Adding to My Current Loop

[mooommo]

So i have posted a bit ago about my plans to transfer my current build into a 900D. I need some advice because i am new to overclocking, and i like a quiet computer.

 

Currently my rig it a 3770k and a 680 watercooled. I have a 240 thin (pull) and a 120 thick rad (push pull). I have been noticing some high temps at even a low overclock. I used asus' overclocking utility to auto do it and it got like 4.2, which is great. The problem is the heat; at load it runs in the 80-90C range which is high i think.

 

Now i am planning to upgrade my case, so if there is a time to upgrade my rig for future overclocking it would be now. So what should i do to make it cooler and quieter. (recommendation for rads or fans)

 

TLDR

Computer is too hot with a 120 rad and a 240 rad wat do?

 

 

I am thinking adding a 480 just to keep it quiet, but it seems a bit overkill. Maybe some other fans (noctua maybe but they are poop brown).

[DrainDoc]

Noctua's can't hurt(if you don't mind the color).  And if you are moving into a 900D? I would say a 480 is almost mandatory.

 

I have never heard about having to much cooling. I can understand, at some point, the over abundance of Rad.'s are going

 

to cancel out and would not be doing much more for performance. Adding more Rad.'s isn't creating more overall performance,

 

as it is just giving more expansion of further water cooling, for more water cooled components.

 

Get the 480. At the very least, you need to take up some room in that beast.

[B NEGATIVE]

The problem you have there would have to start with the Auto OC,they normally push too much voltage for the clocks attained.

 

The rad face you have is acceptable for the heat you should be putting in the loop,360 worth is the usual baseline amount for a CPU/GPU loop,regardless of thickness.

 

Can you provide the loop spec or a photo showing the layout?

[Ghost]

360 is not really enough for a high end GPU and CPU.

 

Get rid of the 120, add a 480 and put the 480 and 240 in the basement.

[B NEGATIVE]

360 is not really enough for a high end GPU and CPU.

 

Get rid of the 120, add a 480 and put the 480 and 240 in the basement.

 

Really?

I would have to disagree,360 worth of rad is very normal for a CPU/GPU loop,regardless of high end or not.

The 680 is not a hot card at all,the main issue here is the 3770,they run hot and its nothing to do with the loop,they need delidding for any reasonable thermal performance.

 

You are advising a 480 upgrade without even covering the basics like a botched mount? No way on earth would you need that amount of rad,its just wasting money.

The H100 (which is pretty weak) can cool a 3770k but his custom kit is not enough and requires 3 times the face surface?

 

No.

 

OP: What GPU temps are you getting and what is your ambient temp? Can you also provide the voltage your chip is getting?

[Ghost]

Really?

I would have to disagree,360 worth of rad is very normal for a CPU/GPU loop,regardless of high end or not.

The 680 is not a hot card at all,the main issue here is the 3770,they run hot and its nothing to do with the loop,they need delidding for any reasonable thermal performance.

 

You are advising a 480 upgrade without even covering the basics like a botched mount? No way on earth would you need that amount of rad,its just wasting money.

The H100 (which is pretty weak) can cool a 3770k but his custom kit is not enough and requires 3 times the face surface?

 

No.

360 is the bare minimum for a CPU + GPU set-up. You need about a 240 to get reasonable temps on any CPU, as proved by the H100i vs the H80i.

 

If that is true and the 3770k TDP is 77W at stock, then adding a 680 with a TDP of 195W at stock it logically follows that the 680 requires more cooling than the 3770k. Why would you only add half of what something with a lower TDP requires?

 

The OP wants a cool and quiet system. For that you need more than the minimum. I did not say add a 480 because he needs it, I said add a 480 because it fits and buying a 360 now and having no option to increase that late if he ever needs it is pointless. Especially as buying rads separately is more expensive than buying a big rad.

 

A loop with 360 rad space will run loudly compared to what I imagine the Op wants, which is just a near silent hum of fans. In fact he could disable some fans with a 480 and activate them when he needs more cooling. It gives more options, both while the PC is running, and if he ever wants to upgrade his rad space, if for example he gets another GPU, or adds a component with a higher TDP.

 

So; Yes.

[B NEGATIVE]

You are making a mistake confusing TDP with actual thermal output,the CPU is what's important here. Going by your outlook,the much lower TDP of the Haswell chip for example would give much better temps when its very obvious this is not the case,the Haswell is a furnace of a chip.

If you had watercooled GPUs before then you would know that they run much colder than CPUs,35-40c is normal for a fully loaded GPU,including the added heat input from the VRM and VRAM.

The CPU in comparsion runs much hotter and is benefitting from modern impingement style blocks.

 

The CPU runs much closer to TJMax than the GPU,hence the importance of tailoring your loop to it.

 

 

I didn't see the 900d as a confirmed purchase,I did see the concern over temps in his existing build however.

 

Addressing that is my interest,if he buys a 900d then I will tailor my reply to that

[mooommo]

I really just want the 900D for the future. With a system like mine i could probably optimize it in a much smaller case. 

 

The problem with my current fans is that they are way too loud when plugged directy into the power supply, and they tend to not ramp up when the computer gets hot if they are plugged into the MOBO.

 

Asus does have fan control software but it is a huge pain for me because it installs in the wrong language for some reason and i can't find a solution.

 

If the Noctua fans are in the basement of the case it isn't as big of a deal if they look so bad, but what fans should be running for cooling the rest of the case. Would the corsair AF-120s work? I am not too familiar with the fans on the market.

[B NEGATIVE]

I really just want the 900D for the future. With a system like mine i could probably optimize it in a much smaller case. 

 

The problem with my current fans is that they are way too loud when plugged directy into the power supply, and they tend to not ramp up when the computer gets hot if they are plugged into the MOBO.

 

Asus does have fan control software but it is a huge pain for me because it installs in the wrong language for some reason and i can't find a solution.

 

If the Noctua fans are in the basement of the case it isn't as big of a deal if they look so bad, but what fans should be running for cooling the rest of the case. Would the corsair AF-120s work? I am not too familiar with the fans on the market.

Very true

The mobo wont be able to control the fans unless they are PWM,unless there is a voltage control point on your board?

AF's are good for straight case airflow but fall over badly for rads,the gaps between the blades are not ideal to promote static pressure as part of the PQ curve.

I would consider doing a remount of your CPU,the temps you have are a bit on the high side.

 

Can you tell me what voltage your chip is at and the GPU temps on load?

 

Also,and this is important,do you have all fans as intakes? Recycling air from rads to feed other rads is a big no-no,you can lose a lot of performance.

[mooommo]

Very true

The mobo wont be able to control the fans unless they are PWM,unless there is a voltage control point on your board?

AF's are good for straight case airflow but fall over badly for rads,the gaps between the blades are not ideal to promote static pressure as part of the PQ curve.

I would consider doing a remount of your CPU,the temps you have are a bit on the high side.

 

Can you tell me what voltage your chip is at and the GPU temps on load?

 

Also,and this is important,do you have all fans as intakes? Recycling air from rads to feed other rads is a big no-no,you can lose a lot of performance.

i mean fans for the airflow in the top part of the case, and noctuas in the bottom part. 

The CPU is most likely way overvolted so i can check that out later. 

The case i have is standard ATX. It has a 240 mount only in the top and top intake isn't optimal either.

[MG2R]

OP, those temps are waaaaay too high for a 4.2GHz OC, what voltage are you running at? Never ever trust the auto OC features on motherboards.

 

 

360 is the bare minimum for a CPU + GPU set-up. You need about a 240 to get reasonable temps on any CPU, as proved by the H100i vs the H80i.

I'm running a 3570K @ 4.5GHz and a GTX670 at stock on a slim 240mm with low FPI in a Bitfenix Prodigy. It's practically inaudible (and I really mean that you have to stick your head inside the case to hear it)

 

You are making a mistake confusing TDP with actual thermal output,the CPU is what's important here. Going by your outlook,the much lower TDP of the Haswell chip for example would give much better temps when its very obvious this is not the case,the Haswell is a furnace of a chip.

TDP=thermal output. There is no discussion about that: https://linustechtips.com/main/topic/28638-linus-on-the-topic-of-tdp/

 

You're confusing thermal output with thermal conductivity. The CPU does put out only 77W of heat when run within spec, but it is inefficient at transfering that heat to the CPU block because of the crappy TIM inside the IHS, hence the high temps.

[Ghost]

TDP=thermal output. There is no discussion about that: https://linustechtips.com/main/topic/28638-linus-on-the-topic-of-tdp/

The TDP rating on components is how much the stock cooler can dissipate.

[B NEGATIVE]

OP, those temps are waaaaay too high for a 4.2GHz OC, what voltage are you running at? Never ever trust the auto OC features on motherboards.

 

 

I'm running a 3570K @ 4.5GHz and a GTX670 at stock on a slim 240mm with low FPI in a Bitfenix Prodigy. It's practically inaudible (and I really mean that you have to stick your head inside the case to hear it)

 

TDP=thermal output. There is no discussion about that: https://linustechtips.com/main/topic/28638-linus-on-the-topic-of-tdp/

 

You're confusing thermal output with thermal conductivity. The CPU does put out only 77W of heat when run within spec, but it is inefficient at transfering that heat to the CPU block because of the crappy TIM inside the IHS, hence the high temps.

You are missing the point,TDP isnt indicative of how much actual heat is released,The Haswell chip is a good example,regardless of the manufacturing issues the glue layer being too thick (Not the TIM,its not good but its not awful,consult the extensive testing in OCN delid club),the output is not in line with the TDP expectation.

GPU's have a much higher TDP yet ,once watercooled,the actual heat output is greatly reduced,more so than the 180w plus TDP quoted would lend you to believe.

 

Above all this,once you OC a chip,the TDP is totally meaningless. It has no value in anyway.

 

TDP isnt even a standard guideline,AMD,Intel and nVidia have different spec as to what qualifies TDP spec.

 

Source

Lars Weinand, Senior Technical Marketing Manager, EMEA, told us that, ‘The problem with TDP is there is no “standard” for this. So everyone is measuring TDP in a different way and TDPs are only really comparable within the same manufacturer… We are using TDP in a way that makes most sense for us.’ Even Lars ultimately suggests that we put more faith in our power consumption tests than a number written on a spec sheet.

[MG2R]

The TDP rating on components is how much the stock cooler can dissipate.

I think not, look at Intel for example. Their stock cooler is delivered with chips of varying TDP. The TDP usually denotes a spec for system builders. It gives them a minimum amount of power the cooling solution they implement should be able to handle. (put the relevant bit in italics)

 

The thermal design power (TDP), sometimes called thermal design point, refers to the maximum amount of power the cooling system in a computer is required to dissipate. The TDP is typically not the most power the chip could ever draw, such as by a power virus, but rather the maximum power that it would draw when running "real applications". This ensures the computer will be able to handle essentially all applications without exceeding its thermal envelope, or requiring a cooling system for the maximum theoretical power (which would cost more but in favor of extra headroom for processing power).

http://en.wikipedia.org/wiki/Thermal_design_power

 

 

 

You are missing the point,TDP isnt indicative of how much actual heat is released

The actual heat (power) output? Nope. Rather, as stated above, it indicates the minimum power your cooling solution should be able to dissipate.

 

The Haswell chip is a good example,regardless of the manufacturing issues the glue layer being too thick (Not the TIM,its not good but its not awful,consult the extensive testing in OCN delid club),the output is not in line with the TDP expectation.

Whatever the cause of the bad thermal conductivity might be (thick layer of glue, crappy TIM,... ), it doesn't change the fact that TDP in itself doesn't give you any indication of how hot a chip should run (and you shouldn't base expectations concerning the operating temperatur upon the TDP alone). What determines the actual temperature is a function of (1) the dissipation capability of your cooling solution, (2) the thermal conductivity of the path between your heat source (CPU) and your heat sink (water block), (3) the ambient temp and (4) the power output of the heat source.

 

It is true that, since (4) is variable, we usually take the TDP as the heat output, but your expectation clearly shouldn't be based on that alone. A very important factor is (2).

 

Above all this,once you OC a chip,the TDP is totally meaningless. It has no value in anyway.

Well, it isn't completely meaningless. As shown above, it is very usefull to outline the cooling setup you'll need. Yes, you'll have to overdesign it when you start OCing, but even than you can pretty much extrapolate based on the TDP what power dissipation capability your cooling setup needs to have.

 

You can do this based on P=C * v² * f with

• P, the power dissipated by the CPU
• C, the internal capacitance of the CPU, a constant
• v, the voltage applied to the CPU
• f, the clock frequency applied to the CPU

Keep in mind that this equation is only valid for ICs that operate on one given voltage Vcc. A CPU is very complex and has differen voltages, so working with this equation is, of course, an approximation.

The TDP is useless as with is, you can (roughly) define C. From their, you can inserted the frequency and voltage you intend to run the chip at (worst case scenario). That gives you a new power rating, which is the approximated power output of your OCed CPU.

 

Again: this is a rough estimation, but it works-ish.

[Ghost]

I think not, look at Intel for example. Their stock cooler is delivered with chips of varying TDP. The TDP usually denotes a spec for system builders. It gives them a minimum amount of power the cooling solution they implement should be able to handle. (put the relevant bit in italics)

http://en.wikipedia.org/wiki/Thermal_design_power

TDP measurement is very inaccurate.

 

Have a look at this: http://www.silentpcreview.com/article169-page3.html

 

By definition you are right but Intel measure TDP as an average. Not the maximum heat the component can produce in a worst case scenario. The cooler provided is typically enough for the listed TDP. Also TDPs vary from chip to chip.

[B NEGATIVE]

I think not, look at Intel for example. Their stock cooler is delivered with chips of varying TDP. The TDP usually denotes a spec for system builders. It gives them a minimum amount of power the cooling solution they implement should be able to handle. (put the relevant bit in italics)

http://en.wikipedia.org/wiki/Thermal_design_power

 

 

 

 it doesn't change the fact that TDP in itself doesn't give you any indication of how hot a chip should run

 

Well, it isn't completely meaningless. As shown above, it is very usefull to outline the cooling setup you'll need. Yes, you'll have to overdesign it when you start OCing, but even than you can pretty much extrapolate based on the TDP what power dissipation capability your cooling setup needs to have.

 

You can do this based on P=C * v² * f with

• P, the power dissipated by the CPU
• C, the internal capacitance of the CPU, a constant
• v, the voltage applied to the CPU
• f, the clock frequency applied to the CPU

 

I have been saying this from the start.

 

Seeing as TDP is a meaningless spec that changes from manufacturer to manufacturer with very different specs from each,using it as a basis for anything is pointless.

The standard rule of 120 then 120 for each block in the loop still stands.

 

Every major WC site has this as the standard,its well known to be the standard rule of thumb for loop design.

No need for TDP to even be considered.

[MG2R]

Seeing as TDP is a meaningless spec that changes from manufacturer to manufacturer with very different specs from each,using it as a basis for anything is pointless.

The standard rule of 120 then 120 for each block in the loop still stands.

 

Every major WC site has this as the standard,its well known to be the standard rule of thumb for loop design.

No need for TDP to even be considered.

So, on that basis, watercooling an Atom would need just as much rad surface as watercooling a 9590?

TDP isn't meaningless at all.

[B NEGATIVE]

So, on that basis, watercooling an Atom would need just as much rad surface as watercooling a 9590?

TDP isn't meaningless at all.

A passively cooled chip which cant easily be watercooled v a power hungry 5 ghz 8 core?

 

Both would be ok on a single 120 rad,just the AMD chip would actually be using it fully.....

[MG2R]

A passively cooled chip which cant easily be watercooled v a power hungry 5 ghz 8 core?

 

Both would be ok on a single 120 rad,just the AMD chip would actually be using it fully.....

I think  this dude will disagree, on the 'single 120mm is enough for a 9590' thing... His 9590 maxed out a dual 120mm rad...

[B NEGATIVE]

I think  this dude will disagree, on the 'single 120mm is enough for a 9590' thing... His 9590 maxed out a dual 120mm rad...

The h100 isnt a custom kit tho,a skinny Alu rad with a glycol/water mix and minimal pump performance just doesnt perform the same.

 

Im not saying the 120 will do an outstanding job but it will work,I could cheat here and use the Alphacool Monsta 120 which is nearly 3 times the thickness. This would certainly be up to the task.

 

 

 

 

I do like a MO-RA tho....nothing beats a big slab of rad like that.

[skline00]

If you plan on buying a 900D case USE it for what it was designed for "tons of radiators" I'd keep the 240 and add a 480 in the bottom. Cool it all with Scythe Gentle Pyphoon AP15s on a fan controller and be done. I'm using a CM HAF 932 Adv case (wish I had bought a bigger case) with a 3930k OC'd to 4.3Ghz and a PNY GTX680 OC. I sought input from water cooling guys and since I already had a XSPC RX 360 added a single Swiftech MCR120-QPK to replace the rear fan exhaust. Trust me with the 3930k you can't over rad it puts out heat. I also own a 3770k that I cool with a Kraken X60 and OC to 4.5Ghz. My Video cards in it are not water cooled. If you are going to spend the coin on a big case like a Corsair 900D I would DEFINITELY have a 480 rad below and use your 240 above. Gives you the ability to add a second gpu without stress.