Notifications
Clear all

New OWC Express 4M2 enclosure performance results

Page 4 / 5
(@softraid-support)
Member Admin

The cable issue is part of the Thunderbolt spec. Nothing anyone can do about it.

If you are careful to put the M2/4 Express on different buses, you should get double bandwidth with two.

ReplyQuote
Topic starter Posted : 25/07/2019 11:43 am
(@tpack)
New Member Customer

macOS Mojave 10.14.6
- OWC Express M42
- 4x Intel 660p 1TB SSD
- TB2 Apple adapter
- SoftRAID XT
- Blackmagic v 3.2

1. configured RAID 0
Speed W 1238 MB/s R 1242 MB/s

2. configured RAID 1+0
Speed W momentarily peaks at 630 MB/s but settles between 143 MB/s to 209 MB/s while R 1202 MB/s

What is the problem?

Also, when opening SoftRaid XT a message flashes momentarily RAID 1+0 degraded 4 missing disks (see attachment)

What is the problem? Is this related to the above performance?

ReplyQuote
Posted : 09/09/2019 12:17 pm
(@softraid-support)
Member Admin

I use AJA System Test, as it gives more consistent results.
(5K video, RGBA, 64GB file size)
I ran AJA on an M2/4 with 4 Samsung blades in it:
2500 reads/1200 writes (rounded off)

I moved it to a 2013 Mac Pro, with Apple Adapter and get:
1350 reads/680 writes

The intels are very reliable, but a little slower than other blades.

Your performance on writes should be in the 600 range, however. See if AJA gives you that.
Else it is an issue. I am assuming a fresh set of blades (certified with SoftRAID with 2 passes) and no data on an HFS volume.

ReplyQuote
Topic starter Posted : 09/09/2019 3:16 pm
(@tpack)
New Member Customer

I use AJA System Test, as it gives more consistent results.
(5K video, RGBA, 64GB file size)
I ran AJA on an M2/4 with 4 Samsung blades in it:
2500 reads/1200 writes (rounded off)

I moved it to a 2013 Mac Pro, with Apple Adapter and get:
1350 reads/680 writes

The intels are very reliable, but a little slower than other blades.

Your performance on writes should be in the 600 range, however. See if AJA gives you that.
Else it is an issue. I am assuming a fresh set of blades (certified with SoftRAID with 2 passes) and no data on an HFS volume.

Fresh blades, certified with 3 passes, no data. HFS volume
Tested also in Raid 5, getting stable 985 read and 723 write (black magic)

Testing with AJA
Raid 5 - 985 read 779 write
Raid 10 - 1273 read 646 write (and seems stable)

ReplyQuote
Posted : 09/09/2019 4:31 pm
(@softraid-support)
Member Admin

intel is selling them based on transactions per second (IOPS), rather than "throughput", I believe.
At least these numbers make some sense.

ReplyQuote
Topic starter Posted : 09/09/2019 6:22 pm
(@j-a-duke)
Active Member Customer

The cable issue is part of the Thunderbolt spec. Nothing anyone can do about it.

If you are careful to put the M2/4 Express on different buses, you should get double bandwidth with two.

I've got a question regarding active cables (for TB3 40Gbps operation). IIRC, the TB3 spec allows longer lengths (> .5 meter) for faster operation, but those need to be "active" rather than the passive cables that either are the .5 meter length running at 40Gbps or run at only 20Gbps for lengths greater than .5 meters.

Have you had a chance to test the active cables to see if they perform as well as the short passive cable?

The price difference from passive to active is significant (~$50 and up for 1 or 2 meter) but if they work, I'm in as they do give me placement flexibility for the enclosure.

Thanks.

Cheers,
Jon

ReplyQuote
Posted : 12/09/2019 9:21 am
(@softraid-support)
Member Admin

I have not tested longer Active cables personally.

Theoretically, if they are certified as 40Ghz for data, they will perform the same as the .5m cables. But all I have are standard passive long cables which do slow performance as expected.

ReplyQuote
Topic starter Posted : 12/09/2019 10:45 am
(@jhoney)
New Member Customer

I have one drive in it at the moment, Samsung 970 PRO 1TB, formatted using Softraid. The speed I'm getting is 657 MB/sec write, and 793 MB/sec read measured using AJA 64GB file. The device's the only one on its Mac Pro Thunderbolt 2 bus.

ReplyQuote
Posted : 28/09/2019 7:38 am
(@softraid-support)
Member Admin

Each PCI lane (Blade) gets about 700MBs. So that result is about correct.

ReplyQuote
Topic starter Posted : 28/09/2019 3:32 pm
(@scstraus)
New Member Customer

2. configured RAID 1+0
Speed W momentarily peaks at 630 MB/s but settles between 143 MB/s to 209 MB/s while R 1202 MB/s

I'm having the same problem. my write speeds are around 100-150 whereas my read speeds are a respectable 1700. Any idea why this would be? I'm using the official cable.

ReplyQuote
Posted : 19/02/2020 7:35 pm
(@softraid-support)
Member Admin

Do you have Secure Boot disabled, if this is a T2 chip equipped computer?
(newer Mac) This seems like the 5.6.8 driver is loading.

ReplyQuote
Topic starter Posted : 19/02/2020 8:57 pm
(@crhendo)
Active Member Customer

Hi, I have the same enclosure on my Mac Mini 2018 and while I am running RAID 0 across all the drives I can at least report that the throughput is W 1993MB/sec and R 2292MB/sec. When I was running RAID 5 around 6 months ago I remember it to be about 20% less than this.

ReplyQuote
Posted : 20/02/2020 12:24 am
(@softraid-support)
Member Admin

You did not reply if you have Secure Boot disabled.

Here is another way to check. Run this in terminal:

kextstat -b com.softraid.driver.SoftRAID

what is the output?

ReplyQuote
Topic starter Posted : 20/02/2020 9:03 am
(@scstraus)
New Member Customer

You did not reply if you have Secure Boot disabled.

Here is another way to check. Run this in terminal:

kextstat -b com.softraid.driver.SoftRAID

what is the output?

Here is the output:

$ kextstat -b com.softraid.driver.SoftRAID
Index Refs Address Size Wired Name (Version) UUID
117 0 0xffffff7f83ed2000 0x3c000 0x3c000 com.softraid.driver.SoftRAID (5.8.2) 4877E914-63E9-3499-95EE-CF13C7D9FB23

It's strange, today I was getting write speeds earlier in the day around 700 MB/Sec, but now again down to between 50-200.

ReplyQuote
Posted : 19/03/2020 12:28 pm
(@softraid-support)
Member Admin

Run the command again. If you just updated Catalina, it may have reinstalled the 5.6.8 driver on you. It should show 5.8.2 (or .3)

ReplyQuote
Topic starter Posted : 19/03/2020 1:49 pm
Page 4 / 5
Share:
close
open