[gentoo-user] ECC-ram, it is worth it.

[gentoo-user] ECC-ram, it is worth it.

[Volker Armin Hemmann]

[894019.770058] [Hardware Error]: MC4 Error (node 0): DRAM ECC error
detected on the NB.
[894019.770084] EDAC MC0: 1 CE on mc#0csrow#2channel#0 (csrow:2
channel:0 page:0x2aa6ce offset:0xc60 grain:0 syndrome:0x63e1)
[894019.770090] [Hardware Error]: Error Status: Corrected error, no
action required.
[894019.770098] [Hardware Error]: CPU:0 (10:4:2)
MC4_STATUS[-|CE|MiscV|-|AddrV|CECC]: 0x9c70c00063080a13
[894019.770105] [Hardware Error]: MC4_ADDR: 0x00000002aa6cec60
[894019.770110] [Hardware Error]: cache level: L3/GEN, mem/io: MEM,
mem-tx: RD, part-proc: RES (no timeout)

and this, my children, is why I am using ECC ram.

Using zfs showed me, that there are errors that the system does not
catch but corrupts data.

And this evening, with a thunderstorm outside I got that beauty above...

[gentoo-user] Re: ECC-ram, it is worth it.

[Nikos Chantziaras]

On 26/07/14 20:39, Volker Armin Hemmann wrote:
>[...]
> and this, my children, is why I am using ECC ram.

I don't really care if the porn I'm watching has one frame with 
corrupted pixels on it.

Re: [gentoo-user] Re: ECC-ram, it is worth it.

[Volker Armin Hemmann]

Am 26.07.2014 19:58, schrieb Nikos Chantziaras:
> On 26/07/14 20:39, Volker Armin Hemmann wrote:
>> [...]
>> and this, my children, is why I am using ECC ram.
>
> I don't really care if the porn I'm watching has one frame with
> corrupted pixels on it.
>
>
>

but you will care when your kernel writes the next file right over the
partition boundary.

Re: [gentoo-user] Re: ECC-ram, it is worth it.

[Mick]

[-- Attachment #1: Type: Text/Plain, Size: 579 bytes --]

On Saturday 26 Jul 2014 19:23:20 Volker Armin Hemmann wrote:
> Am 26.07.2014 19:58, schrieb Nikos Chantziaras:
> > On 26/07/14 20:39, Volker Armin Hemmann wrote:
> >> [...]
> >> and this, my children, is why I am using ECC ram.
> > 
> > I don't really care if the porn I'm watching has one frame with
> > corrupted pixels on it.
> 
> but you will care when your kernel writes the next file right over the
> partition boundary.

Ooh!  Scary!  O_O

Isn't there some kind of kernel/fs check mechanism that ought to check this 
doesn't happen?

-- 
Regards,
Mick

[-- Attachment #2: This is a digitally signed message part. --]
[-- Type: application/pgp-signature, Size: 473 bytes --]

Re: [gentoo-user] Re: ECC-ram, it is worth it.

[J. Roeleveld]

On 26 July 2014 20:27:14 CEST, Mick <michaelkintzios@gmail.com> wrote:
>On Saturday 26 Jul 2014 19:23:20 Volker Armin Hemmann wrote:
>> Am 26.07.2014 19:58, schrieb Nikos Chantziaras:
>> > On 26/07/14 20:39, Volker Armin Hemmann wrote:
>> >> [...]
>> >> and this, my children, is why I am using ECC ram.
>> > 
>> > I don't really care if the porn I'm watching has one frame with
>> > corrupted pixels on it.
>> 
>> but you will care when your kernel writes the next file right over
>the
>> partition boundary.
>
>Ooh!  Scary!  O_O
>
>Isn't there some kind of kernel/fs check mechanism that ought to check
>this 
>doesn't happen?

There is. But all that happens in memory...

--
Joost
-- 
Sent from my Android device with K-9 Mail. Please excuse my brevity.

[gentoo-user] Re: ECC-ram, it is worth it.

[walt]

On 07/26/2014 10:39 AM, Volker Armin Hemmann wrote:
> [894019.770058] [Hardware Error]: MC4 Error (node 0): DRAM ECC error
> detected on the NB.
> [894019.770084] EDAC MC0: 1 CE on mc#0csrow#2channel#0 (csrow:2
> channel:0 page:0x2aa6ce offset:0xc60 grain:0 syndrome:0x63e1)
> [894019.770090] [Hardware Error]: Error Status: Corrected error, no
> action required.
> [894019.770098] [Hardware Error]: CPU:0 (10:4:2)
> MC4_STATUS[-|CE|MiscV|-|AddrV|CECC]: 0x9c70c00063080a13
> [894019.770105] [Hardware Error]: MC4_ADDR: 0x00000002aa6cec60
> [894019.770110] [Hardware Error]: cache level: L3/GEN, mem/io: MEM,
> mem-tx: RD, part-proc: RES (no timeout)
> 
> and this, my children, is why I am using ECC ram.
> 
> Using zfs showed me, that there are errors that the system does not
> catch but corrupts data.
> 
> And this evening, with a thunderstorm outside I got that beauty above...

Is ECC memory a drop-in replacement for ordinary RAM, or does it need
a special motherboard?

Re: [gentoo-user] Re: ECC-ram, it is worth it.

[Volker Armin Hemmann]

Am 27.07.2014 00:55, schrieb walt:
> On 07/26/2014 10:39 AM, Volker Armin Hemmann wrote:
>> [894019.770058] [Hardware Error]: MC4 Error (node 0): DRAM ECC error
>> detected on the NB.
>> [894019.770084] EDAC MC0: 1 CE on mc#0csrow#2channel#0 (csrow:2
>> channel:0 page:0x2aa6ce offset:0xc60 grain:0 syndrome:0x63e1)
>> [894019.770090] [Hardware Error]: Error Status: Corrected error, no
>> action required.
>> [894019.770098] [Hardware Error]: CPU:0 (10:4:2)
>> MC4_STATUS[-|CE|MiscV|-|AddrV|CECC]: 0x9c70c00063080a13
>> [894019.770105] [Hardware Error]: MC4_ADDR: 0x00000002aa6cec60
>> [894019.770110] [Hardware Error]: cache level: L3/GEN, mem/io: MEM,
>> mem-tx: RD, part-proc: RES (no timeout)
>>
>> and this, my children, is why I am using ECC ram.
>>
>> Using zfs showed me, that there are errors that the system does not
>> catch but corrupts data.
>>
>> And this evening, with a thunderstorm outside I got that beauty above...
> Is ECC memory a drop-in replacement for ordinary RAM, or does it need
> a special motherboard?
>
>
>
>
>

depends on your motherboard. ASUS boards support ECC officially. You
just put it in. With Gigabyte some boards support it, some don't - and
they don't advertise it. But on their forums are threads about it. Rest:
I have no idea.

Re: [gentoo-user] Re: ECC-ram, it is worth it.

[Rich Freeman]

On Sat, Jul 26, 2014 at 6:55 PM, walt <w41ter@gmail.com> wrote:
>
> Is ECC memory a drop-in replacement for ordinary RAM, or does it need
> a special motherboard?
>

It requires both CPU and motherboard support I believe.  The RAM
itself isn't much more expensive - really just reflecting the cost of
the extra capacity required.

If you were already going to buy a CPU or motherboard that supports
ECC then the incremental cost is almost certainly worth it IMHO.  The
problem is that if you weren't otherwise going to buy either then the
incremental cost to upgrade all the components to support it can add
up quite a bit.  So, when people say that it is just an extra $10-20
for the RAM, that bay be a rather misleading figure.

Rich

Re: [gentoo-user] Re: ECC-ram, it is worth it.

[Volker Armin Hemmann]

Am 27.07.2014 01:07, schrieb Rich Freeman:
> On Sat, Jul 26, 2014 at 6:55 PM, walt <w41ter@gmail.com> wrote:
>> Is ECC memory a drop-in replacement for ordinary RAM, or does it need
>> a special motherboard?
>>
> It requires both CPU and motherboard support I believe.  The RAM
> itself isn't much more expensive - really just reflecting the cost of
> the extra capacity required.

AFAIK all AMD cpus support it.

Re: [gentoo-user] Re: ECC-ram, it is worth it.

[Edward MN]

On 07/26/14 15:55, walt wrote:
> On 07/26/2014 10:39 AM, Volker Armin Hemmann wrote:
>> [894019.770058] [Hardware Error]: MC4 Error (node 0): DRAM ECC error
>> detected on the NB.
>> [894019.770084] EDAC MC0: 1 CE on mc#0csrow#2channel#0 (csrow:2
>> channel:0 page:0x2aa6ce offset:0xc60 grain:0 syndrome:0x63e1)
>> [894019.770090] [Hardware Error]: Error Status: Corrected error, no
>> action required.
>> [894019.770098] [Hardware Error]: CPU:0 (10:4:2)
>> MC4_STATUS[-|CE|MiscV|-|AddrV|CECC]: 0x9c70c00063080a13
>> [894019.770105] [Hardware Error]: MC4_ADDR: 0x00000002aa6cec60
>> [894019.770110] [Hardware Error]: cache level: L3/GEN, mem/io: MEM,
>> mem-tx: RD, part-proc: RES (no timeout)
>>
>> and this, my children, is why I am using ECC ram.
>>
>> Using zfs showed me, that there are errors that the system does not
>> catch but corrupts data.
>>
>> And this evening, with a thunderstorm outside I got that beauty above...
>
> Is ECC memory a drop-in replacement for ordinary RAM, or does it need
> a special motherboard?
>
>
>
>
   yeah, requires a motherboard that supports ECC ram.

Re: [gentoo-user] Re: ECC-ram, it is worth it.

[Marc Stürmer]

Am 26.07.2014 20:23, schrieb Volker Armin Hemmann:

> but you will care when your kernel writes the next file right over the
> partition boundary.

That's why I do have backups of all my relevant data on an external 
storage medium.

Re: [gentoo-user] Re: ECC-ram, it is worth it.

[Dale]

Marc Stürmer wrote:
> Am 26.07.2014 20:23, schrieb Volker Armin Hemmann:
>
>> but you will care when your kernel writes the next file right over the
>> partition boundary.
>
> That's why I do have backups of all my relevant data on an external
> storage medium.
>
>

Just watch that you don't backup bad data.  I've seen that happen before
and it really sucks.  :-(

Dale

:-)  :-) 

Re: [gentoo-user] Re: ECC-ram, it is worth it.

[Frank Steinmetzger]

[-- Attachment #1: Type: text/plain, Size: 1641 bytes --]

On Sat, Jul 26, 2014 at 11:00:26PM -0700, Edward MN wrote:
> On 07/26/14 15:55, walt wrote:
> > On 07/26/2014 10:39 AM, Volker Armin Hemmann wrote:
> >> [894019.770058] [Hardware Error]: MC4 Error (node 0): DRAM ECC error
> >> detected on the NB.
> >> […]
> >> and this, my children, is why I am using ECC ram.
> >> […]
> >> And this evening, with a thunderstorm outside I got that beauty above...
> >
> > Is ECC memory a drop-in replacement for ordinary RAM, or does it need
> > a special motherboard?
> >
>    yeah, requires a motherboard that supports ECC ram.

Big was my surprise to learn that our old Pentium 3 PC from 1999 has ECC
support in its three RAM sockets. The problem today is the artificial
paritioning of the market.

It seems nigh impossible (at least in the Intel world, please correct me
regarding AMD) to have ECC RAM in a normal Home PC these days, especially in
an ITX form factor, as I am currently investigating. There are Xeons for the
1150 “consumer socket”, but ECC is only supported by server chipsets such as
the C series. Those come either on ITX boards with abysmal I/O capabilities
for home use or on high-power workstation ATX boards that cost a small
fortune. *sigh*

I would have liked the aspect of a system that tells me when something goes
wrong, but there seems no such thing for my requirements. So I must help
myself with file checksums when dealing with my archive disks.
-- 
Gruß | Greetings | Qapla’
Please do not share anything from, with or about me with any Facebook service.

Can you give me a cigarette?  Mine are still in the vending machine.

[-- Attachment #2: Digital signature --]
[-- Type: application/pgp-signature, Size: 819 bytes --]

Re: [gentoo-user] Re: ECC-ram, it is worth it.

[Volker Armin Hemmann]

Am 29.07.2014 20:18, schrieb Frank Steinmetzger:
> On Sat, Jul 26, 2014 at 11:00:26PM -0700, Edward MN wrote:
>> On 07/26/14 15:55, walt wrote:
>>> On 07/26/2014 10:39 AM, Volker Armin Hemmann wrote:
>>>> [894019.770058] [Hardware Error]: MC4 Error (node 0): DRAM ECC error
>>>> detected on the NB.
>>>> […]
>>>> and this, my children, is why I am using ECC ram.
>>>> […]
>>>> And this evening, with a thunderstorm outside I got that beauty above...
>>> Is ECC memory a drop-in replacement for ordinary RAM, or does it need
>>> a special motherboard?
>>>
>>    yeah, requires a motherboard that supports ECC ram.
> Big was my surprise to learn that our old Pentium 3 PC from 1999 has ECC
> support in its three RAM sockets. The problem today is the artificial
> paritioning of the market.
>
> It seems nigh impossible (at least in the Intel world, please correct me
> regarding AMD) to have ECC RAM in a normal Home PC these days, especially in
> an ITX form factor, as I am currently investigating. There are Xeons for the
> 1150 “consumer socket”, but ECC is only supported by server chipsets such as
> the C series. Those come either on ITX boards with abysmal I/O capabilities
> for home use or on high-power workstation ATX boards that cost a small
> fortune. *sigh*
>
> I would have liked the aspect of a system that tells me when something goes
> wrong, but there seems no such thing for my requirements. So I must help
> myself with file checksums when dealing with my archive disks.

I don't know about AMD's APUs but AFAIK all CPUs using the AM2/AM3
socket support it.

Re: [gentoo-user] Re: ECC-ram, it is worth it.

[Edward M]

On 07/29/14 11:18, Frank Steinmetzger wrote:
> On Sat, Jul 26, 2014 at 11:00:26PM -0700, Edward MN wrote:
>> On 07/26/14 15:55, walt wrote:
>>> On 07/26/2014 10:39 AM, Volker Armin Hemmann wrote:
>>>> [894019.770058] [Hardware Error]: MC4 Error (node 0): DRAM ECC error
>>>> detected on the NB.
>>>> […]
>>>> and this, my children, is why I am using ECC ram.
>>>> […]
>>>> And this evening, with a thunderstorm outside I got that beauty above...
>>>
>>> Is ECC memory a drop-in replacement for ordinary RAM, or does it need
>>> a special motherboard?
>>>
>>     yeah, requires a motherboard that supports ECC ram.
>
> Big was my surprise to learn that our old Pentium 3 PC from 1999 has ECC
> support in its three RAM sockets. The problem today is the artificial
> paritioning of the market.
>
> It seems nigh impossible (at least in the Intel world, please correct me
> regarding AMD) to have ECC RAM in a normal Home PC these days, especially in
> an ITX form factor, as I am currently investigating. There are Xeons for the
> 1150 “consumer socket”, but ECC is only supported by server chipsets such as
> the C series. Those come either on ITX boards with abysmal I/O capabilities
> for home use or on high-power workstation ATX boards that cost a small
> fortune. *sigh*
>
> I would have liked the aspect of a system that tells me when something goes
> wrong, but there seems no such thing for my requirements. So I must help
> myself with file checksums when dealing with my archive disks.
>


   Unfortunately, I think it would be difficult finding a home-user board
   with ECC support. since, nowadays appears ECC is mostly for systems
   where data corruption is unacceptable, such a bank,etc

Re: [gentoo-user] Re: ECC-ram, it is worth it.

[Volker Armin Hemmann]

Am 30.07.2014 11:14, schrieb Edward M:
> On 07/29/14 11:18, Frank Steinmetzger wrote:
>> On Sat, Jul 26, 2014 at 11:00:26PM -0700, Edward MN wrote:
>>> On 07/26/14 15:55, walt wrote:
>>>> On 07/26/2014 10:39 AM, Volker Armin Hemmann wrote:
>>>>> [894019.770058] [Hardware Error]: MC4 Error (node 0): DRAM ECC error
>>>>> detected on the NB.
>>>>> […]
>>>>> and this, my children, is why I am using ECC ram.
>>>>> […]
>>>>> And this evening, with a thunderstorm outside I got that beauty
>>>>> above...
>>>>
>>>> Is ECC memory a drop-in replacement for ordinary RAM, or does it need
>>>> a special motherboard?
>>>>
>>>     yeah, requires a motherboard that supports ECC ram.
>>
>> Big was my surprise to learn that our old Pentium 3 PC from 1999 has ECC
>> support in its three RAM sockets. The problem today is the artificial
>> paritioning of the market.
>>
>> It seems nigh impossible (at least in the Intel world, please correct me
>> regarding AMD) to have ECC RAM in a normal Home PC these days,
>> especially in
>> an ITX form factor, as I am currently investigating. There are Xeons
>> for the
>> 1150 “consumer socket”, but ECC is only supported by server chipsets
>> such as
>> the C series. Those come either on ITX boards with abysmal I/O
>> capabilities
>> for home use or on high-power workstation ATX boards that cost a small
>> fortune. *sigh*
>>
>> I would have liked the aspect of a system that tells me when
>> something goes
>> wrong, but there seems no such thing for my requirements. So I must help
>> myself with file checksums when dealing with my archive disks.
>>
>
>
>   Unfortunately, I think it would be difficult finding a home-user board
>   with ECC support. since, nowadays appears ECC is mostly for systems
>   where data corruption is unacceptable, such a bank,etc
>

pretty easy actually. When I looked for ECC support, ALL Asus boards
supported it officially - and a whole bunch of Gigabyte boards according
to their forums.

Re: [gentoo-user] Re: ECC-ram, it is worth it.

[Volker Armin Hemmann]

Am 30.07.2014 11:14, schrieb Edward M:

I just went to Alternate, clicked on their cheapest ASUS Am3 board: ECC yes.

See? Easy.

Re: [gentoo-user] Re: ECC-ram, it is worth it.

[Rich Freeman]

On Wed, Jul 30, 2014 at 6:14 AM, Volker Armin Hemmann
<volkerarmin@googlemail.com> wrote:
>
> pretty easy actually. When I looked for ECC support, ALL Asus boards
> supported it officially - and a whole bunch of Gigabyte boards according
> to their forums.
>

Interesting, the Gigabyte board I'm using makes no mention of it.
Just something that needs to be considered up-front.

It is still a constraint though - if only 10% of the boards support
ECC on the AMD side, then you're committing to an AMD CPU, and you may
find it harder to find the other features you're looking for (number
of RAM slots, PCI(e) ports, SATA ports, SLI/crossfire, clock control,
good price, etc).

Rich

Re: [gentoo-user] Re: ECC-ram, it is worth it.

[Volker Armin Hemmann]

Am 30.07.2014 13:25, schrieb Rich Freeman:
> On Wed, Jul 30, 2014 at 6:14 AM, Volker Armin Hemmann
> <volkerarmin@googlemail.com> wrote:
>> pretty easy actually. When I looked for ECC support, ALL Asus boards
>> supported it officially - and a whole bunch of Gigabyte boards according
>> to their forums.
>>
> Interesting, the Gigabyte board I'm using makes no mention of it.
> Just something that needs to be considered up-front.
>
> It is still a constraint though - if only 10% of the boards support
> ECC on the AMD side, then you're committing to an AMD CPU, and you may
> find it harder to find the other features you're looking for (number
> of RAM slots, PCI(e) ports, SATA ports, SLI/crossfire, clock control,
> good price, etc).
>
> Rich
>
>
Asus M5A99X Evo 2.0

had all I ever wanted. Plus something.

My old gigabyte does not mention ECC either. But the Gigabyte Forums
said yes.

Seriously, it looks like you guys are making excuses ...

Re: [gentoo-user] Re: ECC-ram, it is worth it.

[Edward M]

On 07/30/14 03:16, Volker Armin Hemmann wrote:
> Am 30.07.2014 11:14, schrieb Edward M:
>
> I just went to Alternate, clicked on their cheapest ASUS Am3 board: ECC yes.
>
> See? Easy.
>

   Yes, easy. I looked at many Asus boards and they do
   mention ECC in the Specs page. I stand corrected.

   Thanks for the info.

Re: [gentoo-user] Re: ECC-ram, it is worth it.

[Frank Steinmetzger]

[-- Attachment #1: Type: text/plain, Size: 1012 bytes --]

On Wed, Jul 30, 2014 at 12:50:54AM +0200, Volker Armin Hemmann wrote:

> >>> Is ECC memory a drop-in replacement for ordinary RAM, or does it need
> >>> a special motherboard?
> > [...]
> > I would have liked the aspect of a system that tells me when something goes
> > wrong, but there seems no such thing for my requirements. So I must help
> > myself with file checksums when dealing with my archive disks.
> 
> I don't know about AMD's APUs but AFAIK all CPUs using the AM2/AM3
> socket support it.
> 
Hm... unfortunately, if I were to go the AMD route, I would get an FM2(+)
APU. Also, I'm looking for something smaller than ATX. Interestingly enough,
the product comparison site that I use does not mention ECC for the Asus
board that was stated in the other recent post. :'(

Oh well, what is a world without dreams.
-- 
Gruß | Greetings | Qapla’
Please do not share anything from, with or about me on any social network.

Get your grubby hands off my tagline! I stole it first!

[-- Attachment #2: Digital signature --]
[-- Type: application/pgp-signature, Size: 819 bytes --]

Re: [gentoo-user] Re: ECC-ram, it is worth it.

[Volker Armin Hemmann]

Am 31.07.2014 12:37, schrieb Frank Steinmetzger:
> On Wed, Jul 30, 2014 at 12:50:54AM +0200, Volker Armin Hemmann wrote:
>
>>>>> Is ECC memory a drop-in replacement for ordinary RAM, or does it need
>>>>> a special motherboard?
>>> [...]
>>> I would have liked the aspect of a system that tells me when something goes
>>> wrong, but there seems no such thing for my requirements. So I must help
>>> myself with file checksums when dealing with my archive disks.
>> I don't know about AMD's APUs but AFAIK all CPUs using the AM2/AM3
>> socket support it.
>>
> Hm... unfortunately, if I were to go the AMD route, I would get an FM2(+)
> APU. Also, I'm looking for something smaller than ATX. Interestingly enough,
> the product comparison site that I use does not mention ECC for the Asus
> board that was stated in the other recent post. :'(
>
> Oh well, what is a world without dreams.

so go to the asus site and look up the specs there.

AM3 processors support ECC, FM2 AFAIK don't, Opterons, no matter what
socket do. Easy.