Gentoo Forums
Gentoo Forums
Gentoo Forums
Quick Search: in
System is very slow since a few month (complex setup, RAID)
View unanswered posts
View posts from last 24 hours

Goto page Previous  1, 2  
Reply to topic    Gentoo Forums Forum Index Kernel & Hardware
View previous topic :: View next topic  
Author Message
doublehp
Guru
Guru


Joined: 11 Apr 2005
Posts: 472
Location: FRANCE

PostPosted: Fri Dec 11, 2015 10:00 pm    Post subject: Reply with quote

Code:
$ ls /boot/*`uname -r`*
/boot/config-2.6.34-xen-r4-Gentoo-uranus-1-50
/boot/System.map-2.6.34-xen-r4-Gentoo-uranus-1-50
/boot/vmlinuz-2.6.34-xen-r4-Gentoo-uranus-1-50
$ uname -r
2.6.34-xen-r4-Gentoo-uranus-1-50


Where 50 is the part I always update manually.

Building a kernel is not just build the kernel. Need to update all misc modules and drivers (including ... some not handeled by Gantoo official scripts). Updating grub.cfg ... considering the fact my kernel changes name on each rebuild. Plus, /boot is not my bootable partition; scripts who try to be smart and assume that /boot is bootable may break things. But, there is no way to have my /boot on a non RAID volume; and there is no way to boot a kernel from a raid volume from BIOS; so a caching system / sync mechanism is required to get benefit of RAID security, and push vmlinuz on a place accessible to BIOS. In the end, I don't need initrd (what is rare for people doing raid).

So ... now I have the good flags in my kernel, what's next ?

Thinderbird did not really bore me today; maybe did not spend enough time working with it.

But there is an issue I can easily reproduce. Open a directory, with Rox-Filer, that contains hundreds of pictures or movies. In 2011, it used to generate previews very fast; way more than 10 pictures per second, or 5-3 movies per sec. Now, it does a bit more than one picture per second, and less than one film per second. Yes, I am also having cache issue with ROX, but, that is my fault; I disabled rox preview cache somewhere, and can't find where I dit that. But, preview generation used to be very fast. And, this is easy to reproduce and track. I just open any large folder; can predict the moment, and repeat action. It's a better tracking process than waiting for TB to freese.

For example, I have been looking iotop and top while opening a picture folder: iotop reported total disk rate of 32K/s, and actual disk read 125K/s, traffic generated by ROX. While I had 75% iddle CPU, where heaviest process was X at 75%. It's a quad core processor, so the 75% iddle is average over the whole processor (value between 99% and 0%), while 75% used by X is for one core (value between 0% and 400%). During iddle time, my iddle is about 90%, so X 75% of core used by X is in fact only 15% of the whole processor. This means, dorung the preview generation, X does not even saturate one core; very far from it. So, the issue is not about thumbnails algo not being multi-tasked, or multihtreaded. CPU is not an issue, algorythm is not limited by core; and disk is hardly an issue (still, I wonder how it can generate previews of pictures that are about 900KB each, one per second, while reading disks are 32 or 125K/s ... )
_________________
DEMAINE Benoît-Pierre (aka DoubleHP ) http://www.demaine.info/
>o_/ Coin coin coin \_o<
to contact me (MSN,ICQ, JABBER, Skype ... ) http://benoit.demaine.info/contact.png
Back to top
View user's profile Send private message
krinn
Watchman
Watchman


Joined: 02 May 2003
Posts: 7197

PostPosted: Sat Dec 12, 2015 2:13 pm    Post subject: Reply with quote

if you want a fs that need no defrag, it must copying always file in sectors free and adjacent to each other, something doable with a lot of free adjacent space, something that will get hard with a few (not free space, but free adjacent sectors space, you can have 10g of free space and 0 adjacent sector free).

and doing such work is of course done by the fs, and your cpu is handling it, so writing any file on such fs in software raid, mean that for each sectors the cpu is balancing on all disks and fs is looking for free adjacent sector space to avoid fragmentation....

you should avoid adding work to work on things that should be done fast, it's good to add some workload to get raid software feature or to get compression (or no defrag feature in your case), but if you start stacking them, you're making it too complex for an operation that should remain quick.

If i were you, i would goes with a simpler fs (ext3, ext4) with no extra feature, and count only on raid for speed up, many fs features are wanking feature, that works with simple scheme, but in complex scheme like you (lot of disks and files) show their limit to their claims.
fs should do fs task (read/write file to disk), it should not goes into doing anything complex that would slow its handling for the fun of claiming "we don't use defrag", "we do mirroring", "we compress datas" or "we use md5 hash to make sure content is 100% valid". As while your fs is doing that it, other queries are waiting for it to end.
It's a balance of feature vs time to do the feature, of course non fragment files mean a disk will take less time to reach sector1 and sector2 for a file instead of sector1 and sector100 to read the same file, but on raid, disk1 answering to sector1 and disk3 answering to sector100 query change the rule and is less of a problem, making such non fragment fs feature more a problem than a good feature to use when you use software raid.
Back to top
View user's profile Send private message
doublehp
Guru
Guru


Joined: 11 Apr 2005
Posts: 472
Location: FRANCE

PostPosted: Tue Dec 15, 2015 5:53 pm    Post subject: Reply with quote

How can I have fragmentation with 50% free space, and talking about 33G free for system, 400G free for home, and 2.5T free for storage disk ?

So, after a few days with iotop, I have found that, at some times, thunderbird or Firefox may be responsible for very large disk usage. It tells about 2.6GB/s for 1 to 3 seconds. During this time, the whole system is slow. And top keeps talking about 75% iddle CPU, and very low sy and wa values, and the top process is usually ... none of those two.

Next step: how to know which files they are reading, writing, or copying ?

This does not happen just after boot. It starts happening after at least 2 to 4h of use.

Since I started this thread, I did not have any very long freese (mouse refusing to move for 30s to 10mn). Never been more than 5s not moving at all, or 10-20s moving slowly (during a very heavy write, when iotop reports between 2GB and 3GB read rate).

Now, I can be more specific about what I want:
1: which files are Thunderbird messing with ? in particular, I want to know if they are system files (/usr), user files (/home) or temp files (/tmp).
2: why is rox-filer so slow to generate miniatures (thumbnails) of pictures and movies ?
3: offtopic: why doesn't rox keep thumbnails over time ? As soon as process is killed, all miniatures are lost, and need to be recomputed. I know I messed with this 2 years ago; maybe I have changed some settings, or changed some folder to point to /dev/null or /dev/shm. I just can't remember what I did. I have recently checked the settings, and cache folders, and they all seem fine to me.
_________________
DEMAINE Benoît-Pierre (aka DoubleHP ) http://www.demaine.info/
>o_/ Coin coin coin \_o<
to contact me (MSN,ICQ, JABBER, Skype ... ) http://benoit.demaine.info/contact.png
Back to top
View user's profile Send private message
krinn
Watchman
Watchman


Joined: 02 May 2003
Posts: 7197

PostPosted: Tue Dec 15, 2015 6:58 pm    Post subject: Reply with quote

doublehp wrote:
How can I have fragmentation with 50% free space

record file 1, 2 & 3 of 10 sectors.
remove file2
add file4 of 12 sectors
file4 get record at free sectors 11-20 and the 2 that remains get records after file3, so at sector 41,42
file4 is now fragment ; and nobody cares how much space you have left.
Back to top
View user's profile Send private message
doublehp
Guru
Guru


Joined: 11 Apr 2005
Posts: 472
Location: FRANCE

PostPosted: Tue Dec 15, 2015 8:04 pm    Post subject: Reply with quote

I am not using FAT12. ext4 and ZFS are much more clever than this.
_________________
DEMAINE Benoît-Pierre (aka DoubleHP ) http://www.demaine.info/
>o_/ Coin coin coin \_o<
to contact me (MSN,ICQ, JABBER, Skype ... ) http://benoit.demaine.info/contact.png
Back to top
View user's profile Send private message
doublehp
Guru
Guru


Joined: 11 Apr 2005
Posts: 472
Location: FRANCE

PostPosted: Wed Dec 16, 2015 7:56 pm    Post subject: Reply with quote

I have been watching my system very carefully those last hours. It's about 4h my machine is deadly slow; every 5 or 10mn, it gets slow. Mouse moves very slowly (about 70% of movement information is lost; I have to move my hand 4x more than expected). When mouse is slow, or frosen, disk blinks slowly, iotop reports no traffic (below 300kB/s), and top usually reports X to use 25% of the system (if it saturated one core, it would use 100%; if it saturated the whole CPU, it would use 400%), iddle time is about 70%, wa below 3%.

And this is happening a dozen times per hour, since 4h. Doing desktop work (Firefox, Chrome, OOo).

3 days ago, I could identify TB to be responsible for a specific issue: at some times, the whole box is completely frozen for 3 to 5s. Not slow: frozen. And then, HDD led is fully ON, and iotop reports traffic to be between 2 and 3GB/s. => I know what my machine looks like when some process want to have fun with my storage. So, this is not what have been boring me today.

And, after finishing my work, I still have 9GB RAM free. So, I am very far from running out of RAM. And my used SWAP is 0k out of 21GB.

I need more ideas to dig.
_________________
DEMAINE Benoît-Pierre (aka DoubleHP ) http://www.demaine.info/
>o_/ Coin coin coin \_o<
to contact me (MSN,ICQ, JABBER, Skype ... ) http://benoit.demaine.info/contact.png
Back to top
View user's profile Send private message
schorsch_76
Guru
Guru


Joined: 19 Jun 2012
Posts: 450

PostPosted: Thu Dec 17, 2015 6:49 am    Post subject: Reply with quote

Hm, that sounds strange.

Just some ideas:
- cpu cooler. If the cpu gets too hot, it clocks down. Very common problem on old hardware (dust on the fans) and in the case.
- check PSU: Try to exchange it against a new one
- memtest
- SATA cables could be bad. Bad contact (oxyd on connectors over time)

As you said, it shows up after a few hours of working, my best bet is dust on the coolers. The CPU clocks down to prevent damage.

https://www.google.de/search?q=dust+on+the+cooler&client=ubuntu&hs=iFp&channel=fs&source=lnms&tbm=isch&sa=X&ved=0ahUKEwi4uM-5p-LJAhVHtBoKHVItAIcQ_AUICSgD&biw=1680&bih=921
_________________
// valid again: I forgot about the git access. Now 1.2GB big. Start: 2015-06-25
git daily portage tree
Web: https://portage.schorsch-tech.de
git clone https://portage.schorsch-tech.de/portage.git
Back to top
View user's profile Send private message
C5ace
Apprentice
Apprentice


Joined: 23 Dec 2013
Posts: 296
Location: Brisbane, Australia

PostPosted: Thu Dec 17, 2015 10:01 am    Post subject: Reply with quote

schorsch_76 wrote:
Hm, that sounds strange.

Just some ideas:
- cpu cooler. If the cpu gets too hot, it clocks down. Very common problem on old hardware (dust on the fans) and in the case.
- check PSU: Try to exchange it against a new one
- memtest
- SATA cables could be bad. Bad contact (oxyd on connectors over time)

As you said, it shows up after a few hours of working, my best bet is dust on the coolers. The CPU clocks down to prevent damage.

https://www.google.de/search?q=dust+on+the+cooler&client=ubuntu&hs=iFp&channel=fs&source=lnms&tbm=isch&sa=X&ved=0ahUKEwi4uM-5p-LJAhVHtBoKHVItAIcQ_AUICSgD&biw=1680&bih=921


Add to this:
CPU thermal paste, miss matched RAM, corroded motherboard ground posts. dried electrolytic capacitors, undersized powersupply.

I has a similar problem in 2009 with a new Gigabyte MB. Can't recall the model number. Fixed by replacing the MB with a later model GA-MA790FXT-UD5P using 4x2GB matched Kingston RAM sticks and 5x 500GB SAMSUNG HD502IJ drives configures as RAID-1 for /boot and RAID-5 for /, /home and /VirtualBox on EXT4. The box need vacuming once a year and is still running like new.

I personally would stop wasting my time and nerves with something that is very complex and probably not fixable and instead get a new box with matched memory sticks and a good graphics card with HDMI output. Then get the largest LCD TV you can afford and use this as a monitor. Then sell your old stuff on e-bay or to a pawn shop.
Back to top
View user's profile Send private message
doublehp
Guru
Guru


Joined: 11 Apr 2005
Posts: 472
Location: FRANCE

PostPosted: Thu Dec 17, 2015 5:44 pm    Post subject: Reply with quote

At last, good ideas.

I disagree, but I will need to proove you are wrong. So, which tests could proove you wrong ?
- machine is 5 years old; but I did a complete checkup when I changed my disks in mid 2014: changed several fans, removed all dust everywhere. Machine was not very dusty; I don't think it could accumulate more dust in 1 year, than during the first 4.
- CPU cooler is oversized. My processor is a black edition: all people who bought thismodel bought it to overclock it by 5%, without touching voltages. After adding a second fan to my heat sink (Noctua NH-U12P) and adjusting voltages, people usually do +10%
- my Noctua is oversized; and my fan profile is to keep fan speed as low as possible to reduce aging of ball bearings, total air flow, and dust accumulation (and noise). I would have a hotter CPU.
- I draw my CPU temp since years; and honestly ... 42°C average over the year, with an absolute max at 58°C during the summer, hardly call that hot. 49° max over the last week, 54 the last month. Si, I am pretty sure my paste is good.
- cpu-cooler thermal paste: I would have a hotter CPU
- CPU clocking down for any misc reason: don't know how to check this. Could be an idea.
- dead PSU ... that is probably the least stupid idea. It's a very long time I know my +5V is very low (4.85V), and +12 not really better: 11.4V.
- memtest was good recetly. My RAM is not matched; I mean, I have 2x2G + 2x8G. But sticks are inserting correctly, to respect the topology required for my AMD64. When sticks are not inserted correctly, my BIOS complains about it. They are different models (2G and 8G are not the same model at all) but BIOS is usually responsible to downclock the whole system to the best possible configuration. Any way, I have slowliness issues since mid 2014; and bought 2x8G only 2 months ago; so, it may be a contribution, but can not be the root problem.
- bad Sata cables; worth speaking about it; I had problems due to sata cables, HDD firmwares and so on; but the problem does not sound like this to me: I have very good read/write rates on raw disks, and on raid blocks; perf come down when writing files on filesystem. I could kill my swap if you want, and show you that at disk, and MDADM levels, it's all good (tell me if you prefer JBOD, raid0, raid5 ... I have 4G swap partition on each disk; tests should not take more than 2mn if disks are good). Plus, bad sata cables usually produce messages in syslog or dmesg or SMART.
- RAM/HDD/cables do not seem to me a good explanation: when X is slow, I have 2 led flashes per sec and about 500k/s; when an app freeses completely, iotop reports 3GB/s for 2s. But, the first case could match a CPU frequency problem ... due to ... undervoltage. Undervoltage could be dead PSU, or dead VCU (voltage control unit inside the MB). This is a point which I can not counter-argue.
- sata cables: you say they could have oxydation. Which part could get oxyded ? why oxydation should have to be on cables, cable plugs, and not MB or HDD plug ? is there a way to clean that ? I am not going to change my MB for just a layer of oxyde on sata plugs.
- what I had once ... on a dual P3 800MHz machine was that one CPU had a bad contact on L2 cache line; some times CPU started with 512k, sometimes with 128k. Booting with 128k was always fine. But if Linux started with 512k, and, later the cache got away, loosing 372k of cache on the run was pretty bad. At the beginning, I had kernel panics; because the faulty CPU was CPU #0; because kernel had severe memory corruptions; after inverting CPUs, I only had app crashes; kernel could continue to manage the system from CPU #0, but most apps having tasks on CPU #1 had core dumps. After the partial crash, once the cache had gone away, I just had to kill all old process running on #1, and restart them. Once cpu had lost the bad cache line, it could not go worst; and app started after the loss would work 100% untill bed time. By that Linux, desktop Linux kernel did not yet have the capability to migrate process over cores. It was an electrically bad connection inside the die.
- dried capacitor: inside MB, I hope not; this model claims to have special capacitor that don't dry. PSU ... can hardly be so proud.
- undersized PSU ... I don't think so. I will dig.

Things are way better since I have made a huge cleanup in Thunderbird last week. No more "10mn freese".

What I am going to do:
- record CPU freq
- print CPU temps on screen permanently
- memtest, costs nothing
- cpuburn
- deeper write tests of disks, after killing my swap space; but 4GB per disk is not really that much; I will look for larger free partitions. Depending on results, maybe buy new cables.
- will measure input power with an Amp meter, at 220V side; will give me the total input power. Then, with clamp meter, measure 3V, 5V, and 12V currents, to see if any of them is exceding the PSU specs. I hope I can clamp wires by voltage; will be impossible if tubing of cables come too close to PSU. Ex this: http://www.overclock3d.net/gfx/articles/2009/06/05070513238l.jpg i can't clamp it. Or I will have to cut all tubings (dirty, not nice, but should not affect performances).
- I will setup a permanent external monitoring of all voltages (easy and cheap for me). I don't trust values measured by MB.
- not sure how to dig CPU faults on long term. The P3 was easy to study, and reading /proc/cpuinfo gave different results with time. My actual Athlon may not be as cooperative about details as P3 was.
- change PSU But if +5 and +12V are low, maybe 3.3 and 1.8 are also too short.

I don't need a more powerfull machine. This one is already much more powerfull than I need. And during freeses, disks don't work, and CPU is 75% iddle. Changing the hardware would be useless if issue is in software. And I don't have time to reinstall and configure the whole software now.
_________________
DEMAINE Benoît-Pierre (aka DoubleHP ) http://www.demaine.info/
>o_/ Coin coin coin \_o<
to contact me (MSN,ICQ, JABBER, Skype ... ) http://benoit.demaine.info/contact.png
Back to top
View user's profile Send private message
C5ace
Apprentice
Apprentice


Joined: 23 Dec 2013
Posts: 296
Location: Brisbane, Australia

PostPosted: Fri Dec 18, 2015 11:13 am    Post subject: Reply with quote

doublehp wrote:
At last, good ideas.

I disagree, but I will need to proove you are wrong. So, which tests could proove you wrong ?
- machine is 5 years old; but I did a complete checkup when I changed my disks in mid 2014: changed several fans, removed all dust everywhere. Machine was not very dusty; I don't think it could accumulate more dust in 1 year, than during the first 4.
- CPU cooler is oversized. My processor is a black edition: all people who bought thismodel bought it to overclock it by 5%, without touching voltages. After adding a second fan to my heat sink (Noctua NH-U12P) and adjusting voltages, people usually do +10%
- my Noctua is oversized; and my fan profile is to keep fan speed as low as possible to reduce aging of ball bearings, total air flow, and dust accumulation (and noise). I would have a hotter CPU.
- I draw my CPU temp since years; and honestly ... 42°C average over the year, with an absolute max at 58°C during the summer, hardly call that hot. 49° max over the last week, 54 the last month. Si, I am pretty sure my paste is good.
- cpu-cooler thermal paste: I would have a hotter CPU
- CPU clocking down for any misc reason: don't know how to check this. Could be an idea.
- dead PSU ... that is probably the least stupid idea. It's a very long time I know my +5V is very low (4.85V), and +12 not really better: 11.4V.
- memtest was good recetly. My RAM is not matched; I mean, I have 2x2G + 2x8G. But sticks are inserting correctly, to respect the topology required for my AMD64. When sticks are not inserted correctly, my BIOS complains about it. They are different models (2G and 8G are not the same model at all) but BIOS is usually responsible to downclock the whole system to the best possible configuration. Any way, I have slowliness issues since mid 2014; and bought 2x8G only 2 months ago; so, it may be a contribution, but can not be the root problem.
- bad Sata cables; worth speaking about it; I had problems due to sata cables, HDD firmwares and so on; but the problem does not sound like this to me: I have very good read/write rates on raw disks, and on raid blocks; perf come down when writing files on filesystem. I could kill my swap if you want, and show you that at disk, and MDADM levels, it's all good (tell me if you prefer JBOD, raid0, raid5 ... I have 4G swap partition on each disk; tests should not take more than 2mn if disks are good). Plus, bad sata cables usually produce messages in syslog or dmesg or SMART.
- RAM/HDD/cables do not seem to me a good explanation: when X is slow, I have 2 led flashes per sec and about 500k/s; when an app freeses completely, iotop reports 3GB/s for 2s. But, the first case could match a CPU frequency problem ... due to ... undervoltage. Undervoltage could be dead PSU, or dead VCU (voltage control unit inside the MB). This is a point which I can not counter-argue.
- sata cables: you say they could have oxydation. Which part could get oxyded ? why oxydation should have to be on cables, cable plugs, and not MB or HDD plug ? is there a way to clean that ? I am not going to change my MB for just a layer of oxyde on sata plugs.
- what I had once ... on a dual P3 800MHz machine was that one CPU had a bad contact on L2 cache line; some times CPU started with 512k, sometimes with 128k. Booting with 128k was always fine. But if Linux started with 512k, and, later the cache got away, loosing 372k of cache on the run was pretty bad. At the beginning, I had kernel panics; because the faulty CPU was CPU #0; because kernel had severe memory corruptions; after inverting CPUs, I only had app crashes; kernel could continue to manage the system from CPU #0, but most apps having tasks on CPU #1 had core dumps. After the partial crash, once the cache had gone away, I just had to kill all old process running on #1, and restart them. Once cpu had lost the bad cache line, it could not go worst; and app started after the loss would work 100% untill bed time. By that Linux, desktop Linux kernel did not yet have the capability to migrate process over cores. It was an electrically bad connection inside the die.
- dried capacitor: inside MB, I hope not; this model claims to have special capacitor that don't dry. PSU ... can hardly be so proud.
- undersized PSU ... I don't think so. I will dig.

Things are way better since I have made a huge cleanup in Thunderbird last week. No more "10mn freese".

What I am going to do:
- record CPU freq
- print CPU temps on screen permanently
- memtest, costs nothing
- cpuburn
- deeper write tests of disks, after killing my swap space; but 4GB per disk is not really that much; I will look for larger free partitions. Depending on results, maybe buy new cables.
- will measure input power with an Amp meter, at 220V side; will give me the total input power. Then, with clamp meter, measure 3V, 5V, and 12V currents, to see if any of them is exceding the PSU specs. I hope I can clamp wires by voltage; will be impossible if tubing of cables come too close to PSU. Ex this: http://www.overclock3d.net/gfx/articles/2009/06/05070513238l.jpg i can't clamp it. Or I will have to cut all tubings (dirty, not nice, but should not affect performances).
- I will setup a permanent external monitoring of all voltages (easy and cheap for me). I don't trust values measured by MB.
- not sure how to dig CPU faults on long term. The P3 was easy to study, and reading /proc/cpuinfo gave different results with time. My actual Athlon may not be as cooperative about details as P3 was.
- change PSU But if +5 and +12V are low, maybe 3.3 and 1.8 are also too short.

I don't need a more powerfull machine. This one is already much more powerfull than I need. And during freeses, disks don't work, and CPU is 75% iddle. Changing the hardware would be useless if issue is in software. And I don't have time to reinstall and configure the whole software now.


Use a precision digital voltmeter and adjust the voltages to the correct values and verify with a precision oscilloscope that the noise levels on all motherboard supply lines are less then 0.1 microvolts. Especially with the memory supply voltages. Or borrow a top quality powersupply.
Back to top
View user's profile Send private message
doublehp
Guru
Guru


Joined: 11 Apr 2005
Posts: 472
Location: FRANCE

PostPosted: Sat Dec 19, 2015 4:33 pm    Post subject: Reply with quote

C5ace wrote:
Use a precision digital voltmeter and adjust the voltages to the correct values and verify with a precision oscilloscope that the noise levels on all motherboard supply lines are less then 0.1 microvolts. Especially with the memory supply voltages. Or borrow a top quality powersupply.


Please, only quote the part you reply to.

I have 3-1/2 and 4-1/2 multimeters, and an old scope; and ... I have already notived that many projects get poluted by the PSU of my computer: have never been possible to have a flat line on my scope of my computer is plugged (even ATX power off: noise goes away when I remove the plug). It's an Enertec Schlumberger 5224, 4 ways 100MHz. But, after reading your message 4 times, I thing you exagerate a bit: all devices should accept a 0.1 microvolt ripple. I think you meant 0.1 mV (mili), or 100uV (micro). And even 100µV should be accepted by all devices. Plus ... 0.1µV ripple is not the same on 3.3 and 5.0V. My wonder is: what are the official acceptable extreme values for each voltage line ? An ATX PSU is performing DCDC conversion, which by definition generates squares waves at some point, and thus high frequency ripple. I don't think any PSU can have a ripple as low as 0.1µV (and even 0.1mV would seem amazing to me).

The only value I have now is: USB voltage 4.8V (same in front and back plugs, which are not on the same root). No phone can charge with so few.

But I add your point to my list: check ripple.
_________________
DEMAINE Benoît-Pierre (aka DoubleHP ) http://www.demaine.info/
>o_/ Coin coin coin \_o<
to contact me (MSN,ICQ, JABBER, Skype ... ) http://benoit.demaine.info/contact.png
Back to top
View user's profile Send private message
C5ace
Apprentice
Apprentice


Joined: 23 Dec 2013
Posts: 296
Location: Brisbane, Australia

PostPosted: Sat Dec 19, 2015 4:55 pm    Post subject: Reply with quote

doublehp wrote:


The only value I have now is: USB voltage 4.8V (same in front and back plugs, which are not on the same root). No phone can charge with so few.

But I add your point to my list: check ripple.


Sorry, my error. I wanted to say 0.1 millivolt.
Some time ago I had a intermittend HD and tested the voltaged on the motherboard. They where OK with a noise level of less than 0.1 millivolt. The fault was caused by a bad SATA cable.

The output of the DC-DC converter in the power supply is usually filtered by high capacity electolytic capacitors and the powerlines on the mother board by tantalium and ceramic capacitors they are for suppressing low and high frequency noise genrated by the chips on the motherboard. Mesure the voltages and noise direct on the memeory chips sockets, CPU, Host Controller, SATA and HD power sockets when your system is hot and runs slow.
Back to top
View user's profile Send private message
doublehp
Guru
Guru


Joined: 11 Apr 2005
Posts: 472
Location: FRANCE

PostPosted: Sat Dec 19, 2015 6:59 pm    Post subject: Reply with quote

C5ace wrote:
The output of the DC-DC converter in the power supply is usually filtered by high capacity electolytic capacitors and the powerlines on the mother board by tantalium and ceramic capacitors they are for suppressing low and high frequency noise genrated by the chips on the motherboard. Mesure the voltages and noise direct on the memeory chips sockets, CPU, Host Controller, SATA and HD power sockets when your system is hot and runs slow.


Will be complicated to not poluate the signal with ambiant noise (CRT, LCD, power lines ...).
_________________
DEMAINE Benoît-Pierre (aka DoubleHP ) http://www.demaine.info/
>o_/ Coin coin coin \_o<
to contact me (MSN,ICQ, JABBER, Skype ... ) http://benoit.demaine.info/contact.png
Back to top
View user's profile Send private message
C5ace
Apprentice
Apprentice


Joined: 23 Dec 2013
Posts: 296
Location: Brisbane, Australia

PostPosted: Sat Dec 19, 2015 10:30 pm    Post subject: Reply with quote

doublehp wrote:

Will be complicated to not poluate the signal with ambiant noise (CRT, LCD, power lines ...).


The other option I can think off is to run stress testing software like described here:
https://wiki.archlinux.org/index.php/Stress_Test
Back to top
View user's profile Send private message
Display posts from previous:   
Reply to topic    Gentoo Forums Forum Index Kernel & Hardware All times are GMT
Goto page Previous  1, 2
Page 2 of 2

 
Jump to:  
You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot vote in polls in this forum