Fusion Drive performance: 1 TB vs 3 TB

[dukee101]

To those who are so lucky as to have their brand-new 2012 iMacs already delivered to them:

Can you help us compare the performance of 1 TB vs 3 TB Fusion Drives?

Obviously, odds are you have one iMac with only one of these configurations. So just contribute your feedback about your particular config (either 1 TB or 3 TB) and we'll compare it amongst those with the other config.

I'm shopping around for the new iMac and although I don't have a need for 3 TB of capacity, I know that the 3 TB drives have 3 x 1 TB platters, which perform better than the 1 TB drives with 3 x 333 GB per platter. (The more capacity per platter, the faster the hard drive can access data).

Performance benchmarks for regular (non-Fusion) 1 TB vs 3 TB drives clearly put 3 TB drives ahead by about 20%. So this normally would make a difference, but I'm curious to see how the spread is affected when you throw an SSD in the mix as part of a Fusion setup.

And since storage I/O is such a critical way to "feel" the performance of a system (it's today's bottleneck), I'm willing to get the 3 TB Fusion Drive for the increased speed alone, IF IT'S DEMONSTRABLY THERE.

So please, let us know

 

[Metal Dice]

Hmm interesting, according to the article I linked below smaller drives might be faster. However, I might be missing something.
http://www.silentpcreview.com/WD_Red

 

[Slow Programmer]

Why couldn't the 1tb drive have only a single platter?

 

[atteligibility]

Hmm interesting, according to the article I linked below smaller drives might be faster. However, I might be missing something.
http://www.silentpcreview.com/WD_Red

Maybe you're missing most of the article you're referring to, such as the conclusion?

Performance-wise the 1TB model is nothing to write home about, more or less equivalent in speed to your typical 5,400 RPM desktop drive. The 3TB variant however, is much faster, outclassing older 7200 RPM models like the 2TB WD Caviar Black and Seagate Barracuda XT in our real world application tests.

 

[dukee101]

Why couldn't the 1tb drive have only a single platter?

That's a good question, and there's no doubt that most mainstream 1TB drives will move in this direction in the coming year. But at the moment, it's most economical to mass-produce 1TB drives in the 3 x 333 density or 2 x 500 density.

according to the article I linked below smaller drives might be faster

That silentpc review of the WD Red drive isn't as definitive on 3TB performance as this one:

http://www.hardwarecanucks.com/foru...b-review-1tb-platter-monster-unleashed-4.html

On their tests, the Seagate Barracuda 3TB does 155.8 MB/s read while the fastest 1TB does 115.8 MB/s read, a 38% increase!

I just wanted to know how that would affect Fusion Drive performance. It would be interesting to see whether the 128GB SSD mitigates the 3TB's speed advantages.

 

[Metal Dice]

What about this then:
https://forums.macrumors.com/threads/1512580/

 

[dukee101]

What about this then:
https://forums.macrumors.com/threads/1512580/

As I said on that thread as well (since you brought it to my attention), any claim about higher platter densities making drives slower is simply untrue.

Higher areal densities mean there's more data crammed into the same amount of space, which means that once the read-write head scans any given block, it will read/write more in a given moment than if the density were lower.

Increased areal density therefore leads to increased speeds. The more gigabytes you can pack onto a platter, the faster your hard disk drive should become.

Assuming 3 TB drives are made with platters with greater areal densities, they will in almost all cases be faster than 1 TB equivalents.

 

[aiming]

As I said on that thread as well (since you brought it to my attention), any claim about higher platter densities making drives slower is simply untrue.

Higher areal densities mean there's more data crammed into the same amount of space, which means that once the read-write head scans any given block, it will read/write more in a given moment than if the density were lower.

Increased areal density therefore leads to increased speeds. The more gigabytes you can pack onto a platter, the faster your hard disk drive should become.

Assuming 3 TB drives are made with platters with greater areal densities, they will in almost all cases be faster than 1 TB equivalents.

But how do you explain Barefeats' results? They clearly show that the 1TB HDD is faster.

 

[torana355]

But how do you explain Barefeats' results? They clearly show that the 1TB HDD is faster.

There are two 1TB drives apple are using, the Seagate 1TB is faster then the 3TB Seagate, both of those are faster then the WD 1TB.

 

[aiming]

There are two 1TB drives apple are using, the Seagate 1TB is faster then the 3TB Seagate, both of those are faster then the WD 1TB.

That explains a lot.
May I ask you how you came to this conclusion, is there a comparison benchmark between the HDDs that you could link to?

 

[gnasher729]

But how do you explain Barefeats' results? They clearly show that the 1TB HDD is faster.

Link to their results would be nice, but things are never too simple.

One very important factor for rotating hard drives is this: Data stored on the outside of the platter can be read at much higher speed than data closer to the inside. For data on the innermost tracks, transfer speed can be as low as 40% of the maximum transfer speed. As a result, with two identical disks, one 30% full, one 90% full, the 30% full will be on average a lot faster than the 90% full. That means if you compare a 1 TB and 3 TB drive, both with the same 900 GB data stored, the 3 TB is only 30% full instead of 90% full and that gives it a huge speed advantage. With the 50 GB apparently used in the silentpcreview article, you won't see that difference. Fill the drive up, and the 3 TB will be faster for no other reason than the bigger size. (And it fits more data as well).

 

[aiming]

Link to their results would be nice, but things are never too simple.

One very important factor for rotating hard drives is this: Data stored on the outside of the platter can be read at much higher speed than data closer to the inside. For data on the innermost tracks, transfer speed can be as low as 40% of the maximum transfer speed. As a result, with two identical disks, one 30% full, one 90% full, the 30% full will be on average a lot faster than the 90% full. That means if you compare a 1 TB and 3 TB drive, both with the same 900 GB data stored, the 3 TB is only 30% full instead of 90% full and that gives it a huge speed advantage. With the 50 GB apparently used in the silentpcreview article, you won't see that difference. Fill the drive up, and the 3 TB will be faster for no other reason than the bigger size. (And it fits more data as well).

Ahh, that made a lot of conflicting information connect inside my head! Thank you.

 

[torana355]

That explains a lot.
May I ask you how you came to this conclusion, is there a comparison benchmark between the HDDs that you could link to?

I have the 3TB fusion and split the drives up and did a blackmagic speedtest, the 3tb drive only got 175MB/s where the 1TB gets over 200MB/s as shown it that other thread, the WD 1TB is known to be a slow drive, its closer to 100MB/s.

 

[dukee101]

One very important factor for rotating hard drives is this: Data stored on the outside of the platter can be read at much higher speed than data closer to the inside.

Yes, this is exactly the kind of reason that would make getting the 3 TB drive smarter performance-wise than the 1 TB (assuming you have a good amount of data to store).

Someone should run tests with 1 TB FD and 3 TB FD iMacs with at least 500 GB of data and share the benchmarks.