Feedback

PlayStation 4 (PS4) Blu-ray and Download Storage Performance, Boot Times

Author:
Manufacturer: Sony

Does downloading make a difference?

This is PART 2 of our testing on the PlayStation 4 storage systems, with the stock hard drive, an SSHD hybrid and an SSD.  Previously, we compared performance based on Blu-ray based installations though today we add downloaded titles from PSN to the mix.  Be sure you read PART 1, PlayStation 4 (PS4) HDD, SSHD and SSD Performance Testing.

I posted a story earlier this week that looked at the performance of the new PS4 when used with three different 2.5-in storage options: the stock 500GB hard drive, a 1TB hybrid SSHD and a 240GB SSD.  The results were fairly interesting (and got a good bit of attention) but some readers wanted more data.  In particular, many asked how things might change if you went the full digital route and purchased games straight from the Sony's PlayStation Network.  I also will compare boot times for each of the tested storage devices.

You should definitely check out the previous article if you missed it. It not only goes through the performance comparison but also details how to change the hard drive on the PS4 from the physical procedure to the software steps necessary. The article also details the options we selected for our benchmarking.

View Full Size

Today I purchased a copy of Assassin's Creed IV from the PSN store (you're welcome Ubisoft) and got to testing.  The process was the same: start the game then load the first save spot.  Again, each test was run three times and the averages were reported. The PS4 was restarted between each run.

View Full Size

The top section of results is the same that was presented earlier - average load times for AC IV when the game is installed from the Blu-ray.  The second set is new and includes average load times fro AC IV after the installation from the PlayStation Network; no disc was in the drive during testing.

Continue reading our story on the performance testing of HDD, SSD and SSHD with downloaded and Blu-ray installed games on PS4!!

First, the load times across the board were faster for PSN installs than they were for Blu-ray installs, which actually kind of boggles the mind.  It seems obvious that read from the optical drive is taking some time even when Sony is installing the necessary content on the hard drive itself.  Some kind of authentication is likely taking place in a pattern (or even continuously) to make sure the disc hasn't been removed.  Even with the stock hard drive we see load times drop from 35 seconds to 28 seconds, a decrease of nearly 20%.

Both the SSHD and the SSD results decrease as well though: 20% and 14% respectively.  Keep in mind that the SSHD load times are SLOWER for the first run but decrease in the 2nd and 3rd run as more of the data becomes cached. 

View Full Size

If you read the previous article, you should remember that we perceived a see-saw pattern of load times with the AC IV test when installed from Blu-ray drive.  That no longer seems to take place with the PSN downloaded version of the game.  With the same SSHD at work, the load times were consistent once the data was cached. Very little difference was seen between runs 3-7.  It would appear that some of the disc read times might be randomly inserted, causing the variance we saw with the disc-installed setup.

View Full Size

Finally, users were asking about boot times with the three storage options so I did some quick testing.  Clearly the SSD is the fastest of the three once again getting to the menu 5.5 seconds faster than the stock drive (nearly 25%).  The Seagate SSHD does have better performance as well but carefully consider if those 5 seconds you save are worth it to you. 

Downloading versus Blu-ray Installing

It surprised me, but there is definitely a performance advantage to installing your games from the PSN (and going all digital) compared to installing from a Blu-ray.  Obviously the PS4 is doing some kind of checks with the disc in the drive on a semi-regular basis and not simply checking it once during game load time.  It it were, and if the data downloaded were the same as the data copied from the disc, then load times should be identical; they aren't. 

Potential SSHD Issues

Even though we saw a lot of the oscillation of load time performance go away with the downloaded version of Assassin's Creed IV, there are still some obvious concerns with using a hybrid HDD + SSD solution as they currently exist.  First, though our repeated loading tests show improvements in performance, keep in mind we were loading the same level over and over, not loading new levels or different ones.  I would imagine that frequent changing of games (starting up a game of NBA 2K14 for example) would start to throw the cache for loop and the performance advantages would come and go depending on your particular gaming habits. 

A solution to this (or at least a reduction to the ailment) would be to increase the size of the flash on the drive.  Current Seagate options are limited to 8GB of flash. Increasing that to 32GB would likely allow a lot more caching to occur and make your gaming experiences better overall.  I'm actually surprised neither Sony or Microsoft thought of this option (or at least failed to act on those good insights) for their storage implementations. 

View Full Size

TRIM Support?

Several people have asked about using SSDs with the PS4 without knowing if TRIM is supported on the platform.  While I still don't know for sure if it is, a quick chat with Allyn, our storage guru, was a bit more reassuring.

I don't think the installation process is going to be doing random writes to an SSD, and sequential writes do not impede performance over time.  Remember, it is the sequential write process that actually can restore an SSD to full read/write performance without TRIM in use.

Essentially, because of the installation process going on with PSN installed games, it could be "fixing" any issues that the SSD would suffer without a TRIM-aware operating system anyway.  Random writes, the action that can actually fragment the flash, would be pretty minimal in a console environment with exception of saved games.  Even captures of video content on the PS4 would long, extended (and sequential) writes. 

Why is there no 7200 RPM drive tested?

Honestly, I just couldn't find a 500GB or 1TB 7200 RPM laptop drive in stock today, though, again talking with Allyn, we decided that if we had one to test, the performance would be much closer to that of the stock drive than the SSHD.

Closing Thoughts

My testing testing of the storage performance of the PS4 has been interesting but is no means complete or all-encompassing.  We'll likely be trying out some new games in the coming days, both digital and disc-based, but until then...anyone want to buy a Blu-ray copy of Assassin's Creed IV?

This is PART 2 of our testing on the PlayStation 4 storage systems, with the stock hard drive, an SSHD hybrid and an SSD.  Previously, we compared performance based on Blu-ray based installations though today we add downloaded titles from PSN to the mix.  Be sure you read PART 1, PlayStation 4 (PS4) HDD, SSHD and SSD Performance Testing.

November 18, 2013 | 08:58 PM - Posted by foo (not verified)

Time to buy that $500 1TB 840 EVO to use in the $400 PS4 so I can play the $60 games times X. Ouch my wallet hurts.

November 18, 2013 | 10:56 PM - Posted by SkldRblds (not verified)

After all that cost...
Ouch,.. Money for Games.. lol
Pretty much buy a decent PC for that...
If THIS publication and relative ones get more attention...More better

BluRay seeking performance hit VS SSD/SSHD gets more public, they may actually fix it..
Share the article boys!!!

November 18, 2013 | 11:39 PM - Posted by unlovedhomie (not verified)

Thank you for taking the time (and money ) to run these 2nd wave of test Ryan. When Sony said that psn and disc game can't use the same install I felt that the files can't be the exact same. As far as buying ac 4 , what's your asking price ;)

November 19, 2013 | 01:10 AM - Posted by Phat (not verified)

LOL! I think that you should not buy a PS4 or Xbox One at the moment and of course I have reasons for saying that. You can see it here: http://lovingtheclassicsreviewsite.net/trend/why-you-ought-to-wait-befor...

November 20, 2013 | 03:48 PM - Posted by StewartGraham (not verified)

I suspect of the arguments made will be found weak or compelling only to those who have never purchased any major next-generation hardware... ever. Generally speaking the author is writing: "Wait and your patience shall be rewarded with a greater selection of games, value, more console support, et. al." Well... Duh.

November 19, 2013 | 01:28 AM - Posted by Hunter926 (not verified)

Hey, just wanted to say thank you for taking our input and making a part 2.

Thanks to this I've decided to get a SSD again this generation =)

I will be watching for the further testing that you said you may do.

thanks,

Hunter

November 19, 2013 | 01:30 AM - Posted by Anonymous (not verified)

Excellent meticulous work, kudos!

I am very disappointing by Sony's decision to go with SATA II, limiting the value of SSDs.

November 19, 2013 | 05:03 AM - Posted by Robert23655124 (not verified)

Make a 7200rpm hybrid w/ 32gb of cache and you've got the perfect console upgrade.

There isn't anything else to be done for performance upgrades so I could see a lot of people paying a premium price :)

November 19, 2013 | 08:52 AM - Posted by Anonymous (not verified)

FYI, if you want to find out about TRIM support then you could fill the drive with data in the PS4, delete it, and plug it into WinHEX in a Windows PC. If the blocks get zeroed out then that would indicate TRIM support.

Sony's disk encryption shouldn't be a factor in whether the blocks get zeroed or not. I would also be quite curious.

November 19, 2013 | 12:51 PM - Posted by Anonymous (not verified)

Any chance that you might consider testing the western digital black hybrid drives?

They come with 16gb of flash vs 8gb on the seagate drives. You can buy them on ebay brand new for a similar price. Yes they are OEM drives but it would be interesting to see benchmarks on if that extra 8gb of flash makes a difference. More options are always a plus especially if you get extra performance for the same price. Search WD10J12X on ebay to find them.

November 19, 2013 | 02:12 PM - Posted by Anonymous (not verified)

There is a factual error in the piece, the 750GB and 1TB SSHD by Seagate have 32GB of NAND cache already. It was the original first series 500GB Momentus XT that only had 8GB.

November 23, 2013 | 10:25 AM - Posted by Ryan Shrout

I keep hearing these mentioned but have yet to see a product link anywhere. And trust me I have looked. I don't think they exist...

June 24, 2014 | 10:01 PM - Posted by Synbios (not verified)

The 32 gig nand will be out end of july

November 19, 2013 | 04:02 PM - Posted by Anonymous (not verified)

What about heat and power consumption numbers?

November 19, 2013 | 04:52 PM - Posted by Calinct (not verified)

I got a good deal on a 1tb 7200rpm hgst travel star on amazon a few days before launch and installed on in my ps4 on launch day.

I''ve done a bit of testing, replicating your methods, with a blu ray version of ac4.
My times run a little more than 3 seconds (3.58 to be exact) quicker than your hgst 500mb 5400 rpm drives tests.

November 19, 2013 | 04:59 PM - Posted by Anonymous (not verified)

Very much thank you for these tests. Just ordered mine Seagate for 110 €, better be enough for rhe coming years :)

November 19, 2013 | 05:41 PM - Posted by KittenMasher (not verified)

I wonder if enough power is sent to the HDD sata power connector to run a 3.5" desktop HDD. You would need extenders and the willingness to make your system look hideous, but the idea of putting a 4tb drive on there isn't unappealing. (I suppose you could just use dedicated power from another psu for the drive, but you gotta draw the line somewhere.)

November 20, 2013 | 06:09 PM - Posted by aparsh335i (not verified)

Thats a cool idea, similar to having a 4TB external but should be a faster connection.
You must have a vast collection of 3D Blu-Ray Pr0nz to need that much space ;).

December 16, 2013 | 12:46 PM - Posted by Screech (not verified)

Actually, I've did this with my 60G PS3. I got a Thermaltake Blacx SATA drive dock with its own power supply. Got a cable to convert the hard drive connection in the PS3 to an eSATA connection for the drive dock and installed a 3.5" 10k rpm 74G drive. I did this since I was running Linux on my box and having the faster drive helped with that. Also, this removed all heat generated by the drive from the console as well as removing that power load from the console's PSU.

For my PS4 I'm considering an external RAID enclosure with an eSATA interface. Thinking a pair of those 1T SSHD in RAID-0 would be nice.

November 21, 2013 | 12:56 PM - Posted by Quimozabi (not verified)

Excellent and very informative post, I've decide to go with a Toshiba SATA 2TB 7200RPM 64MB for $110.00, and if you are still selling the AC disk contact me. nelsonspcrepair@yahoo.com

November 26, 2013 | 06:59 PM - Posted by Anonymous (not verified)

Is there a chance Sony on the mid cycle fee fresh in a year or so, move beyond SATA II and give an SSD option? Is the interface they went with that limiting?

November 27, 2013 | 02:26 PM - Posted by Belpela (not verified)

Its highly probable that the current ps4 already support sata III but unless we see 500 gb ssd price fall to around 50-100 $ in the next 7 years, I doubt that we will see an ssd option in the ps4 packs.

November 27, 2013 | 08:46 PM - Posted by mAxius

need the new black 2 drive to be tested :D

November 28, 2013 | 09:48 AM - Posted by Belpela (not verified)

The black2 wont work, as it needs a software install for it to work and that software is windows only so far.

February 12, 2014 | 01:28 AM - Posted by Shane (not verified)

The assertions in the article regarding TRIM are incorrect. Random versus sequential writes have nothing to do with the benefits of TRIM usage.

There are two primary reasons for using TRIM or similar commands such as UNMAP in SCSI.

1. Performance:

Due to the difference between how a flash memory block can be written if the block contains no data, versus how it can be written if it contains so much as a single bit of data.

For a block of flash memory to be read, it is simply read like any other typical addressable block device.

However for a block of flash memory to be written, it can only be written simply like any other typical addressable block device IF the block being written contains no data according to the SSD itself and not the system using it.

If the block of flash memory to be written does contain data, even if it is marked as unused at the filesystem level by the OS, that data needs to be erased before new data can be written. So an erase operation needs to be performed before a write operation can be performed, which impacts performance.

Worse, if the OS addresses the block device at the typical and traditional block size of 512 bytes, but a block of data on the flash device is larger than that (typically 4K), it may first read the flash block into on-board disk cache memory, then erase the flash block, amend the copy of the block in memory with the 512 byte change and then write that full 4K block to flash.

Dumb old SSD's could then do that same full operation another 7 times if 512 byte blocks were sequentially written to the full block, causing terrible performance and unnecessary wear to the flash memory.

Thankfully a smart modern SSD will avoid this as much as possible, however if a flash block contains any data, it does still need to perform an erase operation before a write operation can be performed, due to the unavoidable way flash memory works.

A similar situation can occur if larger blocks are used but not aligned to physical blocks, even if the addressed or written block size is the same as the physical size. Again, a good SSD controller can help with this situation, but not completely cure it.

TRIM assists with performance by allowing the OS to tell the underlying storage device that a block is no longer being used and so is safe to erase when I/O rate is low or idle, so that the next time it is written, an erase operation will not be needed as it was already performed during a more appropriate time when the drive was largely idle.

The requirement to erase a previously used flash block before it can be written again, has nothing to do with whether data was written randomly or sequentially.

Some SSD drives will mark a block unused if that block is completely filled with zeroes, which might be where the idea that sequential writes fixes performance problems comes from.

2. Capacity:

TRIM and SCSI UNMAP commands can be used to reclaim blocks of data for storage systems which provide shared data at the block level. Particularly with SAN's and storage virtualization systems which provide "thin provisioned" storage.

Without the use of TRIM or UNMAP, a used block of data from a thin provisioned storage system cannot be given back to the storage system once it is no longer being used, as the underlying storage system does not have knowledge of the importance of any block of data in the filesystem it is being used for to store. Without TRIM or UNMAP, the underlying storage system cannot tell the difference between current data and deleted data and so must treat all provisioned blocks of data as equally important.

HUGE CAVEAT: Full Disk Encryption:

With real literal Full Disk Encryption, ALL blocks of data on a block device should be in use with data written from the perspective of the storage device.

Even long runs of zeroes will encrypt to data on disk that should theoretically be statistically indistinguishable from a sampling of white noise.

The consequence of a PROPER implementation of Full Disk Encryption with flash based storage is that there is never an opportunity to gain any benefit from TRIM or SCSI UNMAP commands, as all blocks contain data at the storage system level, even if not at the decrypted filesystem level.

So with a proper implementation of Full Disk Encryption, TRIM and UNMAP cannot be used to regain write performance or unused blocks from thin provisioning.

So if it is true that the PS4 uses Full Disk Encryption and it is a proper full implementation of it, then TRIM support is completely pointless.

However, note that there are some vendors which are erroneously calling their disk encryption systems FULL Disk Encryption, when they patently are NOT. Microsoft FDE in Windows 7 is supposed to be FULL, but in Windows 8 (or was it 8.1?) they introduced FDE that only encrypted blocks that were used and future writes to unused blocks. This is NOT FULL Disk Encryption, however some vendors are willing to stoop to claiming that it is in order to gain some performance.

Post new comment

The content of this field is kept private and will not be shown publicly.
  • Lines and paragraphs break automatically.
  • Allowed HTML tags: <a> <em> <strong> <cite> <code> <ul> <ol> <li> <dl> <dt> <dd> <blockquote><p><br>
  • Web page addresses and e-mail addresses turn into links automatically.

More information about formatting options

By submitting this form, you accept the Mollom privacy policy.