Tech myth debunk thread

[Bombastinator]

9 minutes ago, Sauron said:

Myth: this thread is not yikes.

DEBUNKED

This thread is, in fact, yikes.

Try the one on the particulars of PSU design or the one on the new Mac Pro.  Those are my “Yikes” threads right now.  I really wish I could follow the PSU one but it just got too deeply technical for my ability and I had to drop it.  Too bad really.  It was clearly useful stuff.

[Lady Fitzgerald]

1 hour ago, Bombastinator said:

How about “perceived quality and quality are not always the same thing”?

I'll cheerfully give you that. Actually determining which is actually perceived quality and which is actually quality might be a touch tricky, though.

[The1Dickens]

4 hours ago, Bombastinator said:

How about “perceived quality and quality are not always the same thing”?

 

3 hours ago, Lady Fitzgerald said:

I'll cheerfully give you that. Actually determining which is actually perceived quality and which is actually quality might be a touch tricky, though.

Okay, then how does one perceive quality?

[TempestCatto]

44 minutes ago, The1Dickens said:

 

Okay, then how does one perceive quality?

Not sure of the context but I'll say this.

 

Look at Apple. It's the opposite of that.

[PlayStation 2]

Oh, I've got one: Interlaced resolutions are not half their progressive scan resolutions (ex. 1920x1080i not being 1920x540) because of how interlacing works.

Conversely, half of an interlaced resolution being a progressive scan resolution only works if that resolution in context is intended to be displayed in a different way (like 720x240p vs. 720x480i).

[Trik'Stari]

On 6/29/2019 at 2:05 PM, floofer said:

Macs being bad for the money is a huge myth. 

Fine, 95% true.

 

Take it or leave it.

[Spotty]

This thread is for TECHNOLOGY related myths only. The LTT forum is not the place for conspiracy theories about politicians and aliens. Thread cleaned. If the thread goes off topic again it will be locked and warnings may be issued.

[straight_stewie]

On 6/29/2019 at 11:14 AM, Mira Yurizaki said:

While I'm hoping there's more data available about testing between the two architectures, the ISA doesn't make a processor faster or slower, it's the implementation of it.

I would tend to agree, to a point. The design of an ISA does affect how certain algorithms are implemented on that platform, and that can cause performance differences even if all other things are the same.

For example, let's take an addition operation on two theoretical processors, one is accumulator based, and the other is register-register based.

In the first, the add operation might look something like:

1. Load operand₀ to the A register
2. Load operand₁ to the B register
3. Add
4. Store the Accumulator in memory.

This method requires three memory accesses, and all three are conducted one at a time. The other architecture might have multi channel memory, so:

1. Add operand₀ and operand₁
2. Store the accumulator in memory.

If we assume that all other things about the processor are the same, for example memory times, clock speed, and clock cycles per memory access and add operation are equivalent, then the first processor takes twice as long as the second processor to complete an addition.

I'm not arguing that this is the case with ARM and x86, I'm just arguing that it is possible that ISA differences could lead to performance differences, even if the processors are otherwise comparable.

[rcmaehl]

Myth: ESD kills components instantly
 

ESD usually does not instantly kill a device. Static electricity is extremely weak current shock of upwards of 1000 V. It is unlikely to kill a device completely but can definitely damage individual circuitry causing all sorts of strange, abnormal, and seemingly untraceable issues.

[wasab]

On 12/15/2019 at 3:17 AM, Kisai said:

But many applications simply can not be multithreaded because of bad programming practices. For example "forking" a process is an extremely inefficient use of CPU and Memory resources, and this continues to be the default settings in many programs, including ones that are capable of multi-threading because of plugins/extensions that are not thread-safe. For example, Apache HTTPD continues to be prefork on Linux and BSD systems because of mod_perl or mod_php still being used by developers instead of the socket process manager servers (eg php-fpm.) What makes it worse, is that people continue to believe the myth that "forking" is the correct model when it's the worst option.

Forking a process takes advantage of multicore cpu no differently than threads. It is not a bad programming pratice in any way, shape, or form, for webservers especially. If bussiness care about memory usage and efficiency, they wouldn't be picking scripting langauge and instead use C/C++ whose alogrithms of say O(n^3) would already outperform something like O(nlongn) in JavaScript or python by several magnitude. 

[Geoff35674567]

From respected youtubers I have heard heat radiates upwards. No it doesn't. In the Earth's atmosphere hot gas and liquid rises. In your PC hot air goes where the fans blow it.

 

Linux and Mac can't get viruses: sure but this is not 1990. They can get malware and ransomware. Your ISP provided modem/router is likely running an old linux kernel that can be exploited and have its DNS hacked.

 

RAID0 is dangerous. Sure it's dangerous for your business. For your games that you downloaded from steam? Na.

[dalekphalm]

4 minutes ago, Geoff35674567 said:

From respected youtubers I have heard heat radiates upwards. No it doesn't. In the Earth's atmosphere hot gas and liquid rises. In your PC hot air goes where the fans blow it.

These are not mutually exclusive statements. Hot air does rise. But the fans blowing it apply more force to essentially override the aids natural tendency to rise. 
 

and if your airflow isn’t good enough, hot air can indeed rise and create hot pockets in the case. 

4 minutes ago, Geoff35674567 said:

Linux and Mac can't get viruses: sure but this is not 1990. They can get malware and ransomware. Your ISP provided modem/router is likely running an old linux kernel that can be exploited and have its DNS hacked.

Agreed here. 

4 minutes ago, Geoff35674567 said:

RAID0 is dangerous. Sure it's dangerous for your business. For your games that you downloaded from steam? Na.

Depends on what you’re using it for. To store games that can be redownloaded? Sure no problem. Just as long as it’s not data you can’t afford to lose. 

[Geoff35674567]

14 minutes ago, dalekphalm said:

These are not mutually exclusive statements. Hot air does rise. But the fans blowing it apply more force to essentially override the aids natural tendency to rise. 
 

and if your airflow isn’t good enough, hot air can indeed rise and create hot pockets in the case. 

 

Has anyone measured this? I doubt the force is strong enough to measure. Look how big a hot air balloon needs to be to lift people.

[Mira Yurizaki]

1 hour ago, straight_stewie said:

-Snip-

With modern processor design, the ISA is largely a moot point now because the ISA that software is exposed to gets broken down into micro-ops that are then re-arranged and scheduled.

 

Also, if a processor doing the first listing had fancy features like OoE and and superscalar pipelining while the second one was basically a super basic processor with a three-stage pipeline, the first one could potentially execute code faster even thought there's more lines of code.

 

10 minutes ago, Geoff35674567 said:

Has anyone measured this? I doubt the force is strong enough to measure. Look how big a hot air balloon needs to be to lift people.

What @dalekphalm is saying is going to apply no matter what. Hot air will try to find a way to rise. However, I agree with you that the fans will also just make it a moot point anyway because fans generate something else that also moves air: an air pressure differential.

[Kisai]

1 hour ago, wasab said:

Forking a process takes advantage of multicore cpu no differently than threads. It is not a bad programming pratice in any way, shape, or form, for webservers especially. If bussiness care about memory usage and efficiency, they wouldn't be picking scripting langauge and instead use C/C++ whose alogrithms of say O(n^3) would already outperform something like O(nlongn) in JavaScript or python by several magnitude. 

Demonstrably false. When the same quadcore machine can deal with 2000-4000 threads, but only deal with 100 prefork sessions, there is something certainly wrong with forking processes that has a lot more to do with programmer incompetence and/or laziness. You can not secure a web server. Period, and all you gain from preforking is being able to ignore memory leaks/stalls from crappy scripting languages.

 

If a threaded process is 100MB and has 100 threads, and a prefork process is 100MB per process, the latter is using 10GB of RAM to accomplish the same thing, which if the system doesn't have that much RAM or CPU time to spin up new processes, starts to be impaired by the virtual memory system. That's why DDoS's are successful all the time, because they target this lazy attitude toward resource conservation. Meanwhile if you spin up 4000 threads that take up 1MB each, you're still far ahead of the prefork model with 100 processes.

 

Writing a site in C, is just not going to be a thing, and writing a site entirely in python, ruby, php or any other scripting language that doesn't have the capability to do threading kneecaps the performance of the system. One Wordpress site consumes as much cpu time as 100 small sites, but consumes as much CPU power as infinite sites on the same machine without scripting language support. How else do you think bulk-hosting sites operate? And you know how WP sites try to solve it? By adding yet more scripting language plugins that try to turn it back into a static site. What a mess.

 

I'm not going to tell a news site to stop using WP if WP is what they designed their site for, but good grief the amount of money just being flushed down the toilet on hosting costs by using WP is nothing to sneeze at.

[Vacras]

Here is one:

Windows is way too much hassle to setup with drivers compared to macOS or Linux.

[Mira Yurizaki]

8 minutes ago, Vacras said:

Here is one:

Windows is way too much hassle to setup with drivers compared to macOS or Linux.

If you're using hardware from a not-obscure vendor, Microsoft typically pushes driver updates through Windows update. Though the driver isn't likely to be up-to-date, not that it matters most of the time anyway.

[straight_stewie]

24 minutes ago, Mira Yurizaki said:

Also, if a processor doing the first listing had fancy features like OoE and and superscalar pipelining while the second one was basically a super basic processor with a three-stage pipeline, the first one could potentially execute code faster even thought there's more lines of code.

Hence why I said:

1 hour ago, straight_stewie said:

If we assume that all other things about the processor are the same

 

[Mira Yurizaki]

18 minutes ago, straight_stewie said:

Hence why I said:

And hence why I said the ISA largely doesn't matter, it's the implementation of it.

 

As long as instruction outputs what's expected, who cares how it's done?

 

On a side note, when people were looking into RISC, they found in a CISC CPU, even doing simpler instructions rather than the single instruction that achieves the same result could be faster.

 

EDIT: Also optimization is often counter-intuitive.

[wasab]

1 hour ago, Kisai said:

Demonstrably false. When the same quadcore machine can deal with 2000-4000 threads, but only deal with 100 prefork sessions, there is something certainly wrong with forking processes that has a lot more to do with programmer incompetence and/or laziness. You can not secure a web server. Period, and all you gain from preforking is being able to ignore memory leaks/stalls from crappy scripting languages.

Even if you are running multithreaded server, you still would need to run more process of these servers in order to scale out to other machines. This would be no different than forking beaucse that is what forking is, starting more process of the same program. A single computer, multithreaded or not, can not handle all the users in the world without poping up more of the same server processes on more computers. Multiprocess application can also have one process crashed, without taking the entire system offline, can multithreaded application not crashed if like say one thread segfaulted or raised any other unchecked exceptions? And security is not the reason why programmers fork. 

 

1 hour ago, Kisai said:

I'm not going to tell a news site to stop using WP if WP is what they designed their site for, but good grief the amount of money just being flushed down the toilet on hosting costs by using WP is nothing to sneeze at.

Who cares? companies deal with it by simply adding more machines and scaled out. One reason they choose horribly slow and inefficient scripting languages like JavaScript and php over c++ or c for web development is because of productivity and scalability. It is easier to implement new features in these languages and make them work with other technologies like extension or what nots as you said earlier than in C or C++ and easier to scale them out to more computers if performance is becoming an issue due to increased users, traffics, or what nots. In web devs, productivity and scalability outweighs whatever performance you think you may gain from multithreading vs forking or scripting vs native code in C/C++. Whatever increased cost for more hardware or servers due to slow performance can be compensated by the increased productivity and reduce cost of development. 

[Vacras]

2 hours ago, Mira Yurizaki said:

If you're using hardware from a not-obscure vendor, Microsoft typically pushes driver updates through Windows update. Though the driver isn't likely to be up-to-date, not that it matters most of the time anyway.

Exactly

especially since windows 10

[Mira Yurizaki]

2 hours ago, Vacras said:

Exactly

especially since windows 10

I forgot what the title of the thread was for a moment

[Kisai]

7 hours ago, wasab said:

Even if you are running multithreaded server, you still would need to run more process of these servers in order to scale out to other machines.

Come back to me when you have experience running a server.

 

You missed the entire point that this lack of threading is why a computer needs more memory, because it has to duplicate the process memory for every fork. Every tab in Chrome is chewing up a few hundred MB of memory when it could be using kilobytes. And your answer is "get another computer per tab"

 

So each time that involves large memory copies, virtual memory allocation, and so forth. Machines are only getting slower because of this incompetence.

[wasab]

54 minutes ago, Kisai said:

Come back to me when you have experience running a server.

I do

54 minutes ago, Kisai said:

You missed the entire point that this lack of threading is why a computer needs more memory, because it has to duplicate the process memory for every fork. Every tab in Chrome is chewing up a few hundred MB of memory when it could be using kilobytes. And your answer is "get another computer per tab"

I told you that no one cares and it doesnt matter. Saying forking is bad because it consumes more memory is like saying windows 10 is worse performing OS than windows 98 or chrome is a worse performing websrowser than internet explorer because they eat up more ram. It makes no sense. 

 

54 minutes ago, Kisai said:

So each time that involves large memory copies, virtual memory allocation, and so forth. Machines are only getting slower because of this incompetence.

Modern day data centers still run on tape drive. I seriously do not understand why you keep talking about speed, memory, and whatever when these arent an issue while productivity,feasibility, reliability, and availability are.

[Kisai]

16 minutes ago, wasab said:

I do

I told you that no one cares and it doesnt matter. Saying forking is bad because it consumes more memory is like saying windows 10 is worse performing OS than windows 98 or chrome is a worse performing websrowser than internet explorer because they eat up more ram. It makes no sense. 

 

Modern day data centers still run on tape drive. I seriously do not understand why you keep talking about speed, memory, and whatever when these arent an issue while productivity,feasibility, reliability, and availability are.

I doubt you do, and you're just posting your usual nonsense. 

 

16 servers with 4 cores/w 64GB each ($1575USD/ea Poweredge R240) = $25200

4 servers with 16 cores and 256GB ram ($7400 USD/ea Poweredge r6515) = $29600

 

Guess which you're allowed to put on a 15A circuit? 4 if you're lucky.

42U rack with 15A and 1000Mbps, $400/mo

 

So the monthly costs for the bottom configuration is $400/mo, the monthly cost for the top configuration is $1600/mo. If you're going to operate servers, you want the lowest costs, and that means if your requirements can be had with 64 cores in one 15A circuit rather than 4 15A circuits, you're going to do that.