If top blowing coolers are so good, why they are so uncommon?

[pavle]

I will tell you from my own experience - when I used my nForce2 mainboard (good ol' Epox) and had Athlon XP 2500+@3200+ - at first I had that Thermaltake orb cooler that I had since my Duron600@900/1000 days, but it was a bit wonky from all that mounting and dismounting, I thought I'd buy a "tower" cooler. At the time I didn't think much about importance of downdraft for the VRM's.
So yeah, I had my little tower cooler and CPU temps were great, never knew what VRM temps were till I upgraded my system to my Core2 Quad setup and then - my oh my - the poor old Epox nF2 board was literally toasted around the VRM's - brown as it were. And that with many fans in the case.
From then on I hate tower coolers. Always use downdraft (towards the CPU socket) coolers for optimal cooling of CPU and surroundings. If factory one is too small, no problem, there are alternatives, as mentioned in the original/first post.
Poor old Epox board - still works though. How's that for tough? Not TUF or whatever else but TOUGH!

 

[Vayra86]

ah, a thread that made me remember how my Scythe Grand Kama Cross (Rev.B) came to be my favorite "top down" (and HSF) ... aaahhhhh if only i did not sell it ...
View attachment 214218

well i love my tower ETS-T50 nonetheless ... but man ... i want to find a Grand Kama Cross again now ... to add it to my HSF collection

This thing looks absolutely brutal, I gotta say. Reminds me of Zalman for being so out there

 

[Blue4130]

We don't need to. You are arguing against common knowledge and common sense, so the burden of proof is on you.

Positive pressure doesn't remove stagnant air. That is what he posted. That is common knowledge? Blowing is positive pressure. Your coffee is hot. You blow on it. It cools down. In his quote, that would not happen.

 

[lexluthermiester]

Positive pressure doesn't remove stagnant air. That is what he posted. That is common knowledge? Blowing is positive pressure. Your coffee is hot. You blow on it. It cools down. In his quote, that would not happen.

Seems some context is being missed. Go back to the photo of that heat sink, see the space between the fins right down the center axis of the fin array? Yeah, some of the airflow travels into and through the gap providing airflow to the area that you're talking about. Not stagnant air space. Basic fluid physics there.

 

[Blue4130]

Seems some context is being missed. Go back to the photo of that heat sink, see the space between the fins right down the center axis of the fin array? Yeah, some of the airflow travels into and through the gap providing airflow to the area that you're talking about. Not stagnant air space. Basic fluid physics there.

And that positive pressure removes the air/heat. Thus making his claim invalid.

 

[mtcn77]

And that positive pressure removes the air/heat. Thus making his claim invalid.

It is not working 'efficiently'. Just air presence doesn't have to be good conduction happening. You are thinking of open spaces, I'm speaking of axial fans blowing on a fin array and missing spots. Not generating equal flow throughout the stack should be obvious to anyone otherwise we wouldn't have turbulent designs, every fin array would have identical stamped fins, do they?

 

[delshay]

I have a very, very large collection of downdraft coolers from all copper tower type to the flower type. Some have IHS soldered to the coldplate.
I carried out a number of experiments so that I can understand better the transfer of heat.

 

[The red spirit]

There are a lot of reason. As has been said, physical space around the socket is one reason, but another is difficulty of manufacturing. It's greatly easier to make a tower style cooler than a cooler the bends over above the socket area. This requires a lot more effort in the design phase, more tooling and processing then normal "stamp & clamp" types. This makes them more expensive to build and the profit margin lower. From a business perspective it's not as appealing.

And still somehow Choten costs the same as CM 212 EVO. Well at least in Europe. The only design difference I see is that you need longer heatpipes and to bend them more. Other than that, I don't see the difficulty. And maybe it's just me, but when buying Choten, space around socket wasn't a consideration to me. As long as you don't have some unusual VRM heatsinks, it should fit. It's not like Godhand or Susanoo, which make top PCIe slots useless

There's nothing wrong with irrational preferences or dislikes as long as they're not trotted out as valid critiques. There are plenty of empirically tested product stack reviews out there and that's what's useful to a builder.

It's not really an irrational preference. I have read motherboard manuals in CPU installation sections and often they mention that you need downdraft cooler for motherboard cooling, unless you don't mind your board not working as long. And very good OCN guide claims that 71% of all motherboard failures are from overheating VRMs. That's quite significant.

 

[Vayra86]

It is not working 'efficiently'. Just air presence doesn't have to be good conduction happening. You are thinking of open spaces, I'm speaking of axial fans blowing on a fin array and missing spots. Not generating equal flow throughout the stack should be obvious to anyone otherwise we wouldn't have turbulent designs, every fin array would have identical stamped fins, do they?

This. Which is also sort of the point I was making earlier.

Negative pressure basically creates a 'pull' of air and the temperature determines what gets pulled out, while the dust bunnies get forced into the intake airflow before any other holes that aren't filtered. This is also why you want a simple, fixed and well optimized airflow direction in the case. Yes, you can exhaust heat anywhere, but the best way to do it, is to make sure it all gets dragged into that flow from front to back of the machine.

To the coffee example. You can blow for ages on coffee and still find it hot. But you can also have it standing there for 10 minutes and find it cooler, because the heat was removed from all ends of the cup instead of just the top surface you blew on, if it was for instance just standing in front of an open, drafty door.

And still somehow Choten costs the same as CM 212 EVO. Well at least in Europe. The only design difference I see is that you need longer heatpipes and to bend them more. Other than that, I don't see the difficulty. And maybe it's just me, but when buying Choten, space around socket wasn't a consideration to me. As long as you don't have some unusual VRM heatsinks, it should fit. It's not like Godhand or Susanoo, which make top PCIe slots useless


It's not really an irrational preference. I have read motherboard manuals in CPU installation sections and often they mention that you need downdraft cooler for motherboard cooling, unless you don't mind your board not working as long. And very good OCN guide claims that 71% of all motherboard failures are from overheating VRMs. That's quite significant.

1. Motherboard manuals take the safest approach, not the most performant one or efficient one. For the same reason, OEMs use blower GPUs. I'm sure you can agree that in any regular ATX case, a blower is really not the optimal choice. The same goes for topflow and bundled CPU coolers. They serve a purpose, yes. But it is not efficient cooling, its playing it safe when all else fails. You have to imagine that a manufacturer of a CPU will always opt to get the majority of temp variables under its own control. A topflow cooler is the better way of doing so, you just worry about your area around your CPU socket, fuck the rest and overall quality of life, such as the noise level or the required case fanspeeds to keep things tip top outside of the CPU socket.

2. Yeah... 71% of mobo failures. How many mobo's fail? The miracles of stats, eh. And then... a test performed on predominantly a subset of users doing stuff at their own risk (OCing) in a niche of the market. VRM is indeed something you want to cool, but mobo's do provide (as per above example, same thing applies in terms of control - control also over your flow of RMAs ) cooling for the VRM - heat sinks.

Now, you also have a niche of users (and cases ) here that is very keen on trying stuff that is pretty damn stupid. Note my sig. Bad case cooling and wild ideas about airflow can easily be added there.

I'm afraid that I have to concede that after 20 odd years of PC enthusiasm I'm pretty much cured of all bullshit methods people invent to get more performance or do things better than the 'norm' most people consider sane. You're looking at humongous stacks of diminishing returns there, in other words, tons of effort for extremely minor advantages, that often carry disadvantages along with them. I know, boring af, but it is what it is. Most stuff has been invented and reinvented a few dozen times over, and simplicity tends to win the day.

 

[The red spirit]

1. Motherboard manuals take the safest approach, not the most performant one or efficient one. For the same reason, OEMs use blower GPUs. I'm sure you can agree that in any regular ATX case, a blower is really not the optimal choice. The same goes for topflow and bundled CPU coolers. They serve a purpose, yes. But it is not efficient cooling, its playing it safe when all else fails. You have to imagine that a manufacturer of a CPU will always opt to get the majority of temp variables under its own control. A topflow cooler is the better way of doing so, you just worry about your area around your CPU socket, fuck the rest and overall quality of life, such as the noise level or the required case fanspeeds to keep things tip top outside of the CPU socket.

Safest? No, some cheap motherboards are designed with stock cooler in mind, as it provides quite a lot of active airflow for VRMs. It's not even optional, if you wanted FX 8320 on Asrock 970 Pro 3, you had to use stock cooler or some other top flow cooler as otherwise board would overheat. It was completely necessary. And if you used FX 6300 Asrock said that it should be fine without stock cooler, but I noticed that I had some random throttling, so most likely VRMs weren't happy with high temperatures. Obviously, if you buy only ROG shit, then you probably think that stock cooler recommendation is silly, but FX 6300 buyers was certainly not the one who would be looking at motherboards costing several times more than their CPU. And guess what, that board basically lasted warranty period despite using Mugen 4 PCGH. Asrock wasn't playing it safe, they just didn't want lawsuits about burned down houses.

2. Yeah... 71% of mobo failures. How many mobo's fail? The miracles of stats, eh. And then... a test performed on predominantly a subset of users doing stuff at their own risk (OCing) in a niche of the market. VRM is indeed something you want to cool, but mobo's do provide (as per above example, same thing applies in terms of control - control also over your flow of RMAs ) cooling for the VRM - heat sinks.

I'm certain that despite my horrendous experience with various boards, most of them last, but they do fail. Just released boards have RMA rate of around 3% with particular models being at 15% or more. And that's just first year. More failures happen later and many more after warranty period. Motherboard failures really aren't rare and I can say that out of all components in computer they break down more than others.

Now, you also have a niche of users (and cases ) here that is very keen on trying stuff that is pretty damn stupid. Note my sig. Bad case cooling and wild ideas about airflow can easily be added there.

No guidance is given in regards of cooling, so we can assume that board makes are fine with anything and thus their stuff must last.

I'm afraid that I have to concede that after 20 odd years of PC enthusiasm I'm pretty much cured of all bullshit methods people invent to get more performance or do things better than the 'norm' most people consider sane. You're looking at humongous stacks of diminishing returns there, in other words, tons of effort for extremely minor advantages, that often carry disadvantages along with them. I know, boring af, but it is what it is. Most stuff has been invented and reinvented a few dozen times over, and simplicity tends to win the day.

There's nothing complicated about downdraft coolers. It's the most intuitive and oldest cooler design. They make towers look like toddlers. I don't know what you are talking about here anymore.

 

[Operandi]

Safest? No, some cheap motherboards are designed with stock cooler in mind, as it provides quite a lot of active airflow for VRMs. It's not even optional, if you wanted FX 8320 on Asrock 970 Pro 3, you had to use stock cooler or some other top flow cooler as otherwise board would overheat. It was completely necessary. And if you used FX 6300 Asrock said that it should be fine without stock cooler, but I noticed that I had some random throttling, so most likely VRMs weren't happy with high temperatures. Obviously, if you buy only ROG shit, then you probably think that stock cooler recommendation is silly, but FX 6300 buyers was certainly not the one who would be looking at motherboards costing several times more than their CPU. And guess what, that board basically lasted warranty period despite using Mugen 4 PCGH. Asrock wasn't playing it safe, they just didn't want lawsuits about burned down houses.

Thats just a poorly designed and/or undersized VRM then. No board design should require dedicated airflow over the VRM to run a supported CPU at stock speeds. Overclocking is a different story but at stock speeds a VRM should be able cool itself as long as the heat its generating is being removed from the system.

 

[The red spirit]

Thats just a poorly designed and/or undersized VRM then. No board design should require dedicated airflow over the VRM to run a supported CPU at stock speeds. Overclocking is a different story but at stock speeds a VRM should be able cool itself as long as the heat its generating is being removed from the system.

AMD shouldn't have lied about their FX series processor real power consumption. Selling 160 watt chips advertised as 95 watt doesn't look very honest to me. And for good measure including inadequate cooler to cope with it as if they truly believed that it will be 95 watt chip.

 

[TheinsanegamerN]

I have read a lot about how top blowing coolers are good and with certain boards mandatory. They seemingly help to cool VRMs and at the same time they are usually as good as single tower 120mm coolers. And despite seemingly being great particularly for budget overclockers, they are really uncommon. So far, today only competent downdraft coolers are Scythe Choten, Noctua NH-C14 and BQ Shadow Rock TF2. If they are good and help to cool down motherboard (VRM heat is still a big problem), why they are still so uncommon? It's not like they are new or anything, Scythe made Andy Samurai Master in 2003, so they are very well known and likely one of the first big upgrade over stock Athlon XP or Pentium 4 (socket 423) heatsink. I'm pretty sure that they can be manufactured rather well and they aren't really more expensive than single tower 120mm heatsinks.

Biggest part IMO is limited space. A blow down cooler is limited by the size of VRM heatsinks and board configuration. That limited space can only allow so much metal and thus so much heat dissipation. With CPUs pushing over 200 watts the top down cooler, while effective, is incapable fo keeping up.

Their design also makes it harder to take advantage of heatpipe tech. You see some, like noctuas, with heatpipes, but they're limited to 65 watt CPUs. Running a K series under them usually leads to uncomfortable temps,a nd that was back in the 6 core days.

AMD shouldn't have lied about their FX series processor real power consumption. Selling 160 watt chips advertised as 95 watt doesn't look very honest to me. And for good measure including inadequate cooler to cope with it as if they truly believed that it will be 95 watt chip.

The entireity of bulldozer was one gigantic misfire.

 

[Operandi]

AMD shouldn't have lied about their FX series processor real power consumption. Selling 160 watt chips advertised as 95 watt doesn't look very honest to me. And for good measure including inadequate cooler to cope with it as if they truly believed that it will be 95 watt chip.

Fun fact manufacture stated TDPs are not particularly useful figures for real world power consumption. More to the point I don't think motherboard engineers should be designing their VRMs based on the specs printed on the side of the box.

 

[The red spirit]

Fun fact manufacture stated TDPs are not particularly useful figures for real world power consumption. More to the point I don't think motherboard engineers should be designing their VRMs based on the specs printed on the side of the box.

Well, yes, but TDP at least gives you some kind of ballpark estimate of how much chip will consume in some high loads. Rating FX 6300 as 95 watt chips, while it can reach 160 watts is straight up lawsuit material. And I'm sure that board makers get more data from AMD or Intel about CPU capabilities. Maybe AsRock was cheaping out too much or maybe they were given too conservative data, I don't know, but Bulldozer and Piledriver era boards were all dying left and right, doesn't matter to which manufacturer. And it wasn't just for FX chips, but for FM2 and FM2+ chips too. And they lied about 6300 a lot, but they also made 125W FX 8320(50/70), which could consume over 200 watts. And if that wasn't enough, AMD also released 9370 and 9590, which were rated for 220 watts, but consumed a lot more than that. Those chips are just higher clocked 8300 series chips, not even better bins, so essentially they are as good as overclocked FX 8300 chips. Board manufacturers wisened up there and made FX 9000 series only compatible with their best boards, but the damage was already done.

Their design also makes it harder to take advantage of heatpipe tech. You see some, like noctuas, with heatpipes, but they're limited to 65 watt CPUs. Running a K series under them usually leads to uncomfortable temps, and that was back in the 6 core days.

My Choten can cope with i5 10400F with power limits removed. That's around 100 watts. CPU stays at 70s. So if Choten is at full fan speed, which is just 1200 rpm and acceptable temperature of chip is 85C, I think that it could handle 115 watt chips too. 65 watts are nothing for big downdraft cooler.

 

[Operandi]

Well, yes, but TDP at least gives you some kind of ballpark estimate of how much chip will consume in some high loads. Rating FX 6300 as 95 watt chips, while it can reach 160 watts is straight up lawsuit material. And I'm sure that board makers get more data from AMD or Intel about CPU capabilities. Maybe AsRock was cheaping out too much or maybe they were given too conservative data, I don't know, but Bulldozer and Piledriver era boards were all dying left and right, doesn't matter to which manufacturer. And it wasn't just for FX chips, but for FM2 and FM2+ chips too. And they lied about 6300 a lot, but they also made 125W FX 8320(50/70), which could consume over 200 watts. And if that wasn't enough, AMD also released 9370 and 9590, which were rated for 220 watts, but consumed a lot more than that. Those chips are just higher clocked 8300 series chips, not even better bins, so essentially they are as good as overclocked FX 8300 chips. Board manufacturers wisened up there and made FX 9000 series only compatible with their best boards, but the damage was already done.

I have little doubt that it is more than 95 watts but how was that 160 watt number derived though? If that was measured AC at the wall then you have to figure 15-25% loss in the PSU and another 5-10% depending how good the VRM is. My overclocked 8350 draws 250 watts but thats the entire system and with a 80+ Bronze PSU so overclocked maybe my 8350 is responsible for 200 of those watts but thats before PSU and VRM losses.

Not really the point though, its on the board manufacture to properly support the chip. I think what happened in that era is AMD tried a big powerful design that didn't end up working out and they were stuck with it. AMD still had to sell CPUs but way cheaper than they wanted to and that put board manufactures in a tight spot trying to design a price appropriate board to support those power hungry CPUs.

 

[lexluthermiester]

Fun fact manufacture stated TDPs are not particularly useful figures for real world power consumption.

Not true. As a general rule, Intel and AMD both have been listing TDPs to reflect both heat output and power usage. See below;

My 130W TDP Xeon E5-2667V2 is using 121ish watts at full load, full boost clocks, not OC'd. Most(if not all) CPU's made in the last 10 years will consume power wattage very close to their stated TDP wattage.

 

[mtcn77]

Not true. As a general rule, Intel and AMD both have been listing TDPs to reflect both heat output and power usage. See below;
View attachment 214424
My 130W TDP Xeon E5-2667V2 is using 121ish watts at full load, full boost clocks, not OC'd. Most(if not all) CPU's made in the last 10 years will consume power wattage very close to their stated TDP wattage.

I think AMD is particularly susceptible to this Poole Frenkel Effect due to their monitoring software's approximation algorithms, at least their gpus. Their cpus have an HPC bios switch(a primordial pbo hack job) that enable full tdp at rated temperature threshold, so they still might have restrictions to fall back to.

 

[jayjr1105]

Not true. As a general rule, Intel and AMD both have been listing TDPs to reflect both heat output and power usage. See below;
View attachment 214424
My 130W TDP Xeon E5-2667V2 is using 121ish watts at full load, full boost clocks, not OC'd. Most(if not all) CPU's made in the last 10 years will consume power wattage very close to their stated TDP wattage.

Won't a 9900K pull like 220+ watts under certain loads? It's a 95w stated TDP btw.

 

[Shrek]

By top blowing, do we include holes in the top or the case? I worry over spills and would prefer if the top was sealed.

 

[lexluthermiester]

Won't a 9900K pull like 220+ watts under certain loads?

Only when OC'd. At stock speeds and voltages, it should be right in the area of it's TDP.

By top blowing, do we include holes in the top or the case? I worry over spills and would prefer if the top was sealed.

Don't store/keep liquids near your case. Problem solved.

 

[jayjr1105]

Only when OC'd. At stock speeds and voltages, it should be right in the area of it's TDP.

Don't store/keep liquids near your case. Problem solved.

Bone stock it still hits 140-200w under a variety of productivity workloads. Intel real world hasn't been near their TDP's for a long while.

 

[lexluthermiester]

Bone stock it still hits 140-200w under a variety of productivity workloads.

Screen shots? I'm not buying that unless someone shows the CPU package power sensors displaying that data.

Intel real world hasn't been near their TDP's for a long while.

W1zzard's own testing shows that to be untrue.

Intel Core i9-9900K Review

Today, Intel released their new flagship processor for the LGA 1151 platform. The Core i9-9900K finally comes with eight core and 16 threads, reaching parity with AMD's Ryzen offerings. Maximum Boost Clock has been increased as well, now to a staggering 5 GHz.

www.techpowerup.com

And those are whole system measurements, GPU included.

 

[The red spirit]

Won't a 9900K pull like 220+ watts under certain loads? It's a 95w stated TDP btw.

Depends. Due to PL2 it will use more watts for like 58 seconds for short bursty loads and will likely boost to maximum frequency, but once Tau is expired (PL2 allowance period) it will consume no more than PL1 stated watts, which is 95. Due to workloads consisting of various difficulty and their execution time, CPU might be switching between PL2 and PL1 a lot. Also motherboard vendors are seemingly obsessed to raising PLs way above Intel recommended values and and thus some people will see that their i9 always consumes a ton of power and is hard to cool. There are also PL3 and PL4 specs, but those are power limits for fractions of second and generally aren't accessible, modifiable or disclosed in BIOS or in monitoring tools. They also aren't publicly know either.

This power logic has been in Intel processors since Haswell era and maybe since Sandy Bridge, I don't know precisely, but it is quite mature.

I have little doubt that it is more than 95 watts but how was that 160 watt number derived though? If that was measured AC at the wall then you have to figure 15-25% loss in the PSU and another 5-10% depending how good the VRM is. My overclocked 8350 draws 250 watts but thats the entire system and with a 80+ Bronze PSU so overclocked maybe my 8350 is responsible for 200 of those watts but thats before PSU and VRM losses.

Prime95 excluding possible losses by guessing and approximating. Reviewers had equipment to measure that more directly.

Not really the point though, its on the board manufacture to properly support the chip. I think what happened in that era is AMD tried a big powerful design that didn't end up working out and they were stuck with it. AMD still had to sell CPUs but way cheaper than they wanted to and that put board manufactures in a tight spot trying to design a price appropriate board to support those power hungry CPUs.

I see where the logic is here, but I highly doubt that AMD didn't know that. They are designing chips, testing prototypes and doing a lot of work. For such a massive deviation to slip up is near impossible. AMD also revised original Bulldozer with Piledriver architecture and if that mattered to them, they could have reduced power usage. Later there was Carrizo, which was shockingly low power bulldozer derivative, but it was launched right near Ryzen launch, but it shows that AMD could do that to bulldozer or piledriver. But for some reason didn't. Previously AMD had poor Phenom launch, which couldn't clock high enough, but they didn't push power usage for that, instead they waited for better bins. Maybe they didn't want that to happen with FX?

 

[Shrek]

Don't store/keep liquids near your case. Problem solved.

No offense, but good Engineering accounts for the klutz (the klutz being me in this case)