Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.

SteveZee

macrumors member
Original poster
Nov 24, 2011
56
2
I'm running a 250GB OWC SSD in mid-2010 MacPro.
My write speed is a mere 43 MB/s based on Black Magic speed test.
50% unused space. What should I do?
Thanks anyone.

sierra.png
 
TRIM enable?
Yup...

OP > Open Terminal and run this command to turn on TRIM the reboot after.

Code:
sudo trimforce enable

Then reboot to single user mode by holding command-s at startup. Once it boots enter this at the command line.

Code:
fsck -fy

You should see a message at the end about trimming unused blocks. Now reboot and try your speed test again and let is know.
 
Trim should have little impact with a 50% empty disk. There is something else going on.

Not true, TRIM has very large impact. 50% is empty space in OS point of view, without TRIM, unless the remaining 50% never been used, the controller cannot see it as empty space (that's why we need TRIM).

Unless OP secure erase the SSD (zero all cells), and then only use 50% capacity to create a partition for OS to use. The remaining 50% never partitioned. Then the controller now able to see that remaining 50% is "free space" and use GC effectively even without TRIM.
 
I mean their ssds have been old news since day one, but the on board garbage collection should be adequate barring extreme read/write scenarios.
 
The black magic test is slow on my drives whether TRIM enabled or not.
I can't remember which way around it writes data, but it is either compressed or uncompressed. Most ssds are slow at writing whatever the black magic app tests with.

I've seen speeds like yours on my OWC, Samsung and Sandisk drives. I wouldn't worry about it.
 
I mean their ssds have been old news since day one, but the on board garbage collection should be adequate barring extreme read/write scenarios.

My understanding is like this.

The SSD is functioning properly without TRIM, but just at much lower writing performance.

GC cannot replace TRIM, in fact, GC cannot work effectively without TRIM.

TRIM is just to allow the SSD's controller to sync with the OS. So that, the free space in OS (garbage) can be effectively collected, and turned into over provision by GC.

Without TRIM, the controller can't see those free space, and GC cannot collect and garbage. Because it doesn't know which bit is "garbage". It only know that's not zero.

OWC didn't lie. Their SSD can workout TRIM, just like any other SSD. But what they did't tell you is that their SSD will work much worse without TRIM, just like and other SSD.

Unless their SSD has super large build in over provision (which they don't). There is no room (free space) for GC to work effectively. And without pre-emptied cell, writing performance will be greatly affected.

What GC does is grouping all unused bit together, and zero them out.

e.g. cell 1 has 1 bit useful data stored, and cell 2 also has 1 bit useful data stored. All other bits was used, but now belongs to the "deleted" files (garbage).

If TRIM available, then the controller will know that only 1 bit in each cell is used, all other bits are actually "free space". Therefore, GC can group all the "garbage" into cell 2 (move the useful bit from cell 2 to cell 1), and allow the controller to zero out the whole cell 2. So that, on the next write operation, the SSD can write data onto the pre-emptied cell 2 in order to deliver high writing speed.

Without TRIM. In the controller's point of view. Cell 1 is used, and cell 2 is used. GC cannot move any data, because no empty space for data move.

When a new writing operation demanded. The controller has to move the useful data from cell 2 to cell 1 + zero out cell 2 + writing new data into cell 2. This "full process" is much much slower than just writing new data in.

The black magic test is slow on my drives whether TRIM enabled or not.
I can't remember which way around it writes data, but it is either compressed or uncompressed. Most ssds are slow at writing whatever the black magic app tests with.

I've seen speeds like yours on my OWC, Samsung and Sandisk drives. I wouldn't worry about it.

Some OWC SSD used compression to increase the writing speed (on paper), therefore, incompressible data will write at a much slower rate (usually somewhere around 100-200MB/s). However, for 43MB/s, it's clearly not just because of incompressible data, but due to lack of TRIM.

As the others pointed out, purely enable TRIM won't get the performance back straight away, you have actually TRIM the empty space before running the benchmark. I don't know which Samsung SSD you are talking about, but my TRIMed 840 Evo obviously can stay at around 500MB/s in black magic. Flowrider occasionally post his SSD's black magic result. I never ever see any SSD write at 43MB/s range (when everything working properly). In fact, most of them has very good writing speed in black magic test.
 
  • Like
Reactions: Weaselboy
My understanding is like this.

I would disagree with this. Proper garbage collection on the ssd itself should be comparable to a trim enabled ssd so long as the (slower) garbage collection practices have time to do their job. In high write scenarios, gc can be overwhelmed resulting in slower writes and theoretically lost space until it catches up.

Bit lazy to cite another source but this is owc explaining it themselves:

https://blog.macsales.com/21641-with-an-owc-ssd-theres-no-need-for-trim
 
I would disagree with this. Proper garbage collection on the ssd itself should be comparable to a trim enabled ssd so long as the (slower) garbage collection practices have time to do their job. In high write scenarios, gc can be overwhelmed resulting in slower writes and theoretically lost space until it catches up.

Bit lazy to cite another source but this is owc explaining it themselves:

https://blog.macsales.com/21641-with-an-owc-ssd-theres-no-need-for-trim

OWC's explaination is busted long time ago. Again, their SSD don't need TRIM, and as any other SSD does, but won't work as good as TRIM supported, also same as any other SSD does.

If you believe a good GC can work properly without TRIM, please tell me the mechanism of "How to". As I stated, unless there is a super large over provision available inside the SSD. There is simple no room for GC to work, no matter the GC logic is good or bad.

So far, regardless what OWC said, no non-TRIMed OWC SSD can perform better than (even same as) a TRIMed SSD is also a fact.

Also, they newer SSD also support TRIM. This is also a fact. They just made up something to make theie no TRIM Sandforce controller SDD looks better in the old days. As the explanation stated. Their GC has to move BOTH valid and invalid. How can it work better than if TRIM avail (only need to move valid data)? Also, if the SSD is full of invalid data, where they get the empty space to move all those data? I suspect OP is exactly in this case, the SSD is full of invalid data (those deleted files), therefore, on the OS point of view, there is more than 50% empty space. But in the controller's point of view, there is no empty space for GC to work.

The combination of "less room to move data" + "have to move much more data then normally required" is definitely not good for SSD (in both perforamance and life span).
 
Last edited:
  • Like
Reactions: Weaselboy
Is it an early generation SSD from them? I had one of their drives when SSD was pretty new. I remember it performing poorly in the BMD disk test too.

I think the speed of those early SSDs was a lot more use case sensitive. Try another benchmarking app and you might get different results. The Blackmagic app is targeted towards measuring video performance with large files. This sounds like a simple enough scenario for a hard drive to handle (large sequential reads/writes), but I remember seeing the same with an OWC drive.

The same app shows different results with more current drives. If your drive is new... well....
 
Is it an early generation SSD from them? I had one of their drives when SSD was pretty new. I remember it performing poorly in the BMD disk test too.

I think the speed of those early SSDs was a lot more use case sensitive. Try another benchmarking app and you might get different results. The Blackmagic app is targeted towards measuring video performance with large files. This sounds like a simple enough scenario for a hard drive to handle (large sequential reads/writes), but I remember seeing the same with an OWC drive.

The same app shows different results with more current drives. If your drive is new... well....
Note that the Sandforce controller compresses the data stream. It gets much of its performance from compression (the advertised specs are for compressible files), and the compression amounts to additional over-provisioning which helps the garbage collection.

Most media files are already heavily compressed, and don't get much advantage from Sandforce compression.
 
OWC's explaination is busted long time ago. Again, their SSD don't need TRIM, and as any other SSD does, but won't work as good as TRIM supported, also same as any other SSD does.

If you believe a good GC can work properly without TRIM, please tell me the mechanism of "How to". As I stated, unless there is a super large over provision available inside the SSD. There is simple no room for GC to work, no matter the GC logic is good or bad.

So far, regardless what OWC said, no non-TRIMed OWC SSD can perform better than (even same as) a TRIMed SSD is also a fact.

Also, they newer SSD also support TRIM. This is also a fact. They just made up something to make theie no TRIM Sandforce controller SDD looks better in the old days. As the explanation stated. Their GC has to move BOTH valid and invalid. How can it work better than if TRIM avail (only need to move valid data)? Also, if the SSD is full of invalid data, where they get the empty space to move all those data? I suspect OP is exactly in this case, the SSD is full of invalid data (those deleted files), therefore, on the OS point of view, there is more than 50% empty space. But in the controller's point of view, there is no empty space for GC to work.

The combination of "less room to move data" + "have to move much more data then normally required" is definitely not good for SSD (in both perforamance and life span).

The drives have over provisioning. I didn't say they work better than if trim were also enabled, only that during normal use instances they should perform comparably, which they generally do. If you read the link I don't need to explain how the process works.
 
The drives have over provisioning. I didn't say they work better than if trim were also enabled, only that during normal use instances they should perform comparably, which they generally do. If you read the link I don't need to explain how the process works.

The drive has over provision, but I believe is not large enough. To let a SSD keep you it's performance. It require ~20% "free space" for the controller to work with, but over provision usually ~7% of the total capacity. Way too less.

I read the link. As I pointed out. It says, their GC has to move BOTH valid and invalid data. Which means much more work then with TRIM. End result, it can only perform worse then with TRIM, not even comparable (in both performance and life span).

I can't see anything in the link explain how their GC can make their SSD work up to the level that a normal SSD with TRIM can do.

But I totally agree that in general. It doesn't really matter. The SSD is fast enough. And the user may only feel significant slowdown when writing very large incompressible file. Which is not a normal user's daily use on a SSD. Most users use SSD because of its low latency on small files read / write (especially raading). Not large sequential write.

But I just couldn't agree that their SSD (without TRIM support) can work as good as with TRIM. Simply no evidence and theory can support this argument. (I am not talking about "work better", but just up to the same level).
 
No, that is not the consensus.
Sorry for forgetting about you. I just wanted to move this forward and provide the OP some help.

Even without TRIM a SSD should not be running that slow. Early on the reviewers tested OWC drives without TRIM and did not see such bad performance. They filled the drives first so the only difference is the OP's drive is older so may be failing. That is why I suggest to run SSDReporter or similar.

The default setting for Blackmagic is to write a single 5GB file, not random writes. That is way bigger than any block/page size so the drive should not run into the write amplification issues. With GC and over-provisioning blocks don't need to be erased either.
 
You are mistaken. Read this thread for just one example. In that thread the drive went from 25mbps write speeds to over 400 just by running TRIM as I suggested in post #3 of this thread.
That's why I said "should." That thread is about a particular Samsung that has issues. In the same thread another provided performance data showing no degradation of their SSD running for years without TRIM.
 
That thread is about a particular Samsung that has issues.
What do you mean "has issues"? The drive slowed down over time and a quick TRIM run fixed it just like I have suggested the OP try here.

There are many posts just like that one where a TRIM run fixed this. It is not a one off issue with that particular Samsung drive.
 
Last edited:
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.