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

breiti

macrumors newbie
Original poster
Nov 1, 2013
29
1
Hi folks,

i actually searched for benchmarks and the late 2013 iMac's 1TB SSD options "just" get ~700MB/s.

The 1TB SSDs in the MacBook Pro Retinas get ~1000MB/s.

Iam about to purchase an 27" with an 1TB SSD but this is sad? So much money but worse performance than a notebook. :-(
 
My guess is your iMac performance will still blow you away.
 
There is virtually no point to all this SSD speed. It doesn't matter. The CPU will in almost all cases lag behind. The only thing that is really noticably faster is unpacking non compressed archives or direct copy operations. In many cases copying files is limited by the source or target drive which is no where near as fast. Application performance is usually CPU limited and only really cares for random file access speed not so much bandwidth.

1TB/s SSD max sequential performance past a certain point is really mostly for masturbating to benchmark numbers.
 
You won't even notice a difference in real world usage. Don't worry about it.
 
There is virtually no point to all this SSD speed. It doesn't matter. The CPU will in almost all cases lag behind. The only thing that is really noticably faster is unpacking non compressed archives or direct copy operations. In many cases copying files is limited by the source or target drive which is no where near as fast. Application performance is usually CPU limited and only really cares for random file access speed not so much bandwidth.

1TB/s SSD max sequential performance past a certain point is really mostly for masturbating to benchmark numbers.


No it doesn't. Where did you come with this?

Every made a ram disk? Programs load instantly

win.png
 
Last edited:
There is virtually no point to all this SSD speed. It doesn't matter. The CPU will in almost all cases lag behind. The only thing that is really noticably faster is unpacking non compressed archives or direct copy operations. In many cases copying files is limited by the source or target drive which is no where near as fast. Application performance is usually CPU limited and only really cares for random file access speed not so much bandwidth.

1TB/s SSD max sequential performance past a certain point is really mostly for masturbating to benchmark numbers.

This is simply untrue. It depends entirely on what you're doing and how you're doing it. As I just noted in another thread, disk I/O is the bottleneck for many database operations.
 
I would be very surprised if it wasn't the exact same SSD used in both computers. I bet they just used different benchmark software or some other reason for the difference.
 
This is simply untrue. It depends entirely on what you're doing and how you're doing it. As I just noted in another thread, disk I/O is the bottleneck for many database operations.
Disk I/Ops has very little to do with max sequential read/write. Not that database operations matter much on a notebook but the Apple (or Samsung) SSDs used with those highe 700+MB/s read speeds don't really do any better in IOPS than much much slower drives. Actually there are quite a few drives that are much faster in the IOPS department because that is clearly not what Apple optimized for. For database operations these aren't ideal drives. If you want to setup a database cluster you'd probably use different ones with possible much slower sequential speed but higher iops. It is a firmware optimization thing.
Disk I/O isn't just bandwidth and pure bandwidth is rarely the problem. Run any serious database benchmark and check what the actual MB/s read/written is and it will come in way way lower than the max seq read/write speeds which are the matter of this discussion.
 
There is virtually no point to all this SSD speed. It doesn't matter. The CPU will in almost all cases lag behind. The only thing that is really noticably faster is unpacking non compressed archives or direct copy operations. In many cases copying files is limited by the source or target drive which is no where near as fast. Application performance is usually CPU limited and only really cares for random file access speed not so much bandwidth.

1TB/s SSD max sequential performance past a certain point is really mostly for masturbating to benchmark numbers.

This is not true. If the CPU is lagging behind the CPU would show spikes to 100% when the SSD is maxed out and it doesn't. The SSD can transfer the data the the RAM which is way faster the any SSD.
 
Disk I/Ops has very little to do with max sequential read/write. Not that database operations matter much on a notebook but the Apple (or Samsung) SSDs used with those highe 700+MB/s read speeds don't really do any better in IOPS than much much slower drives. Actually there are quite a few drives that are much faster in the IOPS department because that is clearly not what Apple optimized for. For database operations these aren't ideal drives. If you want to setup a database cluster you'd probably use different ones with possible much slower sequential speed but higher iops. It is a firmware optimization thing.
Disk I/O isn't just bandwidth and pure bandwidth is rarely the problem. Run any serious database benchmark and check what the actual MB/s read/written is and it will come in way way lower than the max seq read/write speeds which are the matter of this discussion.

I wasn't contending that sequential read/writes were the barometer of DB performance. But these drives are better than their previous generation in both sequential and random -rw.
 
The 1TB SSD in the rMBPs are 4-lane PCIe channels. All other capacities are 2-lane. That's why you can get read/writes of around 1.2GB/s on a 1TB SSD rMBP.1

The SSD in the iMacs are all 2-lane PCIe, regardless of capacity.
 
Are there any IOPS benchmarks for the both SSDs available anywhere?
 
Are you sure the 1TB uses 4 lanes?

CPU only supports 16. 8 for GPU, 2 for SSD, 4 for TB, 1 for USB3 and one for slow devices such as USB2 (keyboard, trackpad, isight, audio) networking and bluetooth.
 
I've had two rMBP 15" with 1TB SSD and neither went over 1GB/s - PCIe 2.0 is 500MB/s per lane, so two lanes would be 1GB/s.

I've not seen any that went to 1.2GB/s as claimed in an earlier post in this thread so I would assume it is in-fact PCIe 2.0 with 2 Lanes.

As for the iMacs lower performance? I have no idea.
 
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.