SATA vs SCSI

  • Thread starter Thread starter Rob Nicholson
  • Start date Start date
Both Intel and AMD sell CPU trays without heatsink
and fan, and boxed CPUs with heatsink and fans.
Click to expand...

Irrelevant to what happened with the volume retail sales
when the AMD cpus had no over temp protection.
If you decide on a tray CPU, don't blame the manufacturer.
Click to expand...
BTW, we once had problems with cooling. The CPU was an Intel Celeron
1000. The shop had mounted the heatsink the wrong way round, and it
was obvious pretty quickly that there was a problem with the machine
(don't remember what it was). Anyway, the CPU survived. So, the
overheating protection is a defense against sloppy assembly (who knows
how many people are running permanently heat-throttled CPUs because
the heatsink is mounted badly or wrongly).
Click to expand...

I said that in what got snipped off the quoting.
Some older AMD CPUs (AFAIK Athlon before XP) fry
after a few seconds without a heatsink. With a properly
mounted heatsink, the CPU takes much longer to warm
up, and I guess even the old unprotected ones
won't fry (but they will probably fail temporarily).
Click to expand...

They did fry on fan failure.
When I turn on my Athlon 64 system with regulated fan, the fan does
not spin the first few minutes. I have set the spin-up temperature to
35 degrees Celsius, so you see that it takes quite a while to warm up
the heat sink (ok, it is probably much faster, if the CPU is under load).
Click to expand...

Separate issue to fan failure when that failure doesnt get noticed.
 
Arno Wagner said:
Previously Sambo said:
Arno Wagner said:
:

[...]

Rita,


I was "sort of" on your side - until you slagged off AMD
processors.

Five years ago, I would not have dreamt of using AMD processors
in a server. (Inasmuch as I would not consider using VIA CPUs in
even a games machine.)

However, AMD is absolutely, most definitely, the way to go for
practically every application on the planet. Opteron for
servers, the rest for less important machines.

I totally agree. The only thing I did not like on the Athlon XPs
was their lack of heat-spreader and thermal protection. Now that
AMD has fixed that, I think the last disadvantage of these CPUs
has gone. On the other hand the rest if the CPU architecture
is far superiour to Intel. Just look at the thermal
characteristics, the prices and the Multi-CPU support. I think
this is an example of a dominatnt player (Intel) having gotten
lazy.

I have near-mission-critical systems here - and I am as
comfortable using AMD as I am Intel. However, the AMD chips
grind the Intel chips into the dust where performance is
concerned. And I'm talking performance advantages to a very
high degree - not just 5% or so.

I think Intel never had really much going for it that would have
made it significantly more reliable than AMD. There was a time
where on the low-end PC makers did not really have enough
experience with AMD CPUs and botched things like heatsink
selection, but
that is not AMDs fault.

Yes it was, AMD should have supplied everything, cpu, heatsink,
fan. And had decent protection for when something went wrong like
the heatsink not properly mounted on the cpu or the fan stops etc.
A CPU is an engineering product, intended to be mounted by experts
only.
Click to expand...
Click to expand...
No its not, and you need protection for fan failure anyway.
Click to expand...

Of course it is. It is an electronics component and strictly
''experts only''.

Obviously there is not point in cointinuing this discussion.

Arno
Click to expand...
 
Arno Wagner said:
Previously Sambo said:
Arno Wagner said:
:

[...]

Rita,


I was "sort of" on your side - until you slagged off AMD
processors.

Five years ago, I would not have dreamt of using AMD processors
in a server. (Inasmuch as I would not consider using VIA CPUs in
even a games machine.)

However, AMD is absolutely, most definitely, the way to go for
practically every application on the planet. Opteron for
servers, the rest for less important machines.

I totally agree. The only thing I did not like on the Athlon XPs
was their lack of heat-spreader and thermal protection. Now that
AMD has fixed that, I think the last disadvantage of these CPUs
has gone. On the other hand the rest if the CPU architecture
is far superiour to Intel. Just look at the thermal
characteristics, the prices and the Multi-CPU support. I think
this is an example of a dominatnt player (Intel) having gotten
lazy.

I have near-mission-critical systems here - and I am as
comfortable using AMD as I am Intel. However, the AMD chips
grind the Intel chips into the dust where performance is
concerned. And I'm talking performance advantages to a very
high degree - not just 5% or so.

I think Intel never had really much going for it that would have
made it significantly more reliable than AMD. There was a time
where on the low-end PC makers did not really have enough
experience with AMD CPUs and botched things like heatsink
selection, but
that is not AMDs fault.

Yes it was, AMD should have supplied everything, cpu, heatsink,
fan. And had decent protection for when something went wrong like
the heatsink not properly mounted on the cpu or the fan stops etc.
A CPU is an engineering product, intended to be mounted by experts
only.
Click to expand...
Click to expand...
No its not, and you need protection for fan failure anyway.
Click to expand...
Click to expand...
Of course it is. It is an electronics component and strictly ''experts only''.
Click to expand...

Wrong, as always.
Obviously there is not point in cointinuing this discussion.
Click to expand...

Yep, you never ever could bullshit your way out of a wet paper bag.
 
However, I believe SCSI is on its way out. Partly because of marketing
hype, but also because SCSI is incredibly expensive to set up from
scratch, and the vast majority of users are scared of these costs - to
their detriment, admittedly.
Click to expand...

Too right - Adaptec SCSI controller has failed and I nearly fell off my
chair when I looked at the cost of a replacement. It's real overkill though
as all that's connected to the SCSI controller is an external LVD LTO-1 tape
drive. A lot of functionality just sits doing nothing.

Rob.
 
But the time has come to plan for the future and upgrade where
appropriate. We're looking at switching to Windows 2003 and buying another
server to host SQL Server and Exchange. We're at about 40 users at the
moment but have plans to grow to 100+
Click to expand...

A poynant little anecdote to this discussion... We have a secondary storage
server running an dual Athlon MP 2800+ with lots of RAM and SATA RAID-0
sub-system. It's full of non-critical information such as archives, CD-ROM
images etc. It's backed up once a month onto 5 LTO tapes.

Anyway, we managed to fill the 600GB volume so bought another Sil3112 SATA
RAID controller and two 400GB Samsung SATA drives. I have wasted an
afternoon trying to get it working:

o Creating RAID-0 array in the BIOS froze but appeared to create the array
o Attempting to extend existing volume ended up with broken volume as the
BIOS hadn't actually created the array
o Updated the BIOS on the Sil3112 and voila - creating array was instant and
didn't freeze
o Attempting to use the array though in Windows 2000 Server caused server to
freeze, hang and generate lots of disk errors
o Attempting to use drives non-RAID and still lots of errors in the log
about timeouts, bad blocks etc.
o Adaptec SCSI adapter decided to pack up as well so can't restore the data
on the same server
o BackupExec's facility to store a tape catalog on the tape nevers seems to
work so it'll take 10+ hours just to catalog
o Then another 48 hours to restore original 600GB :-)

Computer's aren't easy sometimes <grin>

Low down is that it's either a) broken RAID controller, b) incompatability
with Samsung drives, c) incompatability with Tiger motherboard, d) faulty
drive, e) it's the phase of the moon.

So not exactly glowing success for SATA RAID...

Cheers, Rob.
 
Rob Nicholson said:
Too right - Adaptec SCSI controller has failed and I nearly fell off my
chair when I looked at the cost of a replacement. It's real overkill though
as all that's connected to the SCSI controller is an external LVD LTO-1 tape
drive. A lot of functionality just sits doing nothing.
Click to expand...

So who is holding a gun to your head preventing you from buying
a 5 dollar UW SCSI controller?
 
Rob said:
A poynant little anecdote to this discussion... We have a secondary
storage server running an dual Athlon MP 2800+ with lots of RAM and SATA
RAID-0 sub-system. It's full of non-critical information such as archives,
CD-ROM images etc. It's backed up once a month onto 5 LTO tapes.

Anyway, we managed to fill the 600GB volume so bought another Sil3112 SATA
RAID controller and two 400GB Samsung SATA drives. I have wasted an
afternoon trying to get it working:

o Creating RAID-0 array in the BIOS froze but appeared to create the array
o Attempting to extend existing volume ended up with broken volume as the
BIOS hadn't actually created the array
o Updated the BIOS on the Sil3112 and voila - creating array was instant
and didn't freeze
o Attempting to use the array though in Windows 2000 Server caused server
to freeze, hang and generate lots of disk errors
o Attempting to use drives non-RAID and still lots of errors in the log
about timeouts, bad blocks etc.
o Adaptec SCSI adapter decided to pack up as well so can't restore the
data on the same server
o BackupExec's facility to store a tape catalog on the tape nevers seems
to work so it'll take 10+ hours just to catalog
o Then another 48 hours to restore original 600GB :-)

Computer's aren't easy sometimes <grin>

Low down is that it's either a) broken RAID controller, b) incompatability
with Samsung drives, c) incompatability with Tiger motherboard, d) faulty
drive, e) it's the phase of the moon.

So not exactly glowing success for SATA RAID...
Click to expand...

Someone at Silicon Image must be a member of the Radical Right who went
overboard on the morality stuff--their hardware doesn't play with itself,
at least their early SATA chips didn't. Try a RAID controller with a
different chip and I suspect that it will work fine--watch out for Raidcore
boards though--the Tiger 2875 has a specific problem with them that
required a couple of tries before they fixed it in the BIOS. Before you do
that though, make make sure that both the add-in board and the motherboard
are running the same revision of the Silicon Image firmware, even if you
have to backrev one of them to get there.

Your mention of the Adaptec host adapter packing up suggests that it might
be an interrupt problem--no matter how much Microsoft and Intel hype the
notion that Plug and Pray works and that PCI can share interrupts, in the
real world it _still_ doesn't work cleanly and I wish those twits would get
over themselves and put manual overrides in the default installation
instead of requiring that you reinstall the OS with a non-default HAL in
order to turn them on. Might be worth a try at musical boards to see if
there's an arrangement that makes it happy.
 
Rob Nicholson said:
A poynant little anecdote to this discussion... We have a secondary storage
server running an dual Athlon MP 2800+ with lots of RAM and SATA RAID-0
sub-system. It's full of non-critical information such as archives, CD-ROM
images etc. It's backed up once a month onto 5 LTO tapes.

Anyway, we managed to fill the 600GB volume so bought another
Click to expand...
Sil3112 SATA RAID
Click to expand...

A cheaper than cheap toy.
controller and two 400GB Samsung SATA drives. I have wasted an afternoon
trying to get it working:

o Creating RAID-0 array in the BIOS froze but appeared to create the array
o Attempting to extend existing volume ended up with broken volume as the
BIOS hadn't actually created the array
o Updated the BIOS on the Sil3112 and voila - creating array was instant and
didn't freeze
o Attempting to use the array though in Windows 2000 Server caused server to
freeze, hang and generate lots of disk errors
o Attempting to use drives non-RAID and still lots of errors in the log
about timeouts, bad blocks etc.
o Adaptec SCSI adapter decided to pack up as well so can't restore the data
on the same server
o BackupExec's facility to store a tape catalog on the tape nevers seems to
work so it'll take 10+ hours just to catalog
o Then another 48 hours to restore original 600GB :-)

Computer's aren't easy sometimes <grin>

Low down is that it's either
a) broken RAID controller,
b) incompatability with Samsung drives,
c) incompatability with Tiger motherboard,
d) faulty drive,
Click to expand...
e) it's the phase of the moon.
Click to expand...

f) Or just you.
So not exactly glowing success for SATA RAID...
Click to expand...

Nor SCSI.
 
Ron said:
Oh bullshit. The Northwoods were very decent cpus and ran a lot
quieter than the AMDs.
Click to expand...

Yes, but that doesn't change the fact that the P4 design was a
dead-end.
 
Sil3112 SATA RAID
A cheaper than cheap toy.
Click to expand...

Hmm, our experience doesn't back that up. This server had an existing Si3112
card in there and it has run 24/7 for a year now.
f) Or just you.
Click to expand...

Ahh rudeness, that often works.

Cheers, Rob.
 
Someone at Silicon Image must be a member of the Radical Right who went
overboard on the morality stuff--their hardware doesn't play with itself,
at least their early SATA chips didn't. Try a RAID controller with a
different chip and I suspect that it will work fine--watch out for
Raidcore
boards though--the Tiger 2875 has a specific problem with them that
required a couple of tries before they fixed it in the BIOS. Before you
do
that though, make make sure that both the add-in board and the motherboard
are running the same revision of the Silicon Image firmware, even if you
have to backrev one of them to get there.
Click to expand...

Good idea. Adaptec have the 1210SA for around £40 which might fit the bill.
Your mention of the Adaptec host adapter packing up suggests that it might
be an interrupt problem--no matter how much Microsoft and Intel hype the
notion that Plug and Pray works and that PCI can share interrupts, in the
real world it _still_ doesn't work cleanly and I wish those twits would
get
over themselves and put manual overrides in the default installation
instead of requiring that you reinstall the OS with a non-default HAL in
order to turn them on. Might be worth a try at musical boards to see if
there's an arrangement that makes it happy.
Click to expand...

You could well be right as we suspect something very weird going on with the
cards. I told IT to bite the bullet and buy a replacement Adapatec SCSI
adapter (we could do with one as a spare anyway) and it arrived. Plugged in
the tape drive worked fine. Plugged in the existing SATA RAID and it hung on
boot up. After a considerable time unplugging, plugging and moving around
the server is back up and running. But as far as we can tell, we didn't
actually *do* anything except maybe change the order.

Cheers, Rob.
 
Rob Nicholson said:
Hmm, our experience doesn't back that up. This server had an existing Si3112
card in there and it has run 24/7 for a year now.
Click to expand...

Thanks for debunking your previous SATA-RAID sob story.
Ahh rudeness, that often works.
Click to expand...

Yeah, blaming everything else but yourself for your own stupity, that
works so much better, that really calls the goodness out of people.
 
Rob Nicholson said:
Good idea. Adaptec have the 1210SA for around £40 which might fit the bill.
Click to expand...

Good choice, Nicholson. Not a Silicon Image controller in sight on that card.

Or is there? Oh look:
http://www.adaptec.com/worldwide/su...10SA_board.jpg&imageHeight=440&imageWidth=600
You could well be right as we suspect something very weird going on with the
cards. I told IT to bite the bullet and buy a replacement Adapatec SCSI
adapter (we could do with one as a spare anyway) and it arrived. Plugged in
the tape drive worked fine. Plugged in the existing SATA RAID and it hung on
boot up. After a considerable time unplugging, plugging and moving around
the server is back up and running. But as far as we can tell, we didn't
actually *do* anything except maybe change the order.
Click to expand...

And chanded PCI interrupts by doing that.
 
Rob said:
Good idea. Adaptec have the 1210SA for around £40 which might fit the
bill.


You could well be right as we suspect something very weird going on with
the cards. I told IT to bite the bullet and buy a replacement Adapatec
SCSI adapter (we could do with one as a spare anyway) and it arrived.
Plugged in the tape drive worked fine. Plugged in the existing SATA RAID
and it hung on boot up. After a considerable time unplugging, plugging and
moving around the server is back up and running. But as far as we can
tell, we didn't actually *do* anything except maybe change the order.
Click to expand...

It's surprising how often that works. Another issue that often arises is
that it is common for two PCI slots to be hard-wired to the same interrupt
trace. If possible those should have low-traffic devices or one device
that does not use interrupts installed.
 
Yeah, blaming everything else but yourself for your own stupity, that
works so much better, that really calls the goodness out of people.
Click to expand...

This person usually this approachable?

Cheers, Rob.
 
Back
Top