Slashdot Mirror
86% of Windows 7 PCs Maxing Out Memory
CWmike writes "Citing data from Devil Mountain Software's community-based Exo.performance.network (XPnet), Craig Barth, the company's chief technology officer, said that new metrics reveal an unsettling trend. On average, 86% of Windows 7 machines in the XPnet pool are regularly consuming 90%-95% of their available RAM, resulting in slow-downs as the systems were forced to increasingly turn to disk-based virtual memory to handle tasks. The 86% mark for Windows 7 is more than twice the average number of Windows XP machines that run at the memory 'saturation' point, and this comes despite more RAM being available on most Windows 7 machines. 'This is alarming,' Barth said of Windows 7 machines' resource consumption. 'For the OS to be pushing the hardware limits this quickly is amazing. Windows 7 is not the lean, mean version of Vista that you may think it is.'"
If all that RAM was simply being used for a filesystem cache, the system would not have to "increasingly turn to disk-based virtual memory to handle tasks" - it would just drop some cache when it needed to start a new task, as you said.
It seems that something else is going on.
Just another proletarian malcontent.
Let's start from the story (which I *did* read) - 'Barth acknowledged that XPnet's data couldn't determine whether the memory usage was by the operating system itself, or an increased number of applications,"'
Right there I'd be suspect whether this is even an issue or not. Given Windows (which I generally regard as inferior) as an OS having lots of functionality, I wouldn't be suprised if it takes up all available RAM prior to utilizing swap. I'm on my 2GB Ubuntu system right now and am running at 18% of 2GB with just Mozilla (with two tabs) and Thunderbird. But there's also my network layer (Network Monitor), KTorrent, and my bluetooth daemon running in the background. All told, System Monitor says i have 31 processes running.
Let's do a like comparison - run the exact number of apps and processes before declaring a memory leak.
Sheesh!
The Kai's Semi-Updated Website Thingy
I don't know how caching works in W7 but on Linux, if the system has to "turn to disk-based virtual memory to handle tasks" then the memory utilization isn't caused by buffers because buffers are never swapped out to disk. If W7 behaves in a similar manner the it's either a memory leak, system bloat or the caching mechanism is broken.
The new memory models in recent OS's try to utilize all the available RAM (as they should) to speed up things otherwise. It makes a lot of sense to cache things from hard-drive in low-peak usage points, and in such such way that it doesn't interfere with other perfomance. When the things that are most often used are already cached in RAM, their loading works a lot faster. This doesn't include only files, icons or such, but everything the OS could use or do that takes time.
If theres a sudden need for more RAM, the cached data can be "dropped" in no time. It doesn't matter if it averages at 25% or 95%, just that the perfomance overally is better when you utilize all the resources you can to speed up things in general.
Assuming your claims of how Windows 7 is implemented are true, then the claims from the person who actually collected all the empirical evidence must be false:
resulting in slow-downs as the systems were forced to increasingly turn to disk-based virtual memory to handle tasks.
If the memory was freed up dynamically as needed then no processes would ever be forced to resort to disk-based virtual memory. So either you work at Microsoft and are assuring us that the implementation protects against this or you're speculating against someone who has claimed to gathered a large enough to make such accusations.
... the guy who is offering me numbers.
No offense but I'm going to side with the guy who appears to make his living testing these sorts of things
My work here is dung.
According to their website blog, they are "rethinking windows performance."
So instead of thinking about what actually effects OS performance, they are rethinking things so that they don't have to sell real solutions to their customers, "where [they] maintain several large installations of our commercial DMS Clarity Suite performance analysis solution."
"His name was James Damore."
1. "CWMike" (Computer World Mike) submits Windows bashing link (to Computer World) on Slashdot /. /. blather to follow.
2. Slashdot "editor" blindly shoves it on the front page
3. Page hits to CW from
4. CW profits!
5. Mind numbing
I guess Devil Mountain or whoever don't know about SuperFetch. Or need publicity.
And I guess slashdot editors don't know about SuperFetch. Or maybe an article like this gets them more traffic, revenue, etc.
The fucking bullshit that passes for articles these days..
I think the issue here is that the system is turning to swap. Caching stuff that may be referenced again is fine and dandy, but if the system regularly turns to swap just to keep itself afloat, then you have a problem.
The metric to count is the number of page faults, an indicator of the number of times that the OS addresses memory that isn't in RAM.
As others point out, measuring just the fraction of memory consumption is stupid. I have 6GB of RAM ; my processes are using about 1.7GB of that, but the OS is claiming that 3.8GB is consumed. So that's 2.1GB of cached data that I no longer have to wait for from disk. Hooray.
TFA hints that they may be measuring page faults, and does mention that Win7 is hitting the disk for virtual memory more often. But they should make that clearer if it's the case.
Here we see why /. needs a "-1, Wrong" mod.
Yes. I have a similar problem when people running servers complain that the CPU is at 100%.
If you're seeing an actual slowdown in performance, fine, worry about it.
Otherwise, 100% CPU usage is a good thing: it means there's a process that's not IO bound.
Ahh fair enough. "Colour me learned something today". :-)
Ive been running Windows 7 since the BETA release. I have never experienced any issues that result in I/O thrashing against the hard drive as result of all my ram being utilized. I also have numerous friends running Windows 7 none have reported any issues like this, if anything its been praise for the operating system. So I am amazed to see a 86% number being thrown out there yet never seen this problem before.
No offense but I'm going to side with the guy who appears to make his living testing these sorts of things ... the guy who is offering me numbers.
And now lets quote something from the article...
Barth acknowledged that XPnet's data couldn't determine whether the memory usage was by the operating system itself, or an increased number of applications, but said that Devil Mountain would start working on finding which is the dominant factor in increased memory use.
This single sentence makes the article rubbish. They have no clue whats causing the heavy memory usage, its just an assumption that the OS is causing it and they're yelling fire before looking through all the data or completely analyzing the problem.
Actually in modern operating systems RAM can be used as a disk cache is such a way as it can be "freed" at little to no cost when needed by programs. This in effect means you get quick access to often used programs as they are already cached in RAM.
The only situation in which "RAM is filled with crap and the OS goes to disk" (paraphrased) is when RAM is full - with programs. In this situation yes, it has to go to disk.
You're making an apples-to-oranges comparison, and you don't mention what software is running on either of your machines...
I've found Vista and Win7 to generally work better, giving decent enough hardware. Because of SuperFetch, Visual Studio loads faster on my Vista64 dualcore laptop with 7200rpm drive and 2 gigs of memory than on XP64 quadcore workstation with 10000rpm raptor drive and 8 gigs of memory. But this isn't a fair comparison either, since the rest of the software suite on those two machines are specced differently.
Coffee-driven development.
Definitely the part about HDD caching slowing things down. Even in the DOS age it was well known that hdd caching utilities (I forgot the names, too long ago) improve disk performance tremendously.
/dev/null
Linux does the same things as Windows: it caches as much stuff from disk into main memory as possible. Try running:
cat large_video_file.avi >
You'll see that after running the command, your memory usage jumps up by the size of the video file. Now try running the same command again, it's now an order of a magnitude faster.
On Linux things like this are stored in main memory in the form of caches and buffers. I don't know about Windows, but Linux clears some caches and buffers if applications need real memory. Caches and buffers show up in memory usage reporting tools like 'free', so it's quite normal to see Linux systems using 90% or more RAM, most of which go to caches and buffers. It seems that most people who complain about memory usage don't know how memory is managed on modern operating systems, so they go all apeshit about "OMG HELP linux is using so much memory it sux0rz!!!" and I have to explain again and again how they're not getting it. Same goes to you. Now, Windows is suffering from the same problem.
FYI, here's the memory usage of my Linux server:
total used free shared buffers cached
Mem: 720 702 17 0 55 510
-/+ buffers/cache: 136 583
Swap: 399 0 399
It says 702 MB of used memory. Now look at "-/+ buffers/cache", it says 136 MB. That's the amount of memory *actually* used by applications.
Read TFA. It just claims that Windows 7 consumes all available RAM. That is the "empirical evidence." System slowdown was NOT measured.
Utilizing all available RAM is a pretty well understood technique at this point. All web browsers do this now, as do many other applications. One would expect a well-designed modern OS to do this. Consuming all memory itself is not a sign of poor programming itself, so long as disk caching of things that should be in RAM doesn't occur. This is not something that the people in the article has measured.
I'm going to side with the guy who appears to make his living testing these sorts of things.
Bad science is bad irrespective of the person conducting it. And whatever the original tester said is getting filtered through the viewpoint of the gentleman writing the article. Considering that he says that "Windows 7 is not the lean, mean version of Vista that you may think it is," yet never once compares statistics to Vista (or even mentions it outside of this statement), I'd take the conclusions from these stats with a grain of salt.
The ______ Agenda
Which part is wrong?
Your old P4 is a minimum 3 - 4 times slower at raw CPU tasks than the Athlon X2 (http://techreport.com/articles.x/18448/13). The story about your 512Mb XP machine running faster than the Athlon X2 3Gb machine simply does not ring true. Perhaps you were only playing solitaire on it? No-one is denying that Puppy Linux is fast, or small, or that it fits in memory. Try running Open Office on it with only 30Mb of Ram and then tell me that it is fast.
Maybe the part about HDD caching slowing things down?
YES the part about HDD cacheing slowing things down. HDD cacheing speed things up! Memory Swapping slows things down - which is simply a question of how big your application + operating system's "working set" of memory is (i.e. the memory that simply has to be held in Ram because it is being actively read and written). 3Gb with Windows 7 is plenty for "normal" apps, and it won't be any slower than your 512Mb XP machine running the SAME applications - in fact it'll be a load faster because it won't have to swap as much and the disk is likely faster.
>>>You're making an apples-to-oranges comparison, and you don't mention what software is running on either of your machines...
Nothing exotic. The Windows OS plus Firefox browser. My P4-XP-512MB machine runs faster than my brother's AMD X2-WIN7-3GB machine. XP is more responsive.
.
>>>I've found Vista and Win7 to generally work better, giving decent enough hardware.
Well my Vista install was running Microsoft's minimum recommendation (512 megabytes) and ran like a snail through molasses. When I later upgraded it to 2 gig it did run better, but still nowhere as fast as my old 0.5 gig XP machine, which is why I eventually went back to XP.
My overall impression is that Vista is like a Beetle with a lot of chrome/pinstriping/et cetera. Looks good. Runs like crap.
FOX NEWS.com should be BANNED from television and internet. Have the Congress take it over and give us Truespeak.
The xpnet site does say they factor in "how often it relies on virtual memory", but not how they do this
Even that isn't ideal. It makes sense to swap out some library that hasn't been used in hours or days. The more ram available for disk cache the better. The only solution is to look at Memory used minus disk cache (like the 'free' command on Linux does).
"Knowledge is the only instrument of production that is not subject to diminishing returns" -Journal of Political Econom
You missed the most important emphasis:
Ostensibly means "to all outward appearances." In other words, they're admitting they don't really know the inner workings, the true cause of the delays. They're just supposing it's due to RAM swapping, as opposed to increased networking activity, aero glass, more concurrent programs being run on average, or any other number of other wag'd reasons. Basically, they picked two measurements that are both higher in W7, and just said, "Well, it stands to reason that A causes B." What's that phrase that internet smarty-pantses use all the time about this?
There are two kinds of people: 1) those who start arrays with one and 1) those who start them with zero.
Comment removed based on user account deletion
The 'anecdote' to which you refer is completely plausible. Remember that the info wasn't based on any sort of raw processing power, frames per second, or disk throughput: it was based upon human perception. Don't ever throw the two together and think that they will always (or even mostly) match up.
I can tell you that a computer that has 4x the specs of another doesn't necessarily end up running better, regardless of the (Windows) OS. This can have a lot to do with hard drive fragmentation: one system that has low specs but is defragmented routinely can appear to run faster than a high-end machine that has 50% file fragmentation. I know it sounds like I'm picking on Windows, but the issue with data fragmentation--and the need to defragment--wholly resides on the shoulders of Windows users. And again, the speed difference, especially when it comes to perception, can be hugely affected by how much the hard drive is thrashing when a program is started.
Another issue in perception is with video performance. Two identical machines with the same model of high-end video card can have completely different apparent speeds if one has problematic drivers. A number of other things might be measured. I remember back when a person would say that their high-end, I-just-spent-more-on-my-computer-than-my-car computer was running slow simply because their dial-up connection was slow: the computer was running perfectly, but the user perceived it to be an issue with the computer (instead of the slow connection) because the data wasn't loading fast.
Sanity is like a condom: rather have it and not need it, than need it and not have it.
If stuff slows down due to that swap out, then it's still accurate enough for me.
Maybe the O/S could get it right and swap out and swap in Firefox in a way so I won't notice any slow downs. Give me an example of such an O/S please.
So far in my experience, if Windows or Linux swaps out Firefox for whatever reason, if I then switch to Firefox, I have to wait for it to be swapped back in.
Why "page out" and not "page in"?
"Page in" doesn't necessarily mean that I'll have to wait if I switch to different programs- the O/S is bringing stuff from disk to ram - I believe in some cases the O/S pages in stuff as part of running a new program - so it's not such a useful metric for "not enough memory".
But "page out" means something in RAM is going to disk - if I ever want it back in RAM, I'll have to wait.
If stuff in RAM is going to disk needlessly and causing unnecessary waits then the O/S virtual memory algorithm is getting things wrong.
Doesn't require advanced knowledge to make sense of that ouput,
I'd argue that knowledge of what "wired" means in that context is advanced knowledge, required to make sense of that output. I've been working with computers for a hell of a long time, and I have absolutely no clue what "wired" means in regards to allocation of memory.
I also don't have whatever knowledge (advanced or otherwise) to understand the difference between "buffer" and "cache." And I'm only guessing that "inact" means "inactive", although I don't really get the distinction between "inactive" and "free."
My first guess would be "wired" is total amount of RAM, but from the numbers that's clearly wrong. In fact, that report doesn't seem to tell you your total amount of RAM at all... which is less than useful.
Anyway, the point is, if you think that readout is obvious, you've completely lost contact with the general public. I hope to God you don't write UIs that you expect other people to use.
Comment of the year
Ok that's retarded.
Using 100% of your RAM can easily make your computer slower. Firstly, if a page is "dirty", it needs to be saved to disk before it can be used for another activity. This causes a write to disk before a read, automatically making the read slower.
While that's true, it doesn't matter because those pages would be in-use no matter how much RAM you have.
The "free" RAM (that is, RAM not being used for applications, that is, RAM pages that won't ever be marked dirty) is all caching-- disk and DLL caching. It'll never get marked dirty, and therefore never need saving to disk before being reused.
The problem is "which page do you overwrite next?" If a program is about to read from a portion of its program memory, or is about to read data sitting in the disk cache, you don't want to delete that page. It is just that the operating system doesn't know which piece of data the program will need next.
That's... somewhat true.
First of all, while the OS doesn't know what piece of data the program will need next, it certainly can (and does) make a guess based on what the program has requested in the past. Most programs allocate most of their memory when first starting up, and the OS is going to be able to predict these bits almost 100% of the time. (Unless it's the first run of that program. But it's not magic.)
Secondly, even if the guesses are bad, even if only one guess in a thousand is correct you're still better-off. Tossing away RAM from a bad guess takes zero time, for all practical purposes. In reality it takes time, but it takes less time than the disk takes to load up the data to fill that page, so in practice it takes no time.
Therefore, the penalty for a wrong guess is zero. The reward for a correct guess is significantly higher than zero. It's a win-win situation.
The result is that having empty RAM is almost always beneficial.
Wrong, for the reasons listed above.
Blindly swapping program memory to disk is not a good idea. It never really was.
True, but Windows doesn't do that and that's not what we're talking about. That would never happen unless RAM was 100% full of application data. Even if that was the situation, it's hardly "blind" swapping, as the same heuristics that allow Windows to predict what an application will load next can help it predict what an application will never touch again.
But to repeat: this isn't the situation we're talking about. We're talking about a computer filling unused RAM with disk and DLL cache. You're talking about a computer whose RAM is 100% full of application memory. Apples to oranges.
On today's computers, a memory leak can effectively stop/bring down the computer. The real function of the page file is to soften this problem. Instead of dying relatively quickly, today's computers will slow to a crawl, and then die. The page file causes this.
Ok... and? What's your point? Are you saying that if we didn't cache to RAM this wouldn't happen?
If you have good, well behaved software, and 4 GB of RAM, you will likely experience better system performance under Windows XP by turning the page file off.
Unless that software wants to memory-map a large file, in which case your performance will go straight to hell. Also, in practice, there's a ton of stuff Windows loads initially that can be (more or less) swapped-out immediately and stay there.
For example, the login screen is likely to only be displayed once, ever. And when it's needed again, there's plenty of time to load it-- why leave it in RAM? (And Windows is smart enough not to.)
The only time I'd consider turning VM off entirely is if you're running MacOS 8. Which had worse than useless virtual memory support. You're not, you're using an OS made in the 21st century, turn on the fucking swap file.
Comment of the year
Wrong.
PAE