Discussion Ryzen 3950X Overclocking/undervolting benchmarking data: Temps and CB20 scores.

[tiberium87]

Hello Friends,

I am new to the overclocking world, and so far it's been a fun and educational ride.

I wanted to share some stats I collected on my Ryzen 3950X, and hear your thoughts on them. Maybe people can even use this as a very rough estimate / comparison of what they might be able to expect for their own 3950X in their OC adventures. I know I would have liked something like this a few days ago!

I think mine runs a tiny bit hot, but not too bad. Using a 360mm water cooler (120mmx3 fans), 3 bequiet chassis fans, and a bequiet Pure Base 500 case. Mobo is the X570 Taichi.

Here is the data:
spreadsheet_1
spreadsheet_2

Best,
tiberium

 

[drea.drechsler]

Hello Friends,

I am new to the overclocking world, and so far it's been a fun and educational ride.

I wanted to share some stats I collected on my Ryzen 3950X, and hear your thoughts on them. Maybe people can even use this as a very rough estimate / comparison of what they might be able to expect for their own 3950X in their OC adventures. I know I would have liked something like this a few days ago!

I think mine runs a tiny bit hot, but not too bad. Using a 360mm water cooler (120mmx3 fans), 3 bequiet chassis fans, and a bequiet Pure Base 500 case. Mobo is the X570 Taichi.

Here is the data:
spreadsheet_1
spreadsheet_2

Best,
tiberium

Not very useful without CB20 Single Thread scores too. ST performance is most important for games that only hit a single thread very heavily and all-core overclocking Zen2 usually kills ST scores.

Also...that chart's got some good data to determine how much your CPU degrades after a few months of fixed voltage overclocking. Pretty sure the extreme overclockers on Reddit will appreciate you donating your CPU to the cause if you're willing. Those guys are recommending (strongly) a VCore of 1.2V MAX but they're always eager for well-documented real-world performance.

 

[zx128k]

EDC bug on the 3950x is pretty insane as well on water but you will get more performance with RAM side overclocking. Still the 3950x and 3900x is an expensive cpu to degrade.

 

[tiberium87]

I had no idea I was doing something so potentially damaging to my CPU. Your replies were eye opening. Will read up much more now, especially on the EDC bug, etc.

I, like many others, had assumed that if I went up to 1.25-1.28V and 80-83C just for testing something like 4.15GHz, it would be OK, and for 24/7 use, I was originally thinking of sitting at something like 1.225 to 1.25V.

But I hear your warning. From now on, I am thinking of something like this:

1. stock settings for normal use / gaming, and

2. for work related simulations (which need to run 24/7 for days or weeks sometimes), manual control with a max of 3.9GHz, and 1.15 - 1.18V. The CPU temp is 65C for the simulations, even though in the spreadsheet the CB20 temp is 68C-72C). Right now I am running these simulations at 3.6GHz all core, 1.125V. Its sitting at 63C.

I am hoping this time I have gotten things right, and these choices will not degrade my CPU? I am hoping to find a setpoint where I can be reasonably certain that if I ran the CPU almost 24/7 all-core for 3 - 4 years straight, it would be no problem.

The reason I don't want to use stock settings for my simulations is that the PC runs at 77C - 82C, which even to my pre-eye-opened self was way too high for something that needed to run 24/7.

 

[zx128k]

Most of the overclocking performance is RAM side. IF 1800-1900MHz RAM 3600-3800 MT/s and tighten the timings. You an overclock the cores but most game performance will come from overclock the RAM and IF. If you improve cooling then you will boost higher at stock.

Basically I would get a chiller for a 3950x if I had the money. https://www.aquatuning.co.uk/water-...an-1500-hc500-790watt-cooling-capacity?c=2745

Gaphics card, motherboard and CPU. You will get some core overclocking with that setup.

Temperatures down to 4°C are possible. https://www.bit-tech.net/reviews/tech/cooling/hailea-hc-500a-water-chiller-review/4/

 

[tiberium87]

Alright, yeah. Will look into IF / RAM OC as my next mini project, and maybe learn more about the whole ecosystem of GPU-RAM-CPU.

As for the chiller: good to know this exists (was wondering if something like this was a thing) Maybe one day! Baby steps right now, haha.

Suppose I don't actually need to overclock, and am more interested in some sort of reasonable-ish 16 core performance (can be anything around, say, 3.6GHz to 3.9GHz), as long as I can safely run things 24/7, would you say sub 65C is safe for 24/7?

The whole reason I did the tests above was not overclocking, but actually undervolting. I want to bring the temps down from the stock 80C to something that can run 24/7 while still getting some medium - medium high level of performance.

 

[zx128k]

Alright, yeah. Will look into IF / RAM OC as my next mini project, and maybe learn more about the whole ecosystem of GPU-RAM-CPU.

As for the chiller: good to know this exists (was wondering if something like this was a thing) Maybe one day! Baby steps right now, haha.

Suppose I don't actually need to overclock, and am more interested in some sort of reasonable-ish 16 core performance (can be anything around, say, 3.6GHz to 3.9GHz), as long as I can safely run things 24/7, would you say sub 65C is safe for 24/7?

The whole reason I did the tests above was not overclocking, but actually undervolting. I want to bring the temps down from the stock 80C to something that can run 24/7 while still getting some medium - medium high level of performance.

http://www.ldcooling.com/shop/14-phase-change

Then push to 5GHz if you can afford it.

 

[tiberium87]

Didn't know about this either. I think if I do go this route, it would be purely for the kick of it. Might have to build a cheap-ish PC to test it on, maybe a ryzen 3 or 5.

Going to town sounds like fun.

 

[drea.drechsler]

...
Suppose I don't actually need to overclock, and am more interested in some sort of reasonable-ish 16 core performance (can be anything around, say, 3.6GHz to 3.9GHz), as long as I can safely run things 24/7, would you say sub 65C is safe for 24/7?
...

There's been some discussion about it but nobody really has the answers outside of AMD and they aren't talking. You can start your own inquiries on that Reddit forum I posted earlier....also on http://www.overclock.net.

The thinking is that at the tiny feature sizes of this 7nm TSMC process it's core currents that are at least equally important to control as temperature. Core current is driven by voltage, hence the standard guidance to run no more than 1.2V fixed. Many people have kept very low temperatures and have still experienced degradation within a few months, some running as low as 1.30V. They've realized there's doubtless a very good technical reason the boost algorithm will pull voltage as low as 1.2V so that's what you have to consider as 'safe' if you're not going to let the algorithm keep it safe.

The reasoning comes from how the Ryzen boost algorithm works: it allows fairly high temperatures, well into the 80C range, even up to 95C Tjmax, but pulls voltage (and therefore frequency too) to keep it safely within the FIT margins when it gets there. It's really quite effective too, getting all that can be safely got out of the processor in all operating ranges short of running at sub-zero temperatures.

To be honest, I think what you're doing probably isn't even getting performance as good as it will do on it's own if left on AUTO. I believe a 3950X will hold around 4.2-4.25Ghz in a heavy all-core work load (like CB20) when put under decent liquid cooling and PBO optimized.

But more importantly, it can boost single cores up to 4.7Ghz, most important for game performance to avoid CPU induced stuttering. Or even hold multiple cores at 4.3-4.5Ghz in moderate workloads, by far the most common sort of workloads we'll see. With what you're doing you'll never see that kind of performance with those settings, and once you've degraded it in a few months that will become permanent when you return settings to default.

 

[drea.drechsler]

...
As for the chiller: good to know this exists (was wondering if something like this was a thing) Maybe one day! Baby steps right now, haha.
...

The problem with a chiller will be the same problem they've always presented: going subambient means you'll have condensation around the socket that eats up the motherboard in very short order.

No matter how good the insulation and board preparation it always fails within a few months. Either that or daily maintenance. The one and only solution worth considering long term is total immersion in chilled mineral oil. Or some other chilled inert fluid such as Fluorinert since mineral oil has a fairly narrow range of temps it works well in.

 

[zx128k]

The problem with a chiller will be the same problem they've always presented: going subambient means you'll have condensation around the socket that eats up the motherboard in very short order.

No matter how good the insulation and board preparation it always fails within a few months. Either that or daily maintenance. The one and only solution worth considering long term is total immersion in chilled mineral oil. Or some other chilled inert fluid such as Fluorinert since mineral oil has a fairly narrow range of temps it works well in.

Chillers are good because you run them above the dew point, You get a steady low temperature. This means your coolant can be lock at something like 10-16c or at ambient temperature or just below. You can't really go sub zero with a chiller because the coolant will become viscous and full of tiny ice crystals which will block the small micro fins in the water block reducing cooling. You only have to insulate if you go below the dew point.

Dew pint can be worked out https://en.wikipedia.org/wiki/Dew_point

Some people have been running a chiller for years without any problems. https://www.techpowerup.com/forums/...-cooling-now-over-3-years-operational.222083/

Ambient cooling would not support the 3770K overclocked to 5ghz as I run it at a 10c coolant temperature which is 15c below ambient room temperature.
I run at 10c coolant temperature to avoid condensation which begins forming at 8c as a frosty look and worsens as the temperature drops lower where at 5c drops are forming and that is the danger zone, but as long as it is operated in the 10c range there is zero condensation.

No insulation needed if you run above the dew point.

You can safely enjoy about a 13c below ambient without any condensation worries at all, but you take full responsibility with what you do with this information.

https://forums.tomshardware.com/threads/tec-peltier-cpu-chilled-water-cooling.1281003/

TEC/Peltier Water Cooling
Water Temp. is 12c which is also, 12c below ambient room temperature, (zero condensation), Idle = 14c, Load = 31.25c
Water Temp. is 9c which is, 15c below ambient room temperature, (zero condensation), Idle = 12c, Load = 28.5c

You have to stay above the dew point but that won't mean that the cooling will be bad.

Phase Change and you 100% have to cover everything with insulation. Frost point is covered in that wiki link.

Insultation pictures here https://www.overclock.net/forum/62-peltiers-tec/1633988-chilled-water-cooling-vs-3-0-build-log.html As drea.drechsler clearly implies you have to get the insulation right and maintain that insulation. With TEC/Peltier Water Cooling you are aiming for very low temperatures.

The main goal of a chiller is to replace ambient limited radiator cooling. You can still push temps lower than ambient within reason safely.

 

[drea.drechsler]

Chillers are good because you run them above the dew point ...

Ahhh...didn't realize they were controlling temperature of the cooling block so precisely now. But that also means temp at the block is probably not going to be much better than 10C in most regions outside of high desert and frequently much worse.

Better than the 30-40C from air inside a case, but even so I can't imagine that being very effective with Zen2. The reason comes down to the 7nm geometry. Die size is really small, providing way to little surface area to transfer heat effectively. So it needs a much larger temperature differential to make a significant difference. That's a significant reason driving the lack of conventional overclocking ability since passing through 12nm, and really since 14nm.

Those vids by Der8auer and GN back at Zen2 release were extremely revealing. They just could not get better clock performance than the boosting algorithm - keep in mind that the boosting algorithm follows temperature so it enables better performance too - until they pushed temp into sub-zero region with LN2. The architecture is capable of higher clock speed operation (5.4Ghz on 3100's I seem to recall seeing!) it just seems to be the process geometry that holds it back if not heavily cooled.

It would be interesting to try though...if way costly and bulky in operation...put one of those chillers on a 3950 with fully optimized PBO and see if it could hold clocks in the 4400-4600Mhz range, or maybe even 4.7Ghz max. That would be a significant performer!

 

[zx128k]

Ahhh...didn't realize they were controlling temperature of the cooling block so precisely now. But that also means temp at the block is probably not going to be much better than 10C in most regions outside of high desert and frequently much worse.

Better than the 30-40C from air inside a case, but even so I can't imagine that being very effective with Zen2. The reason comes down to the 7nm geometry. Die size is really small, providing way to little surface area to transfer heat effectively. So it needs a much larger temperature differential to make a significant difference. That's a significant reason driving the lack of conventional overclocking ability since passing through 12nm, and really since 14nm.

Those vids by Der8auer and GN back at Zen2 release were extremely revealing. They just could not get better clock performance than the boosting algorithm - keep in mind that the boosting algorithm follows temperature so it enables better performance too - until they pushed temp into sub-zero region with LN2. The architecture is capable of higher clock speed operation (5.4Ghz on 3100's I seem to recall seeing!) it just seems to be the process geometry that holds it back if not heavily cooled.

It would be interesting to try though...if way costly and bulky in operation...put one of those chillers on a 3950 with fully optimized PBO and see if it could hold clocks in the 4400-4600Mhz range, or maybe even 4.7Ghz max. That would be a significant performer!

There is a graph.

https://www.gamersnexus.net/news-pc/3492-ryzen-cpu-thermals-matter-coolers-and-cases

As you go below 10c the boost increases so slowly that its almost flat down to -75c. At some point between 0-25c you get most of the frequency increase via temperature reductions. The cpu performance increases from 3120 to 3289 I would from reducing temperature from +75c to approx. 15-20c (tdie) but you only gain from 3289-3319 by reducing temperature to less than -75c (LN2 pot temp).

This means that the closer to 15c (tdie) you can get the more the CPU will boost all cores. Below 0c the performance increase is nearly flat, so its not worth it cooling that low.

Pushing the temperature lower does not look like it is worth it if you use the stock boost.

The maximum scale, not that it’s particularly relevant to users, is 6.4% from 3120 points to 3319 points, scaling from -80 to 84 degrees Celsius. Even so most of that 6.4% is to the left of the graph.

The chart starts at about 84 degrees Celsius, which is where you might be sitting with a 3900X with the stock cooler and with the average, modern case. At 84, we measured frequency between 3975 and 4000, sticking closer on average to 3975MHz on the 12-core CPU. As we manually dialed our temperature to reach 78 degrees, we averaged 4050MHz all-core frequency. 71 degrees Celsius for Tdie puts us at 4075MHz all-core, with a step down to 61 degrees Celsius, a 10-degree drop, improving all-core frequencies notably to 4150MHz. The next step is to 55 degrees, where we saw an improvement to 4200MHz.

This is not delta T over ambient, as the only important thing here is actual operating temperature. Ambient was 21 degrees, but also, we’re showing CPU temperature anyway, so that’s not relevant at the moment. What is relevant is that an ambient temperature of 21-25 degrees is common for an air conditioned house, and that case ambient is often approaching 30 degrees even on good cases, so a 55-degree load temperature is achievable primarily with high-end cooling solutions and well-ventilated cases. It’s achievable, but this is the end of our common real-world scenarios, whereas the 78-84 range would better represent a stock cooler load conditions with a warm case, though note we’re not actually showing stock cooler performance here.

So by dropping load temperatures from 84c to 55c. The CPU will increase boost from 3975 to 4200Mhz which is doable on water. You get 225MHz increase just from better cooling. A chiller keeps the temperature low so that the CPU will boost better and the temperature won't increase with heavier loads.

Down to 36 degrees, we see scaling continue to 4225, then again at 18 to average roughly 4260, with individual cores bouncing around more. Eventually, we stop scaling at -80 degrees LN2 pot temperature, so the CPU is somewhere between 0 and -80, where the all-core frequency hits 4300MHz.

So you only get <40Mhz for going sub ambient on a 3900x. So the lower you can get Tdie the better but only up to a point. I have seen Time spy cpu hit 48.5c on my water cooling on a cold day and boost to 4.325MHz with a 3800x. With a chiller this would be permanent. I would never go above 50c in time spy cpu. It would be a really cold day every day and hopefully that would mean that the frequency would never drop in time spy cpu.

Normally I would hit 56c. This is way cooling is important with Ryzen 3000 series cpu's. In games like Destiny 2 with water cooling, the cores all boost above 4.4GHz. Between 4.4-4.25GHz with some cores clocking up and down. I have PBO off in BIOS but core voltages will sit at 1.4 volts when aida64 is running. There has been no degradation so far over the 6 months my system has been running.

Each so a chiller is very expensive for a little more performance over normal custom water cooling. By having a temperature below ambient you can control how you will boost. Having a cooler 16c day or a warmer 21c day wont matter. The coolant temps will be the same because of the chiller.

Hopefully this should allow greater and safe manual all core overclocks. Or higher boosts with pbo enabled.

 

[drea.drechsler]

...
Each so a chiller is very expensive for a little more performance over normal custom water cooling. By having a temperature below ambient you can control how you will boost. Having a cooler 16c day or a warmer 21c day wont matter. The coolant temps will be the same because of the chiller.
...

I do recall those charts...so even a chiller really isn't worth it. Just as I remember their tests though...you really have to push it deeply sub zero to get big improvements even for fixed overclocks. That means LN2. I have wondered, though, how those charts would change if they had the BIOS/AGESA and chipset driver revisions AMD subsequently came out with...1003abba and 1004b mainly.

I've been running my 3700X on PBO for somewhere around 8 mo's now...ever since I got it almost. Before I got PBO well optimized (EDC=10 bug), but after AMD got AGESA fixed, it would boost in games 4.25-4.3 @ 1.4-1.43V. Now it's boosting 4.35-4.425G with the same voltage. In heavy sustained all-core loads (Folding) it's sitting around 4.2-4.25G with volts about 1.30-1.33V. In Prime95 it goes as low as 4.15G / 1.275V. That's with a 240 AIO cooling it, it wouldn't do that with the Wraithe Prism I'm pretty sure.

I'm not really certain what manual all-core overclocking could get for this. I think I'd have to do 4.15Ghz all the time to keep voltage lowered to 1.275V at most to be safe. That means I'd lose so much performance with games and even all-core Folding at Home. Just not worth it.

 

[tiberium87]

Very nice discussion.

The maximum scale, not that it’s particularly relevant to users, is 6.4% from 3120 points to 3319 points, scaling from -80 to 84 degrees Celsius. Even so most of that 6.4% is to the left of the graph.

I agree that the extra cooling does not seem to be worth it..

I was afraid that the Eco-mode function in Ryzen Master would cause me to loose too much performance, but it looks like it's basically perfect for achieving what I was going for with my manual tuning. I was manually tuning down to 3.8GHz at 1.1375V to get a nice 56 - 58C temp after hearing your initial 1.2V cautions (for my all-core workloads, which is the primary purpose of this PC).

Looks like Eco mode does exactly this. See: Eco Mode RM screenshot. So 3.8 - 3.9GHz at 1.12-1.14V with temps of 56C. Most critically, I imagine much of the careful current management (and even transient single core boosting) that the smart people at AMD designed is also still in place if I use the eco functionality (presumably its considered stock)? That is a huge safety plus. Just a quick question here: those red PPT, TDC and EDC dials aren't a cause for concern right? Since the limits have been shifted way down for these metrics?

More precisely, I am referring to what you said here:

The thinking is that at the tiny feature sizes of this 7nm TSMC process it's core currents that are at least equally important to control as temperature. Core current is driven by voltage, hence the standard guidance to run no more than 1.2V fixed. Many people have kept very low temperatures and have still experienced degradation within a few months, some running as low as 1.30V. They've realized there's doubtless a very good technical reason the boost algorithm will pull voltage as low as 1.2V so that's what you have to consider as 'safe' if you're not going to let the algorithm keep it safe.

The reasoning comes from how the Ryzen boost algorithm works: it allows fairly high temperatures, well into the 80C range, even up to 95C Tjmax, but pulls voltage (and therefore frequency too) to keep it safely within the FIT margins when it gets there.

I was initially put off by the 33% performance hit quoted for the eco mode, but now have come around to it since it really does seem to be a great way to keep the CPU healthy if one wants to run all-core loads 24/7 without a fancier cooling setup.

 

[zx128k]

I do recall those charts...so even a chiller really isn't worth it. Just as I remember their tests though...you really have to push it deeply sub zero to get big improvements even for fixed overclocks. That means LN2. I have wondered, though, how those charts would change if they had the BIOS/AGESA and chipset driver revisions AMD subsequently came out with...1003abba and 1004b mainly.

I've been running my 3700X on PBO for somewhere around 8 mo's now...ever since I got it almost. Before I got PBO well optimized (EDC=10 bug), but after AMD got AGESA fixed, it would boost in games 4.25-4.3 @ 1.4-1.43V. Now it's boosting 4.35-4.425G with the same voltage. In heavy sustained all-core loads (Folding) it's sitting around 4.2-4.25G with volts about 1.30-1.33V. In Prime95 it goes as low as 4.15G / 1.275V. That's with a 240 AIO cooling it, it wouldn't do that with the Wraithe Prism I'm pretty sure.

I'm not really certain what manual all-core overclocking could get for this. I think I'd have to do 4.15Ghz all the time to keep voltage lowered to 1.275V at most to be safe. That means I'd lose so much performance with games and even all-core Folding at Home. Just not worth it.

Chiller is worth it if the goal is maximum performance. Also the 3700x is the lower binned silicon.

Check out this 3800x score, CPU temp 54c 5690 points


View: https://www.reddit.com/r/overclocking/comments/f24egl/3800x_cinebench_r20_5517pts_1319v_daily_oc/

https://www.techpowerup.com/forums/attachments/screenshot-202-png.132608/

https://imgur.com/QADy1WV

View: https://i.imgur.com/QADy1WV.png


Some 3700x's do 4.4GHz. This one is 4.5GHz
https://i.redd.it/k0gu7355p6z31.jpg

 

[drea.drechsler]

Chiller is worth it if the goal is maximum performance. Also the 3700x is the lower binned silicon.

Check out this 3800x score, CPU temp 54c 5690 points


View: https://www.reddit.com/r/overclocking/comments/f24egl/3800x_cinebench_r20_5517pts_1319v_daily_oc/

https://www.techpowerup.com/forums/attachments/screenshot-202-png.132608/

https://imgur.com/QADy1WV

View: https://i.imgur.com/QADy1WV.png


Some 3700x's do 4.4GHz. This one is 4.5GHz
https://i.redd.it/k0gu7355p6z31.jpg

Golden silicon is still golden silicon...I have one of the first lots of 3700x's so all brass here but I'm content anyway.

Also..that 3800x...all core 4.5G? at 1.33V fixed? wow...so long as he's happy watching a $400 dollar processor melt away I guess. I'm not all that bothered about multi-core performance, my lowly little 3700X single thread score, 512, is pretty much identical to his. Single thread helps the games more so that's the important metric to me.

And yah, chasing big 'maximum performance' numbers is a game played for it's own sake. The marginal cost of a 'chiller' for getting a couple hundred more points on CB20 multithread becomes astronomical and utterly impractical IMO. But people do it...so whatever floats your boat, I guess.

 

[zx128k]

Golden silicon is still golden silicon...I have one of the first lots of 3700x's so all brass here but I'm content anyway.

Also..that 3800x...all core 4.5G? at 1.33V fixed? wow...so long as he's happy watching a $400 dollar processor melt away I guess. I'm not all that bothered about multi-core performance, my lowly little 3700X single thread score, 512, is pretty much identical to his. Single thread helps the games more so that's the important metric to me.

And yah, chasing big 'maximum performance' numbers is a game played for it's own sake. The marginal cost of a 'chiller' for getting a couple hundred more points on CB20 multithread becomes astronomical and utterly impractical IMO. But people do it...so whatever floats your boat, I guess.

If you are going to get a really big custom loop with multiple big rads and lots of expensive fans. Then just getting one chiller can be 'cheaper'. The gpu will be bought with its own water block and is not a part of the loop but I still have bought its rad and the minimum to get it running with the CPU. I also avoided expensive distro plates to keep the cost down. Also I got a bay reservoirs because it was cheaper, another no-no.

Just put together for ball park price.

Quantity Stock ID Description Price Total
WC-01Q-CS Corsair Hydro X Series XR7 480mm Quad Fan Water Cooling Radiator (CX-9030006-WW) £129.95 £129.95
WC-842-EK EK Water Blocks EK-CoolStream XE 360 (Triple Fan) Radiator £98.99 £98.99
7 WC-9GB-EK EK Water Blocks EK-Vardar EVO 120ER Black BB (500-2200rpm) Fan - 120mm £14.99 £104.93
2 WC-9AK-EK EK Water Blocks EK-Cable Y-Splitter 3-Fan (10cm) £2.99 £5.98
WC-551-EK EK Water Blocks EK-Cable Y-Splitter 2-Fan PWM (10cm) £2.99 £2.99
WC-443-XS XSPC Raystorm NEO CPU Water Block (AMD) - Black Chrome £79.99 £79.99
TH-020-TG Thermal Grizzly Kryonaut High Performance Thermal Paste - 1 Gramm £5.99 £5.99
WC-686-EK EK Water Blocks EK-DBAY D5 PWM (incl. pump) £134.99 £134.99
8 WC-255-XS XSPC G1/4" to 1/2" ID, 3/4" OD Compression Fitting (Black Chrome) V2 £3.95 £31.60
2 WC-46L-XS XSPC FLX Tubing 1/2" ID, 3/4" OD (19/13mm) - 2m Clear £7.99 £15.98
3 WC-9P9-EK EK Water Blocks EK-CryoFuel Clear 1L Premix Watercooling Fluid - 1 Litre £7.99 £23.97 WC-995-EK EK Water Blocks EK-ATX Bridging Plug (24 pin) £1.99 £1.99
WC-9EJ-EK EK Water Blocks Filling Bottle (1000mL) £6.95 £6.95
Total £644.30

Not cheap. Now a chiller. Like the water cooling just the CPU.

waterchiller Hailea Ultra Titan 300 (HC250=265Watt cooling capacity) £362.98
Brennenstuhl Power socket adapter for UK/DE £3.53
2 connection adaptor for Ultra-waterchiller 300 500 1500 to G1/4 £5.80 £11.60
WC-995-EK EK Water Blocks EK-ATX Bridging Plug (24 pin) £1.99
WC-443-XS XSPC Raystorm NEO CPU Water Block (AMD) - Black Chrome £79.99
2 WC-46L-XS XSPC FLX Tubing 1/2" ID, 3/4" OD (19/13mm) - 2m Clear £7.99 £15.98
EK D5 Vario Pump & Black Monsoon D5 Cover Combo £99.95
4 WC-255-XS XSPC G1/4" to 1/2" ID, 3/4" OD Compression Fitting (Black Chrome) V2 £3.95 £15.80
TH-020-TG Thermal Grizzly Kryonaut High Performance Thermal Paste - 1 Gramm £5.99
5 WC-9P9-EK EK Water Blocks EK-CryoFuel Clear 1L Premix Watercooling Fluid - 1 Litre £7.99 £39.97
Total £637.78

Used sources.

PC Systems & Components | Overclockers UK

Leading the way in bespoke PC systems and PC gaming, any and all of your components & peripherals needs are met with our vast range of products and services.

www.overclockers.co.uk

Water cooling

liquid cooling for PCs from Aquatuning: ✚ Huge selection ✚ Points consultation & support ✚ Fast dispatch ✚ Super price-performance ratio!

www.aquatuning.co.uk

Costs change with the addition of the graphics card. The two rads are so the water block can be changed and an Intel waterblock added for 9900K overclocking. For 9900k watercooling the rad should be changed to a
Watercool MO-RA3 360 PRO stainless steel £199.48 https://www.aquatuning.co.uk/water-...tercool-mo-ra3-360-pro-stainless-steel?c=2742 The amount of fans increased. 9 x WC-9GB-EK EK Water Blocks EK-Vardar EVO 120ER Black BB (500-2200rpm) Fan - 120mm £14.99 each fan or 18 x WC-9GB-EK EK Water Blocks EK-Vardar EVO 120ER Black BB (500-2200rpm) Fan - 120mm £14.99 each fan. For the AM4 system I would go for a monoblock covering cpu and motherboard which should be more expensive and add a GPU with its own waterblock (cost for gpu part of gpu price). You can increase the chiller to waterchiller Hailea Ultra Titan 1500 (HC500=790Watt cooling capacity) to add the GPU (£571.72 https://www.aquatuning.co.uk/water-...an-1500-hc500-790watt-cooling-capacity?c=2745)

As soon as you go for an extreme water loop, the cost of a chiller is not so bad. Remember most good looking water loops use a good looks distro plate. EK-Quantum Reflection Evolv X D5 PWM D-RGB Distro Plate £229.99 https://www.overclockers.co.uk/ek-w...5-pwm-d-rgb-distro-plate-plexi-wc-9y6-ek.html Also its wise to have two pumps. A case designed for the water cooling loop which is very expensive. You can save this money with a chiller, you have to use soft tubes and the chiller can only go outside the case. A distro plate wont make things look any better.

If your insanely expensive cooling hits over £1000, then just get phase change.

 

[drea.drechsler]

...

If your insanely expensive cooling hits over £1000, then just get phase change.

I'd only do a custom closed loop for a show-case system to wow the guests with colored LED displays and liquid loops. Which is to say...never.

A chiller won't wow many guests at all...just the nerds.

I'm content with my $79 240mm AIO. Pushing an order of magnitude cheaper and a lot less periodic maintenance. Works just as effectively at it's primary function since I don't need to finish a 30 minute video render 10sec's sooner and I'm definitely not interested risking loss of gaming performance.

I just don't see the value proposition for Ryzen 3000 processors.

 

[zx128k]

I'd only do a custom closed loop for a show-case system to wow the guests with colored LED displays and liquid loops. Which is to say...never.

A chiller won't wow many guests at all...just the nerds.

I'm content with my $79 240mm AIO. Pushing an order of magnitude cheaper and a lot less periodic maintenance. Works just as effectively at it's primary function since I don't need to finish a 30 minute video render 10sec's sooner and I'm definitely not interested risking loss of gaming performance.

I just don't see the value proposition for Ryzen 3000 processors.

I would get a 240-280mm AIO for the 3700x and below and a 360mm or custom water cooling for everything above. Chiller is nice but very luxury and its not a looker like you state.

 

[drea.drechsler]

I would get a 240-280mm AIO for the 3700x and below and a 360mm or custom water cooling for everything above. Chiller is nice but very luxury and its not a looker like you state.

To be fair, a BIG air cooler would work just as sweetly on even a 3900X. By that I mean something like an NH-D15 or Dark Rock Pro 4, price tag about the same as most 240mm AIO's. The problem with them, though, is you have to be really careful about balancing case ventilation and case fans or you end up pumping the GPU's hot air into the cooler with screaming fan speeds even though the CPU's not working all that hard. That's not effective at all.

The really good thing about an AIO is you just need a case with mesh front and front mounting for the radiator. Plenty enough of those, and cheap ones too. But then you can be just about as clumsy as you like and CPU cooling will still be effective.

 

[zx128k]

To be fair, a BIG air cooler would work just as sweetly on even a 3900X. By that I mean something like an NH-D15 or Dark Rock Pro 4, price tag about the same as most 240mm AIO's. The problem with them, though, is you have to be really careful about balancing case ventilation and case fans or you end up pumping the GPU's hot air into the cooler with screaming fan speeds even though the CPU's not working all that hard. That's not effective at all.

The really good thing about an AIO is you just need a case with mesh front and front mounting for the radiator. Plenty enough of those, and cheap ones too. But then you can be just about as clumsy as you like and CPU cooling will still be effective.

You do want a custom water loop but some of those 240mm AIO's are no better than good air coolers.
My case has a 200mm fan over the GPU. Having really good air flow in the case is great as well.

A case without air flow issues. Here the cpu gets 55c in aida64 which is nice. Just 90dB sound wise.

Just go phase change. 9900k 6GHz well 5.9GHz.

 

[drea.drechsler]

You do want a custom water loop but some of those 240mm AIO's are no better than good air coolers.
...

Zen 2's unique problems again: that tiny contact patch the 7nm die has with the heat spreader can't transfer the heat fast enough to make a difference. In terms of heat transfer rate to the radiator a 360mm AIO can't really cool any better, nor can a big air cooler. But the 360mm rad can transfer the heat from the water to the air better so that coupled with a larger water volume gives (way) more time till the water thermally saturates. Curiously though, I can't say my 3700x has ever saturated my 240mm AIO so I really don't think one could help at all.

I've seen it with the 120mm AIO on my son's FX6300...temperature's OK for a while but then about 45min to an hour in to a stress test it starts climbing and goes unstable.

 

[zx128k]

Zen 2's unique problems again: that tiny contact patch the 7nm die has with the heat spreader can't transfer the heat fast enough to make a difference. In terms of heat transfer rate to the radiator a 360mm AIO can't really cool any better, nor can a big air cooler. But the 360mm rad can transfer the heat from the water to the air better so that coupled with a larger water volume gives (way) more time till the water thermally saturates. Curiously though, I can't say my 3700x has ever saturated my 240mm AIO so I really don't think one could help at all.

I've seen it with the 120mm AIO on my son's FX6300...temperature's OK for a while but then about 45min to an hour in to a stress test it starts climbing and goes unstable.

I am running a 3800x with a custom water loop. See, https://forums.tomshardware.com/thr...who-makes-the-best-cpus.3596763/post-21726383 where I beat some 9900k/ks systems in games. I never hit 70c.

 

[drea.drechsler]

I am running a 3800x with a custom water loop. See, https://forums.tomshardware.com/thr...who-makes-the-best-cpus.3596763/post-21726383 where I beat some 9900k/ks systems in games. I never hit 70c.

Sounds about right...I'm running a 240mm AIO. A 3800x should get better performance than my 3700x, and being better binned at lower voltage, therefore run cooler. But even as it is I have to push things really hard just to get to 70C and only goes into the low-mid 70's in crazy stupid tests like prime95 small FFT's. Your custom loop probably benefits greatly from high-flow waterblock and pump, transferring heat a little faster off the heat spreader. But to get only 3 or 4 degrees better cooling, I again just don't see that the cost-benefit ratio is there to justify it. I'd need a new, big case (to fit the plumbing, pump and reservoir) and the requisite maintenance is just not something I'm interested in. I'll just toss my $79 AIO and get another one (at inflation adjusted price) in 5 years once the block becomes gunked up.