• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

WARNING: SATA Cautions

ICE Czar-

I'm a Video Support Tech for www.boxxtech.com and we deal in large SATA RAID Arrays. I think that I can confirm this issue. I had an HD system that requires 142MB/s of sustained bandwidth working fine on prebuild, getting close to 400MB/s on writes. I sent the system to production to get cleaned up, and they bundled the SATA wires together. Now it gets like 250MB down to 140MB/s of bandwidth. I've had to do some other workarounds to get the bandwidth back up to around 230MB/s, but they're work arounds that I don't like.

I'll do some more testing on this.

We cut open some cables and they did indeed have dual independent shielding, but it may be inadequate.

There are also major issues with any RAID card using XOR technology, including 4K byte align issues.
 
I am 90% sure that i can confirm this. Unbundling the 10 sata cables and moving them away from the power cables/molex connectors seems to give me sustained 400MB/s on dual RAIDCore controllers with WD 250GB drives. Moving the power cables and holding them close to the SATA cables seems to give me a 100MB/s swing (300MB/s to 400MB/s sustained). This has been escelated up to our CTO and we are doing some pretty vigorous testing.

I am documenting this using IOMETER. If you'd like the IOMETER config file to use my settings, let me know.

In our case it's a two-fold issue. RAIDCore BC4000's (12152) controllers running firmware 1.3/drivers 1.3 do not play well on the same bus. We were running on a Supermicro X6DAE with the two controllers on the PCI-X 66/100 bus (Bus 3). Moving one controller to Bus 4 (PCI-X 133) also helps the bandwidth issue. But that means that one card will have to share a bus with the Bluefish HD|Fury card, a PCI-X 66Mhz card. I need to do further testing to see how this affects our bandwidth, but that is immaterial to the SATA drive/cabling issue.
 
<Homer> Bluefish ..drool</Homer>

Damn I love real numbers, Thanx Morely ;)
Id love that IOMETER config file, Im PMing my email
or if there is somewhere I can fetch it from ;)
 
i have 2 raptors on the way from new egg right now and this thread has me mildly worried. i have never had any issues with any of my other sata drives but i have never had to run 2 so close to each other. I'm hoping my case (LianLi v-1000) isn't gonna screw me because the sata cables have a long way to go. in other words i have no way to use shorter cables.
 
OK, I've been in RAID hell for the past 50 work hours. I'll update this thread with my findings, and a conference call we had with a WD field engineer..

Long story short, we're all barking up the wrong tree...
 
I'm about to pull the trigger on two Seagate 7200.7 or 7200.8 drives of 160-200GB size. I don't get this...is shielding such a huge issue here? I was planning to bundle the cables together. Now you've got me worried that I'm going to need to wrap them in aluminum foil and then sleeve them (to prevent contact between the foil and anything else in my system) to ensure things are okay. I think that if cross-talk was such an issue, wouldn't there be a lot more SATA problems out there?

Of course, I have found that Tripp-Lite makes EMI/RFI shielded 39" SATA cables for not a lot of money (about $4 apiece plus shipping), and the cables I already have have IEEE standards stamped on them, so I'd guess they are already shielded...right?
 
Of course, I have found that Tripp-Lite makes EMI/RFI shielded 39" SATA cables for not a lot of money (about $4 apiece plus shipping), and the cables I already have have IEEE standards stamped on them, so I'd guess they are already shielded...right?
Click to expand...

linkage?

Still going to hold off the panic train until Morley reports back.
 
well Ive said this is something to be aware of,
though not actively concerned about unless your experiencing corruption,
many if not most SATA cables are now sheilded

the performance drop that Morley experienced in the RAID array above, now seems to lay elsewhere according to the cryptic update, the Tripp-Lite cables seem quite long to me, and Id avoid them unless you actually need that length, the IEEE shielded cables should suffice, but doing some testing wouldnt be a bad Idea before you start transfering critical data around enmass

you can take a number of steps for "insurance" or to troubleshoot actual corruption
generally rerouting of power cables, and data cables would be enough
(though sheilding of either doesnt hurt)
consider that the EM Fields put off from any given source, are going to obey the inverse square law
where in twice the distance quarters the field strength

the other concern being running along ground planes with unsheilded cables

the SATA protocol is pretty robust, and errors rarely get through due to CRC,
(again I assume the issues at ATA|ATAPI.com are somewhat dated and that they are no longer "typical" )
but small performance increases are possible, and "insurance" with a bit of tweaking
much like you tweak your security or OS ;)
 
I guess the cliff notes would be that having a "neatly" routed case, might have some penalties if your bundling and hiding cables (power and data) and that testing might benefit you, and further, that if actual corruption is occuring, cables are a factor that has historically been underated
 
Ok guys, lemme give a quick update.

Ice Czar is right, the SATA protocal is very robust, but grounding is 'optional', and that is what that linked website may be referring to.

However, we build RAID arrays with 10-12 drives placed close together and 10 wires being bundled together. We capture large, multi-gig high def files and experience no data corruption.

Those of you looking to purchase a few SATA drives should not be worried if you are using high quality cables.

What we HAVE identified as a definite issue with some SATA drives is their succeptability to vibrations.

Running IOMETER on a RAID 5 array in full tilt, and then even lightly tapping on the chassis, will result in an up to 80%! reduction in bandwidth.

We have tested this in mobile racks, and mounted directly in the chassis.

A teleconference with a WD field engineer confirmed that there is only vibration compensation (via an accelerometer) in the Raptor II (74GB drives) of any manufacture date, and Raptor I (36GB) drives manufactured starting this month.

I have confirmed this with 3 Raptor II's in RAID 5, they are very robust and it takes a lot of heavy tapping to make them drop.

The upcoming 400GB Caviar will also have that technology, as will the RAID edition drives in the near future. Hitachi SATA drives also have this technology.

We are getting 12 RAID edition drives to eval in tomorrow. According to WD, they go through 40 hours vs. 16 hours of testing, and their internal components are held to a tighter engineering spec. They are meant for 24/7 heavy load use, something that desktop drives from any manufacturer are not meant to do.

This is good for us because a typical RAID 5 array of 10 drives will ALWAYS have 1 or 2 bad drives if we're using WD 250GB or 120GB Caviar drives.

We do not see this issue with Hitachi. It is my opinion that Hitachi manufactures a more robust, stable drive. I highly recommend that for storage arrays, everyone dumps WD for about 6 months until they start implementing some of the technology that Hitatchi already has.

I will post back with my findings with the RAID edition drives as soon as I can.

So for now, don't worry about shielding, what I thought was the cables knocking down the bandwidth was really my jarring the drives and making the heads recalibrate.
 
:eek:
...wow...
oki so now I know I wasnt just being paranoid when I always cushioned my drives via any means i could ^^;
 
not at all, recalibaration for a mild vibration
and a head crash if the tap is just a bit more severe

with the flyheight of a head at all of 2 microns it doesnt take alot, even with vastly improved HDD rigidity design
Head Slap
Head slap is when the actuator arm/read-write assembly impacts the platter due to a shock such as a tip-over, a tap with a screwdriver, or an overly aggressive shove to get the drive into a bay. This can also occur if you try to scoot the box across your desk while it is running.

As a result of the impact, tiny indentations can be formed. The material ejected from this impact is scattered about the disc, and when the drive is powered up the heads will pass over this indentation and the ejected material. This can be the equivalent of running over a bowling ball in a go-cart traveling at Mach 813.
Click to expand...

thats for a 15k Cheetah but its the same difference :eek:
luckily when the drive is unpowered the heads are typically raised and locked
or parked, its why I dont recommend 3.5" form factor "mobile" HDDs (externals) without cautioning people
I literally treat my HDDs like they are full of Nitro
 
basically...

hd_cushion.jpg

this is a 200GB brand new (2 days out from the egg)
on top of a 2 year old 120GB SE. that green is a soda
bottle 2 liter cut up. I am going to redo this soon so
they are hanging from the top...use the same 2 liter
material...
see more at http://jaqiejean.mine.nu/pics/server/

hd_cushion.jpg

this is main system. I am going to swap the two... that
is metal, cheap steel, bent in a < > with tabs for the screws...
the top is a compaq OEM WD 20GB 2MB cache 7200 RPM
in my main rig (fileserver hosts all the big files this runs the
os and game files) bottom drive is a 4GB, I need another
20GB to run games from...
see more at http://jaqiejean.mine.nu/pics/demia2/​

Yes, all WD (cept the seagate 4GB), I am a WD fangirl.
 
I have a 74GB Raptor II and a 120GB WD SE in my main rig at home. No complaints here.

However, for large RAID arrays, I am 100% convinced that today's WD Caviar drives are inadequate.
 
typhoon43 said:
linkage?

Still going to hold off the panic train until Morley reports back.
Click to expand...

Here for one...Note, Tripp-Lite makes shorter shielded SATA cables than 39", but if I want to hidden-route them I may need them this long (I need to so some measuring first)

http://www.provantage.com/buy-7trpa...ite-cables-connectivity-p940-39i-shopping.htm

EDIT: Morley, what is your opinion on Seagate's 7200.7 and 7200.8 SATA drives? I'm looking into these because of the reduced noise and Seagate claims they are quite robust; and they do have the five-year warranties. Plus, their performance so far has been rated highly and they have native NCQ (although I currently won't be using that, it's always nice to plan ahead).
My Antec case's removable drive trays mount the drives from the bottom mounts rather than the sides, and the drives are mounted on rubber grommets to reduce vibration. Until now, I've been using WD Caviar SE's (PATA) because I've had good performance and reliability with them, but the specs on the Seagates have tempted me.
 
THANK GOD i read this...i never realized there such issues assocaited with SATA. I guess i can count out setting up a 3TB SATA system.
 
killerasp said:
THANK GOD i read this...i never realized there such issues assocaited with SATA. I guess i can count out setting up a 3TB SATA system.
Click to expand...

PATA is worse, especially with flat cables

SCSI offers advantages in that its terminated, SCSI RAID arrays that you buy in that SCSI is actually designed to support hotswap and that an EE has gone through the appliance to verify its signal quality, and that SCSI HDDs have been developed with 24\7 as the primary use

I would point out that the initial releases of SATA drives didnt fair to well in comparision to SCSI enterprise drives in UICompE02's torture chamber (SAS development) suffering a substantially higher failure rate, but then he was tryin to bust things up
(through heavy access not environmental stress)

Morley said:
However, for large RAID arrays, I am 100% convinced that today's WD Caviar drives are inadequate.
Click to expand...

then its a good thing I run yesteryears WD Caviars in my PATA RAID 5 array :p
they are past their three year warranty now, but still goin strong, seems the WD400BBs where a robust design
as its not a performance array its not benchmarked much (SX6000 w\ 128MB PC133)
 
LoneWolf said:
Here for one...Note, Tripp-Lite makes shorter shielded SATA cables than 39", but if I want to hidden-route them I may need them this long (I need to so some measuring first)

http://www.provantage.com/buy-7trpa15v-39in-serial-ata-sata-signal-cable-7pin-lifetime-warranty-tripp-lite-cables-connectivity-p940-39i-shopping.htm

EDIT: Morley, what is your opinion on Seagate's 7200.7 and 7200.8 SATA drives? I'm looking into these because of the reduced noise and Seagate claims they are quite robust; and they do have the five-year warranties. Plus, their performance so far has been rated highly and they have native NCQ (although I currently won't be using that, it's always nice to plan ahead).
My Antec case's removable drive trays mount the drives from the bottom mounts rather than the sides, and the drives are mounted on rubber grommets to reduce vibration. Until now, I've been using WD Caviar SE's (PATA) because I've had good performance and reliability with them, but the specs on the Seagates have tempted me.
Click to expand...
We don't even have any Seagate SATA drives in house. But, if I were to buy a large disk for data today I'd buy either a Hitachi or Seagate 7200.8.
 
killerasp said:
THANK GOD i read this...i never realized there such issues assocaited with SATA. I guess i can count out setting up a 3TB SATA system.
Click to expand...
You'd be safe if you went with Hitachi 400GB drives to make that array. It also depends on what you're going to use it for. We use arrays in a full-tilt environment, massive amounts of data transfer. 2K playback can be upwards of 300MB/s.

As it stands, most of our video arrays are used for hi-def 720p or 1080i capture. That's 142-156MB/s sustained bandwidth. The arrays are VERY stable at that data rate. It's when we run IOMETER and it's pegging 400-450MB/s sustained writes that we're seeing issues with vibration. But not so much with Raptor II's or Hitachi drives. If Seagates have some sort of rotary acceleration feed forward implemented in them, you'll see similar performance in regards to vibration control.

I believe that at these arrays will act wholly different at different rates of transfer.
 
yeh there is a huge difference between what a heavilly accessed multiuser environment places on an array (the IOMETER pattern) and even the heavy throughput a single workstation can manage even with High Def graphics

If youve simply got a mildly accessed storage warehouse or fileserver there isnt the need to go overboard

Im hard pressed to generate any substantial traffic on my arrays regardless of how much multitasking Im doing throughout my LAN
but then right now Im not doing realtime video editing either
 
I'm terribly sorry for pointing this out, but this is the first time that I've seen someone consistently misspell 'Hitachi' :)
 
I've got 3 SATA drives (the 2 Raptors and a Maxtor) hooked up using 10" cables I bought...2 of them are wire tied together and I haven't experienced anything out of the ordinary. One thing did get my attention (although I never re-tested it) was the last time I had the Maxtor unplugged I ran HDtach on the Raptors, and then ran it again with it connected, and the graph the 2nd time was a mess.
 
Elledan said:
I'm terribly sorry for pointing this out, but this is the first time that I've seen someone consistently misspell 'Hitachi' :)
Click to expand...
You're right. Didn't flow right off the fingers, just wasn't thinking. ;) I'm nothing if not consistent, however. :)
 
While reading this a thought crossed my mind. (It does happen from time to time) Has anyone run an array with the drives mounted in anything but the typical horizontal mounting position?

Over the years there have been many arguments about drive life, vibration differences etc due to various mounting configurations. Computer cases for the most part seem to suggest (force?) the horizontal mode, but I have seen quite a few companies go out of their way to use vertical, as in top to bottom as well as laying the disk on it’s side.

Just wanted to throw this in for some independent thought(s).
 
My DotHill SANnets are all vertical, and Im currently running my IDE array vertical as well
the "party line" is that it doesnt really matter
but considering the level of access, array density and speed Morley is talking about,
I dont know
 
BillR said:
While reading this a thought crossed my mind. (It does happen from time to time) Has anyone run an array with the drives mounted in anything but the typical horizontal mounting position?

Over the years there have been many arguments about drive life, vibration differences etc due to various mounting configurations. Computer cases for the most part seem to suggest (force?) the horizontal mode, but I have seen quite a few companies go out of their way to use vertical, as in top to bottom as well as laying the disk on it’s side.

Just wanted to throw this in for some independent thought(s).
Click to expand...
In the chassis we sell they are mounted horizontally, but typically get a lot of testing done vertically.

HITACHI

;)

There, that felt better. :)
 
I have never ran SATA and basically until I am forced to I think imma stick with PATA. I see no reeason to go SATA for my own personal use.

you see the pics of my boxen...

the server will acquire some more of those WD 2000BB's as I get the funds (a whole $103.50 for a fairly fast 200GB new drive at newegg is :eek::cool::D )
For my use, it fits the bill perfectly. Im still running 100Mbps network.

Anyways, back on topic: this is SATA caution thread. question: does PATA seem to have same / similar problems in JBOD mode such as I have? in a situation like I will be running soon (3x WD2000BB + 1x WD1200JB) WD high capacity PATA drives stacked loosely [about 3/4 inches space between each]) will I be seeing much performance loss in THE REAL WORLD?
 
LadyJaqie said:
I have never ran SATA and basically until I am forced to I think imma stick with PATA. I see no reeason to go SATA for my own personal use.

you see the pics of my boxen...

the server will acquire some more of those WD 2000BB's as I get the funds (a whole $103.50 for a fairly fast 200GB new drive at newegg is :eek::cool::D )
For my use, it fits the bill perfectly. Im still running 100Mbps network.

Anyways, back on topic: this is SATA caution thread. question: does PATA seem to have same / similar problems in JBOD mode such as I have? in a situation like I will be running soon (3x WD2000BB + 1x WD1200JB) WD high capacity PATA drives stacked loosely [about 3/4 inches space between each]) will I be seeing much performance loss in THE REAL WORLD?
Click to expand...

I don't have that answer but if you type "JOBD mode" into google it brings up a bunch of interesting responses. I almost wish you hadn’t brought it up now ;)
 
LadyJaqie said:
I have never ran SATA and basically until I am forced to I think imma stick with PATA. I see no reeason to go SATA for my own personal use.

you see the pics of my boxen...

the server will acquire some more of those WD 2000BB's as I get the funds (a whole $103.50 for a fairly fast 200GB new drive at newegg is :eek::cool::D )
For my use, it fits the bill perfectly. Im still running 100Mbps network.

Anyways, back on topic: this is SATA caution thread. question: does PATA seem to have same / similar problems in JBOD mode such as I have? in a situation like I will be running soon (3x WD2000BB + 1x WD1200JB) WD high capacity PATA drives stacked loosely [about 3/4 inches space between each]) will I be seeing much performance loss in THE REAL WORLD?
Click to expand...
NCQ and TCQ are enough to get me to stay SATA, and we've seen data corruption with flat IDE cables too. No technology is perfect, but the world is moving to SATA. And that's a good thing.
 
now I have to go wash the drool off my keyboard :p
unfortunately my Video Capture plans have suffered a serious setback


I do have to say I perfer having two external SCSI cables running out the back of my 4U to all those cables running forward
(my PATA internal array is a mess luckily its not the same box being the KR7A)

damn neat power harness to all those HDDs, really like the right angle connectors perfectly spaced ;)

I see your using the OEM heatsinks, those damn things piss me off or more accurately the orientation of the 940s running across the case airflow do, like that Iwill (I assume) my K8W dumps the output into one of my RAM banks

Ive been contemplating what it would take to build a set of Hyper6's that oriented to a passthrough duct scheme at a 90 degree change from the current orientation
force fed from my midcase fans and directly ducted out of the case
 
Yeah, we make a lot of cables in house. I really need to get a shot of a system that has been buttoned up by Production...I have never seen cleaner and more complicated cable routing!

And as to cooling, we have 2 variable speed 92mm's in the back keeping the system cool, along with 3 92's bringing air in from the front over the drives...

And honestly, if you don't mind the lack of an SD downconverter, the Blackmagic Decklink Pro HD is just as good for about $2K.

Of course, the big thing is Cineform's Prosepct and Aspect HD, as the world of DV is quickly moving to HDV.

Check out our Movie In A BOXX in January, you'll like it. ;)
 
holy bandwidth sucking mammoth images batman! :eek:
im on 1.5/1.5 but still those things are an eyesore :(
still that array looks killer :cool:

yah I like the specs on sata much better, but for my income ($550/mo plus $70 FS for 2 people to live on) I have to be budget concious...big time. that's why the BB and PATA.
 
been using a sata WD120GB and a sata 36GB raptor next to each other since march 04, no problems to account for. my case has been open for a while and i frequently use my cell phone around it, still no problems. hopefully it stays that way. very interesting thread though, has put a bit of worry in the back of my mind, but hopefully ill be ok.
 
@Morley: for a moment I tried to imagine what that array would look like with PATA, with tens of flat cables crawling over and under each other, then cringed and pushed the image to the back of my mind :D

Aside from this data corruption scare and lousy connectors which tend to come loose with little effort, SATA is truly an improvement over PATA :)
 
Elledan said:
@Morley: for a moment I tried to imagine what that array would look like with PATA, with tens of flat cables crawling over and under each other, then cringed and pushed the image to the back of my mind :D

Aside from this data corruption scare and lousy connectors which tend to come loose with little effort, SATA is truly an improvement over PATA :)
Click to expand...
On Monday I will take a picture of our old Cineboxx 5U system with a 10 drive PATA RAID 5 array with dual NetCell controllers...
 
Morley said:
On Monday I will take a picture of our old Cineboxx 5U system with a 10 drive PATA RAID 5 array with dual NetCell controllers...
Click to expand...
Something tells me that it's 5U because otherwise there'd be no airflow whatsoever ;)

And yes, please do post pictures :)
 
LadyJaqie said:
quality rounded cables anyone?
Click to expand...
"Parse Error: no match found for 'quality' in 'rounded cables'"

--that or cable oragami, like my box?
http://jaqiejean.mine.nu/pics/server/
(yes yes I know not ten...yet...)
I do plan on adding about 8 total to it eventually.
Click to expand...
That case looks pretty much empty :D

Speaking as someone who has once assembled a system containing two CD-ROM drives, one FDD, and four HDDs, I can state with certainty that 'cable origami' is the only way to go with PATA. Rounded cables, even if you pretend that they're actually not a threat to your data's integrity, are actually far harder to handle, because they're less easy to route.
 
You must log in or register to reply here.
Back
Top