Areca 1880ix-12 issues on Mac Pro 1,1 :(

[Squishy Tia]

As per Monoprice, they do carry what you need here for under $10 (type in "breakout cable"). Cheapest I've ever seen a MiniSAS breakout cable (usually see them for ~$30)

The one you linked is the SFF-8087 breakout cable. I need an SFF-8088 -> 4x eSATA breakout cable (external). That they do not carry and I must purchase elsewhere (might as well be from OWC since I don't get charged tax from them).

Unfortunately, I suspect this issue is with your MP, not the cards themselves.

One thing I can think of (if you have the time and inclination), is the Slot Configuration Utility isn't working properly in newer versions of OS X, so you could try moving back to the original version of OS X, and run that version of the utility. Then see if it works. If so, then upgrade the OS to a newer version, and see if it still works.

Not sure if this will do what you need, but it's something you can try.

Fraid that ain't happening - the current video card I have (and my fallback) both require 10.6.4 + Snow Leopard graphics update minimum for acceleration, and will function in 10.6.0-10.6.4 vanilla, as well as 10.5.8 only, as a 2D unaccelerated raster device with no resolution switching support. Earlier (10.5.7 and below) KPs on boot with these two video cards, a 5870 and 5770 respectively. The original X1900XT has been donated to a family member out of town already.

We'll see if it was my system or if it was the card (Areca) I had before that was the problem. If this card works at x8, then the Areca had either an incompatibility with the 1,1 Mac Pro (that particular card at least) or was defective from the getgo. The only thing I know for sure at this point regarding the Areca card is that it was not one of the initial batches based on two things: It had the new unified heatsink, which beats the everloving crap out of the dual postage stamp heatsinks for dissipating heat, and it didn't make my Mac Pro not boot when installed. The original batches had incorrect capacitance and would cause the MPs to fail to boot at all.

I'm keeping my fingers crossed. $8 shipping, no tax and the discounted price gets the card here on Tuesday. If it works, then I have a nice solution for up to four 6 Gbps SSDs. If not, I send it back and eat a restock fee. Since the price is $279, I eat about $35. I can live with that. Either way, this card will tell me if it was a fluke card (Areca) or my system just isn't gonna take a RAID card and I should use it as a hand me down to my grandparents after getting an SNB Mac Pro later this year/early next year.

 

[nanofrog]

The one you linked is the SFF-8087 breakout cable. I need an SFF-8088 -> 4x eSATA breakout cable (external). That they do not carry and I must purchase elsewhere (might as well be from OWC since I don't get charged tax from them).

Ah, OK.

Unfortunately, those are getting harder to find and thusly more expensive. Highpoint still sells them AFAIK, so you might find it by searching that brand (Highpoint EXT-MS-1MES; newegg has it here, but it's $60 + s/h).

Found one cheaper here, but I'm not sure of the actual build quality (looks like standard internal, non-latching SAS/SATA ends).

We'll see if it was my system or if it was the card (Areca) I had before that was the problem. If this card works at x8, then the Areca had either an incompatibility with the 1,1 Mac Pro (that particular card at least) or was defective from the getgo. The only thing I know for sure at this point regarding the Areca card is that it was not one of the initial batches based on two things: It had the new unified heatsink, which beats the everloving crap out of the dual postage stamp heatsinks for dissipating heat, and it didn't make my Mac Pro not boot when installed. The original batches had incorrect capacitance and would cause the MPs to fail to boot at all.

I think in this case, it was a combination:

1. Slot Configuration Utility isn't actually re-configuring the slots as you set them.
2. Areca you got was bad.

As pet the heatsinks, larger isn't always better, even if it's the same extrusion profile (has to do with the thermodynamics of natural convection; surface area is only good to a point - exceed it, and the additional surface area is useless).

 

[Squishy Tia]

As pet the heatsinks, larger isn't always better, even if it's the same extrusion profile (has to do with the thermodynamics of natural convection; surface area is only good to a point - exceed it, and the additional surface area is useless).

Except that the heatsink is (I believe) dissipating heat from a few other smaller components on the board as well. Not to mention those PPC440/450s run hot. Any added surface area does good with them. Anybody that's ever had a MDD G4 or Quadnostril G4 knows about these puppies - the only thing that ran hotter than them at their time was the Athlons.

Unfortunately, those are getting harder to find and thusly more expensive. Highpoint still sells them AFAIK, so you might find it by searching that brand (Highpoint EXT-MS-1MES; newegg has it here, but it's $60 + s/h).

Found one cheaper here, but I'm not sure of the actual build quality (looks like standard internal, non-latching SAS/SATA ends).

OWC has the HighPoint ones, and since Areca does the HW for HPT (which you can see in the similarities between the 1680ix and HPT $600-$900 cards), the cables are probably pretty good quality. And no tax, very reasonable shipping rates.

 

[nanofrog]

Except that the heatsink is (I believe) dissipating heat from a few other smaller components on the board as well. Not to mention those PPC440/450s run hot. Any added surface area does good with them. Anybody that's ever had a MDD G4 or Quadnostril G4 knows about these puppies - the only thing that ran hotter than them at their time was the Athlons.

There may be other components under it, which would explain why they changed the heatsink design (i.e. voltage regs running hotter than can be tolerated without any cooling techniques).

As per surface area for PPC or any other chip, it comes down to physics (Thermodynamics). You can only go so big, before the additional surface area is totally useless for any given extrusion profile (ends up just a waste of money for the additional material). More efficient extrusion profiles are required in such instances (or better = forced air, heat pipes, ...).

If you're interested in what I'm getting at about extrusion profiles (other cooling solutions as well in terms of real products), take a look here and nose around.

OWC has the HighPoint ones, and since Areca does the HW for HPT (which you can see in the similarities between the 1680ix and HPT $600-$900 cards), the cables are probably pretty good quality. And no tax, very reasonable shipping rates.

Areca doesn't make cables. Those come from other companies such as Foxconn.

As per what they make for Highpoint, it's only some of their RAID products (35xx and up that I know of for sure, and there are design compromises from their standard products due to budget constraints set by Highpoint). Though the hardware may be fine, when you consider the support from Highpoint, I don't recommend this level of product from them unless you don't need any support at all (even updated drivers can be difficult or even impossible).

It's one of the reasons I recommend users go straight to the source (Areca), as they do have good support, and the price/performance ratio is the best out there right now for Mac users (even PC users for that matter in most, if not all cases).

 

[Squishy Tia]

Upon first boot, the NewerTech RAID card shows up as x8 in slot #4 where the Areca card failed to show up as x8. Good start. Now comes the next step: putting my SSDs into the optical bay enclosure and plugging in the SAS cable (already installed) into the RAID card and seeing how it all works and if there's any file transfer problems.

Keeping my fingers crossed!

 

[Squishy Tia]

OK, the NewerTech card (really a highpoint card, even the raid manager is made by highpoint!) does indeed work in my Mac Pro 1,1, with caveats.

Tests were done in JBOD mode (passthrough) to see what I could expect as a minimum for speed outside of a RAID. RAIDs would give greater speeds, though if these numbers are any indication, for a 1,1 Mac Pro, not much greater. See why below.

My test file:

InstallESD.dmg

3.74 GB on disk (3,743,009,943 bytes)

This is Mac OS X Lion's installer .dmg file, the one embedded into the "disposable" installer Apple so lovingly gives us by default.

My "external" destination disk:

RAM Disk (4 GB)

The RAM Disk is created with the following command in Terminal:

diskutil erasevolume HFS+ "RAM Disk" `hdiutil attach -nomount ram://8192000`

File copy from OWC Mercury Extreme Pro 3G 40 GB SSD to RAM Disk: 29.8 seconds (125 MB/sec read...yech)

File copy to OWC Mercury Extreme Pro 3G 40 GB SSD from RAM Disk: 52.75 seconds (ouch!) 67 MB/sec (this number look familiar to you?)

File copy from OCZ Vertex 3 120GB SSD to RAM Disk: 10.75 seconds (344 MB/sec)

File copy to OCZ Vertex 3 120 GB SSD from RAM Disk: 27 seconds (137 MB/sec)

File copy from Intel 8O GB Gen 2 X-25M SSD (on internal SATA2 port) to RAM Disk: 14 seconds (264 MB/sec)

File copy to Intel 80 GB Gen 2 X-25M SSD (on internal SATA2 port) from RAM Disk: 46.75 seconds (79 MB/sec)

File copy from OWC Mercury Extreme Pro 3G 40 GB SSD to OCZ Vertex 3 120 GB SSD: 33.0 seconds (112 MB/sec)

File copy from OCZ Vertex 3 120 GB SSD to OWC Mercury Extreme Pro 3G 40 GB SSD: 50 seconds (74 MB/sec)

As you can see, these are pathetic numbers. Very pathetic numbers even for JBOD mode on a 6Gbps controller.

At x8 my total bandwidth, before overhead is 2.0 GB/sec on a PCI-E 1.0 slot such as the Mac Pro 1,1 has. Unless this also cuts each port's speed from 6Gbps to 3Gbps as well (something I doubt unless HPT really cheaped out in the card's design), I should be getting twice these numbers in both read and write modes.

In fact, the Vertex 3 SSD is showing an abysmal 344 MB/sec sustained read from a single file. We're not talking random 4k files here, we're talking one singular file 3.74 GB in size. That's royally craptastic considering this is on an SF-2200 controller capable of 550 MB/sec reads and 500 MB/sec writes.

RAM Disks can saturate a PCI-E 1.0 slot when transferring data across an x8 link width (it actually hits the 2 GB/sec wall easily). And since I'm copying to only one 6Gbps connection, which is 750 MB/sec max and getting these numbers, I'm really unimpressed.

Just to make sure I wasn't going insane in the membrane, I re-ran the Vertex 3 -> RAM Disk copy test. Just a hair over 10 seconds for the copy to complete. That is about 370 MB/sec. Better than 344 MB/sec, but still nearly 180 MB/sec slower than the Vertex 3 is rated for.

Could it possibly be that the Mac Pro 1,1 just sucks that badly with RAID cards in general?

So from my experience so far with this card, it has potential, but which Mac Pro you use it in is going to make a huge difference.

 

[Squishy Tia]

A further update:

The card's vendor ID is 1103. Anybody that's used a RocketRAID card will recognize that ID instantly. The NewerTech card(s) are really just rebranded RocketRAID 2721 and 2722 cards for the Internal/External and External Only models respectively.

Along with a $30 price hike vs. what you can get the 2721 for at Newegg, you'll (hopefully) get English speaking tech support and something other than the craptacular HPT minimalist web ticket form for tech support.

Another dead giveaway that this was an HPT card in disguise: When I did the NewerTech driver install two of the install entries listed Upgrade: The RAID management software. The only way this should be an "upgrade" is if I'd installed the software before, and I did: when I had the RocketRAID card installed. I didn't uninstall because their uninstaller (HPT's) doesn't remove the RAID management software, hence the "upgrade" when I installed NewerTech's drivers.

The upshot? The auto-passthrough mode is definitely nice for JBOD users like myself that want to eventually update to RAIDs once enough SSDs are procured.

Is it worth $300? We'll find out soon enough.

 

[nanofrog]

OK, the NewerTech card (really a highpoint card, even the raid manager is made by highpoint!) does indeed work in my Mac Pro 1,1, with caveats.

Tests were done in JBOD mode (pass-through) to see what I could expect as a minimum for speed outside of a RAID. RAIDs would give greater speeds, though if these numbers are any indication, for a 1,1 Mac Pro, not much greater. See why below.

My test file:

InstallESD.dmg

3.74 GB on disk (3,743,009,943 bytes)

This is Mac OS X Lion's installer .dmg file, the one embedded into the "disposable" installer Apple so lovingly gives us by default.

My "external" destination disk:

RAM Disk (4 GB)

The RAM Disk is created with the following command in Terminal:

diskutil erasevolume HFS+ "RAM Disk" `hdiutil attach -nomount ram://8192000`

File copy from OWC Mercury Extreme Pro 3G 40 GB SSD to RAM Disk: 29.8 seconds (125 MB/sec read...yech)

File copy to OWC Mercury Extreme Pro 3G 40 GB SSD from RAM Disk: 52.75 seconds (ouch!) 67 MB/sec (this number look familiar to you?)

File copy from OCZ Vertex 3 120GB SSD to RAM Disk: 10.75 seconds (344 MB/sec)

File copy to OCZ Vertex 3 120 GB SSD from RAM Disk: 27 seconds (137 MB/sec)

File copy from Intel 8O GB Gen 2 X-25M SSD (on internal SATA2 port) to RAM Disk: 14 seconds (264 MB/sec)

File copy to Intel 80 GB Gen 2 X-25M SSD (on internal SATA2 port) from RAM Disk: 46.75 seconds (79 MB/sec)

File copy from OWC Mercury Extreme Pro 3G 40 GB SSD to OCZ Vertex 3 120 GB SSD: 33.0 seconds (112 MB/sec)

File copy from OCZ Vertex 3 120 GB SSD to OWC Mercury Extreme Pro 3G 40 GB SSD: 50 seconds (74 MB/sec)

As you can see, these are pathetic numbers. Very pathetic numbers even for JBOD mode on a 6Gbps controller.

At x8 my total bandwidth, before overhead is 2.0 GB/sec on a PCI-E 1.0 slot such as the Mac Pro 1,1 has. Unless this also cuts each port's speed from 6Gbps to 3Gbps as well (something I doubt unless HPT really cheaped out in the card's design), I should be getting twice these numbers in both read and write modes.

In fact, the Vertex 3 SSD is showing an abysmal 344 MB/sec sustained read from a single file. We're not talking random 4k files here, we're talking one singular file 3.74 GB in size. That's royally craptastic considering this is on an SF-2200 controller capable of 550 MB/sec reads and 500 MB/sec writes.

RAM Disks can saturate a PCI-E 1.0 slot when transferring data across an x8 link width (it actually hits the 2 GB/sec wall easily). And since I'm copying to only one 6Gbps connection, which is 750 MB/sec max and getting these numbers, I'm really unimpressed.

Just to make sure I wasn't going insane in the membrane, I re-ran the Vertex 3 -> RAM Disk copy test. Just a hair over 10 seconds for the copy to complete. That is about 370 MB/sec. Better than 344 MB/sec, but still nearly 180 MB/sec slower than the Vertex 3 is rated for.

Could it possibly be that the Mac Pro 1,1 just sucks that badly with RAID cards in general?

So from my experience so far with this card, it has potential, but which Mac Pro you use it in is going to make a huge difference.

SATA III tops out at ~540 - 550MB/s, but the PCIe lane configuration can reduce that, depending on the card design (what you'd find in cheap controllers, as they tend to use 1x lane per port at most).

In the case of Marvell's budget 6.0Gb/s controller (2x port, 1x lane chip), you'd only get 500MB/s max, assuming there's only one drive in operation (PCIe 2.0 lane).

In this case, I'm not sure what they did in terms of the design. It's easy to assume that they've done something similar to the above description, but given some of your performance testing has exceeded 250MB/s (PCIe Gen. 1.0 limit per lane), it doesn't appear to be quite that simple.

To get a better idea of what's going on, you'd need to test the card in a system with an 8x lane PCIe 2.0 slot and the same disks (PC or Mac won't really matter), and compare the results (should give further insight as to what's going on).

A further update:

The card's vendor ID is 1103. Anybody that's used a RocketRAID card will recognize that ID instantly. The NewerTech card(s) are really just re-branded RocketRAID 2721 and 2722 cards for the Internal/External and External Only models respectively.

Along with a $30 price hike vs. what you can get the 2721 for at Newegg, you'll (hopefully) get English speaking tech support and something other than the craptacular HPT minimalist web ticket form for tech support.

Another dead giveaway that this was an HPT card in disguise: When I did the NewerTech driver install two of the install entries listed Upgrade: The RAID management software. The only way this should be an "upgrade" is if I'd installed the software before, and I did: when I had the RocketRAID card installed. I didn't uninstall because their uninstaller (HPT's) doesn't remove the RAID management software, hence the "upgrade" when I installed NewerTech's drivers.

The upshot? The auto-pass-through mode is definitely nice for JBOD users like myself that want to eventually update to RAIDs once enough SSDs are procured.

Is it worth $300? We'll find out soon enough.

That looks too short for an ESN to me (Electronic Serial Number Manufacturer's ID Code Assignment <database>; ESN is the abbreviated version). For example, a code would look something like this: G1103PESA. Unfortunately, the database isn't currently available.

There is another way to find out the maker however, which is by searching the FCC ID code (FCC database).

Keep me posted. Unfortunately, it's not looking good IMO, going by the performance data listed above.

 

[Squishy Tia]

That looks too short for an ESN to me (Electronic Serial Number Manufacturer's ID Code Assignment <database>; ESN is the abbreviated version). For example, a code would look something like this: G1103PESA. Unfortunately, the database isn't currently available.

There is another way to find out the maker however, which is by searching the FCC ID code (FCC database).

Keep me posted. Unfortunately, it's not looking good IMO, going by the performance data listed above.

If you don't believe that this is a 2721, here's the two pages. View them side by side and look at the cards. You'll quickly see what I mean.

NewerTech card vs. RocketRAID 2721.

Viewing the images of the card side by side you can see that they're identical. Only (possibly) the firmware is slightly different. But if you need more proof, here's a pic of the info.plist from the MAXPower RAID driver:

Edit:

With regard to this paragraph:

At x8 my total bandwidth, before overhead is 2.0 GB/sec on a PCI-E 1.0 slot such as the Mac Pro 1,1 has. Unless this also cuts each port's speed from 6Gbps to 3Gbps as well (something I doubt unless HPT really cheaped out in the card's design), I should be getting twice these numbers in both read and write modes.

I was wondering if going PCI-E 1.0 vs 2.0 also caused the card to cut the available bandwidth in half for each individual SAS/SATA connection. SATA3 actually tops out at 750 MB/sec (6Gbps = 750 MB/sec). While I don't expect the Vertex 3 to run at its peak speed except under test conditions, I do expect it to run close to it on a fully open SATA3/SAS2 connection, which it clearly is not.

Also, since this card registers as x8, that leaves two possibilities with the Areca card:

1) It was faulty/defective. This would be the preferrable cause of the x4 trouble, because...

2) The SAS Expander built into the card caused issues with the PCI-E controller, and that means no Areca 1880ix series cards will work, defective or not. That would be the worst possible outcome since that means the card I crave the most will never work.

The big plus this version of the 2721/2722 has is the automatic legacy mode, meaning any preformatted disk/SSD will show up to the OS at the full speed of the controller's ports, minus RAID capability (until assigned to an array). That makes the card truly plug and play instead of plug and pray. The audible alarm being disabled by default means quiet operation for non-enterprise users.

I'm grateful though for one thing: no OS/Finder hangs on file transfers between disks. This is a HUGE relief.

 

[nanofrog]

If you don't believe that this is a 2721, here's the two pages. View them side by side and look at the cards. You'll quickly see what I mean.

NewerTech card vs. RocketRAID 2721.

Viewing the images of the card side by side you can see that they're identical. Only (possibly) the firmware is slightly different. But if you need more proof, here's a pic of the info.plist from the MAXPower RAID driver:

My mention of the ESN and FCC ID suggestion had nothing to do with not believing you or not (expect you're right). But rather trying to figure out who the ODM is, as neither OWC or Highpoint design or manufacture their own products.

Both are Marvell's Reference Design from what I can tell, so they *could* be from different ODM's. Though its more likely that it's the same ODM supplier, and they just changed the PCB solder mask and legend colors for OWC (Highpoint was too cheap to even pay for that... ).

 

[Squishy Tia]

Marvell reference design or not, the software is definitely HPT.

From the uninstall.command file:

#!/bin/sh
#driver uninstall

echo ""
echo ""
echo "This script will attempt to uninstall MAXPower SAS 6G RAID Controller Driver"
echo ""
echo ""
echo "Note: You must be logged on as an administrator to uninstall the software."
echo "  The script will prompt you for an administrator password."
echo ""
echo ""
echo "If prompted for a password please enter your administrator password."
echo ""
echo ""
RVERSION=`/usr/bin/uname -r | /usr/bin/awk -F '.' '{print $1}'`
	if [ $RVERSION -eq '9' ] ||[ $RVERSION -eq '10' ] ; then
		echo "The password cannot be "empty". If the password is empty please create a password and execute the script again."
		echo ""
	fi
echo ""
echo ""
echo "The following driver files will be deleted"
echo "		/System/Library/Extensions/MAXPower_SAS_6G_RAID.kext"
if [ $RVERSION -eq '9' ] ; then
		echo "		/Library/Receipts/maxpoweresas6graid5.pkg"
fi
if [ $RVERSION -eq '8' ] ; then
		echo "		/Library/Receipts/maxpoweresas6graid4.pkg"
fi
echo "		/Library/Receipts/webservice.pkg"
echo "		/Library/Receipts/wwwfiles.pkg"
echo "		/Library/StartupItems/Nwtsvr"
echo "		/usr/share/nwt"
sudo rm -rf /Library/Receipts/webservice.pkg
sudo rm -rf /Library/Receipts/wwwfiles.pkg
sudo rm -rf /System/Library/Extensions/MAXPower_SAS_6G_RAID.kext
sudo rm -rf /Library/StartupItems/Hptsvr
sudo rm -rf /usr/share/hpt
sudo rm -rf /Library/StartupItems/Nwtsvr
sudo rm -rf /usr/share/nwt
sudo rm -f /Applications/NewerTech\ Web\ RAID\ Management\ Utility.webloc
if [ $RVERSION -eq '10' ] ; then
	sudo rm -rf /private/var/db/receipts/com.newer-tech.*.*
	sudo rm -rf /private/var/db/receipts/com.highpoint-tech.*.*
fi
if [ $RVERSION -eq '9' ] ; then
	sudo rm -rf /Library/Receipts/maxpoweresas6graid5.pkg
fi
if [ $RVERSION -eq '8' ] ; then
	sudo rm -rf /Library/Receipts/maxpoweresas6graid4.pkg
fi
echo ""
echo "Process has completed."
exit

 

[nanofrog]

Marvell reference design or not, the software is definitely HPT.

That doesn't surprise me either (cheaper and easier for OWC to get software that's already been written whenever possible).

And HPT doesn't develop all of their software either (i.e. Areca supplies the software that HPT modifies a bit for their Areca manufactured RAID products).

In the case of this particular card, Marvell likely supplied code to HPT as well, and went from there.

 

[Squishy Tia]

Well, it's working without so much as a hiccup so far, so this may be a keeper, for now. I really wish that Areca card hadn't had an SAS expander built in - I have a feeling that's what made things wonky. It'd have been a perfect card with a 1:1 natural port ratio/connection, but meh.

This frees me up to get more HDs into my bays.

 

[nanofrog]

Well, it's working without so much as a hiccup so far, so this may be a keeper, for now. I really wish that Areca card hadn't had an SAS expander built in - I have a feeling that's what made things wonky. It'd have been a perfect card with a 1:1 natural port ratio/connection, but meh.

It's possible, and why testing everything out before-hand is so critical.

Unfortunately, this is harder to do for those without the ability to purchase gear for testing prior to implementation (separate system to test with, not a unit used in day-to-day operations), and the IT personnel to perform the testing (i.e. independents/SMB's with one system per employee, and no dedicated IT staff).

Getting back to the 1880 series... I'm not fond of SAS Expanders when they're not absolutely necessary, as they can introduce additional compatibility issues (not to mention costs). Which is why I'm not pleased with how Areca went about the 1880 series for 12+ port versions without full disclosure of what is actually going on. That sort of deception is wrong IMO in the market they're dealing with (not happy about such tactics at all, but the enterprise market takes a harsh view of products with lots of compatibility issues, as it tends to have notable financial implications).

This doesn't mean that the Areca was the issue, as I suspect rather system and disk incompatibility issues combined. But there was additional testing to prove/disprove certain things (i.e. test out in a newer MP, different brands of SSD's, such as Intel,...). More time, and potentially money to do so.

 

[Squishy Tia]

An update:

I realized (rather late) that I was using what is known as an incompressible data type. That is, the .dmg file is already compressed and can't be compressed further to enhance read/write rates.

So I then created two 2 GB dummy files (2000 MB each) and then concatenated them via Terminal (dd can't create files larger than 2 GB in size, but you can use the 'cat' command in Terminal to combine them to make one ginormous 4 GB file).

Test file: test.txt (all zero data; fully compressible - best case scenario); 4 GB (4000000000 bytes)

I then did a file copy test to my SSDs. Here's how it went:

File copy from RAM Disk to OWC Mercury Extreme Pro 3G 40 GB SSD: 17.25 seconds (231.8 MB/sec)

File copy from OWC Mercury Extreme Pro 3G 40 GB SSD to RAM Disk: 17.0 seconds (235.3 MB/sec)

File copy from RAM Disk to OCZ Vertex 3 120 GB SSD: 9.75 seconds (410.2 MB/sec)

File copy form OCZ Vertex 3 120 GB SSD to RAM Disk: 10.25 seconds (390.2 MB/sec)

File copy from RAM Disk to Intel 80 GB Gen 2 X-25M SSD on SATA2 port: 52.5 seconds (76.2 MB/sec)

File copy from Intel 80 GB Gen 2 X-25M SSD to RAM Disk: 15.8 seconds (253.2 MB/sec)

File copy from OCZ Vertex 3 120 GB SSD to OWC Mercury Extreme Pro 3G 40 GB SSD: 18.0 seconds (222.2 MB/sec)

File copy from OWC Mercury Extreme Pro 3G 40 GB SSD to OCZ Vertex 3 120 GB SSD: 18.15 seconds (220.4 MB/sec)

Now we're getting somewhere. So while the Mac Pro model used with this card will likely influence data read/write rates, what I had above (below?) for my first test run was likely a worst case scenario and I didn't realize it because I was using a file that couldn't be compressed any further, so it would be the "slowest possible case" scenario.

That being the case, this card should do very well for those looking to make a RAID, as you're going to see numbers 1.5-2x faster than either scenario respectively in RAID 0.

I'm glad I sat down and took another look at what I was doing and remembered OWC's postings regarding incompressible data

So there ya have it. A worst case scenario and a best case scenario for JBOD functionality using two commonly used (and preferred) SSDs. You can expect the OWC Electra 6G SSDs to fare on par with the OCZ Vertex 3 as they use the same firmware and SandForce controller. Expect slightly higher speeds for the OCZ MAX IOPS line and OWC's Mercury Extreme Pro 6G line as those are equivalants.

I can't wait to get my next paycheck so I can pay off my cards and get another Vertex 120 GB SSD.

NOTE: The 240 GB+ versions from OWC and OCZ will have still better speeds, as their IOPS are higher than the 120 GB variants this time around, just like what happened with the SF-1200 based SSDs from OWC/OCZ.

 

[mac-hacks]

ARECA 1880ix-8 SSD/HD Settings

NanoFrog,

I updated to beta 07/12/2011 Firmware, i notice few new settings. Could you take look and let me know what you recommend when using SSD and HD on same ARECA Controller.

1.

2.

3.

 

[nanofrog]

NanoFrog,

I updated to beta 07/12/2011 Firmware, i notice few new settings. Could you take look and let me know what you recommend when using SSD and HD on same ARECA Controller.

1. What exactly are you doing?
2. What disks are you using (make and model number)?
3. How are they configured (Pass-Through, JBOD, or what level of RAID)?
4. What other equipment is part of the storage system, and how is it configured (enclosures, mounts, UPS <make and model could be useful>, ...)?
5. And are you having any problems?

I ask, as the answers can help. Given the lack of detailed information, all I can recommend now, is try out the default values, and see if it's stable (includes testing power failure <pull the power cord>, pull disk/s to simulate disk failures, ...).

 

[mac-hacks]

1. What exactly are you doing?
2. What disks are you using (make and model number)?
3. How are they configured (Pass-Through, JBOD, or what level of RAID)?
4. What other equipment is part of the storage system, and how is it configured (enclosures, mounts, UPS <make and model could be useful>, ...)?
5. And are you having any problems?

I ask, as the answers can help. Given the lack of detailed information, all I can recommend now, is try out the default values, and see if it's stable (includes testing power failure <pull the power cord>, pull disk/s to simulate disk failures, ...).

Aerca Controller 1880ix-8
4 x 120 GB OCZ Vertex 3, RAID 0 Triple Boot OSX 10.7.1, Win7, Ubuntu 11.04
8 x 1TB SATA 3 6Gbos HD RAID 6 = 4x Internal Port & 4 x External Enclosure

Highpoint 2310 "I know, it's old but does what i need it to for now"
4 x 120GB OWC RE SATA 2 3Gbps RAID 0 "I use as my SCRATCH DISK for projects"

"Plan on upgrading & getting External Enclosure for 8 HD"

I use Parallels , Xcode, & use to do lot of editing with FCP. Use my MBP with FCP Studio for now. I Haven't made switch to X. Been little slow in department due to recession.

Im little concern on power im pulling. I use Quadro 4000 for Mac, which doesn't pull as much as ATI cards.

 

[nanofrog]

Areca Controller 1880ix-8
4 x 120 GB OCZ Vertex 3, RAID 0 Triple Boot OSX 10.7.1, Win7, Ubuntu 11.04
8 x 1TB SATA 3 6Gbos HD RAID 6 = 4x Internal Port & 4 x External Enclosure

I still don't know what the actual models of the 1TB mechanical SATA disks are (OWC info is enough to go on), or if you're having any problems.

Unfortunately, Areca doesn't publish any SSD's on their HDD Compatibility List (best to buy drives listed on this; they don't test all of them in the series, but if one is there, the others will work).

So if everything is working fine, leave the current/default settings alone.

A couple of notes:

1. RAID 0 only improves sequential throughputs, not random access performance, which is what OS's rely on.
2. Splitting the RAID 0 up into separate disks (one per OS), keeps you from losing all of the OS's in the event the RAID 0 fails (still have at least one OS available to help research a diagnosis/download drivers/purchase new disks online in the event of a failure).

Highpoint 2310 "I know, it's old but does what i need it to for now"
4 x 120GB OWC RE SATA 2 3Gbps RAID 0 "I use as my SCRATCH DISK for projects".

What about the following:

• Backup solution?
• UPS used?

 

[mac-hacks]

I still don't know what the actual models of the 1TB mechanical SATA disks are (OWC info is enough to go on), or if you're having any problems.

Unfortunately, Areca doesn't publish any SSD's on their HDD Compatibility List (best to buy drives listed on this; they don't test all of them in the series, but if one is there, the others will work).

So if everything is working fine, leave the current/default settings alone.

A couple of notes:

1. RAID 0 only improves sequential throughputs, not random access performance, which is what OS's rely on.
2. Splitting the RAID 0 up into separate disks (one per OS), keeps you from losing all of the OS's in the event the RAID 0 fails (still have at least one OS available to help research a diagnosis/download drivers/purchase new disks online in the event of a failure).

What about the following:

• Backup solution?
• UPS used?

- No problems beside my OSX Boot Drives on SSD get fragmented to hell & back in 2 weeks. Part of problem is memory issue with LION causing page-outs. I have this issue on all 3 system . Started using Clusters to see if that would help with fragmention, still need another week before i can tell if it's helping.

- 8 x HD = Seagate Constellation ES ST1000NM0011 1TB 7200 RPM 64MB Cache SATA 6.0Gb/s 3.5" Enterprise Hard Drive -Bare Drive Areca site says they seen issue with Seagate Drives. I have not had any problems with mine.

- 8 x HD i use to backup all 3 systems. I have it partition 4 ways with 2 small partitions that are formatted as NTFS & ext4. I use Acronis, CloneZilla to backup those. SuperDupper for OSX. I have 4x OWC 3G 120GB, 4 x OCZ vertex 2 60Gb, & 2 Intel X25 160GB SSD floating between all 3 CPU for scratch disk dedicated for projects im working on. I normally backup up ever thing to RAID 6 before i call in night. i have timer set, which i hardly ever gets use cause i do it manually 90% of time.

- I have master clone for Triple boot with clonezilla, i keep seperate back up of home dirtectory for my settings on all OS. I can secure erase and re-clone with in hour to get back up running if RAID 0 fails.

- UPS I have Two, BPX-1300G *& 1 small one for my MBP

 

[nanofrog]

- No problems beside my OSX Boot Drives on SSD get fragmented to hell & back in 2 weeks. Part of problem is memory issue with LION causing page-outs. I have this issue on all 3 system . Started using Clusters to see if that would help with fragmentation, still need another week before i can tell if it's helping.

It's because they're configured in a RAID (neither the OS or firmware method such as Garbage Collection work in any RAID level).

To get it to work (and formally at least, OS X only supports TRIM on Apple branded disks; IIRC, there's a hack to get working with other SSD's), is in single disk mode. Yet another reason to eliminate the set, and use the SSD's as individual disks...

- 8 x HD = Seagate Constellation ES ST1000NM0011 1TB 7200 RPM 64MB Cache SATA 6.0Gb/s 3.5" Enterprise Hard Drive -Bare Drive Areca site says they seen issue with Seagate Drives. I have not had any problems with mine.

There have been issues. But those on the HDD Compatibility List will work (Staggered Spin-up value has had to be increased to 0.7 seconds in a couple of cases with Seagate ES's on the ARC1880 series, but this isn't an uncommon issue).

Newer disk firmware may have precluded the need in your case however. But if they do become unstable (see drop-outs, and the set degrades), then change the value and see if that fixes the problem (also seeing a lot of Seagates ES models going bad within 3 years, so you've been warned).

I've been recommending Western Digital for SATA since 2008, as I've seen far fewer failures (consumer <Caviar Black> and enterprise models). I also tend to recommend the Green versions for backup drives, as the cost/GB is the best out there (most users always seem to need more backup capacity ).

- 8 x HD i use to backup all 3 systems. I have it partition 4 ways with 2 small partitions that are formatted as NTFS & ext4. I use Acronis, CloneZilla to backup those. SuperDupper for OSX. I have 4x OWC 3G 120GB, 4 x OCZ vertex 2 60Gb, & 2 Intel X25 160GB SSD floating between all 3 CPU for scratch disk dedicated for projects im working on. I normally backup up ever thing to RAID 6 before i call in night. i have timer set, which i hardly ever gets use cause i do it manually 90% of time.

- I have master clone for Triple boot with clonezilla, i keep separate back up of home directory for my settings on all OS. I can secure erase and re-clone with in hour to get back up running if RAID 0 fails.

So if I understand you correctly, the RAID 6 set is only for backups, not working data?

If so, that's fine. Bit overkill, but fine (could have used JBOD <concatenation> and Green disks in a Port Multiplier enclosure for this; risk level is that of a single drive, and you get the best cost-performance ratio for backup storage this way).

- UPS I have Two, BPX-1300G *& 1 small one for my MBP

I would seriously suggest getting rid of them, as they use stepped inverters. The reason for this, is that a stepped inverter causes damage in an Active PFC based Power Supply, which the MP uses, when the inverter is active (it's fine when the UPS is powering the MP from the wall). BTW, the damage is cumulative, not instantaneous in most cases (so it could appear to be fine, then die "suddenly" one day).

Seen this first hand, so please take it seriously.

You need what's called a Pure Sine Wave inverter type, such as the SmartUPS series (still a Line Interactive type, but the inverter will not damage the MP's PSU). Next step up, is the SmartUPS XL series (inverter is always running, whether there is power on the wall or not; aka Online or Double Conversion type).

Examples:

• SUA1500
• SMT1500 (adds an LCD display)
• SURTA1500XL (Online/Double Conversion unit)

You can find them cheaper if you buy refurbished (other brands besides APC, but the batteries tend to be more expensive <Eaton & Tripp Lite>). Example of an online refurbishing company (I and others here have used them).

In the case of the MBP, it's own battery will act as a UPS (granted, only if it has a sufficient charge to complete a shut down cycle). But continuing to use the BX1300G or skipping it all together can save you some funds (why I mentioned this).

But you do need to get the MP on an appropriate UPS.

 

[mac-hacks]

It's because they're configured in a RAID (neither the OS or firmware method such as Garbage Collection work in any RAID level).

To get it to work (and formally at least, OS X only supports TRIM on Apple branded disks; IIRC, there's a hack to get working with other SSD's), is in single disk mode. Yet another reason to eliminate the set, and use the SSD's as individual disks...


There have been issues. But those on the HDD Compatibility List will work (Staggered Spin-up value has had to be increased to 0.7 seconds in a couple of cases with Seagate ES's on the ARC1880 series, but this isn't an uncommon issue).


Newer disk firmware may have precluded the need in your case however. But if they do become unstable (see drop-outs, and the set degrades), then change the value and see if that fixes the problem (also seeing a lot of Seagates ES models going bad within 3 years, so you've been warned).

I've been recommending Western Digital for SATA since 2008, as I've seen far fewer failures (consumer <Caviar Black> and enterprise models). I also tend to recommend the Green versions for backup drives, as the cost/GB is the best out there (most users always seem to need more backup capacity ).


So if I understand you correctly, the RAID 6 set is only for backups, not working data?

If so, that's fine. Bit overkill, but fine (could have used JBOD <concatenation> and Green disks in a Port Multiplier enclosure for this; risk level is that of a single drive, and you get the best cost-performance ratio for backup storage this way).


I would seriously suggest getting rid of them, as they use stepped inverters. The reason for this, is that a stepped inverter causes damage in an Active PFC based Power Supply, which the MP uses, when the inverter is active (it's fine when the UPS is powering the MP from the wall). BTW, the damage is cumulative, not instantaneous in most cases (so it could appear to be fine, then die "suddenly" one day).

Seen this first hand, so please take it seriously.

You need what's called a Pure Sine Wave inverter type, such as the SmartUPS series (still a Line Interactive type, but the inverter will not damage the MP's PSU). Next step up, is the SmartUPS XL series (inverter is always running, whether there is power on the wall or not; aka Online or Double Conversion type).

Examples:

• SUA1500
• SMT1500 (adds an LCD display)
• SURTA1500XL (Online/Double Conversion unit)

You can find them cheaper if you buy refurbished (other brands besides APC, but the batteries tend to be more expensive <Eaton & Tripp Lite>). Example of an online refurbishing company (I and others here have used them).

In the case of the MBP, it's own battery will act as a UPS (granted, only if it has a sufficient charge to complete a shut down cycle). But continuing to use the BX1300G or skipping it all together can save you some funds (why I mentioned this).

But you do need to get the MP on an appropriate UPS.

-----------------------------------------------------
Appreciate tip & info on UPS. In past i've always kept SSD & HD on separate controllers. I appreciate your help & knowledge.

If so, that's fine. Bit overkill, but fine (could have used JBOD <concatenation> and Green disks in a Port Multiplier enclosure for this; risk level is that of a single drive, and you get the best cost-performance ratio for backup storage this way)

I know, if i had any editing jobs it would be use as my online HD drives. I normally use low res 720x480 in post if there ton of footage. Got good deal on controller and Vertex 3 threw friend shop. I plan on upgrading when Apple release next line MP. I plan to turn my MP 2009 into server so i have added CPU to add Qmaster & to network all my Apple Hardware iPad, iTV, CPU's, & etc...

I've been recommending Western Digital for SATA since 2008, as I've seen far fewer failures (consumer <Caviar Black> and enterprise models).

Wish i would have known, i read where Western Digital says not use Caviar Black only enterprise models with professional controller.

To get it to work (and formally at least, OS X only supports TRIM on Apple branded disks; IIRC, there's a hack to get working with other SSD's), is in single disk mode

.

I'm well aware of this, you know not use TRIMEnabler.app floating around. Just run this any time IOAHCIFamily.kext is updated.

sudo cp /System/Library/Extensions/IOAHCIFamily.kext/Contents/PlugIns/IOAHCIBlockStorage.kext/Contents/MacOS/IOAHCIBlockStorage /System/Library/Extensions/IOAHCIFamily.kext/Contents/PlugIns/IOAHCIBlockStorage.kext/Contents/MacOS/IOAHCIBlockStorage.original

sudo perl -pi -e 's|(\x52\x6F\x74\x61\x74\x69\x6F\x6E\x61\x6C\x00).{9}(\x00\x51)|$1\x00\x00\x00\x00\x00\x00\x00\x00\x00$2|sg' /System/Library/Extensions/IOAHCIFamily.kext/Contents/PlugIns/IOAHCIBlockStorage.kext/Contents/MacOS/IOAHCIBlockStorage

sudo kextcache -system-prelinked-kernel

sudo kextcache -system-caches

sudo shutdown -r now

 

[nanofrog]

-----------------------------------------------------
Appreciate tip & info on UPS. In past I've always kept SSD & HD on separate controllers. I appreciate your help & knowledge.

NP.

I know, if i had any editing jobs it would be use as my online HD drives. I normally use low res 720x480 in post if there ton of footage. Got good deal on controller and Vertex 3 threw friend shop. I plan on upgrading when Apple release next line MP. I plan to turn my MP 2009 into server so i have added CPU to add Qmaster & to network all my Apple Hardware iPad, iTV, CPU's, & etc...

In such a case, you would need to get a backup system for the RAID 6 volume, but that's not too difficult (eSATA controller that supports Port Multipliers, PM enclosure, and Green drives is cost effective).

At least the existing hardware would be able to be transferred to a new MP (PC too, if you decide to go that route when the time comes ; I'm platform agnostic - whatever gets the job done the best, is the right solution for the particular need).

Wish i would have known, i read where Western Digital says not use Caviar Black only enterprise models with professional controller.

Simple controllers (not a hardware RAID controller), work with consumer grade disks. It's hardware RAID controllers that you need to get the enterprise versions, as the Recovery Timings are different (set in the drive's firmware). There's some additional sensors too, but both are added to the consumer drives (in most cases, they cherry pick the platters; all of WD's are 1E15 bit error rated, unlike most consumer disks which are only 1E14). That additional order of magnitude makes a difference...

And it's one of the reasons I like the Caviar Blacks for consumer use.

I'm well aware of this, you know not use TRIMEnabler.app floating around. Just run this any time IOAHCIFamily.kext is updated.

Wasn't sure, so it seemed worth mentioning.

Unfortunately, it won't help you with the SSD's configured in a stripe set (RAID 0). You'd have to break the set up, and run each as a separate disk to take advantage of it.

 

[Spacedust]

Please upgrade firmware to 1.50 and report: http://www.areca.us/support/download/RaidCards/BIOS_Firmware/ARC1882.zip