Setting:
- 3x Samsung QVO 870 8TB
- Thunderbay 4 Mini enclosure
- macOS 12.2.1
- SoftRAID 6.2.1
BlackMagic speed for each drive individually ~520MB/s.
RAID0:~1000MB/s
RAID4: ~700MB/s
I would have expected RAID0 performance of around ~1.5GB/s and had hoped for RAID4 around ~1GB/s. Is this performance expected or is something wrong? What performance could I expect with a 4th drive?
Also, why do I need to restart SoftRAID after disabling safeguard to have it have an effect? Is that intentional?
I would expect better, also.
You can create a RAID 0, then use SoftRAID to convert to Apple format. Then you should see the same performance.
I would use AJA system test (app store), as it is more consistent. Settings would be:
Codev RGBA
Resolution 5K (or higher)
File Size 64GB or 16GB
With AJA you can expand the chart and see if there are dropouts affecting performance.
What do you mean by "create RAID 0, then use SoftRAID to convert to Apple format". The RAID0 volume I created in SoftRAID is already APFS.
AJA (5K, 16GB, 16bit RGBA) results:
SingleDrive:
RAID0 (3 drives):
RAID4 (3 drives):
that is horrible write performance. Do you get the same with RAID 5 volumes?
This is an M1, correct? Write performance is slower as M1 performance on disks is not as good as on intels, but this is way too slow.
Yes, this is on an M1 Pro. Results for RAID5 (3 drives) are here:
What is the model number of your Mini?
If I have one, I can test this, but I have seen better results with 5 year old SSD's.
This would be owctb43mini or owctb3qmsr00gb (not sure).
Interestingly, in RAID4 mode the write speed almost doubles when I connect a second TB cable to the mini?
What speeds would you expect when I add a 4th drive?
Clarify what you mean by a second cable?
With 4 drives writes should be close to 2.5-3X , and reads 3X a single drive.
With 4 drives writes should be close to 2.5-3X , and reads 3X a single drive.
4 drives, RAID4, single cable, 0.75x writes, 2x reads:
4 drives, RAID4, second TB3 cable between Macbook Pro and Thunderbay 4 Mini, 1.3x writes, 2x reads:
Clarify what you mean by a second cable?
What if you only use the second cable?
I would not expect cables to work (interference), I have to check this out.
Makes no difference which cable I choose, results are consistent based on the number of cables.
Let me ask some hardware engineers about this. Its weird, and I would not have believed possible.
What are the two cables? How long is each?
Do both have a "3" on them?
One idea might be one is passive, the other active, but it would not explain the write speed changes.
Run SoftRAID. Zero disks. (100 sectors)
run DiskUtility/RAID Assistant. Create a RAID 0 volume. What numbers are you getting?
If different, "convert" to SoftRAID (volumes menu, restart and retest.
let me know.
both are TB3 cables with 3s on them one, not sure if active or passive. both ports have the same specs in the system profiler.
creating the raid0 in diskutility does not result in different performances. should i open a dedicated support ticket outside of this forum?
No need. Both are same length cables? (.5m 1m or 2m?)
OWC branded cables or other?
this is clearly a hardware issue, as you see. SoftRAID gives the same results as Apple's RAID. I hve some of our SSD's in stock, I can test with. It may be something weird with the Samsung and the mac. If these are the "latest" 870 drives, Samsung is often not mac compatible in first releases of hardware.

