DIY Corsair H50 killer...?

[LOUISSSSS]

i think i made the right decision going with the Noctua D14 over the H50, my idle temps are within 5 degrees C of my ambient temps of 68-70F

 

[Rubycon]

i think i made the right decision going with the Noctua D14 over the H50, my idle temps are within 5 degrees C of my ambient temps of 68-70F

Idle temperatures are utterly useless! What matters is the peak temperatures you see under full load. Always.

Certainly you will be nowhere near the thermal capacity of either cooler with an Athlon II!

 

[LOUISSSSS]

Idle temperatures are utterly useless! What matters is the peak temperatures you see under full load. Always.

Certainly you will be nowhere near the thermal capacity of either cooler with an Athlon II!

yep i can't seem to break 39c full load @ 1.45vCore 3.4ghz loaded with 1hr of linX

 

[BadOmen]

Idle temperatures are utterly useless! What matters is the peak temperatures you see under full load. Always.

Could you please elaborate? I didn't get that, and it's an interesting concept. I don't know, I have always seen idle temps as a nice way to analyze real life situations. After all, who works at full load all the time? Who wants loud noise when just checking email or sleeping close to the computer? Well, but you definitely know something I don't.


And idle temps can be pretty scary when you see something like the H50 performing at idle like this:

http://www.legitreviews.com/article/1025/6/

 

[LOUISSSSS]

Would ruby rather idle at 50c or 25c given that her load temps for both situations are within spec and the same?

I'll take the 25 and yes they matter to me

 

[aigomorla]

idle temps dont mean much because at idle your putting out so little heat.

The only good idle temps will tell you is the mount.

Because even @ little heat, a bad mount will cause increased temps.

After that you need to load your system to determine the capacity of your system.

 

[LOUISSSSS]

yes i know very little heat is put out @ idle, but it does mean something to everyone. Everyone would prefer a lower idle temp over a high idle temp. And thats all i'm saying here.

In the end, the answer is no, its not possible to do a custom water setup that will beat the H50.

But for the price, the Noctua U12 with dual fans (that it comes with) will pretty much tie/beat it with air cooling. And the Noctua D14 should also tie it but most likely beat it.

In exchange for the "smaller" size of the H50, you lose 1x 120mm fan spot in return for less space usage around the MB area.

 

[Schmide]

yes i know very little heat is put out @ idle, but it does mean something to everyone. Everyone would prefer a lower idle temp over a high idle temp. And thats all i'm saying here.

It's a little less noticeable now since the temperature is read on die, but there was a time where water would have higher idle temps due the the lack of airflow around the motherboard socket and VRMs. The extra heat there would bleed to the sensor. Many watercoolers, including myself, still point a fan at the VRMs to make up for this lack of airflow.

"The idle temps don't really matter" saying, probably got its birth in this squabble where even crappy air was beating high end water at idle.

 

[LOUISSSSS]

i'm a water cooler myself and understand that theres less airflow for the MB when WCing. These days we generally don't have overheating and hot chipsets such as the crappy NF4 chipset of 2004/2005. i wouldn't be so worried about our NB's overheating these days...

 

[Gillbot]

So the corsair kit cAn't be beat at $80 with a custom kit? I'd look into this kit for my v351 build... With amd x4 620.

Is the corsair kit better than a true with 1x 1500 fan?

Sure, you can beat the corsair with a custom kit but not with off the shelf parts. If you bargain shop in the FS/FT sections, you can assemble a killer loop for under $80 easily IMHO.

Aside from the Danger Den kit that was on clearance at Fry's, I don't think I've seen an off-the-shelf wc setup that can match the H50 for the price.

See above.

 

[Schmide]

i'm a water cooler myself and understand that theres less airflow for the MB when WCing. These days we generally don't have overheating and hot chipsets such as the crappy NF4 chipset of 2004/2005. i wouldn't be so worried about our NB's overheating these days...

VRMs (voltage regulation modules - the little black boxes near the socket) They eat around 10-20% of the total wattage going to socket and often operate in the 100c range. This heat, if not dissipated, can easily bleed into the socket.

 

[nitromullet]

Sure, you can beat the corsair with a custom kit but not with off the shelf parts. If you bargain shop in the FS/FT sections, you can assemble a killer loop for under $80 easily IMHO.

See above.

Sorry, I thought "new" was implied. You can't really compare used to new because the results are not really reproducible on any consistent basis. Plus, someone could always be selling a used H50 for cheap...

You'd probably still be pretty hard pressed buying a socket 1366 compatible wc loop for $80 simply because there aren't many people selling used blocks yet. If you were looking for a used LGA775 block you'd probably have better luck.

 

[LOUISSSSS]

Yes, new parts only. Used has too many variables. Sure u can beat the h50, u can get all ur parts free too

 

[Rubycon]

"The idle temps don't really matter" saying, probably got its birth in this squabble where even crappy air was beating high end water at idle.

No it most certainly did not.
I base it on experience. A bad mount OR a cooler that's not up to FULL LOAD capacity will reveal this nicely. Idle temps seem fine then BAM! Load it with LinX and it goes through the roof! Yet the idle temps were nearly the same.

This is why I say "where the rubber meets the road" is most important. Load it up for IBT/LinX for 30 minutes using the largest problem size your memory allows for and record the peak temperatures. Note the room temperature if it changes as well.

My idle temps are higher too because I keep the fans running slower and they ramp up as temp increases. So the actual delta between idle and full load is less.

 

[LOUISSSSS]

a bad mount on CPU Cooler ___ will result in bad temps @ idle&load

a CPU operating not up to FULL LOAD with good idle temps will have good temps. Of course, this varies with the heat dissipating potential of the cooler.

Lets take my Noctua D14 for example. i am idling right now, at 25c. Do you think i will have good or bad load temps?

Now lets take that same scenario. Noctua D14 idling at 25c. Now lets take for given that i have a bad mount. It is clear that there will be more of a range between load-idle temps with the bad mount because more hot air is trapped above the CPU and not dissipated.

Hence, the mount is the determining factor, not load.

 

[Rubycon]

a bad mount on CPU Cooler ___ will result in bad temps @ idle&load

a CPU operating not up to FULL LOAD with good idle temps will have good temps. Of course, this varies with the heat dissipating potential of the cooler.

Lets take my Noctua D14 for example. i am idling right now, at 25c. Do you think i will have good or bad load temps?

Now lets take that same scenario. Noctua D14 idling at 25c. Now lets take for given that i have a bad mount. It is clear that there will be more of a range between load-idle temps with the bad mount because more hot air is trapped above the CPU and not dissipated.

Hence, the mount is the determining factor, not load.

Sure if your mount is so bad that transfer cannot keep idle temps down - load temps will result in a shutdown or damage, however.

All I am saying is you pay attention to LOAD TEMPS. It's the worst case scenario. Let's face it! LinX is not very realistic for what your CPU does in the real world. Even when encoding at extremely high loads my temps are every bit of 10C COOLER than LinX. So why do it then? Just to make sure we have enough safety net in case something happens that causes the temp to increase whether it's dust or the room ambient temperature increasing. Last November we had a power outage that affected AC and the ambient temperature was over 34 degrees C!

p.s. If you have hot air trapped above your CPU you don't have a bad mount you have a serious condition that will cause shut down even at idle! Its akin to running with no cooler at all.

Additionally your cooler is very over sized for your CPU. Quads running over 4GHz at high voltage will turn on the heat and very fast. NO air cooler can keep an i7 at 4.5GHz at 1.45V below 85-90C under heavy load unless the air temperature blown through its fins is very cold. Idle temps, OTOH would be in the mid to upper 30s. (again VERY dependent on ambient temps)

 

[Schmide]

Rubycon - I think you may be misunderstand what I was saying.

Long ago especially when the sensors were off die, a watercooled system with lack of air flow around the socket could get idle temperatures a few degrees higher than air cooling.

I can't tell you how many people would come in the oveclockers water forums complaining that they just bought a new WC setup and their idle temperatures are worse than their idle (*edit) temperatures were on air. Once it was pointed out that the load temperatures were that much better, a very common phrase followed "Idle temps do not matter"

This really isn't the case now that temp measurements are now made on die. However, I would bet that if you compared an air cooled vs water cooled (no fan) surrounding socket temperatures, the water cooled would be a bit higher.

edit: *Load didn't really make sense there.

 

[Rubycon]

Rubycon - I think you may be misunderstand what I was saying.

Long ago especially when the sensors were off die, a watercooled system with lack of air flow around the socket could get idle temperatures a few degrees higher than air cooling.

I can't tell you how many people would come in the oveclockers water forums complaining that they just bought a new WC setup and their idle temperatures are worse than their load temperatures were on air. Once it was pointed out that the load temperatures were that much better, a very common phrase followed "Idle temps do not matter"

This really isn't the case now that temp measurements are now made on die. However, I would bet that if you compared an air cooled vs water cooled (no fan) surrounding socket temperatures, the water cooled would be a bit higher.

Oh yes I remember those thermistors under the chip in the middle of the socket, right?

Yes some boards (Asus) even ship with a little blower to attach to the VRM heatsink. They explicitly tell you that if you use water cooling you MUST use this blower or risk damage to the board. This is because with no airflow across the board (from a cpu heatsink/fan - for example) they do get too hot. Personally I like to have LOTS of flow through the case. Case temps should approach ambient room temp IMO.

 

[Gillbot]

a bad mount on CPU Cooler ___ will result in bad temps @ idle&load

a CPU operating not up to FULL LOAD with good idle temps will have good temps. Of course, this varies with the heat dissipating potential of the cooler.

Lets take my Noctua D14 for example. i am idling right now, at 25c. Do you think i will have good or bad load temps?

Now lets take that same scenario. Noctua D14 idling at 25c. Now lets take for given that i have a bad mount. It is clear that there will be more of a range between load-idle temps with the bad mount because more hot air is trapped above the CPU and not dissipated.

Hence, the mount is the determining factor, not load.

not always. I can mount a cooler without TIM at all and my temps are nearly identical at idle with TIM. However, the problem clearly manifests itself as soon as you apply a load.

 

[LOUISSSSS]

You prob got me there. But what idiot doesn't use by tim?

 

[Rubycon]

not always. I can mount a cooler without TIM at all and my temps are nearly identical at idle with TIM. However, the problem clearly manifests itself as soon as you apply a load.

BINGO!

You prob got me there. But what idiot doesn't use by tim?

Me. I met a Tim and he was not very polite to me.

Seriously when in a hurry to test something what I did was re-mount the HS and was surprised how low the idle was. Then when I hit the IBT switch I was like Holy $4!+!!!

 

[Gillbot]

You prob got me there. But what idiot doesn't use by tim?

It isn't always a purposeful omission of TIM, sometimes a bad spread can do the same thing.

 

[Rubycon]

I will tell you this though. I find it rather disturbing how much different loaded temps can vary when using the same TIM, same cooler, just re-mounting. It does not help that no matter how careful I am there is always some motion. Even when we are docked and not moving I still feel like I am and always tend to mess up when bolting down a heatsink.

Eventually I do get it right. Then another CPU comes along and I have to repeat again. Now that is taken care of by having extra boards handy. Ram is still expensive. The newest Dominators are $200 EACH for 2GB but they are FAST!

 

[skinnee]

I do agree with you Ruby, for waterblocks yes, but not for Air Coolers.

Having been through almost 100 mounts on different air coolers the mount variation and result in temp difference is extremely small. I say extremly small only because I compare to the mount variation seen with waterblocks where a remount can swing 2-3 degrees very easily if you completely botched one of the mounts.

And now I need to find my shield for when Ruby comes back at me with guns blazing.

 

[Rubycon]

Why would a waterblock be different though?

No two spread patterns of TIM are alike either - just like snowflakes.

What's frustrating is you have good temps when you think you can do better and never really get them down as good as they were previously.

It's like messing with collimators where one side is not as accessible as the other. The output beam is never devoid of speckle once it gets touched.