Late 2013 iMac slow fusion drive

[jetjaguar]

I bought my 27" ultimate spec iMac from the apple store in december and the specs are:

3TB fusion
3.5 ghz
8gb ram ---> bought 32gb
780m 4gb

I never bothered running the blackmagic speed test before because I had iMacs with fusion drives before and they were always around 500/400 or so. I just tested my current iMac which I guess is supposed to have a faster ssd then the older ones and I scored 580r/ 150-225w .. isn't that really slow ??

 

[jcr918]

I bought my 27" ultimate spec iMac from the apple store in december and the specs are:

3TB fusion
3.5 ghz
8gb ram ---> bought 32gb
780m 4gb

I never bothered running the blackmagic speed test before because I had iMacs with fusion drives before and they were always around 500/400 or so. I just tested my current iMac which I guess is supposed to have a faster ssd then the older ones and I scored 580r/ 150-225w .. isn't that really slow ??

I don't have my Imac yet but that sounds about right. I was getting between 450-500 on my early Macbook pro ssd. I would think a 580 would be fast

 

[quagmire]

I don't have my Imac yet but that sounds about right. I was getting between 450-500 on my early Macbook pro ssd. I would think a 580 would be fast

450-500 would be fine if it was still the SATA III based interface. But other tests on the PCIe based SSD's show about a 700 MB/s read and about 400-500 MB/s write for the Fusion Drive( about equal 700 MB/s read and write for the pure SSD models).

 

[jetjaguar]

yea it is the newer pci-e ssd .. so only getting 200is write is very slow .. my 3tb fusion has about 1tb used. I guess I should call apple

 

[jcr918]

yea it is the newer pci-e ssd .. so only getting 200is write is very slow .. my 3tb fusion has about 1tb used. I guess I should call apple

Let us know what they say. When I spoke to them about speeds they said they didn't keep any record of that then asked a senior adviser and I got the same.

I had a real hard time deciding on a SSD over the fusion drive

 

[quagmire]

Let us know what they say. When I spoke to them about speeds they said they didn't keep any record of that then asked a senior adviser and I got the same.

I had a real hard time deciding on a SSD over the fusion drive

I went with pure SSD because if one component fails on the Fusion Drive, the computer won't work period. It won't continue to operate with either the SSD or HD only operating.

 

[Bear]

yea it is the newer pci-e ssd .. so only getting 200is write is very slow .. my 3tb fusion has about 1tb used. I guess I should call apple

The SSD portion of your Fusion Drive is basically full. There's I think 4GB of space on it that's kept clear for writing. However if you write more than that quickly, it starts to go directly to the hard drive.

Your Fusion Drive is working as designed.

 

[SaSaSushi]

I went with pure SSD because if one component fails on the Fusion Drive, the computer won't work period. It won't continue to operate with either the SSD or HD only operating.

And if your SSD fails, it won't work period either. The moral of the story, of course, is to back up all of your data.

Your Fusion Drive is working as designed.

+1

I get about 570-580MB/s read with BlackMagic on my Fusion Drive. As Bear mentioned, it is relative to how full the drive is and I immediately migrated over my very large, old system on this new machine when I started using it.

If you're using a smaller portion of the SSD I would expect to see 700MB/s+ read results.

 

[IA64]

And if your SSD fails, it won't work period either. The moral of the story, of course, is to back up all of your data.

.

You missed the point. If your SSD or HDD fails, Fusion drive will not work.

There is a risk of failure with pure SSD but with a fusion drive, your chances are doubled. Same applies RAID0 drives which although gives higher speeds, but a disk failure will be catastrophic.

 

[SaSaSushi]

You missed the point. If your SSD or HDD fails, Fusion drive will not work.

There is a risk of failure with pure SSD but with a fusion drive, your chances are doubled. Same applies RAID0 drives which although gives higher speeds, but a disk failure will be catastrophic.

The risk of failure in either component is minimal. The solution to any failures is to backup your data.

The Fusion Drive is an excellent option for those who want a mix of SSD speed with larger storage capacity and those who find it an attractive alternative should not feel any obligation to go for pure SSD based solely on the fact that it has two physical components.

Oh, and of course I also highly recommend AppleCare.

And for myself, if my iMac has a storage failure beyond the coverage period I will open it up and repair it. The Apple double-sided tape for the LCD panel is easily attainable online.

 

[quagmire]

The risk of failure in either component is minimal. The solution to any failures is to backup your data.

The Fusion Drive is an excellent option for those who want a mix of SSD speed with larger storage capacity and those who find it an attractive alternative should not feel any obligation to go for pure SSD based solely on the fact that it has two physical components.

Oh, and of course I also highly recommend AppleCare.

And for myself, if my iMac has a storage failure beyond the coverage period I will open it up and repair it. The Apple double-sided tape for the LCD panel is easily attainable online.

I do back up my data. I also understand SSD's fail too. It also makes sense if the SSD part of the Fusion Drive fails the computer won't boot because the OS is on there.

But if the HD fails, why can't it continue to work?

Overall, it is another thing that can go wrong. So I went with the 512 GB SSD.

 

[SaSaSushi]

I do back up my data. I also understand SSD's fail too. It also makes sense if the SSD part of the Fusion Drive fails the computer won't boot because the OS is on there.

But if the HD fails, why can't it continue to work?

Because it uses a core storage VolumeGroup, concatenating the two drives into one entity.

Overall, it is another thing that can go wrong. So I went with the 512 GB SSD.

I'm jealous but the 512GB SSD option was out of my budget for this machine. I decided to go for the 1TB Fusion Drive over the 256GB SSD and have had no complaints.

There are any number of things that can go wrong on the machine and there are effective ways to mitigate and insure against all of them.

 

[AppleFan360]

I'm jealous but the 512GB SSD option was out of my budget for this machine. I decided to go for the 1TB Fusion Drive over the 256GB SSD and have had no complaints.

There are any number of things that can go wrong on the machine and there are effective ways to mitigate and insure against all of them.

I agree. The Fusion drive is a good balance between speed and cost. When I run Blackmagic, my read speeds sometimes reach up to around 700 and 400 write. More than enough for what I do and I spent a hell of a lot less.

 

[joema2]

...There is a risk of failure with pure SSD but with a fusion drive, your chances are doubled. Same applies RAID0 drives which although gives higher speeds, but a disk failure will be catastrophic.

Your basic point is valid, but the probability of failure is actually the sum of two *different* failure rates, one for SSD and the other for HDD.

Expressed as hours to MTBF (Mean Time Between Failures), this is equivalent to the reciprocal of the sum of the individual MTBF reciprocals.

E.g, say the 128GB SSD has a MTBF of 2 million hrs and the HDD 1 million hrs.

The calculation would be Combined MTBF = 1/(1/2000000 + 1/1000000),
or 666,000 hrs.

Unfortunately nobody really knows the actual MTBF of any drives, since the manufacturers don't release the numbers.

Tom's Hardware did an extensive study of this: http://www.tomshardware.com/reviews/ssd-reliability-failure-rate,2923.html

Their conclusion was SSD is more reliable than HDD, but SSD has significant failure rates.

The 3TB hard drive used in recent iMacs is a 7200 rpm 6gbps Seagate Barracuda ST3000DM001. It has a stated annualized failure rate (AFR) of < 1%. However as the Tom's Hardware article shows, actual AFR can differ widely from the spec -- for both SSD and HDD.

Regardless of type of disk, where SSD or HDD, it should be backed up regularly.

 

[joema2]

my 27"...iMac...the specs are:

3TB fusion
3.5 ghz
8gb ram ---> bought 32gb
780m 4gb

...I just tested my current iMac which I guess is supposed to have a faster ssd then the older ones and I scored 580r/ 150-225w .. isn't that really slow ??

I have the same configuration iMac and the 4GB BlackMagic test shows about 576 MB/sec read, 314 MB/sec write on an empty 3TB Fusion Drive. After filling it with 1TB of data, the numbers sometimes drop off considerably, but it's erratic.

The app iStat Menus (http://bjango.com/mac/istatmenus/) allows monitoring what % of I/O goes to the SSD vs HDD portion of the FD. As already mentioned what's happening is it's doing physical I/O to the HDD, which is slower.

However you shouldn't get obsessed over a single benchmark. E.g, under the same conditions (1TB in use out of 3TB) QuickBench (large test) is still very fast.

Likewise doing actual file copy tests show much less slowdown. On a 5.2GB folder copy to/from an external RAID array, the difference between an empty 3TB FD and one with 1TB used was:

Write speed, empty: 291 MB/sec
Write speed, 1TB used: 254 MB/sec

Read speed, empty: 470 MB/sec
Read speed, 1TB used: 335 MB/sec

Note these were single-shot tests. Over time you'd expect the FD adaptive algorithm to reposition frequently-used data for higher performance.

In your situation you're already using 1TB of storage out of 3TB. A 512GB SSD isn't an option in that case -- it's too small. If you had a 512GB SSD, you'd have to buy an external disk, figure out what data to move, then re-evaluate the performance in that combined environment.

The 3TB FD isn't perfect but it's very flexible, has generally good performance, and avoids the hassle of managing data placement. A pure SSD can be a lot faster, and will have more consistent performance, although the FD performance should be self-correcting over time.

 

[DerekS]

I had this same problem.

Boot to single user mode (Cmd-S during boot) then: fsck -fy

 

[Bear]

I had this same problem.

Boot to single user mode (Cmd-S during boot) then: fsck -fy

You're presuming it's a problem with the disk and not a problem with the testing method.

 

[IA64]

Your basic point is valid, but the probability of failure is actually the sum of two *different* failure rates, one for SSD and the other for HDD.

Expressed as hours to MTBF (Mean Time Between Failures), this is equivalent to the reciprocal of the sum of the individual MTBF reciprocals.

E.g, say the 128GB SSD has a MTBF of 2 million hrs and the HDD 1 million hrs.

The calculation would be Combined MTBF = 1/(1/2000000 + 1/1000000),
or 666,000 hrs.

Unfortunately nobody really knows the actual MTBF of any drives, since the manufacturers don't release the numbers.

Tom's Hardware did an extensive study of this: http://www.tomshardware.com/reviews/ssd-reliability-failure-rate,2923.html

Their conclusion was SSD is more reliable than HDD, but SSD has significant failure rates.

The 3TB hard drive used in recent iMacs is a 7200 rpm 6gbps Seagate Barracuda ST3000DM001. It has a stated annualized failure rate (AFR) of < 1%. However as the Tom's Hardware article shows, actual AFR can differ widely from the spec -- for both SSD and HDD.

Regardless of type of disk, where SSD or HDD, it should be backed up regularly.

I wouldn't really count on MTBF to predict a hardware failure. Too many things to take into account.

A sudden power outage in the past corrupted 2 HDD one of them is external.

HDDs are much more prone to failure from vibration and heat t han SSD.

We are still talking about the iMac in this situation, an AIO desktop.

In the case of MB or any other laptop, HDD lifespan is much lower, no need to explain why. SSD is always the safest way to go.

Backups seem to make this a non issue. But I'd love to see you re-installing everything from scratch or sending your machine for 1 week with no replacement for a new drive.

Theoretically you might be right; but practically, and with all due respect, I can guarantee you that you're far from the truth.

 

[joema2]

I wouldn't really count on MTBF to predict a hardware failure. Too many things to take into account.

I'm not counting on MTBF to predict hardware failure; I'm just showing the right way to calculate it (which isn't just doubling the chance of failure). In fact the Tom's Hardware article I referenced shows how far from reality published MTBF and AFR numbers are. That's one reason I posted it.

In the case of MB or any other laptop, HDD lifespan is much lower, no need to explain why. SSD is always the safest way to go.

HDD lifespan or annual failure rate is generally worse than SSD, but not always. The worst SSDs overlap with the best HDDs; that was shown in the Tom's Hardware investigation.

Theoretically you might be right; but practically, and with all due respect, I can guarantee you that you're far from the truth.

What exactly are you talking about?

 

[Arfdog]

I wouldn't really count on MTBF to predict a hardware failure. Too many things to take into account.

A sudden power outage in the past corrupted 2 HDD one of them is external.

HDDs are much more prone to failure from vibration and heat t han SSD.

We are still talking about the iMac in this situation, an AIO desktop.

In the case of MB or any other laptop, HDD lifespan is much lower, no need to explain why. SSD is always the safest way to go.

Backups seem to make this a non issue. But I'd love to see you re-installing everything from scratch or sending your machine for 1 week with no replacement for a new drive.

Theoretically you might be right; but practically, and with all due respect, I can guarantee you that you're far from the truth.

What's wrong with MTBF? It's a valid statistics tool if you understand what it is and what kind of tool it is, ie comparing between 2 products using the same test method. No one ever said it was supposed to predict the life of a product (at 30 years or whatever).

And if you backup, you don't have to reinstall anything, what are you talking about?

 

[IA64]

I'm not counting on MTBF to predict hardware failure; I'm just showing the right way to calculate it (which isn't just doubling the chance of failure). In fact the Tom's Hardware article I referenced shows how far from reality published MTBF and AFR numbers are. That's one reason I posted it.


HDD lifespan or annual failure rate is generally worse than SSD, but not always. The worst SSDs overlap with the best HDDs; that was shown in the Tom's Hardware investigation.



What exactly are you talking about?

I am talking about Fusion drives not SSD or HDDs.

Chance that your Fusion drive will fail is much higher than an SSD or HDD side by side. It's not that I'm calculating the probability here or anything but there are many downsides with a FD.

The first of these is that there is a longer seek time for data that is stored on the hard drive. Also, there are more frequent spin cycles (which in turn could lead to a reduced lifetime due to wear that comes from this process)

Repeatedly spinning up and spinning down the disk platters wears out the traditional part of the fusion drive faster than normal use of a traditional hard drive where the platters keep spinning.

The SSD part of your hybrid drive is constantly being used, unlike regular SSDs where you use a different part everyday. Since SSDs have a limited write life, the SSD part of your hybrid hard drive will wear out quicker than a full-sized SSD.

Let's put it this way... Your HDD or SSD fails or filesystem gets corrupted; You can recover your files using a 3rd party tool or software.

Now how are you going to do that with a hybrid one ? Specifically the Fusion drive ? Do you have any idea which file is on which part of the disk ? Have you ever wondered why FD doesn't work under Windows and only the HDD is accessible ?

Unless you want to throw several hundreds of dollars on Drivesavers service, which isn't also guaranteed to work as per their statement :

#Recovering data from the Fusion Drive is not for the weak at heart," said Chris
Bross, strategic technology alliance manager. "It's a complex storage solution that combines two completely different
technologies to give Apple customers more speed with higher capacity#

I'd really love to hear your opinion. I'm gonna make myself some popcorn and join you back


-------------------------------------------------------------

What's wrong with MTBF? It's a valid statistics tool if you understand what it is and what kind of tool it is, ie comparing between 2 products using the same test method. No one ever said it was supposed to predict the life of a product (at 30 years or whatever).

And if you backup, you don't have to reinstall anything, what are you talking about?

I think you are not taking into account all Bootcamp users here... are you ?

 

[quagmire]

The SSD part of your hybrid drive is constantly being used, unlike regular SSDs where you use a different part everyday. Since SSDs have a limited write life, the SSD part of your hybrid hard drive will wear out quicker than a full-sized SSD.

Another aspect that made me shy away from the Fusion Drive. With the OS constantly evaluating app usage to determine what apps are stored on the SSD and what apps are stored on the HD, there will be a lot of writing being done on the SSD.

 

[joema2]

...Chance that your Fusion drive will fail is much higher than an SSD or HDD side by side.

You first said the chance of failure was doubled, I showed how that was incorrect, now you're saying it's "much higher" (no longer doubled?)

Repeatedly spinning up and spinning down the disk platters wears out the traditional part of the fusion drive faster than normal use of a traditional hard drive where the platters keep spinning.

As the Tom's Hardware article showed, there is primarily a relationship between in-use hours and Annualized Failure Rate (AFR). It is purely speculative that any relationship exists between spin up/down cycles and AFR. That sounds like an interesting theory -- what is your factual basis for asserting this as a hard fact in such an authoritative manner?

Also, lots of other HDDs besides FD automatically spin down for power saving. Are you saying these aren't conventional HDDs? Are they also wearing out faster?

...The SSD part of your hybrid drive is constantly being used, unlike regular SSDs where you use a different part everyday. Since SSDs have a limited write life, the SSD part of your hybrid hard drive will wear out quicker than a full-sized SSD.

It's true that different generations of SSDs have different write life. But how do you possibly know the histogram access pattern of a Fusion SSD vs a regular SSD? Do you have some special instrumented data that Tom's Hardware and everyone else doesn't have? Or is this just more speculation?

Let's put it this way... Your HDD or SSD fails or filesystem gets corrupted; You can recover your files using a 3rd party tool or software.

I have done low level data recovery and database patching for nearly 30 years. The ability to recover data is often very limited -- regardless of drive type. Everything isn't simple text files. Losing a limited region of a higher level hierarchical data structure can mean loss of the entire file or database, or make repair impractical.

Choosing a hard drive based on a guess about data recovery is overweighting that decision element. It's a situation you want to avoid anyway and for which recovery is unsure -- no matter what drive type. The majority of time re-loading the last backup is the most reliable, time-effective solution. Those backups are needed regardless of whether your drive is SSD, HDD, or Fusion. They can all fail.

If you read my posts, you'll see I'm advocating a balanced, reasonable perspective. I'm not saying FD is the ultimate solution.

 

[hfg]

You missed the point. If your SSD or HDD fails, Fusion drive will not work.

There is a risk of failure with pure SSD but with a fusion drive, your chances are doubled. Same applies RAID0 drives which although gives higher speeds, but a disk failure will be catastrophic.

Hmmmm ...

I was going to buy a PowerBall ticket to win $milliions$ ...

but, if I buy 2 tickets, I will have double the chance of winning!

It's a sure thing then, lets go shopping!

 

[macthefork]

It's likely the testing method...

As Bear mentioned, if the Fusion drive's SSD is full, it has to write to the HDD to make room for more than 4GB of data, which is the amount left free on the SSD portion of the Fusion drive under normal circumstances.

The default setting of BlackMagic Disk Speed Test is 5GB of data, So, your Fusion drive is likely moving data from the SSD to the HDD to make room for the "write" speed testing data.

Try setting the data size to 1 or 2 GB (under the Stress menu in Blackmagic) to see if that speeds things up. Anything less than 4GB should show you a substantial increase in write speed.