5800X3D - Heat fix for me

[kksnowbear]

On the "This is fine", as I recall, Jay was referencing the "this is fine" meme, not saying that was the right way to mount things. 

Well, I would certainly stand corrected if that's the case - and admittedly, I didn't watch it, only part(s) of the first video.  I appreciate the correction. 

To be fair and accurate, though, I did refer to the picture (here) and it appears wrong prima facie.  I also said I wasn't sure why it said that.

Some people just take the wrong thing from somewhere it wasn't intended.

On that point, this reinforces my contention - here and in other threads *ahem* - that you really have to be careful about the impression you give others.  *Some* might know better but others might not.

Now if you'll allow, I'd be happy to address your other questions separately.

EDIT: Sorry, I watched about 4 minutes of the JayZ video and just couldn't take any more.  (Don't get me started).

I will say that it's difficult to imagine how very *little* a person would have to know for some of these videos to actually be useful.  Seriously.

Also, I'll have to stand on my premise about the picture.  It remains that if I saw it, others can.  And of itself it depicts an inaccurate concept.  (I can literally feel it being posted all over the internet by people claiming it's proof that "This is fine", even if maybe some of us know better).

Edited January 19, 2023 by kksnowbear

[kksnowbear]

On the front rad, one thing I wonder is, how it that different from having an internally exhausting sir cooled GPU?

Good question, one that should be discussed...however, at least initially, consider that I don't believe I've said anything about internal air cooled GPUs.  The discussion has been about front mounting a liquid radiator as opposed to top/rear mounting (as you can see, both assume liquid cooling).

Broadly and generally, concerning the woes associated with mounting radiators, I'll say this: It's not smart (to me) to spend a ton of money on a component, and more money still on a cooling system, and then diminish the performance of said arrangement by shoe-horning it all into a case that's a poor fit, just because that's what you have.

    Ya spent all that money on GPUs, CPUs, and cooling, get a case that aids in the goal (cooling) instead of opposing it (introducing heated air).

Unfortunate if you'd have to dismantle anything, but it's either that, or you're intentionally choosing an arrangement that even common sense tells you cannot be optimal.

Not directed at any individual, but it seems apparent here that some people will sacrifice "optimal" to save a little work or cost - while introducing the very thing they're paying to get rid of.

And that's fine, if it's your choice.  But call it what it is.  My only problem is when someone tries to toss the laws of thermal dynamics out the window by acting like there's no difference.  Fact and physics say otherwise.

Edited January 19, 2023 by kksnowbear

[tomcat_driver]

The setting that did the trick for me was Kombo Strike 3. Under my NH-U14S, temps were hitting 89C. After turning this BIOS feature on, it's cool and quiet below 70C.

 

[Dragon1-1]

1 hour ago, kksnowbear said:

And that's fine, if it's your choice.  But call it what it is.  My only problem is when someone tries to toss the laws of thermal dynamics out the window by acting like there's no difference.  Fact and physics say otherwise.

Actual experiments say that you're wrong. It doesn't matter. Experiment trumps theory all the time, and this had been tested. Turns out, there are minor differences between top and front mounted rads, GPU runs a little hotter, mobo runs a little hotter, CPU runs cooler. 
https://www.msi.com/blog/how-to-place-your-liquid-cooler

The only component of my rig that has any thermal issues whatsoever is the CPU, therefore the front mounted configuration is the best for me. There are two important things to take away from those tests, which is to keep the air away from the tubes and pump (so make sure the highest point is not the tubes or the pump), and to have something pulling the air into the case.

[kksnowbear]

That link from MSI shows an article, not test results. There's almost no empirical data, and zero mention of methodology or any other factor that is *crucial* to proper testing.  There is no way at all to verify the findings of their "tests" without actual published data and procedures.

The simple fact is MSI sells liquid coolers.  They are motivated by sales.  If they acknowledge that it's a bad idea to have a liquid radiator exhausting into a case, they're discouraging sales to people with cases that are ill-suited to radiators in the first place.  I don't know how many cases in the world there are that can only house radiators in the front, but for every one of them, that's a potential lost sale unless the owner is convinced that it's not a problem.

So I'm the marketing director over at MSI, I'm gonna tell my staff to get busy showing it's a good idea - without really publishing anything other than an article saying "Trust us...we tested it"

This is a problem with cases people already have, and either don't want to or cannot replace, and that's all.  It absolutely does not prove it's a good idea to exhaust a radiator into a chassis - it's biased marketing, plain and simple.

And, as I've already explained: My points are not theory, they are practice, backed by formal training and over 40 years' experience in professionally-compensated computerized systems maintenance.  I don't need to cite articles from others.  If you want to compare factual data on our own actual first-hand real world experiences, I welcome the opportunity.

It's pretty, but there's not enough space... well, anywhere, really.  So you ended up putting the radiator up front because of your case.  Fairly typical; case manufacturers consider quite a few factors when designing cases, but optimal liquid cooling isn't necessarily one of them.

I'd have mounted the radiator on top, but there's no room up there.  Good; we agree the mounting was driven by available space, as opposed to an arrangement for optimal cooling (which you acknowledge you'd have done if there were room).

One problem with the idea that you're "not blowing hot air into the case" based on this example is that you're cooling the CPU.  Different story if a GPU rad is installed as an intake (that's what the OP's doing I believe), and actually cooling a significant load.  I can't speak about CPUZ as a stress test, I don't use that.  Run Prime 95 for a few hours and then look at the CPU temp, not the coolant temp.  If I fail to properly install a cooler (air or liquid) then the cooler/coolant itself won't get warm, but the device being cooled will be much warmer.  It doesn't really matter what temp the coolant is, the goal is a lower temp at the CPU or GPU, not the liquid in the radiator.

By the way, if your ambient air is 69F and the coolant is 32c...then you most assuredly *are* blowing heated air back into the case.  I don't know about you, but 90F(32c) is pretty warm for most people.  In our context, if the air being introduced isn't the same or near ambient, it's not really "cool".  Result is that the inside of your case is warmer than if it had a direct intake of outside ambient air (69F in your example).

The entire system will absolutely cool better if the intake air is cooler, and that's just physics.  Putting a radiator in the front panel inherently means the typical primary path for introducing cool air is now occupied by something that's heating the air.  Even in your case, there's a ~21F delta in ambient/intake vs. exhaust/internal, and it's strictly because there's a radiator on the front panel which is heating the main source of intake.

You're going to be hard pressed to prove that heated air at the intake doesn't cause warmer temps inside.  At that point, you're not arguing with me - you're arguing with the laws of physics and thermal dynamics.

And yes, it's going to be warmer during the summer.  Can't really count cooling in a room that's 69F ambient as a good test; I doubt it's typical for many situations especially during the summer.  Best practice testing assumes worst case.  When I 'burn in' an overclock, I intentionally make the room 80F (even using a space heater if need be) so that I'm testing what is a likely scenario (upstairs bedroom, game-crazed teenager with door closed), or at least something that is slightly beyond what can be reasonably expected - thus testing will exceed the need.  Tests that don't consider likely extremes are sort of pointless.

You also have to factor in a sort of thermal dynamics, whereby introducing warmer air means the components inside will in turn run warmer, which will of itself (and due to physics) mean the air inside the case is now warmer too, and so on...in a closed box, with air intake that's not enough volume and/or not low enough a temp, the (actual component) temps *will* eventually increase beyond what they'd have been with cooler air coming in - again, just physics.

Your CPU might have actually been cooler with a good air cooler, leaving the front panel for good intake fans - and it might've cost less, too.  (Also everything else, including the motherboard VRMs, would be cooler too - something a liquid cooler simply cannot do).

Your situation is like many: Cooling system performance is impacted due to mechanical constraints.  Nothing personal, it is what it is, but let's just be honest: What makes it acceptable to you is that it fits in your chassis - but that's at the expense of blowing heated air into your system.

I get it, it's a compromise - but let's just call it what it is, and not act as if the laws of thermal dynamics don't apply.

Edited January 19, 2023 by kksnowbear

[edmuss]

My CPU is air cooled sucking cool air downwards and exhausting onto the backplate of the GPU, my GPU is air cooled and vents hot air out of the side and radiates off the backplate up to the CPU. I don't have a case, it's just a motherboard tray and rear panel screwed to the wall, completely open to ambient.

In DCS the 5800x3D sits at around 65-70°C and the GPU sits at 55-65°C with an ambient of around 24-28°C.  On 100% load testing the CPU hits 75°C and the GPU hits around 80°C.

[tomcat_driver]

2 hours ago, edmuss said:

My CPU is air cooled sucking cool air downwards and exhausting onto the backplate of the GPU, my GPU is air cooled and vents hot air out of the side and radiates off the backplate up to the CPU. I don't have a case, it's just a motherboard tray and rear panel screwed to the wall, completely open to ambient.

In DCS the 5800x3D sits at around 65-70°C and the GPU sits at 55-65°C with an ambient of around 24-28°C.  On 100% load testing the CPU hits 75°C and the GPU hits around 80°C.

Do you have Kombo Strike 3 or a custom undervolting?

[edmuss]

Undervolted on both, -27mv on the CPU and 975mv on the GPU running at 2040Mhz clock.
Unrealistic CPU loads like prime95 small FFTs smash it straight up to 90°C but it's still stable and there is no normal usage that will cause such a load; on the other sections of prime95 testing the CPU sits at about 60-65°C.  A 10 minute cinebenchR23 run gets it up to around 72-75°C with no thermal throttling.

[Dragon1-1]

16 hours ago, kksnowbear said:

That link from MSI shows an article, not test results. There's almost no empirical data, and zero mention of methodology or any other factor that is *crucial* to proper testing.  There is no way at all to verify the findings of their "tests" without actual published data and procedures.

null

They even mention ambient temp (25 degrees). I agree, that's not up to research standards, but for PC tuning, it will do. If you have that kind of rig you can probably replicate that with the data provided. Seriously, I'm a scientist by trade. I wouldn't have linked something that didn't meet those standards.

Yes, this is done by MSI to sell coolers. They like to sell coolers, but they also like not to see them returned as defective and/or fix broken pumps on warranty. Thus, they have a vested interest in people treating them right. They didn't word it very strongly for that reason (bottom radiator is "not recommended" and not "the dumbest thing we've ever heard of"), but the findings themselves look valid. 

Your experience is nice, but how the needs of a gaming PC compare to a "professionally-compensated computerized system" anyway? My friend has a threadripper workstation for chemical simulations at home (funnily enough, it'd suck at DCS due to single core clocks being rather modest), and that thing requires a completely different approach to cooling. Not much consideration for noise in those applications, either (I like my PC to be reasonably quiet). I cite a test that I don't have a good way of doing myself, from folks who make gaming hardware. You seem to think that just because you can set up a cooling loop in a datacenter, you know all about how to do it inside a PC case that has to fit on a normal desk in normal living space.

16 hours ago, kksnowbear said:

One problem with the idea that you're "not blowing hot air into the case" based on this example is that you're cooling the CPU.  Different story if a GPU rad is installed as an intake (that's what the OP's doing I believe), and actually cooling a significant load. 

Where did you get that idea? OP mentioned EK Elite 360, which is an all-in-one CPU cooler. There are no AIOs for GPU that I know of, if you're cooling your GPU with water, you probably have a custom rig with a separate reservoir, thus making the problem with air in the pump pretty much moot if you properly degas the loop. The article I linked likewise focuses on AIOs and not on fancy custom setups. If you're cooling both GPU and CPU, then considerations are somewhat different, and you need a lot more radiator area.

Quote

I can't speak about CPUZ as a stress test, I don't use that.  Run Prime 95 for a few hours and then look at the CPU temp, not the coolant temp. 

Didn't try Prime, but with CPUZ stress test (all cores at 100%) the CPU stabilized at 77 degrees, and coolant at 32. Fairly cool day, we'll see how it works in the summer, but if we get 30+ degrees ambient temps, I probably won't be in a mood to do much flying. On the other hand, the fans might actually start turning. The fans are actually tied to the coolant temperature, and if it starts going up, they'll get to work trying to get it back down. Whether this will remain at 32 degrees remains to be seen, it's too cold outside to test right now.

16 hours ago, kksnowbear said:

You're going to be hard pressed to prove that heated air at the intake doesn't cause warmer temps inside.  At that point, you're not arguing with me - you're arguing with the laws of physics and thermal dynamics.

I'm not arguing that. I'm arguing that it doesn't matter a whole damn lot in the greater scheme of things. 10 degrees above ambient isn't something to be worried about. The difference in GPU temperatures, as shown by MSI, exists, but is not large enough. Mine doesn't seem to have any thermal issues so far. The nice thing about my rig is that I've got plenty of cooling capacity to spare, so even if it gets hot, it should be able to keep up. 

You seem to be arguing that the point is to get the overall temps as low as possible. It isn't. The point is to get the rig to run without thermal throttling. My solution does that just fine. I'll be adding additional fans, mostly because I like RGB and I have a lot of free headers at the moment, but right now, it worked out great.

Incidentally, the front radiator started out as a tradeoff due to case space, but after testing it I'm convinced this was the way to go from the start. Cooling that particular CPU is considerably harder than getting the rest of the rig to cooperate. As such, sacrificing some GPU and VRM cooling to keep the CPU low is the right thing. The only real tradeoff is having one mismatched RGB fan at the front, but I believe I can make it look good.

17 hours ago, kksnowbear said:

Your CPU might have actually been cooler with a good air cooler, leaving the front panel for good intake fans - and it might've cost less, too.  (Also everything else, including the motherboard VRMs, would be cooler too - something a liquid cooler simply cannot do).

Not true if we're talking about custom water rigs, like you seemed to assume. There are mobo coolant blocks that can allow a liquid cooler to also cover VRMs. For 1000$ you can get a mobo with one of those integrated, but this is overkill IMO. In fact, decent VRMs can run comfortably at 60 degrees. Mine does, and that's no reason to worry, and there's no performance to be gained from keeping the VRM cool.

And yes, the also sell water blocks for RAM and M.2 disks (RGB included, of course, because why not). An all-liquid PC was likely tried already, by someone with way too much money to spend. I do want to build myself one of those one day, with more glowy bits than you can shake a disco ball at, but I've got other priorities right now.

[kksnowbear]

A. (As regards all your comments about custom cooling loops): The OP talked about cooling his CPU with the EK unit, but here's an idea:  Look at his pictures.  He's got a 'hybrid' GPU and it appears the rad for that is in front (the EK is clearly visible up top).  I'm not talking about custom loops, so again, you may well have been better off (and spent less) with an air cooler and leaving the front open for as much (unheated) fresh air intake as possible.  And that would absolutely, undeniably mean cooler temps inside for things like the VRMs, chipset, M2 drives (which slow down if they get hot) etc...you're just not getting away from the reality that more cool air intake is better overall than just making a difference in one component.  And an air cooler for the CPU - with an otherwise proper case and intake arrangement - could have cooled just as well as the liquid unit, without necessarily dumping heat on everything else.

B. Yes, I am absolutely arguing that the point is (and should *always* be) to manage overall temps - because if done properly, this will almost invariably allow better individual component cooling (but without sacrificing other components as you acknowledge you are).  Pump enough cool (ambient) air into a case to create positive pressure, you're not going to have throttling unless you've jacked something up too high.  Fact is, if I displace heated air from inside the case with an adequate volume of cooler air from outside (which is how positive pressure works), the net is a reduction in temps - and that applies to *everything*.

C.  Far as the summer goes: As a scientist, you (should) know better than trying to introduce your "mood" into this.  We're talking about system performance.  So, mood aside, your system is going to run hotter in the summer, thus the situation with exhausting heated air from a radiator into your case *will* become worse.  The coolant *cannot* remain at 32c and still remove more heat from your system - that's not possible.  If there's more heat (summer) and the coolant temp stays at 32, it's not exchanging heat properly.  Unless ambient stays the same, the intake temp will be higher, and your case will be warmer inside.  This is what you seem to be overlooking: You're increasing the temp inside your case by dumping heated air into it from the radiator.  No matter how small, it's a net increase - and that's more thermal load.  No matter how slight, it's still (at least) 10c warmer than it would've been if you were jacking the front panel with cooler (unheated) ambient air, period.  By the way, "It's too cold right now" is easily overcome - that is, if you're really concerned about properly testing something.

D. My experience with maintenance of computerized systems means I've worked with this stuff 40 years plus.  Yes, a lot in data centers, but at least half that is not.  Over that time, I've built or overseen building and maintenance of hundreds (if not *thousands*) of PCs.  I currently build around 15-20 every year, and most of these are gaming machines.  This includes several builds and consults for members of this forum and others like it, who play one or more of the popular flight sims (DCS, IL2, MSFS, etc).  At least a few of these people are on their second build that I'm doing.  The systems I've built (and support) in just the past 4 years alone are in use by at least 30 gamers and around 6 or 8 streamers, maybe more.   I'm sitting in a room right now with no less than 17 working gaming PCs, half a dozen with liquid cooling.  More in a second room just across the hall, still others in a third room, and yup, even more in yet a fourth room. 

I really dislike having to break out the whole CV, as it were, but I *only* mention all this because you saw fit to question my "theory".  I'm telling you I do this, in reality, every day - and have for several decades.  You want to question my use of "theory", but I'm citing facts.  And every bit of what I'm telling you is verifiable, in case you want to challenge that.  This isn't just theory.  And again, I'd be happy to compare respective, relevant experience any time.

Incidentally, the background outlined above also includes the best formal electronics training in the US Navy, as well as first hand electronics repair and assembly work I still do every day, for local companies, one of which deals with (you'll love this) thermal gas systems.  So while my experience and knowledge certainly includes gaming PCs, I would have to say it goes *well* beyond that of the typical casual gamer.  I understand more than probably most here do about VRMs, heat in electronics, and so forth.  I'm not some self-appointed internet "expert" who drives a bulldozer or sells insurance for a living by day.  (Not that those aren't respectable professions, mind you, they're just not...*this*).

I am also currently using a 5800X3d, so there's not much need to tell me what's involved in cooling one.

E. The article from MSI:  This warrants a separate post.  However, I did say "almost" no empirical data, and I already saw the chart, so no need to throw that up.  Most of the chart doesn't even apply to this discussion, and (as I will illustrate in my next post) the parts that do apply are simply more about marketing than any real testing - as I already explained, but will further discuss in a moment.  In short, it's little more than marketing-based, carnival shell game BS...but let's actually qualify that description...

Edited January 20, 2023 by kksnowbear

[kksnowbear]

On the MSI article: Among other things, how hard were the GPU fans running during these tests?  Factually, GPUs control their own temps (within limits) by altering fan RPMs and/or clock speeds/voltages.  I can also create custom fan curves in software that will cause a GPU to maintain a certain temp across a *range* of internal temps, simply by running the fans differently.  So, I could show the same GPU temps in a chassis that was warmer inside as compared to another.  Point is since GPUs have independent thermostatic fan controls, GPU temp alone is not necessarily a reliable indicator of air temps inside the case, but that's all the "test" shows besides CPU temp.

What about temps other than CPU or GPU? Motherboard VRMs? Chipset? Internal drives (especially M2, which are known to perform better when kept cooler)?

Case fan data (How many? Size, RPMs, CFMs, pressure?).  Controlled by what?

The software used is named, but not settings, time duration etc.  Therefore it's impossible to determine actual load conditions, especially as compared to "real world".

Was the case sealed? Were any openings dampened?

We don't know any of this because it's not disclosed.  The fact that key data is missing means this "test" isn't proof of...anything, really.

Well...I suppose it *does* prove that MSI has convinced at least one person that blowing heated air into a case means better cooling.  After all, there's a reason the shell game still exists, and still works on some people.

Incidentally, the case that they used for this test? It's around $400 far as I can see.  Kinda doubtful their results translate precisely into other case types/configurations, especially depending on ventilation and fans.  And the case's particulars factor into this sort of test significantly.

And the CPU temps?  Well, of course they are lower, you're blowing outside air directly at the radiator that's cooling the CPU.  But as I already discussed, this just means that the air heated by the CPU is being re-introduced to the inside of the chassis, making it warmer than if it weren't heated.    A front mounted radiator *might* be "better" for the CPU...but everything else in the chassis is warmer, without doubt.

And that's just stuff I saw, at a glance, that indicates this is not an objective test.  It's obviously designed to specifically show what they want it to show.  I simply don't have time to waste on rigged "tests" that are clearly lacking the level of detail required to be any sort of meaningful.

As it is, they're showing what they want you to see, while omitting crucial information - in other words, shell game.

[kksnowbear]

Oh, and also (copied directly from the MSI article itself):

TOP Mount
The problem of air accumulation will be improved in the top mounting configuration. The air in the AIO system will accumulate on the top of the radiator, which basically does not affect the coolant flow in the system and does not reduce the thermal efficiency.

This configuration is the most recommended in terms of thermal performance. However, some cases may not have enough space to install a 360mm or 280mm radiator on the top.

Sounds a lot like...exactly what I've said all along.  Top mount is the best performance and doesn't suffer air void problems, but the reason it can't always be used is because of mechanical constraints (a case that wasn't built for it).

So get a better case...or accept something that even they acknowledge is not the "most recommended" (lol)

Even though I don't agree with their approach/methods, they apparently agree with me    

Edited January 20, 2023 by kksnowbear

[tomcat_driver]

2 hours ago, kksnowbear said:

Oh, and also (copied directly from the MSI article itself):

TOP Mount
The problem of air accumulation will be improved in the top mounting configuration. The air in the AIO system will accumulate on the top of the radiator, which basically does not affect the coolant flow in the system and does not reduce the thermal efficiency.

This configuration is the most recommended in terms of thermal performance. However, some cases may not have enough space to install a 360mm or 280mm radiator on the top.

Sounds a lot like...exactly what I've said all along.  Top mount is the best performance and doesn't suffer air void problems, but the reason it can't always be used is because of mechanical constraints (a case that wasn't built for it).

So get a better case...or accept something that even they acknowledge is not the "most recommended" (lol)

Even though I don't agree with their approach/methods, they apparently agree with me    

 

I'm not sure, but I think that after 2 pages, we got it.

[kksnowbear]

15 minutes ago, tomcat_driver said:

I'm not sure, but I think that after 2 pages, we got it.

I am simply responding to those who directed comments toward mine (and apparently, some still don't get it).

Genuinely sorry for any imposition - but, to be fair, no one's forcing anyone to read any of this

 

[Slammin]

Prime95 small fft's is about the only way I can get this rig to throttle now.  I am very happy with the results.

To recap, my CPU was running too hot at idle and under load.  My fix was to rotate the waterblock 180 and that fixed the CPU issue.  I also reversed the six fans on the top radiator to be intakes instead of exhaust, and increased the rear fan exhaust speed to steady 1400rpm. I also have a large fan in the bottom of the case that I also reversed to an exhaust fan.

End result is all system components are running cooler and quieter.  I am happy with the outcome and hope this helps anyone else facing the issue.

[Wali763]

On 1/17/2023 at 8:00 AM, tomcat_driver said:

Yesterday while monitoring it, mine peaked at 89C under a Noctua U14S.

Reflecting on a possible upgrade, does anyone have it running with a Noctua D-15S?

 

Hi.

I have a 5800x3d on a D15S.

I just did some testing. 10 mins of looped CinebechR23 multithread.

MaxTemp was about 75degC. The fan of the Cooler only spun up to 1050rpm.

Normally in idle I can see 30-45 degC.

I need to add though, that my rig is in the basement, where I usually only have about 16-18degC.

 

Edited January 20, 2023 by Wali763

[tomcat_driver]

7 hours ago, Wali763 said:

Hi.

I have a 5800x3d on a D15S.

I just did some testing. 10 mins of looped CinebechR23 multithread.

MaxTemp was about 75degC. The fan of the Cooler only spun up to 1050rpm.

Normally in idle I can see 30-45 degC.

I need to add though, that my rig is in the basement, where I usually only have about 16-18degC.

 

 

Thanks, do you have it undervolt or Kombo Strike 3?

 

[tomcat_driver]

Mine after a 10 min Cinebench R23, with Kombo Strike 3, a U14S + 3x P-14S Redux all running at 100% and an ambient temp of 22C.

It idles at ~39C with the CPU fan at 60% from 0C up to 60C.

Edited January 21, 2023 by tomcat_driver

[Slammin]

And that was quite the adventure.  The 5800X3D is a bad@ss chip, but in my rig with my gear, seems to not be a perfect match.  It's much better in lessor loads, but on heavy sims where you want the highest VR resolutions possible, the 5800X3D does not provide the performance I can get with the 5950.  For all else though, and especially racing sims, the 5800X3D gives me better performance at high res in VR.

So, this chip will go to my other grandson.  Looks like I really am at the end of the line for AM4.  Not a bad run and it ain't over yet.  I expect this rig will hold me for at least a couple more years, hopefully.

[edmuss]

That's odd, everyone else seems to have largely the opposite result in DCS. For productivity tasks where you need high clock speed and core count then yes but for DCS (and MSFS) currently it's about as good as you can get

[Dragon1-1]

On 1/20/2023 at 6:14 PM, kksnowbear said:

B. Yes, I am absolutely arguing that the point is (and should *always* be) to manage overall temps - because if done properly, this will almost invariably allow better individual component cooling (but without sacrificing other components as you acknowledge you are). 

And? A cooler CPU doesn't thermal throttle, and anything that's not a 5800X3D can possibly be overclocked more. However, the latter is not a concern for me, because the X3D does not overclock. Likewise, a cooler GPU won't thermal throttle and will perhaps allow a faster clock speed. A cooler VRM is... cool. For most components, as long as they're kept within their optimal operating temperatures, they'll work fine, and that's all they need to do. It doesn't matter if the VRM, or audio, or even the M.2s are on 60 or 45 degrees. Likewise, I'm happy with the X3D being anywhere between 85 degrees and room temperature, as it starts throttling at 90. Likewise with the GPU, if it doesn't hit its 84 degrees throttle point, it'll be OK. The M.2 disks are a concern, but they will have a separate intake fan, below the AIO cooler (needed for a bunch of reasons, notably to cool the HDDs living in the lower compartment).

The point is. I don't care about overall temps. I care about system performance. If the system is at a point where no further performance can be extracted from the components, lowering temps even more accomplishes nothing. That's what you don't seem to get. It's ultimately about getting it to run well, not keeping the innards as cool as possible. The arrangement with the radiator in front works for that purpose, and given the temperatures I'm seeing, the increase in air temperature doesn't affect the performance of anything important. 

On 1/20/2023 at 6:14 PM, kksnowbear said:

C.  Far as the summer goes: As a scientist, you (should) know better than trying to introduce your "mood" into this.  We're talking about system performance. 

My mood certainly goes into this, because the system is designed for me. This is my PC, and it's my usage patterns that I'm optimizing it for. Yes, the case will be warmer inside, and it might run hot in 30 or 40 degree air (yes, it can get that hot around here). No, it won't cause a problem, because it will not encounter a full CPU load under those conditions. Or, to put it another way, the primary user will thermal throttle long before the rig does.

And yes, I could turn the room into a sauna if I wanted to test this really badly, even just by stressing both CPU and GPU, and letting it run for a good long while. In fact, after flying a lot the temperature in the room gets really nice and warm. I could also stick a space heater in front of the front intake, although it's cold enough at the moment that I'd have trouble ensuring a consistent temperature. Might be worth doing anyway, after I finish building it. The worst summer heat around here was about 43 degrees IIRC (not that it was actually that hot inside), even if it does throttle when blasted with air that hot, I don't care. Now, if anything throttles at 30 degrees, I guess I'll have to eat my words, but I don't think it will.

On 1/20/2023 at 6:37 PM, kksnowbear said:

TOP Mount
The problem of air accumulation will be improved in the top mounting configuration. The air in the AIO system will accumulate on the top of the radiator, which basically does not affect the coolant flow in the system and does not reduce the thermal efficiency.

This is only in context of the air space on top of the radiator. Besides, nobody is arguing that top mounting isn't the best. It is, and that's what the MSI article is saying, for the exact same reason you're quoting. However, that's not the only thing it's saying. I'm arguing that going with the second best configuration is not going to cause a performance decrease in real world usage. All I'd get out of a fancier case and a top-mounted radiator is a few temperature numbers being lower, while the CPU numbers are a touch higher. Nothing that matters. 

On 1/20/2023 at 6:30 PM, kksnowbear said:

Case fan data (How many? Size, RPMs, CFMs, pressure?).  Controlled by what?

The software used is named, but not settings, time duration etc.  Therefore it's impossible to determine actual load conditions, especially as compared to "real world".

Was the case sealed? Were any openings dampened?

We don't know any of this because it's not disclosed. 

This is a comparative test, done on a single rig. Absolute values might vary, their relative relations should not. The former is not the point, the latter is. My whole point is that the tradeoff here compared to the optimal configuration is insignificant. 

Yes, they might have cheated, but honestly, the conclusion checks out. The best arrangement is best, the second best makes for a barely measurable difference. The second part is what you refuse to accept. You don't seem to realize that, in practical terms, there difference between the two is small enough not to matter, and the performance margin that I've got exceeds that difference handily.

 

[Slammin]

16 hours ago, edmuss said:

That's odd, everyone else seems to have largely the opposite result in DCS. For productivity tasks where you need high clock speed and core count then yes but for DCS (and MSFS) currently it's about as good as you can get

The 5800X3D is going back in tonight.  It only takes me about 15 minutes to swap.

What is amusing to me is that before rotating my waterblock, CCD0 was always coolest, always had several cores hitting 5+ghz and CCD1 almost never had any cores hitting 5+ghz.  CCD1 used to always hit higher temps than CCD0.

ANywho, I did discover why my testing of the 5800X3D produced such weird results and I have corrected that, and am back on track here

 

kksnowbear: Not sure what about this thread triggered you so much, but I promise I meant no offense to you.  I was just sharing an adventure, and I do welcome input from folks that know about stuff.  So thanks, I hear ya, but I also know a few things too.

Edited January 23, 2023 by Slammin