Seagate Multi-actuator hard drive breaks throughput record with Sequential performance of 480 MB/s

[cbn]

But that is for professional use, where there is always need for more storage.

Eventually affordable cameras with 8K sensors will come.

Then Atomos (and others) will predictably have a external recorder for it

(If 4K30 DNxHR HQX needs 104 MB/s then (extrapolating upwards) the 8K30 version of that will need a consistent 416 MB/s write. I think SATA 6Gbps should still work for this, but in order to use the entire the drive even the portion of the platter near the spindle will need to be able to write a consistent 416+ MB/s.)

P.S. The 8K60 would need a consistent 832+ MB/s write.....this is well above SATA 6Gbps.

 

[nosirrahx]

instead of multiple actuators, one wonders if they couldn't put multiple heads together on one arm so they could read/write multiple tracks simultaneously

It is far easier to do asynchronous read/write operations if the heads are not physically linked. Drives already have multiple platters and read/write heads, there is only so much you can do to improve performance if all heads move as one unit.

 

[nosirrahx]

But that is for professional use, where there is always need for more storage.

In general you are right. Home user massive storage is for media and media does not need crazy performance to play without issue.

You really only need blazing media access if you are editing it.

On my son's gaming system we went with a 4TB SSD and there is no way it will ever go over 75% full.

 

[tynopik]

It is far easier to do asynchronous read/write operations if the heads are not physically linked. Drives already have multiple platters and read/write heads, there is only so much you can do to improve performance if all heads move as one unit.

for random operations sure, but for bulk i/o, certainly there's going to be plenty of times when data is on adjacent tracks, so instead of read track 1, move to track 2, read track 2, move to track 3, read track 3 if you could simply read all 3 tracks at once, surely that would be faster

 

[nosirrahx]

for random operations sure, but for bulk i/o, certainly there's going to be plenty of times when data is on adjacent tracks, so instead of read track 1, move to track 2, read track 2, move to track 3, read track 3 if you could simply read all 3 tracks at once, surely that would be faster

That is why sequential speed on HDDs is already not terrible. Its the non-sequential stuff that HDDs are terrible at.

 

[tynopik]

That is why sequential speed on HDDs is already not terrible. Its the non-sequential stuff that HDDs are terrible at.

yes, but this entire thread is about improving sequential speed

it just seems less mechanical complexity and more scalable to put multiple heads on each arm

does anyone remember those old Kenwood 72X True-X CD drives that would read 7 tracks at once with a laser splitter?

 

[cbn]

On my son's gaming system we went with a 4TB SSD and there is no way it will ever go over 75% full.

4TB is a lot of space for a gaming drive.

But as these articles point out game sizes are increasing:

https://www.extremetech.com/gaming/256790-game-installation-sizes-still-exploding-upwards

https://www.pcgamer.com/how-game-sizes-got-so-huge-and-why-theyll-get-even-bigger/

With that noted, I do wonder how much the lowest common denominator (console: 8GB RAM for PS4, PS4 Pro, Xbox One (12GB RAM for Xbox One X) + 2.5" HDD) is holding back game sizes?

That (console) is not a lot of I/O.

(In the future) With NVDIMMs (caching the small files and providing a greater amount of system memory) + multi-actuator HDD or SSD do we finally get enough I/O to really push game (and simulation) install sizes upward without making the larger amount of data a bottleneck?

How soon does 8TB become the new 2TB?

 

[cbn]

With more actuators on the way what is the next step for SATA?

SATA Express?

SATA 12 Gbps?

SATA 12 Gbps Express?

(Or could it be that consumer boards begin to use SAS?)

P.S. Some additional info from this article.

Seagate says the drives could use SAS, SATA, or NVMe interfaces, but the company will respond to the needs of its customers to develop the final solutions.

For instance, the venerable SAS interface features dual-port connectivity that provides two separate pathways into a single drive. The additional pathway can be used for multi-pathing or fail-over. The pathways can also be combined into a dual port connection to offer up to twice the performance from a single drive, as we see with some of Seagate's cutting-edge enterprise SSDs. The NVMe protocol also supports similar features.

 

[Billy Tallis]

With more actuators on the way what is the next step for SATA?

Nothing. It doesn't look like dual-actuator drives will need any more than a 6Gbps SATA link. Drives with more than two sets of actuators will be way too expensive for consumers, especially because such drives will probably only be made in capacities that are far beyond what makes sense for the consumer market. Any multi-actuator drives that can use more than SATA bandwidth will be sold into a market where SAS is already widely supported. The closest we could see to SAS making inroads into the consumer market would be consumer NAS devices with SAS ports.

 

[nosirrahx]

4TB is a lot of space for a gaming drive.

But as these articles point out game sizes are increasing:

The system was overbuilt with the plan of only upgrading GFX card for the next 4 to 6 years.

With more actuators on the way what is the next step for SATA?

SATA Express?

SATA 12 Gbps?

SATA 12 Gbps Express?

(Or could it be that consumer boards begin to use SAS?)

P.S. Some additional info from this article.

I don't think we will see SATA 600 going anywhere. If people want crazy fast sequential speed they will use a SSD or SSD cache and for 4KQ1T1 SATA 600 never be the bottleneck for a HDD.

It keeps costs down to use the same interface and people can use existing motherboards.

What I would to see is better tiered storage that is transparent to the user and does not suffer from 4K performance degradation.

It would be cool to put a large HDD, medium sized SATA SSD and small M.2 NVMe drive in a system and have the OS just figure out optimal caching without the user needing to deal with 3 drive letters.

 

[cbn]

The closest we could see to SAS making inroads into the consumer market would be consumer NAS devices with SAS ports.

If people want crazy fast sequential speed they will use a SSD

With SAS being 21.5% faster per PCIe lane than NVMe, it is too bad faster SAS controllers haven't been developed (yet).

But maybe NVDIMMs (caching for small files*) would provide the incentive?

(That could potentially make for one blazing fast NAS. Fastest 4K QD1 Read + Fastest Sequential Read in one box.)

*Optane software does this.

 

[cbn]

Drives with more than two sets of actuators will be way too expensive for consumers, especially because such drives will probably only be made in capacities that are far beyond what makes sense for the consumer market.

Maybe not.

If Seagate gets 12 actuators on a single pivot there is still dual pivot to consider which would hypothetically increase actuators to 24 (for a 12 (.381mm glass) platter HDD) at the cost of some capacity.

This could be scaled down to a lower number of platters (without helium) and used with NVDIMMs (caching the small files).

 

[cbn]

https://www.anandtech.com/show/1282...ive-blog?utm_source=twitter&utm_medium=social

02:27PM EDT - Use Optane for random access, NAND for sequential access

I wonder how much that might change in 5 to 10 years? (This if low rpm multi-actuator drives succeed*)

*The greater the number of actuators (per X number of platters) the greater the performance per watt.

 

[Dasa2]

This would work great in combination with a ssd cache that is smart enough to only cache the small scattered files leaving large files to the hdd

 

[cbn]

This would work great in combination with a ssd cache that is smart enough to only cache the small scattered files leaving large files to the hdd

Besides Optane software (which caches small files), I have read here that Bcache (Linux) can be tuned for file size.

2- sequential read cutoff

The other thing you might wish to tune is the size of the sequential read/write cutoff, we want a size short enough to be worth caching, by default, it is 4MB, so that everything under 4MB sequential will be cached, I personally like to take that down to 1MB judging by the fact that files larger than 1MB do read pretty fast directly from the disk ! but surely, this will depend on your application and on experimentation with your application.

cache 1 megabyte and smaller

echo 1M > /sys/block/bcache0/bcache/sequential_cutoff
cache everything (special value, not the same mathematical logic of less than)

echo 0 > /sys/block/bcache0/bcache/sequential_cutoff
back to caching 4 mega bytes and smaller (default)

echo 4M > /sys/block/bcache0/bcache/sequential_cutoff

I haven't used this software (Bcache) yet, but that does sound like it would be fun to play with.

P.S. I wonder if there is a way to eventually make Bcache a smart cache? A cache that evaluates and optimizes once maximum capacity of the system is reached?