ideas to bring down temps even more

[howzz1854]

hey guys, registered here years ago, but never really bothered to post. always just quietly observed the forum.

recently went through an overhaul and upgraded to the 3770K and was able to clock it up to 4.8ghz at 1.37v. recently retired the OLD D4 pump and finally got rid of that jet engine pitch noise, and got myself a D5, running at power level 5. also got rid of some sub par fans on my rad and replaced them with 7 NB eLoop b-12 P PWM fans. as you can see from my sig, the radiators are still the same (one GTS 360 rad + one GTS 120 rad). the block is koolance cpu380.

before upgrading the pump and the fans, i was getting low 70's (71~72 degrees C) in X264 encoding, which is about the most intensive app i use on my rig, but for stress testing Prime95 V27.9 small FFT 12K with AVX reaches 93~95 degrees C at ambiemt temp of about 18~22C.

after the pump and the new eloop fans, at around the same range of ambient temp 18~22C, i am seeing some improvement of high 60's low 70's in X264 (70~71 degrees C), and small FFT Prime V27.9 @ 12K peaks 90C at roughly 18C ambient and 92C at 22C Ambient.

again, i apologize for the wide range of estimate, but if i have to sum it up, comparing before vs after pump + fan upgrade, in best case scenario, i got a 5 degree performance improvement. but basically i believe i am getting anywhere between 2~5 degree of temp improvement average from upgrading the pump and the fans, factoring in the range of ambient temps. i should mention that i have all the eLoop fans setup through PWM in push + Pull so it throttles to full 2000rpm during load and quietly operating 1000rpm at idle. this is exactly what i wanted.

i understand that the inherent problem with HOT 3770K lies within the TIM used under the HIS. but since i am not ready to take the plunge just yet on delidding. is there anything else i can do to even bring the temp down by a few degrees. do you think adding another 120mm rad will lower the water temp more? i am also looking at getting the EK Supremacy, since the review here showed that in socket 1155, the Koolance CPU380 performed really poor in comparison to the Supremacy. then i am not sure if i will see that 5 extra degrees of improvement according to the review if i were to upgrade.

or i could buy more powerful fans, and buy another round of GT AP-29 PWM fans that run at 3000rpm. how much temp improvement do you think i'll see in that?

[howzz1854]

basically what i am asking is will it bring down temps more if i

1. get a new block (EK Supremacy CSQ)
or
2. add another 120mm rad in the loop
or
3. upgrade to AP29 3000rpm GT fans.

or have i hit the limit on the 3770K without delid.

[k0ST]

I'm no expert, but IMHO you are OK on the temps. Delidding seems like only option to reduce temperatures some more, and adding more stuff to your loop will bring diminishing returns (2-3 C max). I would experiment with lower pump speeds - it may yield some improvement...

[=[PULSAR]=]

I think you've hit the limit of your 3770k. Just a reference I'm running a 3770k @ 4.8ghz 1.315 vcore 100% load from crunching with a h110/stock fans at 7v ranges from high 60's to low 70's.

EDIT: If the cpu is the only thing in your loop you can run it as low at 1 on the pump remember the higher the speed the more heat is dumped into the loop.

[howzz1854]

thanks for the inputs guys.

my radiators are actually very restrictive. i think it's one of the most restrictive rads out there. although i am only cooling the cpu, the rads are rated at something like over 1psi each according to various sources at 1GPM flow rate, so my whole loop is actually somewhat of a medium flow.

however your inputs are valuable, that's exactly what i was looking for. the last thing i wanna do is to go through all the trouble and only see 1 degree of difference. i guess i'll really have to look into deliding in the future. although like i said, about the most intensive thing i do on my comp is encoding with X264, which btw is actually quite a good stress test, and it peaks to 70c, i know that's still a far cry from the safety range of 90~100c. i guess this hobby just has me obsessed with seeking more improvement. i was just glad that i was able to bring the peak temp of prime95 V27.9 small FFT down from 95c to 90c, that was quite an accomplishment. even tho in X264 i only saw roughly a 2~3 degree improvement from 72c to 70c.

if anyone else who have had experience with both CPU380 block and EK supremacy on socket 1155, particularly with the 3770K, i would love ot hear more.

[howzz1854]

I think you've hit the limit of your 3770k. Just a reference I'm running a 3770k @ 4.8ghz 1.315 vcore 100% load from crunching with a h110/stock fans at 7v ranges from high 60's to low 70's.

EDIT: If the cpu is the only thing in your loop you can run it as low at 1 on the pump remember the higher the speed the more heat is dumped into the loop.

btw, that's quite an impressive chip you got there. looks like someone won the chip lottery there. for me 4.7ghz was my wall. 4.7 only required 1.23v, but from there to 4.8 i needed 1.37v to make Prime95 stable. as you can imagine temps pretty much shot up from there. lol

[=[PULSAR]=]

btw, that's quite an impressive chip you got there. looks like someone won the chip lottery there. for me 4.7ghz was my wall. 4.7 only required 1.23v, but from there to 4.8 i needed 1.37v to make Prime95 stable. as you can imagine temps pretty much shot up from there. lol

Yeah got it in the for sale forums here, I guess someone wanted more from it but I'm happy did 4.7@1.28, 4.8@1.315, 4.9@1.36 I've had it over a year now running 24/7 with no hiccups. Then again I have a 4770k that only does 4.7@1.41v so they arent all golden

[howzz1854]

Yeah got it in the for sale forums here, I guess someone wanted more from it but I'm happy did 4.7@1.28, 4.8@1.315, 4.9@1.36 I've had it over a year now running 24/7 with no hiccups. Then again I have a 4770k that only does 4.7@1.41v so they arent all golden

4.7 for 4770K is actually quite impressive, 200mhz higher than most can get out of their 4770K. http://hwbot.org/hardware/processor/core_i7_4770k/

let me know next time you're about to go buy lottery tickets, i'll come with you.

[=[PULSAR]=]

That's with a beefy watercooling setup too though average 70-75C

[OldChap]

If you are able to measure your water temp and your ambient temp as well as current cpu temps, I will say this:

I have run a massive rad and fan which would bring the water temps down to within 1 degree of ambient but the cost of doing such a build is not cheap quite apart from the fact that the rad is no longer in your case. I would guess that it would be cheaper to buy another cpu and de-lid or get someone with previous experience to do so. If you know your water temp now you can see how much gain you could get by subtracting air temp plus say 1-3 degrees from the water temp.

This kind of thing is more suited to those of us that run 24/7 with multiple rigs. For you to see a significant gain you either need to de-lid or to invest in a chiller.

[howzz1854]

If you are able to measure your water temp and your ambient temp as well as current cpu temps, I will say this:

I have run a massive rad and fan which would bring the water temps down to within 1 degree of ambient but the cost of doing such a build is not cheap quite apart from the fact that the rad is no longer in your case. I would guess that it would be cheaper to buy another cpu and de-lid or get someone with previous experience to do so. If you know your water temp now you can see how much gain you could get by subtracting air temp plus say 1-3 degrees from the water temp.

This kind of thing is more suited to those of us that run 24/7 with multiple rigs. For you to see a significant gain you either need to de-lid or to invest in a chiller.

hummmm.... good info. maybe i'll stick a thermometer inside the reservoir and find out. would you recommend just any thermometer would do? i don't wanna cause any adverse chemical/metallic reaction to the loop.

also, when you said add 1~3 degrees, do you mean add 1~3 degrees on top of the water temp and that's the core temp? because i've always been under the impression that the core temp is a far cry from the actual water temp. i thought the two were at least 20~30 degrees apart. because even assuming that my current ambient temp is roughly 20C, which is a known factor, but if i had to take a WILD guess, the water temp is probably 35C or 40C, again i have no idea but the air coming from the rads are actually very cool, and i know that my load temp one the hottest core is 70C during X264 encoding. that leaves 30C of difference.

[howzz1854]

never mind... i just answered my own question.

http://www.xtremesystems.org/forums/...blocks-roundup

according to that review, the delta between cpu and water for a 3770K is 40~50 degrees c. so i am guessing my water temp is probably quite close to ambient, which makes sense because the air coming out of the rad are very cool. i'll see if i can go get a thermometer from Lowes tomorrow and measure the water temp. what kind of thermometer would you guys recommend.

[OldChap]

For something like this sort of measurement cheap and simple that that fits in res..... and scaled to read around ambient temp

I find these useful:

http://www.amazon.co.uk/Digital-Aqua...l+thermometers

[howzz1854]

maybe i should go to petco instead of Lowes. lol

[Justintoxicated]

you could clock down your CPU a little, or use a better TIM.

[howzz1854]

i think there's plans in the future to delid the chip and replace the TIM, but at the meantime, i am just not ready for that just yet.

[howzz1854]

I think you've hit the limit of your 3770k. Just a reference I'm running a 3770k @ 4.8ghz 1.315 vcore 100% load from crunching with a h110/stock fans at 7v ranges from high 60's to low 70's.

EDIT: If the cpu is the only thing in your loop you can run it as low at 1 on the pump remember the higher the speed the more heat is dumped into the loop.

btw, i am just curious, what's your load temp in Prime V27.9 small FFT. since your voltage requirement is so much lower than mine, i am guessing that you probably have a 8~10 degree temp advantage on me.

[howzz1854]

ok so i stopped by the local pet store on the way home from work. and picked up a conventional aquarium analogue thermometer, i know i know.. it's not digital. but at least it'll give me a rough estimate.

so i stuck the thermometer into into the reservoir and recorded the idle water temperature, also took note of the idle core temp in Realtemp, and also took note of the room ambient temp using another analogue thermometer that's 4' away from my comp. and from there on, i fired up X264 render, recorded the load water temperature, recorded the load core temperature; subsequently from there on fired up Prime95 V27.9 Small FFT and let it ran for 20 mins and recorded the load water temperature, and load core temperature from real temp. so here are the results. and i must say i am quite surprised.

based on this, i guess it's pretty conclusive that there pretty much isn't much i can do other than deliding the damn thing and replace the TIM that's underneath the IHS. at least i know that my radiators, fans, and pump are working as they should. my water temp is only 4.5?c higher than ambient, and what's really surprising was that even during load, my water temp only rose by 1~1.5?c, i guess thanks to those PWM fan profile i setup, that all 7 eLoop fans spool up automatically to full 2000rpm in push and pull during load. what's even more surprising was that my idle core temp are almost same as ambient. 22.5?c vs 21?c.

[Martinm210]

A simple test. Feel the rad with your hand when loading. if it just feels mildly warm your rad/fans are doing fine and not much to gain there. measuring water vs air temp would be better, what you can gain is a fration of that difference. Push plus pull 2000 rpm 360 is pretty powerful. most blocks also only a few degrees apart. Delid is probably going to yield the greatest benefit, but if itis stable .. is that needed?

[OldChap]

air temp ~21.5 water temp at load 26.5. difference 5 deg which is already quite good

water is never going below ambient even with a HUGE rad so going the upgrade rads and fans route I do not expect you to gain any more than ~4 degrees.... probably less. A Phobya Extreme Supernova would probably gain you ~3 degrees

Now, as Martin says, you need to decide if you really want lower temps enough to de-lid.

[howzz1854]

thanks guys, appreciate all the input. i guess i'll settle for what i got for now. and maybe, just maybe in the future when i get bored i'll delid the chip. but thanks to you all, i was able to do this experiment, instead of plunging down more cash on upgrades that probably won't make any difference.

[howzz1854]

i thought i'd give everyone an update.

i decided to delid the chip after owning it for 6 months. after hammering it with a hammer and a vise, for a min i thought i was doing it wrong, the damn lid wouldn't come off no matter how hard i hit it. even all four sides of the wood block i used got dented.

eventually the chip went flying and the IHS came off. after applying some Liquid Pro, which btw is a pain. i recommend Liquid Ultra instead.

my prime95 27.9 small FFT load temp dropped more tan 20 degrees, and this is in the summer time, compare to before in my previous post, the experiment was conducted in late winter. so i have no doubt the temp will drop even more in the winter. if i had to guess, the different in temp improvement is probably somewhere around 25~30c.

i can now prime at low 70's in small FFT at 4.8ghz. as a result, i bumped up the voltage to 1.42 and was able to hit 4.9ghz. temperature still remains reasonable at high 70's and resting at around 80. but in everyday tasks like X264, the load temp never go above 63c in this hot summer southern california weather. i expect my typical load temp to be around low 50's in the winter at 4.9ghz.

on a separate note, i did however hit the limit for my motherboard VRM. the other day when the weather was really hot, even tho core temps remained at reasonable high 70's. i noticed the cpu frequency would dip back down to stock speed every few mins during small FFT prime at 4.9ghz 1.42v. then i read somewhere that people with similar Asrock boards, or even Asus boards, mostly lower end, and mid grade boards like my ASrock Extreme6/4. people reported VRM throttling after certain voltage and causing the VRM chip to overheat. i experienced this exact same thing. my VRM heatsink did get very HOT, although not to a point to cause 1st degree burn, but it was pretty hot. so far Prime95 small FFT is the only program that caused this VRM throttling.

thanks all for all your help. and hopefully this post will help others in the future.

[stren]

Yeah delidding is great assuming you can afford to live without a warranty. Thank goodness Intel got their stuff together with Devil's Canyon to finally improve things. I'm glad you didn't bother moving to the supremacy I really don't think that would gain you much if anything over the CPU-380, yes I've seen Moonman's data but I've also seen my 4770K data. The main advantage the supremacy has is with the naked mounting kit so that you can get the cpu block without using the IHS at all.

Martin's advice on feeling the radiator to detect approximately how close you are to ambient is a great idea also if you don't have a temperature sensor. Coolant and ambient temp sensors are pretty necessary though to really debug and improve loop performance!

[howzz1854]

Yeah delidding is great assuming you can afford to live without a warranty. Thank goodness Intel got their stuff together with Devil's Canyon to finally improve things. I'm glad you didn't bother moving to the supremacy I really don't think that would gain you much if anything over the CPU-380, yes I've seen Moonman's data but I've also seen my 4770K data. The main advantage the supremacy has is with the naked mounting kit so that you can get the cpu block without using the IHS at all.

Martin's advice on feeling the radiator to detect approximately how close you are to ambient is a great idea also if you don't have a temperature sensor. Coolant and ambient temp sensors are pretty necessary though to really debug and improve loop performance!

that's the thing. after i started this thread and realized that it's not as much the water block that's limiting me, but the TIM under the IHS. i figured i wouldn't gain much by changing out WB, probably at the most 2 degrees celcius.

when the weather started getting hot, one time my x264 crashed while i was gaming at the same time (the GPU hot air going up through the CPU radiator). that's when i got fed up, and decided to to go big and delid the damn thing.

i am hoping that after Devil's Canyon, Intel will bring soldering back. a better TIM is great, but nothing is better than metal to metal. that 6 degree difference from the old TIM to the new supposedly improved TIM is not all that impressive consider the difference of 25~30 degrees i just got from deliding.