PDA

View Full Version : Who has Sata drives and VT4 working rock solid?



Seti Orion
03-14-2006, 01:20 AM
I have seen a few guys in here who said they had
sata running with vt 4 solid.

I just tried 2 maxtor 300 gig 16 meg cache drives
in a software striped raid. They were showing 120 megs read and write. I was able to record for 1 hour and 25 mins and then it aborted.

My tests have shown that sata just cant cut it.

Maybe a huge sata array?

I tested my scsi array for 2 and half hours intill I ran out of space and it works flawless for 4 tests so far.
I also recorded 2 streams for close to 2 hours
before running out of space with no problems.
Scsi has been solid!

Who is really pushing vt4?
Is anyone recording 2 streams for 5 hours while playing a couple ddrs switching several camera etc?

This is the main reason I got the system

Bobt
03-14-2006, 04:52 AM
All SATA drives are not created equal.
Long time back people bought Maxtor drives
and had the same results. ALL bad.

Also I think you need 4 drives not 2 for VT.
2 Should record all the way through though.

When you run disk tests you need to run the kind that
read and write to the entire drive.

Doing what is called butterfly read write tests dont cut it.
Good for office bad for video.

Bob

Blaine Holm
03-14-2006, 09:29 AM
Maxtor drives are the worst, and 2 of them are not enough. There are other reasons for aborted captures. The biggest one I see is having the capture setting set to capture 4 channel FP audio.

John Perkins
03-14-2006, 10:41 AM
I've put 8 Seagates on a Highpoint RocketRAID 1820a and still not had the performance of 4 SCSI drives. That's the closest I've come to having a 100% working SATA solution.

The single read or write performance of SATA is amazing, but when you try to do a read and a write, or multiple simultaneous accesses (multiple streams), they just can't cope. It may work sometimes and not the next.

If you're only doing DV, or not many layers, SATA can work. The problem is that you trade cash and space for reliable playback and recording. SATA just isn't consistent, but it is plentiful and cheap.

amiga2091
03-14-2006, 10:41 AM
First of all, are these SATA I or SATA II drives? How many RPM are the drives running at? Are they running from a stripe from the motherboard or a PCI SATA raid card? Even though we are not supporting SATA drives yet, I have talked to some people who seem to be running four SATA drives off of a mother board. Since the PCI cards are sharing PCI bandwith, I wouldn't go that route. SCSI is still tried and true and that is why we stick by that standard for now.

IMPERIAL
03-14-2006, 11:39 AM
As far as I know using 2 disks will increase the bandwith, but 4 will keep that bandwith all the time. What happens is that when you fill your disks with data, the transfer goes lower and lower. I use 1 sata disk 160GB (nforce4 onboard sata2 controler) on VT3 and it works great untill is half full.
Try the test with full disks and you will see that 120MB will become less than 30-40MB.
I was working with 4 ata disks on VT2... it was working just fine.

Maxtor was good and fast.. 2-3 years ago. Best stable results I ever got was with Segate ATA, now using WesternDigital SATA1 witch is also good so far.

Bobt
03-14-2006, 08:33 PM
All depends on the slot the Raid controler is in and how many drives you have.
IDE/SATA seek performance is not that of SCSI because SCSI is spinning at 10K RPM. Now you can try the new WD 10K IDE drives but you loose the price performance edge you had.
I am not sure if Newtek tweaking their disk performance to match the Raid drives. You can get quite a bang out of IDE as well as SCSI you just have to read from them differently to get Max performance.

Bob

Eric Pratt
03-14-2006, 09:43 PM
I've been using 8 drive sata stripes for all my VT and HD systems for the last two years, and they've just gotten faster and faster as the various iterations of drives and controllers and host bus interfaces has increased from Sata I to Sata II, PCI-X to PCIe, Seagate 7200.7, then .8, then .9, each with a few more features making it more and more SCSI like.
In short, they always beat out 4 SCSI drives, cost less, and store more data.
The only drawback I've run into, is that when the drives get full, performance degrades less gracefully in SATA than SCSI.
I've been posting benchmarks on BlueRAID, which is a website for an external 8 port enclosure we design and manufacture: http://www.blueraid.com, we've sold a lot of these units as NLE storage and server storage. Haven't had a chance to update with the most recent drives though. About the only place where we have failed is HD on the Mac, because the drivers for the only 8 port internal controller for the Mac aren't as good on the Mac as they are on the PC, and the resulting data rate is just short of 1080i.
I've also tested 16 drives which doubles the data rate again and I'm hoping the next enclosure will hold 16 instead of 8.
Anyway, just thought I'd put that in, don't like to see Sata knocked around by SCSI. Now SAS... that's a whole other kettle of fish.

edmellnik
03-15-2006, 04:10 PM
I am not sure if SATA means IDE.../????
but I have been using 4- 120 gb 7200 RPM Seagate IDE drives striped
all along and they work flawlessly. Now sometimes the stutter a little
if the effects have not rendered yet...but other than that they play great.

Ed

tmon
03-15-2006, 04:25 PM
There are really two questions here. Maybe three.
We use IDE drives in our SCSI RAID arrays. We use SCSI controllers.

We are looking at using RAID arrays with SATA drives, but would probably still use some kind of SCSI controller, maybe PCI-X.

Eric Pratt
03-15-2006, 04:26 PM
No, IDE is what we had before Sata, but Sata grew out of IDE so the technology is sort of the same concept, except instead of a parallel like IDE Sata is serial, which is why there are so many less connectors. In fact 3 of the 7 pins on a sata cable are ground. SAS is the same way (serial attached SCSI).

isaacu
04-06-2006, 12:35 AM
I'm running two seperate SATA striped volumes ( 2 volumes with 2 disks each) and I find them to perform as could be expected. It is definitely faster than strait up Sata, but I do ocassionally have performance problems If I am being abusive. Now, give this a try. If you really need to play from DDRs and record at the same time use one volume for Playback and the other for Recording(Round Robbin). I do this with live sports. I run breaks off of one volume and record the game on another for instant-replay. I have an intel workstation board with 2 SATA channels and two IDE channels and aparrently the disk controllers don't gobble up the PCI(E) bus and the toaster card usually runs totally stable. I have done web-casts that last for hours with good results. Running encoder on the same machine is another story....

John Perkins
04-06-2006, 12:59 AM
That is a great solution and we should put it in the FAQ. If you need to recoed while playing back, add another drive just for recording or playback.

It isn't often you get a good cheap solution to a problem. :thumbsup:

Kevin Brice
04-07-2006, 06:44 PM
This is worrying me... I was going to get the following to upgrade our current Video Toaster:

3 Seagate Barracuda 7200.8 3.5" Serial ATA150 Hard Drive
1 HighPoint RocketRAID 1640 PCI SATA Controller Card RAID 0/1/10 JBOD

As I said a long, long while ago, I was staying away from SCSI drives because of the heat, power draw, cost, etc. We mostly do live stuff, but it would be nice to be able to use ToasterEdit for at least simple stuff... would three striped SATA drives be able to handle basic editing, DDRs during live switching, etc.? How would that compare to a single SCSI drive? (what we have now)

Also, that card is a hardware RAID card. Would it be better to use Windows 2000/XP software RAID? Is that possible with that card?

(Please keep in mind that this is a Dimension 4550, which, according to John, shares 133MBps with all PCI cards)

goodrichm
04-08-2006, 08:35 AM
Can SATA drives be chained together (like IDE)?

My Intel 7525 MOBO (Dell PWS 670) has two SATA ports, but the cables that came with the PC only allow me to have one drive hooked to each port/cable. Thanks for anyone's help...MG

Kevin Brice
04-08-2006, 09:10 AM
Can SATA drives be chained together (like IDE)?

My Intel 7525 MOBO (Dell PWS 670) has two SATA ports, but the cables that came with the PC only allow me to have one drive hooked to each port/cable. Thanks for anyone's help...MG
I don't believe so. You'll have to get a PCI controller card to use more drives.

Brian Peterson
04-09-2006, 08:44 AM
3 Seagate Barracuda 7200.8 3.5" Serial ATA150 Hard Drive
1 HighPoint RocketRAID 1640 PCI SATA Controller Card RAID 0/1/10 JBOD

Also, that card is a hardware RAID card. Would it be better to use Windows 2000/XP software RAID? Is that possible with that card?

I'm using the 1640 card and you can build a software raid with it. However I am only getting 66mbs transfer rates from the blasted thing and still haven't tracked down why.

Bobt
04-09-2006, 09:48 AM
Can SATA drives be chained together (like IDE)?

There is a board that you can connect 8 seperate SATA drives to. Cant remember what its caleed.

>However I am only getting 66mbs transfer rates from the blasted thing and >still haven't tracked down why.
I get more from 2 SATA drives. Is your drives connected via PCI?
Is the PCI shared with VT? Maybe thats why.
I am so confused this year as to what to get or do. I cant even dream of a system I want anymore. Conroe is coming when? I dont know and I am not sure its any good. M2 for AMD is coming. Will these systems offer more PCI -e slots or will they all offer cheap bandwith limiting PCI slots. All I need is ONE PCI slot now. I want everything else to be 8 to 16 channel PCI-e.
or I will take PCI-x as well. As far as hard drives I am just as confused. IDE /SATA just do NOT seek as fast as SCSI and drive manufacturers are protecting SCSI till the end. :( Western Digital wants such a premium for 10K ide that it makes no sense.
If you think I sound frustrated your RIGHT !
MAN what a MESS!!

Bob

Blaine Holm
04-09-2006, 10:04 AM
That is a great solution and we should put it in the FAQ. If you need to recoed while playing back, add another drive just for recording or playback.

It isn't often you get a good cheap solution to a problem. :thumbsup:

It's something we usually do, even with SCSI. The newest generations of SCSI drives can't play back and record as well as they used too, especially if you do not tweak any of the cache settings of the drive. It makes much more sense to dedicate a single drive for that purpose.

Brian, what happens if you shutdown winrtme and then run the disk test?

In our VT systems we get 220-250MBís per second with 4 SATA Seagate drives, and the Super Micro MV-8 PCI-X controller. :thumbsup:

John Perkins
04-09-2006, 10:48 AM
There are hardware RAID controllers and there are what I call "firmware" RAID controllers (I don't know the proper name). A true hardware RAID controller has a CPU of some kind onboard. The others will have enough of a firmware onboard to tell your CPU how to read the drives.

In almost every case, if you pay less than $200, you bought a firmware RAID controller and are using your CPU to do the work. If the company knows what it's doing, this is fine. If not, you can get worse performance than a single SATA drive.

This isn't to say that a hardware RAID controller is always better. Like anything else, it depends on how well done the design is. Most hardware RAID controllers are built for server use. Servers almost never need high bandwidth streaming transfers from disk like we do for video. They are usually sending small chunks of random data over a slow network connection. Unless you need the safety of redundancy, the extra cost of hardware RAID may not give a good return on investment.

One of the things Windows does really well is software striping disks. They have a really low overhead RAID0 implimentation. I can't say that I've ever seen a controller card that could run any faster than letting Windows handle the striping. Buying a simple, low cost SATA or SCSI controller and striping the drives in Disk Management will normally give very good results if the drive controller itself is fast.

Another factor to look for is what bus the controller is on. A normal 32bit PCI card can theoretically transfer up to 133MB per second. You will never get above this during a benchmark with a 32bit PCI drive controller, no matter how fast the drives are. In addition, that bandwidth is shared with every 32bit PCI device on the system. When you add in a firewire card, IDE controller, VT card, etc you have a fraction of that left over for your RAID controller.

PCI-X (64bit slots) and PCI-Express have many times more bandwidth and are not sharing with the PCI bus.

PCI-133MBps
PCI-X 66Mhz - 533 MBps
PCI-X 100MHz - 800 MBps
PCI-X 133MHz - 1064 MBps
PCI-E 1x - 250MBps
PCI-E 2x - 533MBps
PCI-E 4x - 1066MBps
PCI-E 8x - 2133MBps
PCI-E 16x - 4000MBps

From this, you can see that you will have a hard time saturating all but the slowest of these bus types with a RAID controller or even two.

Kevin Brice
04-09-2006, 10:52 AM
It's something we usually do, even with SCSI. The newest generations of SCSI drives can't play back and record as well as they used too, especially if you do not tweak any of the cache settings of the drive. It makes much more sense to dedicate a single drive for that purpose.

Brian, what happens if you shutdown winrtme and then run the disk test?

In our VT systems we get 220-250MB’s per second with 4 SATA Seagate drives, and the Super Micro MV-8 PCI-X controller. :thumbsup:
I think the fact that your controller is PCI-X is the key. The RocketRaid 1640 is a regular PCI card, which means it has to share the bus with the VT card.

Does anybody know how much bandwidth the PCI VT card acutally uses out of the theoretically available 133MBps (when using a BoB)?

Bobt
04-09-2006, 11:14 AM
PCI-133MBps

These are great stats.
PCI is limited though. You are lucky if you get
100MB/sec out of it. Most boards come in at about 90ish.
REAL bad when you need the band width.
Wonder what the reality of PCI-e is? Does it sustain those numbers.
Are the numbers real or just max performance ?
Anyone do any real hard core tests? I know that PCI-X has held up
its bargin by the amount of incredible numbers I have seen with disk I/O.

Anyone have a PCI-e 4x SATA controler yet? Can it sustain big numbers.
What the latency like? Man the things I just dont know..

Bob

Blaine Holm
04-09-2006, 11:36 AM
. I can't say that I've ever seen a controller card that could run any faster than letting Windows handle the striping. Buying a simple, low cost SATA or SCSI controller and striping the drives in Disk Management will normally give very good results if the drive controller itself is fast.



That's very true, I've never seen it either. Another reason to software stripe is that it is possible to recover data from a stripe if a drive goes offline, or fails (but can still be read from), it's pretty much impossible to recover from a hardware RAID based stripe.

We've got a new RAID-3 array that has a SCSI interface, but uses SATA-II drives internally.

Check it out here:

http://www.videohardware.com/html/mediaguard_raid-3.html

John Perkins
04-09-2006, 11:44 AM
Here is a really good article explaining PCI/PCI-X/PCI-E
http://arstechnica.com/articles/paedia/hardware/pcie.ars/1

From the article:
The latest version of the PCI-X spec (PCI-X 266) also double-pumps the bus, so that data is transmitted on the rising and falling edges of the clock. While this improves PCI-X's peak theoretical bandwidth, its real-world sustained bandwidth gains are more modest.

and also:

By centralizing the traffic-routing and resource-management functions in a single unit, PCIe also enables another important and long overdue next-generation function: quality of service (QoS). PCIe's switch can prioritize packets, so that real-time streaming packets (i.e., a video stream or an audio stream) can take priority over packets that aren't as time critical. This should mean fewer dropped frames in your first-person shooter and lower audio latency in your digital recording software.


Here is another on understanding bandwidth and latency
http://arstechnica.com/paedia/b/bandwidth-latency/bandwidth-latency-1.html

Very technical, but well written.

Bobt
04-09-2006, 11:55 AM
This should mean fewer dropped frames in your first-person shooter and lower audio latency in your digital recording software.

Yeah but this can also mess up everything as well. AI has never been a real
strong point for Software. What I mean it never seems to do what you really want.

Bob

Richvideo
04-10-2006, 09:19 AM
I think that I have a problem with my 4ch Raidcore card and I am changing it out with a replacement Raidcore card on Monday.
My problem that I am having is minor file coruption
Very slow write speeds (Write speed is very fast )
I changed the drives from Hitachi deskstars to WD drives (250GB 2500SD) and it did not help so now I am changing the card.
I am
Running WinXP (Service Pack2)
on a
Supermicro X5DAL-G
I have another system with the same Raidcore card running WD drives and that system has much faster write times.That MB is X5DAL-TG. I have the most recent Firmware updates on both systems for the cards.
I have the card in the 133mhz slot.
Raidcore had me lower the speed to 100mhz and that did stop most of the corruption errors.
My question is about formatting the RAID to get the best speed performance. I just wanted to get some advice from someone who has some experience with this card and MB.
Thanks for the help in advance
Rich

Brian Peterson
04-10-2006, 02:51 PM
Blaine thanks for the suggestions. Killing WINRTME helped.

After much fiddeling with the drives. JBOD and using single drives only gave a 50mbs performance. I stuck the card in the 64 bit picx slot the VT card resided in and droped the VT down one to gain 110mbs from the stripped raid... But the VT card is putting off a tremendous amount of heat (what the heck?) and I watched performance drop quickly as the Rocket card heated up...

So solution was I moved the card to the first pci slot on the board and got 96mbs out of the sucker. That dropped my ide 3ware raid card down 1 slot and I'm not half way happy with performance.

Bizzare result of all this, my ide raid is now giving me 115mbs performance, up from 86 which it has long been running at. Go figure. :-(

Blaine, like the raid case, nice.

John Perkins
04-10-2006, 03:12 PM
I normally test without killing winrtme. Winrtme will be running when VT is open, so it's more realistic.

During autoconfig, the "warning, winrtme is running, may cause lower performance, blah, blah" is an error. We actually start winrtme so that we get better real world numbers. (and then complain about it)

If you get unusable numbers with winrtme running, then you have a conflict between the disk controller and the VT card and changing slots may help.

fboulene
04-11-2006, 03:26 PM
I am pretty happy with a 3ware 9550SX and 4 250gb Sata II HItachi/IBM rig.
This week I just finished a 1.40 h program. For the first time I did the whole thing in real time. Zillions of cuts. DVEs.Audio clips all over the place. Splines for every shots. Audio corrections everywhere. Start and end credits made of 4 and 5 layers. Everything you can think of. The whole thing was sent to the DigiBeta in RT with additionnal corrections and DVE on the Switcher without a glitch. Green light all the way, I just have to wait at the beginning 4 or 5 mn before it comes on. I think this is a pretty good setup for the money. I 've been using it for almost 6 month now and I think it is pretty rock solid.