« first day (1750 days earlier)      last day (3589 days later) » 

This number in the task manager tells you how much time disk requests are taking to be handled.
user15026
Okay, that makes sense.
24ms is pretty low (this is an SSD), and even then it's an eternity in CPU time.
@badp how many clock cycles is that?
I was about to ask @AshleyNunn just that.
23:02
Hang on... uh... calculates for self
@GodEmperorDune doesn't that depend on cpu frequency...?
user15026
@badp i dont know :(
Say your netbook has a mere 1 GHz CPU. That means 1 billion cycles per second. So that's 1 CPU cycle per millionth of second. 24 milliseconds is 24,000,000 billionths of second. So that's 24,000,000 cycles.
Woo, I was right! I guess
@Chippies Given, 3.57 GHz
user15026
23:03
@badp Oh, right. I do know how to do that.
Sorry I done goofed up
@Unionhawk The answer is: too many for me to calculate
user15026
(I am still working on getting all of this locked into my head)
(Did you notice I'm kinda tired? >_<)
So even a brand spankin' new SSD with barely any disk activity on it makes my CPU wait for millions and millions of cycles.
It only gets worse.
See, what happens is that since you need to hold more memory than you have room for, this means you have to pick what gets to stay on RAM and what doesn't.
There are a few ways to do that, but naturally none is perfect. When you decide that you accidentally got rid of some memory you did need, you have a "hard page fault" and you have to wait for millions of cycles.
Usually operating systems will privilege the processes that are on the foreground.
If multiple processes however are competing for your RAM, what might happen is that essentially each will step on the other's toes.
And since the operating system is not a wizard, it's not really easy to decide what stays and goes optimally
23:08
So more and more memory gets swapped and more and more hard page faults happen and more and more disk requests queue up and more and more time is required to service them
the cpu ends up spending more time waiting for disk than doing anything productive
so essentially there comes a point where it doesn't matter how much swap your computer has, because it's already completely unusuable way before you've filled it up
If you've ever tried to alt-tab between programs and it takes an eternity and the UI seems to draw bit by bit painfully slow -- this is what is happening.
The computer is literally scrambling to gather the bits of memory from the hard drive swap that tells it how to draw the window.
Windows had thrown all of that away because the program was in the background and it figured it wasn't anything too urgent to hold on to.
So in 99% of computers, the hard drive is the bottleneck. How much RAM you have dictates how much you need to swap and makes the bottleneck that much tight
...but it's still not about how much RAM you have, but how expensive it is to deal with needing more RAM than what you have.
@badp what is the bottleneck on the other 1%?
You just get stuck in the bottleneck more often.
@GodEmperorDune Various, depending on the application
23:12
@GodEmperorDune I guess that if your computer's main task is, say, transcoding videos, then the CPU might be your bottleneck.
If you have enough RAM to swap very infrequently, swapping isn't really your bottleneck
assuming you have enough memory for the task that disk read times don't really become a factor.
@murgatroid99 Windows will swap regardless of how much RAM you have. Linux will only use swap files if absolutely necessary.
@badp you're saying that with half-empty ram, Windows will still swap pages into the hard drive?
Yeah, I'm at 37% swap usage, according to rainmeter
@murgatroid99 Absolutely.
23:14
why? That makes no sense
@murgatroid99 because windows
@GodEmperorDune windows wasn't programmed by idiots
@murgatroid99 It's to speed up swapping in the event RAM does become full. Windows preloads content into the pagefile, that's why.
@DragonLord copying into the page file is not the same as swapping out to the pagefile
@DragonLord Yep, I was looking for the Raymond Chen blog post about it but I can't find it.
284
A: Any benefit or detriment from removing a pagefile on an 8GB RAM machine?

quuxTL;DR version: Let Windows handle your memory/pagefile settings. The people at MS have spent a lot more hours thinking about these issues than most of us sysadmins. Many people seem to assume that Windows pushes data into the pagefile on demand. EG: something wants a lot of memory, and there is ...

Bridge, I bought myself a new figure:
At any rate
Here's a nice thing about Linux. Say a rogue program has balooned up, and you managed to kill it in time.
Now you have a shitton of stuff in swap and lots of free memory.
If you have more free memory than paged (swapped) memory, you can temporarily disable swapping to force it all back to RAM in one go.
It's sudo swapoff -a.
You could run that and go for a snack.
(Windows doesn't have such a flexible swap management and you have to just suck it.)
You can then re-enable swap with sudo swapon -a. Since all of your memory already fits entirely in RAM, that won't do much at all.
That's your free crash course on hard drives and swapping and how it's almost always the bottleneck.
@badp if the memory is associated with a killed program, why wouldn't it just mark the space as usable?
@murgatroid99 It does, but it doesn't actively try and fetch back memory pages from swap. You'd risk making things a hell of a lot worse.
23:22
Anyway, that doesn't mean that swapping is usually the bottleneck
That is unless you tell it to.
Obviously, swapping is the bottleneck in the case where you have to use a lot of swap space
I know it was on my laptop, and it is for my desktop at work, and it 100% is on her netbook.
(It's quite sad really.)
It depends on what programs you're running
Sure, I don't swap a lot on my current desktop with SSDs, 16 GB of RAM etc.
23:24
If you have 8GB of ram, and all of your programs put together only allocate 7GB, swapping should never be a problem
@murgatroid99 oh lol. IBM's version of Eclipse, two copies of Websphere, a couple Chrome tabs, usually Sublime Text, fish in cygwin, a couple oddballs. Also occasionally the virus scanner kicks up and I can't do jack shit to stop it.
4 GB of RAM. Windows 7.
Eclipse and Websphere alone take up more than 2 GB.
@badp Yeah, that's my point. If you're running all of that memory intensive stuff, obviously swap is your bottleneck. But that doesn't make it a general rule
@murgatroid99 Unfortunately most customer grade computers are specced so that swapping is the bottleneck.
@badp if you're running memory intensive stuff
Hell, even 4 GB isn't going to be enough RAM for just moderate Chrome activity.
23:26
@badp Chrome is a memory hog
Well yes browsers are memory intensive stuff.
@badp except laptops. Every laptop under the sun nowadays has more ram than average customer would ever need
@Chippies I had 8 GB in my desktop and I still felt the need to go to 16 GB.
@badp you're not the average user though
@badp The only point I'm trying to make is that when you say that "swapping is almost always the bottleneck", it's misleading, if not flat out wrong, because it depends on how you're using your computer
23:28
@Chippies actually nowadays my load is probably lighter than an average user's. When I finally get home I've only got energy for Chrome and maybe a videogame.
also laptops have weaker parts as well, so they are more likely to bottleneck elsewhere before their 8gb is used up
@badp The primary bottleneck for everyday computer use is the storage, more often than not.
This is why an SSD is such a big deal these days.
@murgatroid99 I rest convinced that in the most general usage patterns, that's where your bottleneck is. Yes, there are more specialized use cases where you might be bottlenecked by the CPU. Another example is bitcoin mining.
@DragonLord my wife has a 120gb SSD on her gaming PC and a tiny 300gb HDD, which is hardly ever used and she's doing fine
Those are not most general usage patterns.
23:30
@badp You're running Eclipse, Sublime, and Chrome. That's not a general usage pattern either
@murgatroid99 It's still close enough that the hard drive and swapping are 100% the bottleneck.
Even if HDD is not usually the bottleneck, it is the slowest part in the computer. SSD makes a HUGE difference
@badp that doesn't even make sense, unless you're trying to make a circular argument
you still have to load things from HDD into RAM even if you're not using swap at all
I am not.
I said "common cases result in this being the bottleneck." Your retort is "yours isn't the common case." I say: "and yet this is still the bottleneck."
There is no circularity.
23:33
@badp I'm saying that the fact that you're getting that bottleneck has no relation to the common case
If you want to go GPU mining by all way skimp on everything in order to get those Titan's in budget.
@murgatroid99 Fine. That does not change the veracity of my claim.
@badp which claim?
You are not disproving my claim with that retort.
So I guess that's settled. I bid you good night.
You say "the common case has swapping as the bottleneck." Your evidence is "my setup has swapping as the bottleneck." I say that your case is substantially different from the common case, which implies that that evidence does not apply to that assertion
@murgatroid99 No. I have plenty of evidence for that. I've spent the last hour explaining my point.
23:38
@LessPop_MoreFizz who should I cheer for in the Washington-Rangers series? My personal feeling is neither.
@badp you asserted it several times, but I can't find any other evidence
I am struggling to remain polite about this, so I will simply accept your right to be wrong on the internet. Bye.
@MBraedley Rags because Lundqvist is a stand up dude who deserves a win.
He is literally the only reason I can cheer for that team, but he's a pretty tremendous reason to cheer.
@badp I've said this before. "You have the right too be wrong on the internet" is probably the most obnoxious and condescending dismissal of a person I've ever seen
@LessPop_MoreFizz Actually, TBQH, I should be cheering for the team that will give me the better chance of winning the office pool. Not quite sure who that is right now though.
23:41
@murgatroid99 Yes, I'm dismissing this argument with roughly the same amount of condescension I'm seeing coming from you. I was hoping you'd realize this is a pattern with you and the arguments you get into.
This is why I am quite strongly refusing to continue.
This is my final "assertion" for the night.
while Idk if swapping plays a huge role in this, but loading programs from an HDD takes much longer than from SSD, therefore storage drive is the bottleneck
considering average user closes their browser every time they're done looking at things online, that is indeed their main bottleneck
@LessPop_MoreFizz I think the Rangers will give me a better chance at winning the pool. Which is good, because I think I'd have a hard time cheering for Washington.
whenever HDD is accessed, everything else has to wait on it, so it's the bottleneck by default
Even though NYR eliminated the Pens.
@badp I definitely didn't intend that, and I'm sorry if I came off that way. But I honestly don't see it. I was just trying to have a conversation about a statement I disagreed with.
23:45
and that being said, back to running grifts!
@mbraedley also the rangers twitter gimmick this week is just top notch.
@LessPop_MoreFizz didn't see it
The game basically decompiles the internal bytecode and displays Lua assembly code for manipulation.
@MBraedley everything on their Twitter is all lowercase.
And will be until the series is over.
23:50
The game is all about tampering with its internals to solve puzzles.
So it's not unexpected to cause the Lua interpreter to generate errors, you're manipulating the game code directly.
That's why there are these "The universe has collapsed" dialogs.
In the game's final battle, the antagonist will in fact try to execute code which will trigger these crashes.
@DragonLord ouch

« first day (1750 days earlier)      last day (3589 days later) »