Wikipedia:Reference desk/Archives/Computing/2014 November 27

Welcome to the Wikipedia Computing Reference Desk Archives
The page you are currently viewing is an archive page. While you can leave answers for any questions shown below, please ask new questions on one of the current reference desk pages.

November 27

L1 cache size question

My Sandy Bridge i7 has 32KB of L1 cache per core and I think the newer Haswell has twice that. But that is a tiny amount. In several tests, I've found that if I can get all of the data into the L1 cache, the program often runs 2-3 times faster. So doing it in a way that would be less efficient if all of the memory was the same speed can be more than twice as fast! Why not make a much larger L1 cache on the CPU? I think it takes six transistors per bit of memory (correct me if I'm wrong). The Haswell has 1.4 billion transistors, and some of them are for things that don't really need to be on the CPU (e.g. graphics). So you can have an L1 caches of 1MB for about 50 million transistors. You could have four of these for 14.3% of the total transistors. In my tests, there are diminishing returns for the third and additional cores, probably because of memory bandwidth. So why not make big L1 caches on the CPU (more for data than instructions)? It seems to me that three cores with big caches would run circles around four cores with small caches. I think that two cores with big caches would beat four cores with small caches. Bubba73 ^{You talkin' to me?} 00:18, 27 November 2014 (UTC)

Hack intel if you really need the information – just kidding. L1 cache is static RAM. In market time of Pentium 3, I guess the Celeron was nothing else than Pentium with defective and deactivated optional units like cache blocks or media extensions like MMX. --Hans Haase (talk) 06:39, 27 November 2014 (UTC)

Memory is low, but cache expensive. See Category:Microprocessor dies on commons. It would be nice, if anybody would add image notes to these pictures. --Hans Haase (talk) 06:50, 27 November 2014 (UTC)

The L1 cache is on the CPU, right? Having a big L1 would make a much bigger improvement in performance than Intel is getting with all of these tiny increases. For instance, Haswell (microarchitecture) says "Total performance improvement on average is about 3% (compared to Ivy Bridge)" - that that is an Intel "TOCK"!

I had a Motorola 6502, then Intel 8088, 80386, Pentium 1, Pentium 2, Pentium 4, etc. (I skipped the 80286, the 486, and Pentium 3). But each of those was a major improvement over the previous one - much larger jumps in performance than the sub-5% increases for a new generation now (even a "tock"). I don't know much about this stuff, but it seems that making a large L1 would result in a big boost in performance. Bubba73 ^{You talkin' to me?} 07:14, 27 November 2014 (UTC)

In modern chips, there are two and sometimes three caches on the chip itself (e.g. L1, L2, L3). Since these are all on the chip itself, the performance difference between them is largely related to their relative complexity and the distance from the cache to the core. Making a cache larger will add complexity and distance, and negate much of the advantage you are hoping to create. That's why modern chips generally use several levels of caches, so they can have both a very small, simple, fast cache, and a slightly slower but still faster than main memory cache. The largest cache on a modern Xeon is ~24 MB and including memory controllers occupies about 20-25% of the die. Dragons flight (talk) 08:18, 27 November 2014 (UTC)

But the size of the caches has generally been increasing, albeit very slowly. I develop software on an i7 and optimize it for the i7 L1 cache. But mostly it executes on i5s, so I have to take it over there and re-optimize it for the smaller i5 L1 cache. L3 cache size is definitely an important factor.

PS. Haswell was never intended to be a major performance release, rather Intel's focus in that design iteration was on reducing power consumption. They dropped wattage by ~40% at the same performance levels. This made it possible to put more cores on a single chip, which increased the possible performance for the ultra high-end many-core chips, and allowed them to expand the usability of their low power designs in tablets and other portables. Dragons flight (talk) 08:34, 27 November 2014 (UTC)

I think they would be better off making significantly larger L1 cache than adding more cores. There is a diminishing return on the number of cores, starting at three. In my tests, running eight threads on a hyperthreaded quad-core i7 is at best 5.5 times running on a single core. Usually it is less than that. Unless I made a mistake, running four threads of prime95 on my i7 is less productive than running three! (Not so on the i5). Bubba73 ^{You talkin' to me?} 08:48, 27 November 2014 (UTC)

As Dragons flight said, a larger L1 cache would be slower. Modern CPUs do have multiple megabytes of cache on the die, but it's called L2 and is somewhat slower than L1 for unavoidable reasons, including signal propagation time. -- BenRG (talk) 19:39, 27 November 2014 (UTC)

He mentioned the distance. It can't take that much longer for the electrons to travel that distance. And historically, haven't they been increasing the sizes of the caches? Wouldn't a larger L1 cache mean that you are more likely to get a hit in that cache? Or maybe the L2 cache is the one that needs to be much larger. Bubba73 ^{You talkin' to me?} 03:12, 28 November 2014 (UTC)

The speed of light is roughly 3*10⁹m/s. Grace Hopper famously illustrated a nanosecond with a foot of telephone wire. If your core runs at 4 GHz, then in one clock cycle, a signal can at most propagate 7.5 cm. For a round trip, 3.75 cm. So if your cache connections are just 4 mm away from the core, you will spend 10% of your overall time budged for signal propagation in the best case, ignoring the difficulties of synchronising different part of the processor with different signal propagation times. --Stephan Schulz (talk) 08:34, 28 November 2014 (UTC)

A cache on the CPU chip is still a lot closer than the DIMMs! And if the software can't get it out of the cache, it has to go to the main memory, which is much slower. Bubba73 ^{You talkin' to me?} 01:59, 29 November 2014 (UTC)

Sure. That's why we build caches and the whole memory hierarchy. But it's a trade-off. The bigger the cache, the slower it is. With current technology (and likely future technologies), even if money were no concern, at any given technological level it's better to build several levels of increasingly larger caches than one giant cache. --Stephan Schulz (talk) 15:04, 30 November 2014 (UTC)

Computer Crashed and probably ruined. Seemingly due to driver install, believing I upgraded from AMD radeon to AMD Catalyst. It all went wrong.

Hey. I was running this new game on my computer yesterday, but in order to run it I was prompted to update my AMD Radeon driver from version 13.9 to at least 13.12. Not being an ace on computers I followed the link given me to AMD.com where I started looking for an update. I found no Radeon, I only found 'AMD Catalyst 14.9' and I thought this is probably a newer and better driver so I started installing it. I was walking a bit to and from the computer in the meantime, and when I came back the one time the computer crashed just as I returned, and there was blue screen with white writing on it, an error message, lots of writing, then PC turned itself off and rebooted. The driver was seemingly not done installing before the crash (can't say for sure as I returned only to get the blue screen up just as I entered the room), but I tested the game and it worked fine, seemingly at least. But the crashes continued sporadically 4-5 times. I could run the computer and the game for a little while, but then it would crash, give me the blue error-screen and then reboot... Then the fourth or fifth time it stopped rebooting - it no longer managed.. Computer constantly tries to start up if I turn it on, but within 1-2 seconds it stops and fall silent, dead... then another 1 or 2 seconds pass and it tries again only to fail after 1-2 seconds. It keeps trying to start up like this until I unplug it and remove its source of electricity. If I try again, then same old story... Its attempts to start up is completely futile, but it never stops trying; start-stop-start-stop-start-stop... In fact, when I try now today, the 1-2 seconds are slightly increased to maybe 3-5 seconds, and the computer starts ringing like an alarm, then it fails, and repeats just as described above.

As I said, I'm no ace on computers, but it seems clear it was the installing (or half installing?) of the Catalyst driver that fu*ked and ruined my computer up.. Maybe my assumption that the Catalyst was an upgrade of the Radeon and would overwrite it and improve it was entirely wrong? Maybe they're simply not compatible?

Anything I can do? Or must I bring it in for repair? It's fairly new and all. Lots of files and documents on it that I need, so I'm a bit worried. 2A02:FE0:C711:5C41:7476:7B3B:50FC:E143 (talk) 08:54, 27 November 2014 (UTC)

You enforced to install a wrong driver. Turn on, wait till RAM test and BIOS/UEFI finished. Before you see anything from Windows, hit F8. Start in "safe more". Login as administrator, uninstall the driver software, reboot. Instead of this, choose instead "safe mode", "last known good". --Hans Haase (talk) 11:47, 27 November 2014 (UTC)

But I can't get to the part where I can start "safe mode"... because the computer don't start up.. it fails at doing so after only barely few seconds. From your reply, should I assume the computer's "start-stop-start-stop" - behavior is due to what you call "RAM test and BIOS/UEFI" and that it will eventually start up and that then is the time to follow your instructions? 2A02:FE0:C711:5C41:7476:7B3B:50FC:E143 (talk) 12:13, 27 November 2014 (UTC)

A few seconds into boot is before the AMD Catalyst driver loads, so it could have caused this only by physically damaging the graphics card. I think that's very unlikely. This may be a coincidence of timing. It could also be that, say, the fan on the card had failed some time previously, and playing this game caused the card to overheat and damaged it (although I think modern GPUs have thermal protection circuitry to prevent that). Or playing the game taxed one of your fans or your power supply to the point of failure and the power-on self test is detecting that and shutting down to prevent damaging other hardware.

I would send it in for warranty service. If the hard drive is removable and removing it doesn't void the warranty, I would remove it before sending the rest for repair, as it's the only component that's irreplaceable and it's unlikely to be the cause of the problem. (You could try turning the machine on again after removing it. If it still shuts down, the hard drive isn't the problem. If it stays on and complains about a missing hard drive, the hard drive may be the problem.) -- BenRG (talk) 19:08, 27 November 2014 (UTC)

BEN RG - there has been no previous failure of anything. It's fairly new, and it may not be the strongest or best computer out there, but I believe it's quite decent. That the fan fails and the thing overheats seems very unlikely for a new computer built for the purpose of playing new, modern games. Besides, as I said, this crashing started during the installing of the new driver...First time it happened, I wasn't even in game. I was on the desktop and the only thing running was the installing of the new driver, and it kept happening after that. According to Hans Haase above I enforced wrong driver to install... He's probably right. I've no warranty as I didn't buy it in the store. I bought through a private person I know, as I always do. Knows everything there is to know about computers - He'll help me, when he finally becomes available, which is rare. Already texted him last night. I just hoped there might be something I could do myself. But I realize trying to remove the hard-drive and stuff on my own might not be the smartest thing to do. This here problem seemingly started due to my incompetence with computers in the first place. I should avoid botching it further. But thanks for replies. ;) 2A02:FE0:C711:5C41:5B9:900C:86D7:EAA9 (talk) 07:13, 28 November 2014 (UTC)

When turning on power, if and charakter or picture, generaded by the computer appers on the screen, the solution above should work. F8 key is caught by BIOS/UEFI pending on the mainboard/BIOS/UEFI manufacturer for what use. When BIOS/UEFI finnished, the bootdrive is accessed and the operating system starts up. Windows listens for F8 key, too. I also have often trouble, to get there the 1st time. Some flat screens also need time to startup before displaying a picture from input. Hit F8 often, to get there. When too early, BIOS/UFEI comes of, asks to select boot drive or similar, when too late, windows starts up and in your case, it fails. If you do not see anyting from your computer, neither BIOS or UEFI messages or spash screen or logo from the mainboard manufacturer, check connectors. If the computer does not turn on, remove it 2 hours from power grid to reset the power supply's controller by waiting for discharge of the primary capacitors. If this does not solve the problem and the monitor display pictures from other computers, it is a hardware failure. Hardware can not be destroyed by drivers. The only affect to hardware by software is to flash (malicious) software into flash memory devices. Have check, it you are not a victim of the capacitor plague or sililar failures of, power supply, mainboard graphic card whre such passive devices are used. --Hans Haase (talk) 08:01, 28 November 2014 (UTC)

If if was caused by the driver, refer this and see section with this URL. --Hans Haase (talk) 12:08, 28 November 2014 (UTC)

It seems like you are using Windows. Is it Windows 8? That behaves differently for boot F8.[1] Thincat (talk) 11:33, 28 November 2014 (UTC)

Thincat - I use Windows 7. Always heard it was better to stick with 7 and that 8 ain't really an improvement from 7. As for other latest replies above, not sure what you're getting at about showing pictures from other computers, Hans Haase. There's absolutely no picture or anything on the screen during the 1-3 seconds start-up attempts, just black, but I'll try a few times to press F8 frequently as explained during start-up attempt. Failing at that, I'll remove the power for a few hours and see. 2A02:FE0:C711:5C41:3985:7AF4:804:3A47 (talk) 09:17, 29 November 2014 (UTC)

It used to be easier with CRTs. Some TFTs take several seconds to display after switching modes. Mash the F8 button as soon as you get the second beep if your BIOS beeps twice, or as early as possible if it doesn't. Sometimes, it's hard to hit the time window (1 second?) 217.255.188.60 (talk) 08:43, 1 December 2014 (UTC)

Random logout in mobile Gmail

When searching a mail in gmail, you get a list of search results. When you click on your desired search result, you are first redirected to a thread with all mails having the same subject as your mail, only from there you can continue to the specific mail. In software testing (using a virtual machine with Windows 7 & IE 11), this step in between (the thread) is leads sometimes a random logout, causing problems for the tests. Is there a way to avoid this problem? --KnightMove (talk) 08:58, 27 November 2014 (UTC)

It won´t sign out

Hello!

I created my account in a website a few days ago, and it was okay, I could sign in and out with no problem. Until today. I logged in and I was around for a while, then, the sign out button wouldnt appear. But when I go to the main page of the site it does appear. The problem is that inside the main site there is a Community and you have to log out there too and I just can´t. I tried to erase the history of my browser, I turned off the PC, I erased the password´s log in the browser and nothing. Every time I open the site it says Im still logged it. Can someone please help me? This is not my PC and I am worried someone can use my information. I have asked at the site and no one knows about it. Miss Bono ^{[hello, hello!]} 14:52, 27 November 2014 (UTC)

The website will have saved an HTTP cookie on your browser; if you use a cookie manager you can erase the cookie, thus effectively making the site forget about your login. I use Cookies Manager Plus on Firefox. -- Finlay McWalterᚠTalk 15:08, 27 November 2014 (UTC)

I installed it just now. What should I do next? Miss Bono ^{[hello, hello!]} 15:13, 27 November 2014 (UTC)

I did it. Thank you, Finlay McWalter. Can someone mark this post as solved? Thank you! Miss Bono ^{[hello, hello!]} 15:17, 27 November 2014 (UTC)

Now the site doesnt let me log in. This happened after I erased the cookies. Miss Bono ^{[hello, hello!]} 18:05, 27 November 2014 (UTC)

Assuming you remember your sign-on and password it may not let you sign on if you are actively blocking cookies or third-party cookies. (I pretty much tell sites that require thrid-party cookies o go to heck, and don't use them.) Check your settings for permissions. μηδείς (talk) 23:40, 27 November 2014 (UTC)

Where the heck should I look for the permissions´ settings? Miss Bono ^{[hello, hello!]} 13:07, 28 November 2014 (UTC)

Which browser are you using? In Chrome, open the menu, click "Settings", scroll to the bottom and click "Show advanced settings...", then under "Privacy" click the "Content settings..." button, then you'll see all of the cookie settings...you may need to get down into "Manage exceptions..." where you can set specific options for specific web sites that can override your default choices. SteveBaker (talk) 15:36, 28 November 2014 (UTC)

Computer Studies

Can someone view the link please (http://en.wikipedia.org/wiki/Decision_tree#Decision_tree_using_flow_chart_symbols), I would like to know what the boxes are called in computer terms.

(Russell.mo (talk) 15:34, 27 November 2014 (UTC))

Flowchart symbols -- Finlay McWalterᚠTalk 15:40, 27 November 2014 (UTC)

Thanks. -- (Russell.mo (talk) 18:54, 27 November 2014 (UTC))

Resolved

ajax form dubmit

I have one form on my web page with two fields for name and e mail, i want to save entered name and email to mysql database that too without reloading the current page and want to show alert whether data was successfully saved or not, I want coding, anybody please help me 106.76.195.252 (talk) 18:31, 27 November 2014 (UTC)

How to stop Adobe Flash videos

I have Firefox 33.1 and when I visit pages with embedded video content, such as BBC or Sydney Morning Herald the whole machine seems to lock solid while the videos - sometimes in adverts, but not always - load and start playing. How do I configure either Firefox or Adobe Flash Player (15.0.0.239) to just show a still frame - or even nothing at all - at the positions where these videos want to run? Most of the time I want to read the text, but I can't scroll through the page until the videos have all loaded and started playing. I tried setting Flash Player to ask me when sites want to store data locally, but now I get a popup window "Local Settings xxxx is requesting permission to store information on your computer. Requested: Up to 10 KB Currently Used: 5KB" which will not go away and give me back control unless I click "Allow" - the "Deny" button does nothing. --Redrose64 (talk) 20:08, 27 November 2014 (UTC)

Is your system harddisk full?

run %temp% to access the local settings folder. This folder is part of the user profile on your computer. It ist beeing created the first time you log on on this computer. When it is not accessable it may caused by user rights of the files and folders in the file system or an inconsistent file system. A hardware damage of the drive would also be possible. If the flash player is not getting the pointing to it, lets reinstall the player:

Get the flash player from origin source: http://fpdownload.macromedia.com/get/flashplayer/pdc/15.0.0.239/install_flash_player.exe

Note: You will have replace 15.0.0.239 by the newest version of the player in the link above!

Ensure you are not creating flash animations yourself on this machine or have a backup of all your work, before you continue!

Check Firefox for Add-ons you did not install and you don't use.

Close Firefox. In the control panel uninstall the Flash Player and active-X flash player.

run %appdata%

In the folder roaming is a folder named Macromedia. In doubt create a backup, then delete the folder Macromedia

Install the downloaded flash player.

Start Firefox when installation has finished, clear the browsers cache by STRG+SHIFT+DEL, only select cache, all else is optional. Note cookies store your logon to Wikipedia and else!

It should work, if not, check for other versions of flash player or wait for update before reinstalling it. You also may reinstall Firefox. If it does not solve the problem, let me know, we check your firefox settings on issues. --Hans Haase (talk) 21:57, 27 November 2014 (UTC)

• Is suggesting click add-ons under settings and getting Adblock and its sisters not a good idea? At least this seems in most cases to block popups and ads that autoplay. μηδείς (talk) 23:36, 27 November 2014 (UTC)

Yes, go for it! When having trouble or not getting entries deaktivated and removed, we restore your Firefox user profile for remove any hidden unneccessary thing. --Hans Haase (talk) 07:34, 28 November 2014 (UTC)

Until a year or so back I used the Firefox extension Flashblock http://flashblock.mozdev.org so that might help not getting into Flash in the first place. Thincat (talk) 11:16, 28 November 2014 (UTC)

Adbock (with emblem of a stop sign) enables you to select which content is loaded. In this issue I guess blocking would be a workaround. When the ad is hosted on the same domain, the workraround will fail. --Hans Haase (talk) 12:05, 28 November 2014 (UTC)