Is water cooling still needed for good overclocks anymore?

Dutt1113

[H]ard|Gawd
Joined
Jun 30, 2005
Messages
1,601
Is custom water cooling still needed for good overclocks anymore? I've been thinking. With all the good CPU AIO water coolers now and the fact that liquid cooling the GPU doesn't seem to give you higher overclocks vs air anymore. What's the point in expensive custom water cooling when an easy convenient CPU water AIO cooler works well and GPU's overclock nearly the same on stock cooler vs on water anymore?
 
We've had a solid 10-15 or so years where high-end air coolers are as good or better than entry-level watercooling systems. And yeah, there's not exactly much of a jackpot for playing the silicon lottery, either.

Watercooling is kick-ass, though, especially if you're into airy glass cases with fancy RGB. It's essentially a hobby all of its own.
 
The air coolers can compete yes but a custom loop with two 360's or two 480's will indefinitely be cooler and much quieter which is quite luxurious of a feature. Plus, the extremely small footprint of the blocks themselves no big coolers in the case everything is streamlined flat single slot nice and neat for maximum airflow and extremely low fan speeds.
 
Large AIOs can certainly compete with custom watercooling setups, but custom setups offer much more flexibility. Also large AIOs approach the cost of entry level watercooling, especially the good AIOs.

I've always said this, from a price-performance standpoint, custom watercooling has never and will never be worth it. Custom watercooling is done to achieve max performance 24/7 setups, low/no noise setups, for the fun of it, for the looks/showing off, or some combination of the above.
 
You need watercooling to try to run boost clocks with 13th gen Intel and AM5 lol

Tl;dr: Diy cooling can be relatively affordable if you go outside the norm too, and/or get used parts.

Long:
Back when I started guys were still using aquarium pumps and car heater cores with soldered or brazed copper for the block. I've got several lang d4 and d5 pumps for dirt cheap, they're the oem for most of the big pumps. I got the d4s for $25 each unused on sale, chose to cut the nipples off of one and tap for g1/4 to use barbs. Used radiators go on sale around here, usually half off. I got 3x360 and a 480 over time with decent savings. Might want to cover the fins and spray paint the outside, they're usually beat up. A good new CPU block is usually $50-65, gpu $120 generally for full cover.

Hose barbs are MUCH cheaper than compressions. You can get 3 chrome barbs for a comp, or get probably 6 ugly nylon barbs. Barbs are so small, who cares? Well you could easily spend $12ea or more. People often overbuy fittings so you might find new ones being resold for a good discount.

Decent hose is a buck a foot maybe? Distilled and AF is good enough, dyes will leach into hose eventually-even the exotic fancy stuff it seems.

So a budgeted cpu&gpu DIY setup with high end performance could cost you $275-350. Cutting out the gpu cuts the cost a lot, and sky is the ceiling if you want glitz and glamor.

I'm currently rehabilitating a 50mm peltier block, back when I was using it it had a 226W pelt, cooling a stock 110W cpu. Overclocked who knows, but that loop had a ton of heat getting dumped in. The block outlet would get VERY hot when pushing suicide runs.
 
Overclocking in general is not what it used to be.

Back in the ~32nm days and before we used to often be thermally limited when it came to overclocking, so lowering temps resulted in huge overclocking benefits.

Heck, I remember back on my 486sx 25, which shipped without a heatsink or fan, just adding a small HSF, lowered the temps enough for me to DOUBLE the performance at 50mhz.

This just doesn't happen anymore.

This is partially because as the process nodes have shrunk we usually run into electrical/voltage limits before we run into thermal limits and because of this lowering temps no longer has quite the same impact as it once did.

It is also partially because the manufacturers use dynamic clocks and dynamic voltage these days and fine tune the silicon such that they are taking almost every bit of performance out of it out of the box, leaving less left to be gained from overclocking.

In fact, for many CPU's, overclocking is all but dead. I've had my Threadripper for almost 3 years now, and I haven't even tried overclocking it, as the small gains just aren't worth the trouble anymore. It's still alive in GPU's, but to a lower extent than it used to be.

I still like water cooling, as it provides benefits such as less fan noise (in a well designed loop) but no, it's just not the same as it used to be, when you could use a water loop to significantly drop temps and see HUGE performance gains from it.
 
Last edited:
Nail on the head. Competition is too stiff and every ounce of performance is being "boosted" out.

The manufacturers don't have to speed bin nearly as much either if the label says base clock 3.6, boost up to 5.8. Up to - meaning you could reach that but might not.

On the flip side fab nodes are giving less gains, so aside from design efficiency improvements the only ways up in "performance" is going to add more wattage: more cores and 3D stacking. Will need AIOs/WC to move that heat, Copper heat pipes can only move so much, only so many jammed in a space. With rumors of 13th gen thermal throttling triple and quad AIOs it'll be interesting to see. I think AIOs are lacking in tube volume, pump volume but would need an IR gun or flir to verify if it's that or the rad not being able to cope. Would be neat to see!

I haven't seen data sheets on my big rads, but I vaguely remember a spec on how many watts they could handle. A black ice gtx 480 could do like 750W with 1/2" ID.
 
When it comes to straight WC rad capacity is everything. Designing systems for best performance in mind once your coolant temperature is two degrees above ambient you need more rad. Or airflow if noise isn't concerning. These outlandish claims of x rad has 1kW thermal capacity don't mean squat if the coolant temperature is twenty degrees above ambient!
 
When it comes to straight WC rad capacity is everything. Designing systems for best performance in mind once your coolant temperature is two degrees above ambient you need more rad. Or airflow if noise isn't concerning. These outlandish claims of x rad has 1kW thermal capacity don't mean squat if the coolant temperature is twenty degrees above ambient!
I disagree on this. I regularly let my coolant get 5-10 degrees over ambient, but it's far quieter than it would be with air coolers (essentially silent) and is still running at lower temps than it would be on air. Doing it per your specs means I would need a bigger case or put up with more noise.
 
I guess it depends on what you're cooling and how far you're pushing things. My builds of yesteryear were nowhere near stable without coolant temperature at ambient or below. Pushing things back a bit, it would be more tolerant. Of course today's stuff is different but I do like to have extra margins in case ambients go higher, etc.
 
I disagree on this. I regularly let my coolant get 5-10 degrees over ambient, but it's far quieter than it would be with air coolers (essentially silent) and is still running at lower temps than it would be on air. Doing it per your specs means I would need a bigger case or put up with more noise.

Yeah, the higher your loop temp is, the higher the Delta-T is to air, and the higher that Delta-T is, the more efficient the heat transfer is, so slower fans

It really depends on what your objectives are.

If you want silence and don't think you'll get much in the way of overclocks, turn the coolant temps up a bit.

If you are right on the edge of desired performance and want to eke every little bit of extra boost clocks out of a system, set the temps lower.

I maintain my loop temps at 38C for regular desktop use. It is dead silent that way. When I run a demanding game that requires the best I can throw at it, I turn those loop temps down to 33C. With my crazy 430W 6900xt and the Threadripper this makes even my overkill radiator capacity (two 480's and one 420 all in push-pull) loud at load.

Back when I had my Pascal Titan X this allowed me to keep my highly overclocked GPU at under 40C core temp. Under 40C seemed to be the magic number for best boost clocks on that GPU. With the new crazy high wattage 6900xt, 33C loop temps wind up resultibg in 47-48C core temps.

With some lighter games I can afford to keep it at the "desktop load" 38C setting. I'm in the midst of a big Civ5 revisit right now, and I can run it at 38C with AMD cool and quiet limiting it to 60fps when I am idle, and boosting up to 120fps when I am active and the system is dead silent.

Most newer titles will require everything I can throw at it though. 4K ultra is a bitch to power.
 
Last edited:
I think it is barely worth bothering with overclocking anymore regradless of what cooler you use. You waste 20-30% more power on a few percent gains at best, or worse.
As Roman (der8auer) have demonstrated with the 4090 even the stock configuration is well above the ideal zone of the power/performance sweet spot. Reducing the power draw by 33% only means 10% performance loss if I remember the numbers correctly. If I had a 4090, I'd be downclocking it not overclocking.

I'd argue that overclocking for daily use is dead, it is only the playing field of professional record seekers. Because I doubt you could notice the gains of overclocking in real world applications without benchmarks.
 
I think it is barely worth bothering with overclocking anymore regradless of what cooler you use. You waste 20-30% more power on a few percent gains at best, or worse.
As Roman (der8auer) have demonstrated with the 4090 even the stock configuration is well above the ideal zone of the power/performance sweet spot. Reducing the power draw by 33% only means 10% performance loss if I remember the numbers correctly. If I had a 4090, I'd be downclocking it not overclocking.

I'd argue that overclocking for daily use is dead, it is only the playing field of professional record seekers. Because I doubt you could notice the gains of overclocking in real world applications without benchmarks.

If I weren't at 4k, struggling to get my min framerates above 60fps in most new titles, I'd agree with you.

I tamp it down when I can, but crank it up when I need it. I need it more often than I don't.
 
If I weren't at 4k, struggling to get my min framerates above 60fps in most new titles, I'd agree with you.

I tamp it down when I can, but crank it up when I need it. I need it more often than I don't.
It's placebo. While I can tell the difference between 30 and 45 fps I definitely can't tell between 43 and 47, that would be 10% gain, but realilstically you can't even hope for 10% gains through OC nowadays.
 
It's placebo. While I can tell the difference between 30 and 45 fps I definitely can't tell between 43 and 47, that would be 10% gain, but realilstically you can't even hope for 10% gains through OC nowadays.

Well, in all fairness it was a bigger deal before G-Sync and FreeSync when dropping below te refresh rate of the screen had pretty big penalties, but even so, I definitely want to keep my mins at no lower than 60fps.
 
Speaking of rad space & temps I just turned all 3 of my rads fans on exhaust and the coolest temps across the board for every thermal in hwmonitor. It is a 780t with a 360 front/ 360 top/ 120 rear all alphacool and push pull exhaust for all fans best configuration if you guys haven't given it a shot the results are shockingly good contrary to what you might think would not be an ideal setup it actually is lol.
 
That is one thing I miss as doing it as a hobby was all the different experiments and findings such as this. All at no cost other than your time. :)
 
That is one thing I miss as doing it as a hobby was all the different experiments and findings such as this. All at no cost other than your time. :)
You know I realized? Lol I like having a main water-cooled PC and a almost as good very worthy secondary air cooled PC. It fills the niche for the different types of cravings for tinkering cause like you said, it is a fun hobby. Also have a nice lightweight XC mountain bike to tinker with but that's for a different forum lol
 
Fast, quiet, cheap - pick 2

I chose fast and silent...

(Eight 200mm fans at 340rpm)
 
Speaking of rad space & temps I just turned all 3 of my rads fans on exhaust and the coolest temps across the board for every thermal in hwmonitor. It is a 780t with a 360 front/ 360 top/ 120 rear all alphacool and push pull exhaust for all fans best configuration if you guys haven't given it a shot the results are shockingly good contrary to what you might think would not be an ideal setup it actually is lol.
What do you have for your intake?

Sure, you might get lower component temps, but generally peripheral temps suffer as a result. Additionally, with negative case pressure, dust tends to build up in all the nooks and crannies of the case. Positive case pressure ensures that dust pretty much only comes in through the intake fans and you can put filters on those.
 
What do you have for your intake?

Sure, you might get lower component temps, but generally peripheral temps suffer as a result. Additionally, with negative case pressure, dust tends to build up in all the nooks and crannies of the case. Positive case pressure ensures that dust pretty much only comes in through the intake fans and you can put filters on those.
The bottom of the Corsair 780T full tower case has room for 2 filtered 120mm case fans (only using one 140mm right now) The rear has a few areas of perforated ventilation (just ordered a few dust filters I was planning on putting a few here) & the front two 5.25 bays have been swapped out for perforated ventilation dust filtered panels. I know it sounds weird and is weird, but I'll be darned if it wasn't the best performing option for this particular case with 3 radiators.

I even found a video on YouTube with Greg Salazar doing a video for corsair about fan configs in a situation similar (different case my 780t is much better) to mine except I have some more radiator than the test rig does. For my 3 rads I have the front 360 exhaust the top 360 exhaust & the rear 120 exhaust all push pull somehow fit. The all exhaust on all radiators was the clear all-around winner.


 
Last edited:
The bottom of the Corsair 780T full tower case has room for 2 filtered 120mm case fans (only using one 140mm right now) The rear has a small area of perforated ventilation (just ordered a few dust filters I was planning on putting a small one here) & the front two 5.25 bays have been swapped out for perforated ventilation dust filtered panels. I know it sounds weird and is weird, but I'll be darned if it wasn't the best performing option for this particular case with 3 radiators.

I even found a video on YouTube with Greg Salazar doing a video for corsair about fan configs in a case similar to mine except I have some more radiator than the test rig does. For my 3 rads I have the front 360 exhaust the top 360 exhaust & the rear 120 exhaust all push pull somehow fit. The all exhaust on all radiators was the clear all-around winner.




What fan directions provide the best temperatures is going yo vary greatly from loop to loop, case to case and setup to setup.

In your case, with all exhausts, what is probably going on is that you are improving temps by avoiding hot air from one radiator passing through another radiator.

In some cases this would harm airflow, but your case must have a good amount of available spots (grille type PCIe slot covers, other vented areas, etc.) open to the outside where air can be pulled in.

Provided the temps are good, the only downside to a setup like this is that when setting all fans to exhaust, you are creating negative pressure in the case, which means air is being pulled in through the afore mentioned venting spots. These spots generally don't have dust filters, so more dust will get into the case.

This is why the most common recommendation is to have more intake than exhaust, as this results in positive pressure, with air being forced out all of these cracks and crevices. You can then control where the air is pulled in, and make sure you have filters there.

Either way, in your setup where this results in better cooling performance, you may just decide that dealing with more dust (or cleaning it more frequently) is worth it.

That's one of the many reasons this hobby is fun. Finding a solution that works best for your priorities and going with it.
 
The bottom of the Corsair 780T full tower case has room for 2 filtered 120mm case fans (only using one 140mm right now) The rear has a few areas of perforated ventilation (just ordered a few dust filters I was planning on putting a few here) & the front two 5.25 bays have been swapped out for perforated ventilation dust filtered panels. I know it sounds weird and is weird, but I'll be darned if it wasn't the best performing option for this particular case with 3 radiators.

I even found a video on YouTube with Greg Salazar doing a video for corsair about fan configs in a case similar to mine except I have some more radiator than the test rig does. For my 3 rads I have the front 360 exhaust the top 360 exhaust & the rear 120 exhaust all push pull somehow fit. The all exhaust on all radiators was the clear all-around winner.



The air won't just be coming in through the obvious gaps, it will be coming in through any and all available gaps. This means corners and crevices that you didn't even know existed could be potential spots for dust accumulation. The unfiltered air would also likely mean eventual dust buildup on your radiators as well. A counter-flow loop setup would greatly reduce the impact of passing air through radiators twice. As said though, you can choose what your priorities are. Just providing my opinion as to the benefits of alternative orientations and setups.
 
Back
Top