Wikipedia:Reference desk/Archives/Computing/2012 February 4

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.

February 4

Making a hard disk non-bootable

Hard disk upgrade time! I love this time. I've left my old Vista hard disk in my Windows PC and set up Windows 7 on the new hard disk, like I always do. I set up the new hard disk in the BIOS to have boot priority. Oddly, today when I walked up and un-slept the computer, it had apparently restarted and booted off the old Vista drive despite the boot priority. I was thinking that maybe the new drive had taken a while to spin up at boot time for some reason, so the BIOS just skipped over to the Vista drive and booted.

I would like to make the old hard disk nonbootable (but I want to keep the partition and all its files intact as I gradually move needed stuff over to the new hard disk). Googling, surprisingly, hasn't really helped with this task that I would have thought was simple — all my Google results helpfully tell me how to repair the MBR or fix problems where the computer is supposed to boot but does not. Thanks for any help. Comet Tuttle (talk) 02:16, 4 February 2012 (UTC)

Can't you set it in the BIOS to not be bootable ? The other alternative, I suppose, is to delete some bits from the hard disk used in the boot process, but that might produce an error at bootup. StuRat (talk) 02:49, 4 February 2012 (UTC)

No, just boot order. I don't like the idea of just corrupting the boot sector; I'd like it wiped, if anything (or otherwise marked as "don't boot this disk"). Comet Tuttle (talk) 18:34, 4 February 2012 (UTC)

There's a "bootable?" entry in each partition record in the master boot record, and various tools can be used to manipulate it. In linux I used gparted's "manage flags" option (which I think calls fdisk to do the work), and I've done it the hard way using dd as a disk editor, to change a single byte at raw address 0x1be from 0x80 to 0x00. I don't know if Windows' disk-management plugin will let you change the "bootable" flag of a partition to off, or if it will do so for the partition you've just booted. Failing that, you can book to a linux liveCD to make the change. -- Finlay McWalterჷTalk 18:47, 4 February 2012 (UTC)

I am adventurous enough to try this and will try to go for it when I have time. Thanks! Comet Tuttle (talk) 02:26, 7 February 2012 (UTC)

This might work; type msconfig at Start, go to the "boot" tab and see if Vista is listed there. If it is, select it and click delete AvrillirvA (talk) 18:57, 4 February 2012 (UTC)

The Vista hard disk isn't listed there, unfortunately. Comet Tuttle (talk) 02:26, 7 February 2012 (UTC)

CPU heating concern

Hello! It's my first time building a computer. I bought an Intel i5 Sandy Bridge processor and am running Ubuntu. I've been compressing a 150+ GB file with pbzip2 which is able to utilize all 4 cores. When I check the CPU temp sensor while compressing, I get concerned because all four cores are at or just below 98.0 °C, which is the listed critical value according to the Ubuntu $ sensors command. The processor seems to be throttling to not exceed that critical threshold, but could I have made a mistake in the build? Is it normal for the processor to get this hot when it is being fully utilized? I double-checked the heatsink fan (came with the processor) and it appears securely fastened above the processor. Oddly, if I put my finger close to the heatsink fan, the air is cool even when the CPU is reported to be quite hot (I was expecting it to feel more like a laptop fan vent, which seem to be habitually quite warm). Even after I kill the compression task, the CPU still reports values 93 - 94 °C 5 to 10 minutes afterward. Is this bad for the processor to be operating at such high temperatures? Thank you for any advice of insight into this problem.--el Aprel (^facta-_facienda) 02:41, 4 February 2012 (UTC)

Sounds to me like the sensor is bad and should be replaced. StuRat (talk) 02:46, 4 February 2012 (UTC)

Just to update, over 40 minutes after cancelling the compression, and having done nothing since then but browse the web, the CPU is still report >90°C temperatures. When I first turned on the PC and immediately called $ sensors in Ubuntu, the temperatures were only 40-50°C. Not sure what temperatures to expect, but I guess those would be believable (but still perhaps a bit high) for startup. For reference, calling # sensors-detect in Ubuntu recognizes what it calls the "Intel digital thermal sensor" and uses the "driver 'coretemp'" to access it.--el Aprel (^facta-_facienda) 03:20, 4 February 2012 (UTC)

I know only a little about this, but those temperatures sound accurate to me, and it is getting too hot. A temp 45-50C is about typical for a motherboard. The Intel i3/5/7 are designed to run hotter than that, up to about 75C. But >90C is way too hot. Windows monitors the temp and scales back the Turbo Boost if it needs to. I don't know about the OS you are using. Bubba73 ^{You talkin' to me?} 03:26, 4 February 2012 (UTC)

Unplug the computer and let it sit overnight. Then plug it in and boot it and read the temp immediately. Presumably the temp should still be around room temp then, which is typically around 22°C. If it's reading much more than that, you know the sensor is bad. (Although, if it reads right then, that doesn't necessarily mean the sensor is good, as it might only go bad at higher temps.) StuRat (talk) 03:33, 4 February 2012 (UTC)

I suspect that by the time a PC has booted, the CPU is likely to be well above room temperature - I can get a temperature change of around 10 degrees C within a few seconds on my overclocked i7, just by loading an image into Paint.net (measured via Core Temp 1.0). At the moment in a distinctly chilly room, it is running at about 40 C idling, and CPU-intensive activities (i.e. games) will easily take it into the mid 60s, even with a heatsink the size of a biscuit-tin (the GPU will hit 80 C or so if pushed hard, but they are intended to run hotter). One possibility is that the heatsink hasn't been installed properly - did you use the conductive grease you are supposed to? Of course, you can really only check this by taking it off - at which point you'd need to reapply the grease anyway... AndyTheGrump (talk) 03:49, 4 February 2012 (UTC)

The CPU thermal sensor is an on-die thermal diode, connected through the CPU socket to logic in the northbridge (e.g. on Socket 764 and Socket 940 it's pins THERMDA/THERMDC). Intel Xeon 5500 processors feature multiple on-die sensors per core (DTS) the values of which can be read in software over the PECI, which the northbridge uses for fan control. As these sensors are built right into the cpu die, replacing them is impossible. Neither is it likely that the relevant connections to these sensors have failed, as this would show as no reading, not as a wrong reading. A common error new builders make is misapplying the thermal interface between the cpu and the cooler - too much paste or too little, or an air bubble in the paste pad. 91.125.186.46 (talk) 04:38, 4 February 2012 (UTC)

The processor's heatsink came with a thermal pad on the bottom, but I attached and removed the heatsink a couple times without thinking much of it. I didn't know that introduces air bubbles and lessens the conductivity of heat to the sink (well, at least that is what I have read about thermal compound, but perhaps the same goes for the cheap pre-applied thermal pad). I'll go out tomorrow and buy some thermal paste and apply it correctly; I'll post back afterwards.--el Aprel (^facta-_facienda) 05:54, 4 February 2012 (UTC)

Yes, you really only get one go. You need to thoroughly remove the existing thermal grease (see this video) from both the cpu and heatsink, and make sure they;re clean and dry before you apply the new layer. Make sure not to add too much either, as a thick layer will also be a problem. This process isn't really very hard, but if its your first time, or if you only build a PC every few years, its a skill you don't get much practice at. 91.125.186.46 (talk) 12:51, 4 February 2012 (UTC)

I agree with 91.125... 98 C is too hot, and yes running it at subcritical, but elevated temperatures over long periods of time can appreciably shorten the life of the processor (overclockers run into this problem occasionally). I also suspect, assuming the sensors are correct, that the problem has to do with the connection between the heat sink and your processor.

All of this goes without saying that it's quite possible to brick your processor, either by not having it cool sufficiently enough, or by getting the paste on a conductive component and shorting it out. There's a certain "expert" element to this kinda task. And as for keeping it off and starting it back up... cpus and their heat sync assemblies, even installed wrong, are remarkably conductive. I've seen videos of cpus go up in smoke within 3 seconds of removing the heat sink (most modern chips should throttle themselves or shut down, which may be why yours hasn't destroyed itself yet... I don't know the details on the i5s). By the time the CPU booted any temp differences would be indistinguishable. Shadowjams (talk) 16:33, 4 February 2012 (UTC)

Update: I went to remove the heatsink to clean and apply thermal paste, and found that one of the latches wasn't touching all the way down to the board, making partial incomplete contact with the CPU (embarrassing). I took it all the way off, cleaned the heatsink and CPU with rubbing alcohol, and applied the thermal paste according to directions. Big difference: at idle, the CPU reading are 35-45°C; at full load (100% on all 4 CPUs) the readings hover between 77-81°C, occasionally reaching 82°C, but no hotter, and I've been running at full load for about 15 minutes now (factory settings, no overclocking). When I kill the task, the temps drop almost instantly by 10°C, and cool off more as more time passes. This sounds much better to me than the >90°C temps with improper heatsink installation, but I'd still like confirmation that 80°C at max load is still okay. The coretemps utility calls 80°C "high" (understandable for full load), but nowhere near 98°C, which it labels "critical." Is it still healthy for the processor to run highly-computational tasks for extended periods of time (a few hours) as long as the processor doesn't get hotter than around 80°C? Thank you for all your helpful information to have come this far.--el Aprel (^facta-_facienda) 01:45, 5 February 2012 (UTC)

I believe that you have solved the problem. Bubba73 ^{You talkin' to me?} 01:47, 5 February 2012 (UTC)

Why does images change?

why this: "||||||||||||||||" sometimes looks like green???? (And there are some images that can change perception!) (for example this (i found it on a forum) change the wole image when you change the angle of you screen? How's this? --190.158.184.192 (talk) 05:37, 4 February 2012 (UTC)

http://www.lagom.nl/lcd-test/subpixel.php ¦ Reisio (talk) 11:48, 4 February 2012 (UTC)

power of a Kobo (or other e-readers)

I am curious as to how the "power" of modern e-readers (i.e. memory, speed, etc.) compare to older computers such as the old Commodore 64 (which got me through university!). While I suspect that there is no way to make a fair comparision (it's probably like comparing apples to oranges), it would be interesting to see how they do compare.

72.136.132.216 (talk) 14:28, 4 February 2012 (UTC)

Well, there are many different e-readers. According to our article Kobo eReader this thing has 1GB of memory (or around 1,000 books) plus up to 32GB on a SD card. According to Commodore 64 it had 64 kB RAM + 20 kB ROM... Von Restorff (talk) 14:44, 4 February 2012 (UTC)

A C=64 had a single 8-bit MOS Technology 6510 microprocessor running at around 1 Mhz; a Kobo Vox has a single ARM Cortex-A8 (a 32 bit ARMv7 architecture) running at 800Mhz. For a lot of real applications, the richer instruction set and wider registers will mean that the modern 32 bit architecture will take a lot fewer instructions to get stuff done. So you're right, there's no way to make any kind of fair comparison, but from a "getting stuff done" perspective, we can maybe handwave at 2000 times faster. -- Finlay McWalterჷTalk 14:47, 4 February 2012 (UTC)