech.com/show/6994/nvidia-geforce-gtx-770-review 1.438 * 140 = 201. So, The Kraken X40 should be able to dissipate 201W of heat and will become heat-soaked after that, so yea, asking a radiator capable of getting rid of 200W of heat to get rid of 230...totally reasonable!

 

http://www.pugetsystems.com/labs/articles/NZXT-Kraken-G10-Review-527/

They get 59C running a comparable test (Furmark vs cryptocoin mining) using the same Kraken X40. 10C for an extra 100W of heat to dissipate, is that reasonable? Let's go to http://www.extreme.outervision.com/tools.jsp and plug in the numbers: CPU wattage: 290, Case temp 22, C/W: .130 (from the first test by anandtech). The result? Expected temp 60C! Wow! Science!

 

http://www.kitguru.net/components/graphic-cards/henry-butt/nzxt-kraken-g10-liquid-gpu-cooling-bracket-review/

Shows 62C for the G10 w/X40 combo. Considering the result above, no need to argue this one, as its result is in line.

 

http://www.tweaktown.com/reviews/6318/nzxt-kraken-g10-liquid-cooling-gpu-mounting-kit-review/index6.html

They don't mention the temperature of their room. Their result is 20C cooler than Kit and Puget got! Oh noes! I'm wrong...wait, wait, no I'm not. They used Uniengine, and in Puget's review that scored at 40C as well. Oh, damn, yet another result that agrees with my statement that a 120mm fan cannot be expected to cool 300W of heat at 50C.

 

I could keep going, but I'm 4/4. Does anyone need me to pick apart the rest? 

 

Now, as for OP @HectiKCookiE , you have an H20 620 cooler, what is reasonable cooling performance? Well, from this test: http://www.tomshardware.com/reviews/h2o-h80-h100-benchmark-overclocking,3084-8.html we can see that on a CPU with a 148W TDP (computed based on 4.0 GhZ @ 1.388V here: http://www.extreme.outervision.com/tools.jsp) that you get a 50C *DELTA*. This means if the room is 22C, then your CPU would be at 72C. Using that same outervision page, we compute that the °C/W for that cooler is 0.3378

 

Quite frankly, that's a pathetic result. From Faceman's first review we know that .33 is acceptable for air coolers, but not for water coolers. But, let's continue our math. 290W, 22C case, .3378 °C/W performance, expected CPU temp: 120C.

 

Hm. Something must be wrong. More reviews! http://www.overclockersclub.com/reviews/antec_kuhler_h20_620_review/4.htm Ah, ok, tomshardware says their temperatures are deltas, but they are, in fact, absolute temps. Let's try our math again.

 

If we recompute the i7-870 using 50C instead of 72C (absolute temp vs delta temp) we get 0.1892 °C/W, this is much more reasonable for a water cooling solution. Not awesome, but not terrible. Let's continue our math. Using 0.1892 °C/W on a 290W heat source, we see our expected temperature in a 22C room is...76.9C. Considering you were getting 75C, I'd say this pretty much says exactly what I said in the beginning: your radiator is too small to remove that much heat from the GPU.

 

Faceman, we all make mistakes, but you are wrong in this case. Lay down your arms and embrace education.

[Faceman]

 

Wow...you really can't be bothered to think can you?

 

http://www.anandtech.com/show/7719/nzxt-kraken-g10-review-liquid-cooling-for-your-gpu

Manages 50C on a GTX 770, which has a TDP of 230W according to http://www.anandtech.com/show/6994/nvidia-geforce-gtx-770-review 1.438 * 140 = 201. So, The Kraken X40 should be able to dissipate 201W of heat and will become heat-soaked after that, so yea, asking a radiator capable of getting rid of 200W of heat to get rid of 230...totally reasonable!

 

http://www.pugetsystems.com/labs/articles/NZXT-Kraken-G10-Review-527/

They get 59C running a comparable test (Furmark vs cryptocoin mining) using the same Kraken X40. 10C for an extra 100W of heat to dissipate, is that reasonable? Let's go to http://www.extreme.outervision.com/tools.jsp and plug in the numbers: CPU wattage: 290, Case temp 22, C/W: .130 (from the first test by anandtech). The result? Expected temp 60C! Wow! Science!

 

http://www.kitguru.net/components/graphic-cards/henry-butt/nzxt-kraken-g10-liquid-gpu-cooling-bracket-review/

Shows 62C for the G10 w/X40 combo. Considering the result above, no need to argue this one, as its result is in line.

 

http://www.tweaktown.com/reviews/6318/nzxt-kraken-g10-liquid-cooling-gpu-mounting-kit-review/index6.html

They don't mention the temperature of their room. Their result is 20C cooler than Kit and Puget got! Oh noes! I'm wrong...wait, wait, no I'm not. They used Uniengine, and in Puget's review that scored at 40C as well. Oh, damn, yet another result that agrees with my statement that a 120mm fan cannot be expected to cool 300W of heat at 50C.

 

I could keep going, but I'm 4/4. Does anyone need me to pick apart the rest? 

 

Now, as for OP @HectiKCookiE , you have an H20 620 cooler, what is reasonable cooling performance? Well, from this test: http://www.tomshardware.com/reviews/h2o-h80-h100-benchmark-overclocking,3084-8.html we can see that on a CPU with a 148W TDP (computed based on 4.0 GhZ @ 1.388V here: http://www.extreme.outervision.com/tools.jsp) that you get a 50C *DELTA*. This means if the room is 22C, then your CPU would be at 72C. Using that same outervision page, we compute that the °C/W for that cooler is 0.3378

 

Quite frankly, that's a pathetic result. From Faceman's first review we know that .33 is acceptable for air coolers, but not for water coolers. But, let's continue our math. 290W, 22C case, .3378 °C/W performance, expected CPU temp: 120C.

 

Hm. Something must be wrong. More reviews! http://www.overclockersclub.com/reviews/antec_kuhler_h20_620_review/4.htm Ah, ok, tomshardware says their temperatures are deltas, but they are, in fact, absolute temps. Let's try our math again.

 

If we recompute the i7-870 using 50C instead of 72C (absolute temp vs delta temp) we get 0.1892 °C/W, this is much more reasonable for a water cooling solution. Not awesome, but not terrible. Let's continue our math. Using 0.1892 °C/W on a 290W heat source, we see our expected temperature in a 22C room is...76.9C. Considering you were getting 75C, I'd say this pretty much says exactly what I said in the beginning: your radiator is too small to remove that much heat from the GPU.

 

Faceman, we all make mistakes, but you are wrong in this case. Lay down your arms and embrace education.

 

You're 4/4 in the fact that all of those tests fall within 50-60C which is what I said would happen.  I said the 780s skew closer to 50 and the 290s skew closer to 60.  Also, those tests are stressing the systems to the absolute max with unsafe tests like Furmark.  Gaming results aren't going to come close to that. I dunno about you, but when I game, I usually have V sync on which puts a limit on the amount of GPU usage.  Or in the games that do stress my system to 99% load, I am not exceeding 52C even after a lengthy play session.  In those tests it sounds like they are only using 1 fan.  Add in a 2nd fan, which is what a lot of users do and temps drop even further.

 

Considering the OP is running Unigine Valley, it isn't stressing his system to the max and shouldn't be hitting that 75C.  He needs to either tighten the screws, or there is a possible gap caused by a midplate, in which case a copper shim and TIM on either side should do the trick.  Maybe his pump isn't working?

[asquirrel]

1. I said the 780s skew closer to 50 and the 290s skew closer to 60. 

 

2. In those tests it sounds like they are only using 1 fan.  Add in a 2nd fan, which is what a lot of users do and temps drop even further.

 

3. Considering the OP is running Unigine Valley, it isn't stressing his system to the max and shouldn't be hitting that 75C.  He needs to either tighten the screws, or there is a possible gap caused by a midplate, in which case a copper shim and TIM on either side should do the trick.  Maybe his pump isn't working?

 

1. Right. Because a Antech H20 620 and a Kraken X40 are similar enough to warrant a direct comparison in terms of thermal performance. *rolls eyes* Nevermind I did the math to prove my correctness on the thermal mathmatics.
2. Push-pull only helps if your fans lack sufficient static pressure to maintain flow through the radiator. I mean, it's totally possible it would help him, but it's not the silver bullet people make it out to be. That's why push-pull with noctuas or other static pressure fans is a waste, in all but 25+FPI thick rads the Noctuas have sufficient static pressure to maintain flow.
3. Ok, I missed that part.

@HectiKCookiE Check that you didn't use too much thermal paste. I assume you are using the pre-applied stuff? If so, then thickness shouldn't be an issue. I say this because I ended up replacing the thermal paste on my 660TI and it got me about 10C (air cooling, stock paste vs 4 year old NH-1). I chalk most of that gain up to the sheer thickness of paste EVGA put on their card.

 

If you need to clean TIM off, use 90% or higher isoproyl alcohol. If you are in the US, a Walmart pharmacy carries this, white/clear bottle with a red label, costs about $1. DO NOT DRINK IT! It's not the 'good' alcohol...in case you got your hopes up.

 

Beyond that, yea, screw tightness, orientation...all valid things to check. Maybe you have a kink in your hosing?

[Faceman]

 

1. Right. Because a Antech H20 620 and a Kraken X40 are similar enough to warrant a direct comparison in terms of thermal performance. *rolls eyes* Nevermind I did the math to prove my correctness on the thermal mathmatics.
2. Push-pull only helps if your fans lack sufficient static pressure to maintain flow through the radiator. I mean, it's totally possible it would help him, but it's not the silver bullet people make it out to be. That's why push-pull with noctuas or other static pressure fans is a waste, in all but 25+FPI thick rads the Noctuas have sufficient static pressure to maintain flow.
3. Ok, I missed that part.

@HectiKCookiE Check that you didn't use too much thermal paste. I assume you are using the pre-applied stuff? If so, then thickness shouldn't be an issue. I say this because I ended up replacing the thermal paste on my 660TI and it got me about 10C (air cooling, stock paste vs 4 year old NH-1). I chalk most of that gain up to the sheer thickness of paste EVGA put on their card.

 

If you need to clean TIM off, use 90% or higher isoproyl alcohol. If you are in the US, a Walmart pharmacy carries this, white/clear bottle with a red label, costs about $1. DO NOT DRINK IT! It's not the 'good' alcohol...in case you got your hopes up.

 

Beyond that, yea, screw tightness, orientation...all valid things to check. Maybe you have a kink in your hosing?

 

I am sure your math is correct, I'm not doubting that because you have far more experience with that than me.  I am the one with the G10 and an avid member of the G10 Club on OCN who has seen and dealt with a myriad of issues regarding the G10.  I have more experience in that department.  The problem I see is that the load placed on the GPU is different in games and in synthetics.  Furmark for example will cause insane temperatures because it stresses the GPU so much, it even breaks cards in some instances.  Compare that to a game that maxes out the GPU and there is a big difference in temperature.  Even though both the synthetic and the game push the GPU to 99% usage, the synthetic causes a lot more heat.  I'm sure that at the absolute maximum amount of load put on a GPU, like Furmark then your math is correct, but games don't push GPUs that much and its rarely, if ever as constant as a torture test like Furmark.

 

Push pull helps tremendously.  I use Cooler Master Blade Masters, a very underrated fan with lots of static pressure.  They are loud at high RPMs, but work wonders in the 1,200-1,600rpm range while being very quiet.  Adding a 2nd fan dropped my temperatures by 4C when they are both running at 1,500rpm compared to one fan at 1,500rpm.  I also forgot to mention that I do use Gelid GC-Extreme, a very good thermal paste. In the G10 Club some do and some don't use aftermarket paste, but it doesn't cause wildly inconsistent results with or without.  I don't know how many C it took off compared to the stock thermal paste because I used Gelid from the start, but we can probably assume 1-2C conservatively.

 

Thickness can be an issue, which is why I asked which specific R9 290X.   Sometimes cards have mid plates, like mine, which interfere with the copper plate of the AIO making proper contact with the GPU die.  A copper shim is then necessary in order to bridge the gap.  TIM must be applied to both sides.

 

First picture is a picture of the mid plate on an EVGA GTX 780 SC.  The 4 prongs that extend towards the GPU die are what causes the cooler to not make proper contact.  You can either cut off the 4 prongs, which voids your warranty, or you can use a $3 copper shim.

Second picture is of the copper shim I use on my EVGA 780 SC to accommodate for the gap between the copper plate and the die.  You can also see the TIM on the underneath.  Very few GPUs require the use of a shim, but in the ones that do, this is an easy and quick fix.

 

@HectiKCookiE  Have you tried the recommendations we have made?

[kainobean]

i have 2 r290x's with the kraken x41 as the core cooler and a modified stock heatsink base without the fan or the fins as the vrm and ram cooling. my core temps at 1100mhz is between 45-50c while my vrms never reach over 60c. im quite happy so far with this setup. but the g10 mounting method is a pain in the butt i had issues with my temps being very high in the 60-80c range and i found that because of the super complicated mounting method that requires more than 2 hands to do properly i had a bad mount and just had to fiddle with ti till i got a proper mount and then finally got super low temps like i was looking for... also careful with the g10 screws as they strip very easily or they did for me and i had to use pliers to tighten them cause the heads are all destroyed after one use... dunno if any of this helps and was to lazy to read everything others said cause they said a ton and mines probably no better since i type like a mess...