Can software cause hardware damage through overheating? (Split from: Kerbal Space Program 1.10.1 is live!)

[steve_v]

2 hours ago, Boyster said:

Is that statement true?

Yes.
 

2 hours ago, Boyster said:

Cant software demand more than what hardware allows and cause problems?

No.
 

2 hours ago, Boyster said:

Is that not the definition of a blue screen back in the day?

No. A "blue screen" (or BSOD) is a generic term covering a multitude of fatal Windows errors.
 

2 hours ago, Boyster said:

It should have the capability to diagnose your system and refuse to run or shutdown if it detects overstress.

Don't be ridiculous. Why should a game, of all things, nanny broken hardware?

Monitoring hardware is the job of the hardware monitoring ASIC or microcontroller (like this one), and sometimes the OS or BMC.
Individual applications have neither the need nor the privileges required to access or control hardware limits. Monitoring hardware temperature isn't the applications job (and would mean including drivers for a multitude of systems), and changing hardware limits or cooling targets would require interfering with the BIOS or CPU microcode, which no game could or would even attempt.
 

2 hours ago, Boyster said:

yes as you said before, its a form of ''hacking'' and bypassing some things it should not

Now you're really grasping at straws...
 

2 hours ago, Boyster said:

i kinda still want to stand by my logic.

And you should probably quit while you're ahead well, something anyway.

 

 

2 hours ago, Lisias said:

this will eventually damage the computer

My overclocked and still perfectly operational machines might beg to differ, a couple of them are going on 20
As long as you're not pushing silly voltages or running silly hot, overclocking is just "re-binning" the CPU anyway.

[Vanamonde]

Ease off on the tempers, please, guys. There's no need to take an accusatorial tone with each other. 

[kerbiloid]

When I was using North Kerbal Dynamics nukes in previous versions of KSP, and was looking at the explosion from close distance, my coolers were terribly roaring during the slideshow, due to enormous amount of generated particles. I guess if I hadn't aborted the mission, something would be damaged irl by the virtual nuclear explosion.

[Deddly]

@kerbiloid, the sound of the coolers roaring means that they are working. You can get applications to show you CPU/GPU temperatures if you aren't sure. 

[kerbiloid]

4 minutes ago, Deddly said:

@kerbiloid, the sound of the coolers roaring means that they are working. You can get applications to show you CPU/GPU temperatures if you aren't sure. 

I know, and I believe they were crying to me: "We are not Indiana Jones fridge! Move this thing away!"

[steve_v]

32 minutes ago, Deddly said:

the sound of the coolers roaring means that they are working.

Sweet music it is

CPU1 Temp        | 60.000     | degrees C  | ok
CPU2 Temp        | 60.000     | degrees C  | ok
System Temp      | 51.000     | degrees C  | ok
Peripheral Temp  | 47.000     | degrees C  | ok
PCH Temp         | 56.000     | degrees C  | ok
P1-DIMMA1 TEMP   | 58.000     | degrees C  | ok
P1-DIMMB1 TEMP   | 54.000     | degrees C  | ok
P1-DIMMC1 TEMP   | 52.000     | degrees C  | ok
P1-DIMMD1 TEMP   | 55.000     | degrees C  | ok
P2-DIMME1 TEMP   | 55.000     | degrees C  | ok
P2-DIMMF1 TEMP   | 59.000     | degrees C  | ok
P2-DIMMG1 TEMP   | 60.000     | degrees C  | ok  
P2-DIMMH1 TEMP   | 59.000     | degrees C  | ok
FAN1             | 2850.000   | RPM        | ok
FAN2             | 2850.000   | RPM        | ok
FAN3             | 1950.000   | RPM        | ok
FANA             | 1200.000   | RPM        | ok
FANB             | 1200.000   | RPM        | ok

A pair of 92mm Nidec UltraFLOs at 2850RPM is, uhh, moderately loud? Flat out at 3800RPM is... Louder. Never seen it get there though.
Still pretty quiet for a dual-socket server though, the 2U coolers I took off it rev to 8400RPM...

Ed. 550W, without a GPU in sight... welp. My power bill will love me.

Edited September 17, 2020 by steve_v

[Lisias]

7 minutes ago, steve_v said:

Still pretty quiet for a dual-socket server though, the 2U coolers I took off it rev to 8400RPM...

Music? This beast has some "music" for you!

[steve_v]

1 minute ago, Lisias said:

This beast has some "music" for you!

The only thing more obnoxiously loud than a 2U cooler is a 1U cooler...
But it's a Dell, isn't it? Ick. Supermicro all the way.

[Lisias]

13 hours ago, steve_v said:

But it's a Dell, isn't it? Ick. Supermicro all the way.

These were cheap around here at that time. I bought that 1425 cheaper than a Raspberry Pi!

-- post edit --

Bringing the discussion back to thermals. @Boyster, that screaming crap I call a server is a perfect example about what we are discussing here.

That thing have 16Gb of server RAM, TWO formidable electric heaters acting as CPU called Xeon with 2 Cores each at 3.6GHz, two SCSI 10K rpm disks spinning crazily trying to accelerate entropy into the World. All of that packed on a case with 1.75 inches high. Less than 5 cm.

And yet, that thing runs cooler than an MacBook - God knows I wish it would run quieter too.

You can't have the cake and eat it too. If you want small form factors, you need to compensate the lack of thermal dissipation by using powerful turbines, I mean, coolers on the system.

Powerful and silent MiniPcs and notebooks are, frankly, pure scams. No one can fight entropy.

 

Edited September 18, 2020 by Lisias
bringing discussion back to topic

[Okhin]

I remember having to convince people that a Macbook Air for a mission in Abidjan was a bad idea, for thermal reasons. They did not understood that the mean heat over there was too high for the fanless computer to manage. I mean, when we reach 30°C and more in the office, in EU, those computer already seemed to have some issues with heat, when the mean temperature over there is a little bit above 32°C I don't want to touch them, I'll get burned (iirc the aluminum case was used as a poor man eat sink, don't put them on bare skin).

Anyway, they got around us and got those mac book air. Two month later I got complaint that the battery life of those are bad (the internal temperature will go above 40°C, which shorten the battery life for chemical reason), and that they sometime shut down for no reason (well, the reason was heat, but they would not admit that).

It's not like I tried to propose toughbook with fans and way more batteries for the same price than those expensive coffee heaters. Sure they were not fancy, but they would have worked. Instead you bring your European and north American centric hardware, not really made for hot countries, and you try to put the blame on me.

I've quit the job sometime after that, for other internal shenanigans.

[Incarnation of Chaos]

44 minutes ago, Lisias said:

These were cheap around here at that time. I bought that 1425 cheaper than a Raspberry Pi!

-- post edit --

Bringing the discussion back to thermals. @Boyster, that screaming crap I call a server is a perfect example about what we are discussing here.

That thing have 16Gb of server RAM, TWO formidable electric heaters acting as CPU called Xeon with 2 Cores each at 3.6GHz, two SCSI 10K rpm disks spinning crazily trying to accelerate entropy into the World. All of that packed on a case with 1.75 inches high. Less than 5 cm.

And yet, that thing runs cooler than an MacBook - God knows I wish it would run quieter too.

You can't have the cake and eat it too. If you want small form factors, you need to compensate the lack of thermal dissipation by using powerful turbines, I mean, coolers on the system.

Powerful and silent MiniPcs and notebooks are, frankly, pure scams. No one can fight entropy.

 

Man, the saddest part is I'm thinking of picking up a similar server to use as a Pfsense router just because they're actually the cheapest thing i can find that has AES instructions and isn't absolute trash....but it's absolute overkill for the application at the same time. It'll be in a closet, so heat isn't too big of an issue. Power is dirt cheap here (2 cents/kwh last i checked), but i ammmmm going to be running a VPN and a gigabit pipe with it.

Oh the things i tell myself to justify my obsession~

Unfortunately i couldn't find a similarly dirt cheap server that supports 3.5 inch disks, everything iv'e seen looks to be 2.5 and the cost per TB is miserable for laptop form-factor HDDs. What was this thread about? Something about power viruses causing actual hardware damage? i forgot xD

 

[Okhin]

For the virus* damaging hardware part, it can be possible. After all, the speed at which your disks turns are controlled by computer, and even the check in place there are software and can be tempered with (as any other piece of software). I mean, we now have multiple processors in hard drive, to the point you can run Linux on one of them (I mean, Malwaretech even wrote a detailed series of article on how to do it, yes it requires physical access). That's why a lot of hardware control point have been implemented those years on motherboards (like shutting down the CPU before it reaches destructive level of temperature, etc). And SSD are a bit harder to mechanically destroy since they don't really have mechanical parts anymore. You can probably over heat them somehow, but then again, a temp fuse will fire up somewhere to shut down the computer.

And then, what's the point of destroying the computer you hacked ? You generally have more profitable things to do. And you can mess up with a system without destroying it, including corrupting backups, altering data and, in case of factory sabotage, mess with the hardware they control. Also burning the computer you've got access means that you'll leave proof of tampering on it (you cannot clean up a shut down computer), while you still have a possibility to reduce the trace you've left when the computer is active.

So yes, I guess it is still possible to burn some hardware with a virus. But it's harder than, say, in the nineties, because of hardware switches and safety. Also it has almost no interest (except for theatrical in movies) for an attacker to do so.

--

* any malicious code

[steve_v]

2 hours ago, Lisias said:

All of that packed on a case with 1.75 inches high.

In a datacenter or dedicated server room, that kind of thermal engineering is highly effective. For home use, the answer is a realistically sized chassis and reasonably quiet (read "large diameter low RPM") fans. In both of those situations it's extremely rare to see any CPU above 70C (Tcore), even under continuous full-load and less than ideal ambient.
Laptops and portables on the other hand often run at 90+, because manufacturers consistently overstate performance and underengineer cooling solutions in the name of form factor. This would be fine if they also tweaked the system to throttle at sane temperatures... which they consistently don't. So when you actually use that advertised performance, they cook.
 

2 hours ago, Lisias said:

God knows I wish it would run quieter too.

Take one surplus Supermicro ATX board. Remove insane CPU coolers and replace with 120mm towers, Noctua fans and custom mounting hardware. Place  in "silent" desktop mid-tower case. Profit.
64GB ECC RAM, 2x 8-core 3.1GHZ Xeons, 8x 3.5" and 8x 2.5" HDDS (~40TB total), all in a package quiet enough for a bedroom or HTPC (Xeon box #2).
The joys of scrounging ex-server parts to take home...

 

2 hours ago, Lisias said:

Powerful and silent MiniPcs and notebooks are, frankly, pure scams.

Thermally limited burst performance certainly has it's place, but if that's what you're going for you need to take continuous load scenarios into account as well.
Personally I blame Apple for starting and perpetuating the thinness wars, it's got to the point that it's extremely difficult to find a laptop with competent thermal design at all, let alone a reasonably priced one.

 

1 hour ago, Okhin said:

a Macbook Air for a mission in Abidjan was a bad idea

A Macbook Air is a bad idea in general.

 

1 hour ago, Okhin said:

I tried to propose toughbook with fans and way more batteries for the same price

Pretty much anything not-Apple will have better thermal and power management for the same price, because you're not paying for the hipster appeal.
Say what you will about innovation and all-American design, Apple have a long history of terrible thermal design, fragile products, and liquid-poor serviceability.

 

1 hour ago, Incarnation of Chaos said:

Unfortunately i couldn't find a similarly dirt cheap server that supports 3.5 inch disks

If you really want rackmount, you're kinda screwed. The only place you're likely to find 3.5" bays is in a 4U server, and those tend to be either very expensive or very old. The industry loves density and all that.
Otherwise, see above regarding supermicro ATX boards and non-server cases/coolers. Maybe hit up ebay for ex-server parts and roll your own? Hell, for a router you don't really need server-grade hardware at all, just slap a couple PCIE NICS in any old post-2012 board.
 

FWIW, my home router is one of these. I wouldn't run PFSense on it though, the BSD intel NIC drivers suck donkey balls. IPFire does gigabit routing/firewalling on it with ease, but VPN at that bandwidth will obviously require more CPU.

I tried to like PFSense, but performance really does suck when compared to a GNU/Linux based router distro, at least on low-power hardware. So does the non-journaling non-power-failure-tolerant UFS filesystem it uses, though I hear it can do ZFS too now... at the expense of even higher system requirements.

Edited September 18, 2020 by steve_v

[magnemoe]

19 hours ago, Lisias said:

In a nutshell: NO.

Overheating can damage the computer, but it is not caused by software. It is caused by problems on the cooling system or by simple bad design of the cooling system.

The cooling system of your computer is like the cooling system of a car: if the car overheats, it's due bad maintenance or because the manufacturer screwed up.

There's a lot of notebooks and mini-pcs dying by overheat lately and the reason is only one: bad design. But since it's cheaper to blame the software to cover up engineering mishaps, this urban legend persists - "let's save some bucks by convincing the customer that he installed a bad software, instead of paying ourselves for our mistakes".

 

This, now modern cpu and probably gpu has thermal throttling,  that is performance is temperature dependent. this is why you can run a single core faster, it produces less heat than something who uses the cpu more efficient. 

Second line in defense is automated shutdown. If your pc start crashing doing any sort of heavy works as crash in multiple games this is probably the reason. 
Once had an graphic card where the fan stopped working, things worked well enough in windows but crashed in games. 

However you have pc components who are not protected this way and might take damage from overheating. This is true for all sort of stuff like the red ring of death on the xbox 360 was caused by overheating. 
Yes putting it on an carped blocked flow in at the bottom but designers should expected this. Better air intakes and / or overheat protection would solve this. 
And no its not the users fault here. Same with laptops who also have air intake at bottom. 

You are far more likely get this running more demanding software like games. 
And the problem was way worse say 15 years ago then cpu lacked overheat protection so if you cpu fan stopped you could burn out your cpu, however usually if stopped working and rebooted 
But we had once server cup who was destroyed because boss turned off air condition in server room as we was working on an second server and it was cold. 
Yes that server could not survive 20 centigrade heat

Tower pc's tend to not have problems here but an friend repaired an pc who tended to overheat. The rear fan was set to blow air into the case but the cpu cooler blow air backward so they countered each other  
Pretty much the only issue is if you have multiple 3.5" disks stacked together and no air flow over them can easy get to hot is used a lot like an file server.  

[magnemoe]

 

16 hours ago, Deddly said:

@kerbiloid, the sound of the coolers roaring means that they are working. You can get applications to show you CPU/GPU temperatures if you aren't sure. 

4x140mm fans in and water cooling. cpu on 35 degree and can not hear the computers unless I play game or render. 
Granted it stand inside an cabinet with an funnel to take air from the top and send it out back of the cabinet. 

[steve_v]

27 minutes ago, magnemoe said:

the red ring of death on the xbox 360 was caused by overheating.

The Xbox 360 was an essay in why finalising the trendy case design before the internal hardware is a terrible idea. MS managed to create a product that was both excessively loud and excessively hot, the result being one of the worst reliability records of any console in history.
Apple was obviously asleep during this class, and are still making machines that prioritise case design over functionality... Paradoxically, people are still buying such products.
 

27 minutes ago, magnemoe said:

the problem was way worse say 15 years ago then cpu lacked overheat protection so if you cpu fan stopped you could burn out your cpu

Once upon a time, the entire CPU heatsink/fan assembly fell off my:

... And absolutely nothing happened.
The thing ran like that for at least a week before I noticed (basement dial-on-demand router and voice-to-email answering machine). Guess those glacial MP3 encoding jobs were infrequent enough to not crash it, or the basement cold enough.

 

27 minutes ago, magnemoe said:

The rear fan was set to blow air into the case

Ahh, the all-in bass-ackwards rookie cooling configuration. A true classic.

Honestly, I don't understand how people (even some "professional" system builders) manage to get cooling so completely wrong so often.
Fans all facing in. Fans all facing out. CPU cooler opposite case airflow. No case airflow. Fans to nowhere. Bottom intakes with no case feet. Common sense does not seem to prevail.

Air goes in at the bottom, out at the top, hot things are in between. Simple, no?

 

Edited September 18, 2020 by steve_v

[magnemoe]

21 minutes ago, steve_v said:

The Xbox 360 was an essay in why finalising the trendy case design before the internal hardware is a terrible idea. MS managed to create a product that was both excessively loud and excessively hot, the result being one of the worst reliability records of any console in history.
Apple was obviously asleep during this class, and are still making machines that prioritise case design over functionality... Paradoxically, people are still buying such products.
 

Once upon a time, the entire CPU heatsink/fan assembly fell off my:

... And absolutely nothing happened.
The thing ran like that for at least a week before I noticed (basement dial-on-demand router and voice-to-email answering machine). Guess those glacial MP3 encoding jobs were infrequent enough to not crash it, or the basement cold enough.

 

Ahh, the all-in bass-ackwards rookie cooling configuration. A true classic.

Honestly, I don't understand how people (even some "professional" system builders) manage to get cooling so completely wrong so often.
Fans all facing in. Fans all facing out. CPU cooler opposite case airflow. No case airflow. Fans to nowhere. Bottom intakes with no case feet. Common sense does not seem to prevail.

Air goes in at the bottom, out at the top, hot things are in between. Simple, no?

 

Impressive with the DX4-120. Granted cooling demand back then was much lower as in a 40 mm fan.  
Had an of the first multi core amd cpus where the cpu fan stopped, air flow in the case was so good nothing happened  

And for laptops its so easy to block the air flow, putting it on top of the duvet in bed is an classic same with carpets or have them inhale half a cat. 
Friend of mine having an old desk, cut a hole in it an put an 120 mm fan in it with an grille on top, now the laptop was silent. 

[Incarnation of Chaos]

1 hour ago, steve_v said:

If you really want rackmount, you're kinda screwed. The only place you're likely to find 3.5" bays is in a 4U server, and those tend to be either very expensive or very old. The industry loves density and all that.
Otherwise, see above regarding supermicro ATX boards and non-server cases/coolers. Maybe hit up ebay for ex-server parts and roll your own? Hell, for a router you don't really need server-grade hardware at all, just slap a couple PCIE NICS in any old post-2012 board.
 

FWIW, my home router is one of these. I wouldn't run PFSense on it though, the BSD intel NIC drivers suck donkey balls. IPFire does gigabit routing/firewalling on it with ease, but VPN at that bandwidth will obviously require more CPU.

I tried to like PFSense, but performance really does suck when compared to a GNU/Linux based router distro, at least on low-power hardware. So does the non-journaling non-power-failure-tolerant UFS filesystem it uses, though I hear it can do ZFS too now... at the expense of even higher system requirements.

That's the sad part, i can find a dual xeon (v2 if I'm not mistaken, so 32nm) server with 16GB RAM, and PSU from a reliable "recycler" for 100 bucks shipped on ebay....

I literally cannot beat that price unless i go for something like a raspberry pi o.o

But yeah, for a NAS i came to the same conclusion. I'll have to rig a ATX box up for what I'm after. 

[Okhin]

Raspberry pi are nice. Low on power, for small servers / routers they do the job (even if they lack ethernet ports for my taste). I mean you can run your personal wordpress + email setup + a VPN, throw a 3G or 4G routers in it, and you now have a fully self hosted solution that you can carry on you (I know a guy who stick one in his hat, with  battery, walking around with hsi blogs and email on top of its head).

If you don't need the graphical chip of the raspberry pi, go for a banana pi, they're more server oriented. but a bit bigger and expensive (plus, i do not think they can run on a battery alone). No need for that much computational power for a server, the main bottleneck will usually be the bandwidth anyway, not your computational power.

[steve_v]

34 minutes ago, magnemoe said:

carpets or have them inhale half a cat

I get an involuntary twitch every time I see a tower case sitting on a carpeted floor.
 

32 minutes ago, Incarnation of Chaos said:

i can find a dual xeon (v2 if I'm not mistaken, so 32nm) server with 16GB RAM, and PSU from a reliable "recycler" for 100 bucks shipped on ebay....

32nm covers everything from Westmere through Sandy Bridge, and you most definitely want the latter. The old LGA771 Westmere (AKA Nehalem-C) machines are dirt cheap now, but they're power hogs and have DDR2 RAM and very special cooler mounts.
The "v2" bit confuses me slightly, I'm guessing you mean E5-xxxx v2, which is a 22nm Ivy Bridge EP (LGA2011) part. $100 is a steal if that's what they are...
 

32 minutes ago, Incarnation of Chaos said:

for a NAS i came to the same conclusion. I'll have to rig a ATX box

I can personally vouch for the Supermicro X9 series (say a X9DRL-IF) and something like the Fractal Define R4 making a very nice, very quiet ZFS NAS box.

 

23 minutes ago, Okhin said:

for a server, the main bottleneck will usually be the bandwidth anyway, not your computational power.

If you're just hosting some blog pages, sure. There are other (and more CPU/RAM intensive) uses for a server than that though.
As mentioned, my primary home (file/web/VM/various other stuff) server is a dual socket (16 physical cores @ 3.1GHz) Xeon with 64GB RAM. Can't really do that on an ARM based SBC. Nor can you realistically connect 16+ HDDs to one...

RPis are cool and all (I still have a Mk1 around here somewhere), but they're more hobbyist toys than real servers IMO, and too slow or I/O limited for a great many applications.

Edited September 18, 2020 by steve_v

[Okhin]

Oh, sorry, I was still on the use of an overpowered server as a router + VPN (which is a not too bad use for a raspebrry pi), yeah, if you wanna serve VM you obviously need more.

[steve_v]

1 minute ago, Okhin said:

a router + VPN (which is a not too bad use for a raspebrry pi)

Router yes, webserver yes, VPN... Maybe. The crypto load for a decently fast VPN server is non-trivial, and I bet the RPi CPU would bottleneck throughput, especially as (AFAIK) it lacks AES-NI or any other acceleration.

[Deddly]

27 minutes ago, steve_v said:

Fans all facing out.

Perhaps they thought a vacuum would help cooling, because space is cold

[Incarnation of Chaos]

10 hours ago, steve_v said:

I get an involuntary twitch every time I see a tower case sitting on a carpeted floor.
 

32nm covers everything from Westmere through Sandy Bridge, and you most definitely want the latter. The old LGA771 Westmere (AKA Nehalem-C) machines are dirt cheap now, but they're power hogs and have DDR2 RAM and very special cooler mounts.
The "v2" bit confuses me slightly, I'm guessing you mean E5-xxxx v2, which is a 22nm Ivy Bridge EP (LGA2011) part. $100 is a steal if that's what they are...
 

I can personally vouch for the Supermicro X9 series (say a X9DRL-IF) and something like the Fractal Define R4 making a very nice, very quiet ZFS NAS box.

 

If you're just hosting some blog pages, sure. There are other (and more CPU/RAM intensive) uses for a server than that though.
As mentioned, my primary home (file/web/VM/various other stuff) server is a dual socket (16 physical cores @ 3.1GHz) Xeon with 64GB RAM. Can't really do that on an ARM based SBC. Nor can you realistically connect 16+ HDDs to one...

RPis are cool and all (I still have a Mk1 around here somewhere), but they're more hobbyist toys than real servers IMO, and too slow or I/O limited for a great many applications.

DDR3, so E5 and 22nm. Man I shouldn't post after a night at work, but yeah it's a bargain.

But I'm actually strangely familiar with socket 771, since it's similar to socket 775 that hosted the Pentium 4 in my childhood computer. 

And the crypto performance for VPN is what I'm after, so it looks like it may not be absolutely overkill after all from the latest comments. And I'll check out those supermicro boards. 

Also I was being sarcastic about the RPi xD

I'd love to get one of them to play with one day, but that's basically what it would be xD.

 

[steve_v]

3 hours ago, Incarnation of Chaos said:

I'm actually strangely familiar with socket 771, since it's similar to socket 775 that hosted the Pentium 4 in my childhood computer.

Yeah, pretty much the same CPUs but with a slightly different socket to make them incompatible with desktop parts. The usual Intel "make Xeon special" shenanigans.
Plus the coolers mount to the case rather than the board, so you have to break out the machine tools to fit them in a non-server case. I know, I've been there. LGA2011 is a far more sane proposition.