Strangely slow Velocity Duo X2 plus RAID 0 Samsung 850 Evos.

[Ph.D.]

Hello all,

I have an Apricorn Velocity Duo x2 loaded with two 1 GB Samsung 850 Evos in a RAID 0 configuration, mounted in Slot 1 in a 3,1 (one of the two faster slots). They were formatted while in the Duo via Disk Utility.

The speeds seem a little disappointing. Black magic reports about 660 read and write scores. I would have thought it would be around 800 or more.

But worse, the subjective responsiveness is unremarkable or even quite poor. Loading a program seems reasonably fast, but when I open a folder containing a very large number of files (photos, etc.), the delay is enormous. It's very significantly slower than 2 older, slower SSD's that I have mounted in normal drive bays. It's so slow that I'd half expect even a hard disk to be faster.

Something feels wrong, no? Any ideas?

Thanks.

 

[DEMinSoCAL]

Hello all,

I have an Apricorn Velocity Duo x2 loaded with two 1 GB Samsung 850 Evos in a RAID 0 configuration, mounted in Slot 0 in a 3,1. They were formatted while in the Duo via Disk Utility.

The speeds seem a little disappointing. Black magic reports about 660 read and write scores. I would have thought it would be around 800 or more.

But worse, the subjective responsiveness is unremarkable or even quite poor. Loading a program seems reasonably fast, but when I open a folder containing a very large number of files (photos, etc.), the delay is enormous. It's very significantly slower than 2 older, slower SSD's that I have mounted in normal drive bays. It's so slow that I'd half expect even a hard disk to be faster.

Something feels wrong, no? Any ideas?

Thanks.

Do you mean Slot 1, the first slot (lowest towards the bottom)?

Did you try a different slot?

 

[Ph.D.]

Sorry, yes, Slot 1, the lowest and closest to the memory slots (edited).

I swapped it to Slot 2, with no change in performance per Black Magic disk speed test.

Slot 3 or 4, by the way, results in drastically reduced performance, down to the level of a hard disk (150). This was somewhat surprising, since I believe the Duo is a 2-lane card, and those slots are 4 lanes.

 

[m4v3r1ck]

Strangely slow Velocity Duo X2 plus RAID 0 Samsung 850 Evos.

Sorry, yes, Slot 1, the lowest and closest to the power supply (edited above).



I swapped it to Slot 2, with no change in performance per Black Magic disk speed test.



Slot 3 or 4, by the way, results in drastically reduced performance, down to the level of a hard disk (150). This was somewhat surprising, since I believe the Duo is a 2-lane card, and those slots are 4 lanes.

Hi Ph.D,



No, the slots of the PCIe's are numbered the other way round. So slot #1 is the one closest to the memory trays.

Perhaps you could try and BM the SSD's in the SATA2 HDD bays first, to check their r/w readings are similar when inserted. The readings you should get on a RAID0 on the DUO in slot#1 or #2 is indeed around 750-800.



1. How long ago did you install the DUO and SSD's and what was your install/restore scheme? Perhaps there's a proces e.g. like Spotlight Indexing havely using resources.


2. What are the other components occupying your PCIe slots? Please share your config, remember slot#1 is the one closest to the memory trays.



Cheers

 

[Ph.D.]

Thanks for the response.

Sorry, as you say, it's closest to the memory slots. (Perhaps I was thinking of my old PPC tower when I wrote that.)

Anyway, I've indeed tried both slots 1 and 2, with the same result, while a try in Slots 3 and 4 were much slower. I still do wonder why the difference is so much given that the Duo is only a 2-lane device.

The Duo and the 850's are only a few weeks old. The speed was the same when first installed. I've checked it many times under various conditions and it's always the same.

The only other device in the PCI slots is an MSI R9 280X, unflashed, but with the R17 resistor removed for 5 Gb/s operation. Perhaps I'll put my Apple 5770 back in it and try it then.

Thanks again.

 

[m4v3r1ck]

Strangely slow Velocity Duo X2 plus RAID 0 Samsung 850 Evos.

Thanks for your feedback, perhaps before changing GPU, try the AJA Systems disk speed tester for your current setup?

Cheers

 

[Ph.D.]

Aja reports approximately the same results (slightly slower, actually, depending on the file size).

Thanks.

 

[crjackson2134]

Not sure if Trim is able to work on a Raid system, but it sounds like you need to test Trim your drives next .

 

[m4v3r1ck]

Strangely slow Velocity Duo X2 plus RAID 0 Samsung 850 Evos.

Not sure if Trim is able to work on a Raid system, but it sounds like you need to Trim your drives.

Never benched without TRIM, so I'm very curious myself if that will make the difference.

EDIT: I have 840 EVO's with the latest firmware and performance update tool.

Cheers

 

[crjackson2134]

Never benched without TRIM, so I'm very curious myself if that will make the difference.

Cheers

It's something to try...

 

[Ph.D.]

Update: No change when I swapped the 280x for a genuine Apple 5770.

(Trim is turned on under Mavericks using Trim Enabler. But anyway, the disks are only half full since new and trim would not do much if anything under this circumstance.)

 

[m4v3r1ck]

It's something to try...

It sure is! ;-)

 

[Ph.D.]

Another update:

Turning Trim Enabler off also made no difference.

Maverick: How did you prepare your disks for the Duo?

 

[m4v3r1ck]

Maverick: How did you prepare your disks for the Duo?

1. Installed the 2 SSD's on the DUO
2. Placed the DUO in slot#2 (then MP 3.1)
3. Used Disk Utility (shown as externals) to format to RAID0
4. Clean installed OSX 10.9.5
5. Used Migration Assistant to migrate users/apps/settings etc.

Worked like a clock for me!

Cheers

 

[crjackson2134]

I would test each SSD, installed one at a time in a standard configuration. For example only one SSD, freshly formatted/Trimmed and installed physically on the DUO at a time. Perhaps one drive isn't working properly and dragging the other one down the rabbit hole.

----------

1. Installed the 2 SSD's on the DUO
2. Placed the DUO in slot#2 (then MP 3.1)
3. Used Disk Utility (shown as externals) to format to RAID0
4. Clean installed OSX 10.9.5
5. Used Migration Assistant to migrate users/apps/settings etc.

Worked like a clock for me!

Cheers

What is your stripe setting?

 

[Ph.D.]

Hi Maverick,

OK, I did the same except I didn't do a clean install, I cloned from my pre-existing OSX 10.9.5 Raid 0 disk array (in normal drive bays). This sort of thing has always worked well for me in the past, and I'm not sure how I'd consider that a likely culprit (it would be a pain to redo it at this point, too).

Can you confirm that in your 3,1 you were getting ~800 on BM?

Thanks.

 

[m4v3r1ck]

I would test each SSD, installed one at a time in a standard configuration. For example only one SSD, freshly formatted/Trimmed and installed physically on the DUO at a time.

Perhaps one drive isn't working properly and dragging the other one down the rabbit hole.

I would suggest testing each SSD in the SATA2 HDD/SSD bays, therefore ruling out the DUO and test only the SSD's!

What is your stripe setting?

Have to check that for you, using iPhone atm and sipping a beer... ;-)

----------

Hi Maverick,



OK, I did the same except I didn't do a clean install, I cloned from my pre-existing OSX 10.9.5 Raid 0 disk array (in normal drive bays). This sort of thing has always worked well for me in the past, and I'm not sure how I'd consider that a likely culprit (it would be a pain to redo it at this point, too).



Can you confirm that in your 3,1 you were getting ~800 on BM?



Thanks.

I got around ~700-710 on MP 3.1 OSX 10.9.5, will post screens later as I'm having trouble posting DropBox pics using TapaTalk! ;-)

 

[Ph.D.]

Maverick,

If so, perhaps ~670 is not unreasonable after all. That would only be a ~5-7% difference, which would be slightly disappointing but not outrageous. The 850's do use a radically-different technology than the 840's (a 3-D version of an older node). Still...

Thanks.

 

[hfg]

When I first received my Velocity DUO x2, I was intending to boot Windows and OS X from it. However, before I did I ran some RAID-0 tests against my Tempo Pro which I was currently using as a RAID-0 boot device.

The issue I was concerned about was that the Velocity DUO x2 was only a 2 lane card, whereas the Tempo Pro was a 4 lane card.

I used the exact same pair of Crucial m4 512GB SSDs, in the same slot, in the same 5,1 Mac Pro to run the tests. As you can see, there was a significant read speed difference between the cards when running RAID-0. (Note: it is normal for the M4 generation SSDs to have slower Write speeds than Read speeds).

Here are the BlackMagic test results:

Velocity DUO x2 --------- Tempo Pro (both in RAID-0)

.

 

[m4v3r1ck]

Strangely slow Velocity Duo X2 plus RAID 0 Samsung 850 Evos.

@Ph,D

Do you perhaps have a bootable backup of your OSX, so you can bench the DUO RAID0 from it?

Cheers

----------

When I first received my Velocity DUO x2, I was intending to boot Windows and OS X from it. However, before I did I ran some RAID-0 tests against my Tempo Pro which I was currently using as a RAID-0 boot device.



The issue I was concerned about was that the Velocity DUO x2 was only a 2 lane card, whereas the Tempo Pro was a 4 lane card.



I used the exact same pair of Crucial m4 512GB SSDs, in the same slot, in the same 5,1 Mac Pro to run the tests. As you can see, there was a significant read speed difference between the cards when running RAID-0. (Note: it is normal for the M4 generation SSDs to have slower Write speeds than Read speeds).



Here are the BlackMagic test results:



Velocity DUO x2 --------- Tempo Pro (both in RAID-0)



.

Your write speeds on the DUO seems not right in resemblance to my benches ~ 700. Your read speeds do differ very much on the Tempo!

Cheers

 

[hfg]

Your write speeds on the DUO seems not right in resemblance to my benches ~ 700. Your read speeds do differ very much on the Tempo!

Cheers

The Crucial M4 series does have slow write speeds which I see on every benchmark I run, so I am used to it. They are very reliable however, which is why I continue to use them.

I also have used Samsung 840 Pro SSDs which are much more balanced with write/read speeds. Here is a test with a pair of 840 Pro 256GB SSDs on the Tempo Pro card ... much better writes!

.

 

[phairphan]

Slots 3 and 4 are four-lane slots, but they're four-lane PCIe 1.0 slots. Even thought the Duo is a two-lane PCIe device, when you plug it into Slot 3 or 4, it uses two lanes, but two PCIe 1.0 lanes. In effect, the card is running at half-bandwidth compared to it being in Slot 1 or 2.

As for the speed, I recall seeing speeds ~700 for an 840 Evo RAID 0 set in the primary Duo thread around here somewhere. I believe m4v3r1ck will be able to comment, as s/he started and contributed greatly to that thread.

Thanks for the response.

Sorry, as you say, it's closest to the memory slots. (Perhaps I was thinking of my old PPC tower when I wrote that.)

Anyway, I've indeed tried both slots 1 and 2, with the same result, while a try in Slots 3 and 4 were much slower. I still do wonder why the difference is so much given that the Duo is only a 2-lane device.

The Duo and the 850's are only a few weeks old. The speed was the same when first installed. I've checked it many times under various conditions and it's always the same.

The only other device in the PCI slots is an MSI R9 280X, unflashed, but with the R17 resistor removed for 5 Gb/s operation. Perhaps I'll put my Apple 5770 back in it and try it then.

Thanks again.

 

[Ph.D.]

Yes, booting from a separate disk did not change the Duo's speed.

Thanks.

@Ph,D

Do you perhaps have a bootable backup of your OSX, so you can bench the DUO RAID0 from it?

Cheers

----------

Yes, I suspected as much about slots 3 and 4 (1.0 vs. 2), but lacked confirmation.

I'm beginning to think that my speeds are all that should be expected from this combination.

Thanks.

Slots 3 and 4 are four-lane slots, but they're four-lane PCIe 1.0 slots. Even thought the Duo is a two-lane PCIe device, when you plug it into Slot 3 or 4, it uses two lanes, but two PCIe 1.0 lanes. In effect, the card is running at half-bandwidth compared to it being in Slot 1 or 2.

As for the speed, I recall seeing speeds ~700 for an 840 Evo RAID 0 set in the primary Duo thread around here somewhere. I believe m4v3r1ck will be able to comment, as s/he started and contributed greatly to that thread.

 

[m4v3r1ck]

The Crucial M4 series does have slow write speeds which I see on every benchmark I run, so I am used to it. They are very reliable however, which is why I continue to use them.

Reliablity is the core! I agree 100%

I also have used Samsung 840 Pro SSDs which are much more balanced with write/read speeds. Here is a test with a pair of 840 Pro 256GB SSDs on the Tempo Pro card ... much better writes!

Great numbers, that's what we want to achive upgrading our cMP! Thanks for sharing your info's!

Cheers

 

[m4v3r1ck]

Strangely slow Velocity Duo X2 plus RAID 0 Samsung 850 Evos.

Slots 3 and 4 are four-lane slots, but they're four-lane PCIe 1.0 slots. Even thought the Duo is a two-lane PCIe device, when you plug it into Slot 3 or 4, it uses two lanes, but two PCIe 1.0 lanes. In effect, the card is running at half-bandwidth compared to it being in Slot 1 or 2.

I used slot #2 in my MP 3.1, but still got better r/w speeds as OP.

As for the speed, I recall seeing speeds ~700 for an 840 Evo RAID 0 set in the primary Duo thread around here somewhere. I believe m4v3r1ck will be able to comment, as s/he started and contributed greatly to that thread.

I'm a he and thanks for your compliment about the DUO thread, being just an end-user and not a tecnician. It was great fun but also a giant leap to throw around PCIe cards in my MP 3.1 and testing different setups. ;-)

Indeed I tested the DUO with ~ 700 r/w speeds then. In my cMP 5,1 the speeds don't differ that much, but still happy with this rock-solid solution.

Cheers