P2B and 40GB HDD

[jerzyorzeszek]

Hello,

I need to know if p2b will work with Seagate ST340810A?
I know that P2B has ATA33 and hdd ATA100...so hdd will work slower.
Is that correct?

Dragon

 

[RSM Sison]

Hi,

You can buy a Promise Ultra ATA controller so you utilize the speed of
your drive.

Regards,

Roy

 

[Daniel Mandic]

Hello,

I need to know if p2b will work with Seagate ST340810A?
I know that P2B has ATA33 and hdd ATA100...so hdd will work slower.
Is that correct?

Dragon

Hi Dragon!


ATA33 on the BX Southbridge, is the best implementation of Ultra DMA I
ever saw.
It´s not SCSI. But very good for IDE.

I think with latest BIOS you can use up to 160GB.
You can attach any kind of U-DMA/IDE Harddisk to the onboard
controller. Maybe also U133 - I haven´t tried.
But an U100 is working fine (Maxtor 60GB/5400).
It will work slower.... It will work just on 33MB/sec instead of 66 or
100MB/sec. The phisically Bandwith of a fast HD is about 10-30MB/sec.
Well, you can count that the HD-cache can (HD onboard RAM) work with
100MB/sec or 66 respectively.

Maybe it depends also which OS you will use. Windows XP have so much
caching strategies you could not benefit from the higher BUS-speed, so
much. How much HD´s are faster than 33MB/sec? 15.000UPM Cheetah??


The ST340810A should work fine with the P2B. IMO.



Best Regards,

Daniel Mandic

 

[Robert Hancock]

Maybe it depends also which OS you will use. Windows XP have so much
caching strategies you could not benefit from the higher BUS-speed, so
much. How much HD´s are faster than 33MB/sec? 15.000UPM Cheetah??

Most hard drives can do sustained transfers over 33 MB/sec these days.
ATA-66 might not be a limitation but ATA-33 may well be.

It would be better to use an add-in controller for a new drive on a
system like that, so that you can get better transfer speeds. You can
even use SATA that way if you wish. This also prevents any problems with
the onboard controller/BIOS not recognizing the full drive size.

 

[Stephan Grossklass]

STRs like that have not been a problem for, like, 5 years. The (not so)
venerable Deathstar 75GXP managed up to ~37 megs a second. Besides, no
hard drive has an interface that's 100% efficient. The best I've seen
was 30 MB/s (UDMA33) / 90 MB/s (UDMA100) for IBM/Hitachi drives (in
coretest). The Barracuda ATA IV was pretty inefficient and only managed
25 MB/s on UDMA33.

Most hard drives can do sustained transfers over 33 MB/sec these days.
ATA-66 might not be a limitation but ATA-33 may well be.

Today, UDMA100 is pretty much necessary to achieve full STR, as current
7200 rpm drives reach up to ~65 MB/s. (A 74 gig Raptor clock in at 71.8
megs a second, current 15k SCSI drives reach 88.5 to 97.4.) And it's not
only this which makes current drives much faster, firmware and
controller hardware has also improved a lot (not to forget the larger
buffer sizes). Thus a current harddrive rarely is a bad idea, even in a
somewhat dated system.

It would be better to use an add-in controller for a new drive on a
system like that, so that you can get better transfer speeds.
Seconded.

You can
even use SATA that way if you wish.

Though for a system like that, a plain Ultra100 TX2 would be just fine.
Current 7200 rpm disks also exist in old-style PATA versions, which are
just as fast and have to dissipate typically ~0.6 W less.

This also prevents any problems with
the onboard controller/BIOS not recognizing the full drive size.

Which would be less of a problem with the P2B, where the 1014 beta 3
BIOS will swallow anything up to 128 GiB just fine.

Stephan

 

[Stephan Grossklass]

Hello,

I need to know if p2b will work with Seagate ST340810A?

A little note: Be sure to try and disable AAM (or set it to fastest) on
the old Seagate U6, as these seem to ship with AAM set to quiet and thus
feature a fairly pathetic access time of more than 19 ms stock. It'll
only be slow instead of very slow afterwards.

Stephan

 

[Gilbert]

You can buy a Promise Ultra ATA controller so you utilize the speed of

your drive.

http://dlsvr03.asus.com/pub/ASUS/mb/slot1/440bx/p2b-d/p2bd(s)-106.pdf
From the P2B Manual:
Intel AGPset: Features Intel's 440BX AGPset with I/O subsystems and
front-side bus (FSB) platform, which boosts the traditional 66-mhz external
bus speed to 100 MHZ.


http://www.promise.com/support/file/manual/u133_manual_Promise_v2.PDF
From the Promise Ultra133 TX2 Manual:
The Ultra133 TX2 is designed to support motherboards with a PCI bus speed of
66MHz
(versus earlier 33MHz models). The faster bus speed provides up to 266MB per
second bandwidth.


http://www.wdc.com/en/products/Products.asp?DriveID=32
From the Western Digital web page for the WD800JB HD (EIDE):
Transfer Rates
Buffer To Disk 602 Mbits/s (Max)
Buffer to Host (EIDE)
Mode 5 Ultra ATA 100.0 MB/s
Mode 4 Ultra ATA 66.6 MB/s
Mode 2 Ultra ATA 33.3 MB/s
Mode 4 PIO 16.6 MB/s
Mode 2 multi-word DMA 16.6 MB/s


.. . . so, the sustained transfer speed would be 100 mb/s with the promise
controller card and 33 mb/s without it?

-g

 

[Stephan Grossklass]

. . . so, the sustained transfer speed would be 100 mb/s with the promise
controller card and 33 mb/s without it?

Given 100% efficiency and no other throughput limits, it would be 133
and 33 MB/s, respectively. In practice this would work out to anything
between 75 and 90 MB/s for UDMA100 and a third of this for UDMA33. With
UDMA133 you could thus expect up to 120 MB/s, but this cannot be fully
reached on an ordinary PCI32/33 PCI bus, which will limit things to
~110...115 MB/s - fair enough. When writing, you'll see a ~55 MB/s
throughput limit imposed by the CPU-PCI bridge, a last remnant of the
PCI performance issues in PPro days.

Stephan

 

[Daniel Mandic]

~55 MB/s throughput limit imposed by the CPU-PCI bridge, a last
remnant of the PCI performance issues in PPro days.

Stephan

Hi Stephan!


I see you follow the things as they are and not as they look.

Yeah, something about 50-60MB/sec you can have from the PCI Bus. ;-)

I remember my Saturn II, SP3G with 512K second level kätsch. The PCI
Bus on this beast was unbeatable. It is a speed standard, till now .
58MB/sec.
Other Chipsets made some thousands of KB throughput with a VGA up to
5000 or so. The Saturn II outfitted PC´s made over 9000/sec. It was a
first choice chipset for gamer, much faster than Pentium 90 or 100/50
derivates at this time. The Pipelined Burst Cache Chipsets like the
mighty FX and the humble HX were faster, of course. But only bez of the
P5 Power and the 64Bit down-architecture, even the genial interleaved
RAM technique cannot compare.

It´s only beaten by the i440bx (so far I saw), which is slightly faster
than a Saturn II. I mean real-DOS, 640KB.
9400 vs 9800KB/sec.But at least, the more modern BX is so fast on PCI
as it can be ;-)
Not to mention the PCI-ISA Bridge, splendid.



Best Regards,

Daniel Mandic

 

[Stephan Grossklass]

Hi Stephan!

I see you follow the things as they are and not as they look.

Well, it's kinda hard to deny this issue if one gets benchmark results
like these:
http://stephan.win31.de/hdt722525dlat80-v44o.ps
http://stephan.win31.de/hdt722525dlat80-v44o.txt
Note the capped writes, while reads are fine. Since I had seen 55 megs a
second for graphics transfers in the past, I only had to put one and one
together.

BTW, I do like to do some more in-depth dabbling with old hardware,
partly because I'm too broke for the new interesting stuff . (These
benchmark results for Intel's new "Sossaman" core CPU look nice, the
power consumption of the beast even more so. It would be great if these
came in socket 604, then one could use a nice würgstation, err,
workstation board with PCI-X and all the goodies without being killed by
(a) a look at the power bill or (b) $PARENTS after (a).) You may want to
read up on my current GA-686KDX adventures, including my first
(successful) attempt at BIOS modding. (Admittedly that BIOS mod wasn't
too hard to do, the right tools are easily to be found for a BIOS of
this vintage and I didn't do anything too outlandish. Sure was fun
though - and the result even works.)

Yeah, something about 50-60MB/sec you can have from the PCI Bus. ;-)
More.

I remember my Saturn II, SP3G with 512K second level kätsch.

^^^^^^
That's what I call an interesting interpretation of "cache"... Sure
beats my "dadaschiet".

The PCI
Bus on this beast was unbeatable. It is a speed standard, till now .
58MB/sec.

The most you can wring out of a 32/33 PCI bus is about 110...115 MB/s
for busmaster transfers (--> tecChannel). I have personally seen more
than 90 on a BX board (Adaptec SCSI Bench reading from an U2W and U66
attached drive at the same time) so I guess it wouldn't fall short here
(at least reading, i.e. PCI -> CPU). How much you can get on a 440
series chipset in CPU -> PCI direction will depend on the exact chipset
and on whether USWC (uncached, speculative write combining, or write
combining in short) is turned on for the graphics memory ranges. With
USWC, even a humble i440FX will manage to transfer more than 90 MB/s to
a Millennium II in LFB VESA modes, provided one has used FastVid (1.10
is the last version, it seems) or ctppro. The latter also is good for
SCSI host adapter tweaking, btw - set the address range with the SCSI
BIOS to WT instead of UC, and the SCSI stuff will operate faster in DOS,
as evidenced by the System Speed Test v4.78 disk benchmark (on the
GA-686KDX with a 53C895 based HA using the last 4.19.00 BIOS, maximum
STR grew from 39 to >41 MB/s - the maximum for the oldish Fujitsu MAJ I
used - , burst transfer from 33xxx to 51xxx MB/s). Without USWC, write
data rates seem to range from 30..35 MB/s (i440FX) to 55 MB/s (i440BX).
On a BX (or LX) board, using USWC allows quite impressive transfer rates
to AGP graphics cards, up to 252 "MB/s" (which probably is more like
MiB/s and would thus translate to ~264 MiB/s or almost 100% of the
bandwidth available in PCI mode), even more with the AGP overclocked
(BX@133). I have tried to enable AGP with 2X transfers by writing to the
chipset and graphics card registers but alas, no change whatsoever.
Maybe some more configuration of the graphics card prior to enabling AGP
mode would be necessary, and I haven't studied the AGP spec intensively
enough to know what this would have to look like.

One could try to circumvent the 55 MB/s write limit for disk controllers
by setting up their memory ranges for USWC, but I have no idea how one
would modify CPU MTRRs in WinNT/2k/XP/S2003 (nor what the cards may
think of this).

Stephan

 

[Daniel Mandic]

Well, it's kinda hard to deny this issue if one gets benchmark results
like these:
http://stephan.win31.de/hdt722525dlat80-v44o.ps
http://stephan.win31.de/hdt722525dlat80-v44o.txt
Note the capped writes, while reads are fine. Since I had seen 55
megs a second for graphics transfers in the past, I only had to put
one and one together.

Still fast.

BTW, I do like to do some more in-depth dabbling with old hardware,
partly because I'm too broke for the new interesting stuff . (These
benchmark results for Intel's new "Sossaman" core CPU look nice, the
power consumption of the beast even more so. It would be great if
these came in socket 604, then one could use a nice würgstation, err,
workstation board with PCI-X and all the goodies without being killed
by (a) a look at the power bill or (b) $PARENTS after (a).) You may

How many Watts do the CPU need?

^^^^^^
That's what I call an interesting interpretation of "cache"... Sure
beats my "dadaschiet".

To be honest. It was at a time, I was in military defense
(Grundwehrdienst), a colleague and I were commanded to clean the cellar
;-). And this guy talked to me for hours about XT, AT, 368, 486 etc...
etc.. But at best I remember his outspeak of the 2nd level cache . I
couldn´t understand what he meant, my AMIGA does not have, nor it needs
a 2ns level cache. At this time, I did´nt have even a 1st-Level Cache
(Plain C= AMIGA Plus with 1MB Chip RAM extension for a total of 2MB)
Now I am a PC-Depp, and I understand that the most speeding thing for
the Intel spaghetti-code, is a fast Kätsch.
Especially with the P3-S, where I really recognize the outstanding
speed of it.
On my older Celeron Cu I could disable the 2nd cache and I did´nt miss
it.
Otherwise, I cannot disable the 2nd, nor the 1st with the 1014.003 beta
and the P3-S, at all. ???
Fortunately I do not need it. Seemingly, my DOS programs are working
even better (more compatible) than on the Celeron P3. Windows NT do the
same. And, Solid Rock stable! Unbelievable. Splendid.

The most you can wring out of a 32/33 PCI bus is about 110...115 MB/s
for busmaster transfers (--> tecChannel). I have personally seen more
than 90 on a BX board (Adaptec SCSI Bench reading from an U2W and U66

Yes, nominal is about 60...

from 30..35 MB/s (i440FX) to 55 MB/s (i440BX). On a BX (or LX)
board, using USWC allows quite impressive transfer rates to AGP
graphics cards, up to 252 "MB/s" (which probably is more like MiB/s
and would thus translate to ~264 MiB/s or almost 100% of the
bandwidth available in PCI mode), even more with the AGP overclocked
(BX@133). I have tried to enable AGP with 2X transfers by writing to
the chipset and graphics card registers but alas, no change
whatsoever. Maybe some more configuration of the graphics card prior
to enabling AGP mode would be necessary, and I haven't studied the
AGP spec intensively enough to know what this would have to look like.

I tried to tweak a little bit the GfX Performance, but USWC did´nt
help. Should I enable it?


Best Regards,

Daniel Mandic

 

[Stephan Grossklass]

Stephan Grossklass wrote:

[i440BX PCI write throughput limit]

Still fast.

Definitely. Can't beat a modern harddrive when it comes to wringing more
performance out of older systems - this one (a Deskstar T7K250) is
noticeably faster than the Cheetah 36ES before, and quieter and cooler
at that, not to mention fairly capacious (and pretty full already -
these FLAC'd radio recordings take up a lot of space, at ~400 MB/h and
~4.5-6 GB/week). Recommended.

How many Watts do the CPU need?

They write that the preproduction 1.5 GHz sample drew a maximum of 31
watts - not a mean feat for a dual-core CPU executing Cinebench 2003
faster than an Athlon 64 X2 4800+, particularly considering that a
Pentium D needs up to 130 watts (OUCH!). The whole system needed less
than 100 W. I really have to measure power consumption for my rig one
day, ideally it would be at =<100 W with monitor.

I tried to tweak a little bit the GfX Performance, but USWC did´nt
help. Should I enable it?

The option in the P3B-F's BIOS setup only turns on WC for the VGA memory
range, not the framebuffer. Use Fastvid. In Windows, the graphics card
driver should set this up.

Stephan

 

[Daniel Mandic]

Definitely. Can't beat a modern harddrive when it comes to wringing
more performance out of older systems - this one (a Deskstar T7K250)
is noticeably faster than the Cheetah 36ES before, and quieter and
cooler at that, not to mention fairly capacious (and pretty full
already - these FLAC'd radio recordings take up a lot of space, at
~400 MB/h and ~4.5-6 GB/week). Recommended.

But SCSI is SCSI. No IDE System can compare to a SCSI.
I saw a Network-Board dual P3 (one 933 installed!) with 3x18Wide
SCSI-Cachecontrolled, installing 2000.
Formating needs some seconds. And installing some minutes. Absolutely
mad.
My Machine is in a distance of one meter, beneath my Ears. I prefer
super-calm Harddisks. Like one, with 5400RPM. 12ms or slower...
I´ve had SCSI before, Mostly 1GB HD´s. My first IDE was a IBM 6.4GB
(U33!), In a BX it is pretty faster than a Fast-SCSI 1GB HD and much
more cheaper. I would need a 7200 9GB W-SCSI HD (inclusive controller,
I have just the plain AHA-2940) to outperform my slow IDE-HD (a 60GB
Maxtor U33). Too loud.
But as often in life, you cannot have both sides. Fast or slow.
I tried many times to find Ultrasilent SCSI HD´s......, the HOST is far
better than IDE :-(

They write that the preproduction 1.5 GHz sample drew a maximum of 31
watts - not a mean feat for a dual-core CPU executing Cinebench 2003
faster than an Athlon 64 X2 4800+, particularly considering that a

Really? The 64bit Athlons are two times faster (latency) in RAM than
the BX with 100MHz.
The BX can compare in latency, up to dual-channel 8xx and 9xx. But not
the 64Bit AMD, they are really faster.

How does the RAM performance looks like, with this Sossaman? Latency?

Pentium D needs up to 130 watts (OUCH!). The whole system needed less
than 100 W. I really have to measure power consumption for my rig one
day, ideally it would be at =<100 W with monitor.

Power measuring at 230V AC, would probably give you an exact result.
Overall....!

The Monitor takes the most power ;-(.
I am maybe about 120-150W Total. + 28" TV = Total 220W
CPU is maybe, at 25W or a little bit lower.

The option in the P3B-F's BIOS setup only turns on WC for the VGA
memory range, not the framebuffer. Use Fastvid. In Windows, the
graphics card driver should set this up.

Stephan

Ah, well. Preferably for DOS.
Thanks.




Kind Regards,

Daniel Mandic

 

[Stephan Grossklass]

But SCSI is SCSI. No IDE System can compare to a SCSI.

That's what you say. SCSI is not "better" per se when desktop
performance is considered, it's the drives that may make a difference -
or may not. You can buy current 15k SCSI drives that measure up with
excellent STR and access time yet barely are as fast as a
last-generation 7200 rpm ATA drive when using them as ordinary
single-user desktop drives. On the other hand, you can buy others that
qualify as the fastest desktop hard drives, period. (See
storagereview.com.) Very, very much is dependent on firmware
optimization (and controller hardware) when you're looking at hard drive
performance. SCSI drives of semi-recent vintage tend to be tuned more
for server performance than desktop performance, which in extreme cases
leads to sucky desktop performance like I wrote. I personally used to
have an U2W HA with a Cheetah 36ES in this machine, a fairly nice older
10k drive that remained the quietest such beast for a few years, thanks
to using FDBs (it also runs fairly hot because of this). Well decoupled,
this drive qualifies as "almost silent", the only annoying thing
remaining is its motor whine. A few months back, I was in urgent need of
more disk space and bought a Deskstar T7K250 PATA, connected it to the
Ultra100TX2, moved everything over, and what can I say? The new drive
not only measures, but also *feels* faster, and on top of that, it's
dead quiet in idle and barely audible in access when decoupled, and it
gets noticeably less warm as well. It has worked flawlessly so far, and
will hopefully continue to do so for a few years. What more could one
want?

My Machine is in a distance of one meter, beneath my Ears. I prefer
super-calm Harddisks. Like one, with 5400RPM. 12ms or slower...

You wouldn't believe what good decoupling can do. I even got the Cheetah
36ES access noise (only somewhat quieter than on my old 1st-gen
Barracuda ATA) down to 2.5" levels, not to mention that the hum was
eliminated nicely as well. I prefer this white layered foam primarily
found in the packaging for devices made in Korea (in this case, my LG
DVD burner). It's soft enough to give good results for decoupling, while
being fairly stable mechanically nonetheless. Two "rails" are now
located under the sides of the T7K250, and a drive case with the Cheetah
36ES in another machine sits on two pieces of the stuff. I can hear the
Samsung SV0802N (yes, the single-platter Samsung V80, Nidec motor) in
idle, which sits in an ordinary drive cage and is not decoupled - but
not so the decoupled T7K250, which is completely inaudible over the two
slow 80 mm fans. The rig sits next to my desk and is only about 1 m away
either. The single loudest part currently is the PSU fan, which has
already been replaced in the past. Maybe some grille cutting and fan
decoupling (not easy to do here) would help, I'm reluctant to part with
the 110 W PSU since a current bigger one would probably be much less
efficient at the low loads encountered.

The Monitor takes the most power ;-(.

Yup, those old CRTs are fairly power hungry. My old 19" was measured at
105 W. My current 19" TFT is spec'd at <40 W, which probably works out
closer to 25..30 W in practice.

Stephan

 

[Robert Hancock]

One could try to circumvent the 55 MB/s write limit for disk controllers
by setting up their memory ranges for USWC, but I have no idea how one
would modify CPU MTRRs in WinNT/2k/XP/S2003 (nor what the cards may
think of this).

Write combining (USWC) is irrelevant for disk controller performance.
That only affects performance of writes from the CPU onto PCI or AGP
memory. Disk controllers use bus mastering to do their reads and writes,
the CPU is not involved.

There is no 55 MB/sec limit either, it is possible to get over 100
MB/sec in certain cases with a disk controller on a standard PCI bus.
Any limit would be due to some PCI burst limits of the controller or one
of the bridges in between the CPU and the bus.

 

[Daniel Mandic]

You wouldn't believe what good decoupling can do. I even got the
Cheetah 36ES access noise (only somewhat quieter than on my old
1st-gen Barracuda ATA) down to 2.5" levels, not to mention that the
hum was eliminated nicely as well. I prefer this white layered foam
primarily found in the packaging for devices made in Korea (in this
case, my LG DVD burner). It's soft enough to give good results for
decoupling, while being fairly stable mechanically nonetheless. Two
"rails" are now located under the sides of the T7K250, and a drive

I like to screw the HD, so strong as it may, in a free 3.5" Mounting.
Letting all vibrations to the Tower.

Iron to Iron.
Not better for the ears, of course, but much better for the mechnic,
speed and lifetime.




Kind Regards,

Daniel Mandic

 

[Daniel Mandic]

There is no 55 MB/sec limit either, it is possible to get over 100
MB/sec in certain cases with a disk controller on a standard PCI bus.
Any limit would be due to some PCI burst limits of the controller or
one of the bridges in between the CPU and the bus.

More than 100MB/sec. The PCI is specified much faster.
But as you can see on a posting from Stephan with a link. The writing
does not work over 55. Reading goes as fast as the HD Can.

I remember the 55-60MB Limit from my 486 times. The PCI is still the
same, and the limit seems not changed!?
So if your HD can Write DATA with over 55MB/sec , you should have a
PCI Express or something faster than PCI to use the full capacity.
Otherwise, no PC can load a program with 55MB/sec. Let´s say a .exe,
50MB Big. Do you think you could load this in one second? Maybe a Super
Duper SGI Reference Workstation.


As I many times said, everything arabic. Just numbers, saying nothing.
Limitations on PC Architecture are not to find in the PCI-Bus. First
and biggest limitation is the Intel or AMDCPU.



Best Regards,

Daniel Mandic

 

[Daniel Mandic]

There is no 55 MB/sec limit either, it is possible to get over 100
MB/sec in certain cases with a disk controller on a standard PCI bus.
Any limit would be due to some PCI burst limits of the controller or
one of the bridges in between the CPU and the bus.

Well, I have done some testing.

The biggest .exe I have found on my Maychine, was mame32.exe (33,7 MB).

The first load was done in 3-4 seconds. The second load (NT caching)
was in a second, maybe slightly less.
My RAM is told to make 800MB/sec. 1st level chache makes 1GB/Sec etc
etc... but I did´nt see the result.

My result is: Approximately 10MB/sec HD to RAM to CPU to RAM
Performance (The PCI could offer 50)
and cached, Approx. 40MB/Sec RAM to CPU to RAM Performance
(Here I don´t need PCI)


Bus-Mastering is fine, but does not say the CPU can sleep a while.
As I wrote above, you see the CPU is always doing the most.

I am sure you can understand that the U-100 on my HD wouldn´t help
nothing at all, to increase the results.
The Cpu is limiting.




Best Regards,

Daniel Mandic

 

[Stephan Grossklass]

Write combining (USWC) is irrelevant for disk controller performance.
That only affects performance of writes from the CPU onto PCI or AGP
memory. Disk controllers use bus mastering to do their reads and writes,
the CPU is not involved.

'K, that makes sense.

There is no 55 MB/sec limit either, it is possible to get over 100
MB/sec in certain cases with a disk controller on a standard PCI bus.
Any limit would be due to some PCI burst limits of the controller or one
of the bridges in between the CPU and the bus.

Apparently it's the i440BX PCI interface that caps write throughput
then. When c't magazine reviewed PCI GbE cards last year, they did a
comparative measurement of the fastest card (the Intel desktop thingy,
unsurprisingly) in a PII-400/P2B system to see how fast an old CPU could
push the data and achieved a throughput of - 55 MB/s (NETIO, Linux 2.4).
Now I'm not exactly sure how NICs get their data, I think it's via MMIO
(certainly not PIO) and thus probably involves busmastering as well.

Stephan

 

[Stephan Grossklass]

You wouldn't believe what good decoupling can do. I even got the
Cheetah 36ES access noise (only somewhat quieter than on my old

[snip]

I like to screw the HD, so strong as it may, in a free 3.5" Mounting.
Letting all vibrations to the Tower.

Iron to Iron.
Not better for the ears, of course, but much better for the mechnic,
speed and lifetime.

This attitude would be right for a server room where it's noisy as heck
anyway, but rather not for a home computer that you may possibly even
want to sleep not too far away from (which is required once a week
here). I have only a single drive which is picky about rigid mounting, a
Barracuda 4XL. If you hold it really tight, it'll manage its 13.6 ms
access time, but mounted inside an ordinary case it'll manage maybe 15
ms (and still is loud as heck). That one would probably love a case made
of concrete . Other drives of more recent vintage proved to be far
less demanding - the MAJ3091MP rather had a slightly *lower* access time
when decoupled, a Cheetah 36ES only loses 0.1 ms (still managing 8.7
ms), and the T7K250 0.3..0.5 ms (13.3 ms still is plenty fast enough).
That's by far not enough difference to influence performance noticeably,
certainly not in a desktop setting with fairly localized access where
firmware/cache optimization tends to have a great effect. [1] And if you
care about hard drive longevity, the most important part is to keep 'em
reasonably cool. With a 23°C room temp a few days ago, the T7K250 never
got any warmer than 41°C during sequential reading (long self test) and
was idling at 35°C (current drives tend to run fairly cool on average,
with the difference being most noticeable with Maxtors - while the DM+9
was fairly hot, the DM10 is one of the coolest ones around when idling),
the SV0802N has never claimed anything about 35°C anyway (I guess you
can safely add 4°C or somesuch to that, Samsungs are quite notorious for
underreading a bit; the sensor in the T7K250 appears to be fairly
accurate), at least since I improved case ventilation to the point that
it actually works.

Stephan

[1] If you still prefer "rigid mounting", nothing keeps you from using a
drive cage with some sort of weights attached which then in turn is
decoupled as a whole. Sure you'd need a sufficiently spacious case for
that.