Skip to main content

Long read: The beauty and drama of video games and their clouds

"It's a little bit hard to work out without knowing the altitude of that dragon..."

If you click on a link and make a purchase we may receive a small commission. Read our editorial policy.

PS3 Hard Drive Upgrade Guide

Value, performance and emerging storage tech tested by Digital Foundry.

Game install testing

Four game installs are carried out in turn: Gran Turismo 5, Mass Effect 2, BioShock and Devil May Cry 4. GT5, DMC4 and ME2 are infamous for their lengthy install times, while BioShock was chosen for other reasons: it dumps 5GB of game data onto the hard disk in just 10 minutes. We've certainly not been backward about criticising this lacklustre port in the past, but the efficiency of its install and the speed of its loading times make it an intriguing game to compare against GT5 in particular.

For the reasons we've talked about earlier concerning how drives naturally become slower the more they are filled, the tests are carried out twice: first with an empty drive, and then with the disk full of data. Yes, we did fill three drives with over 400GB of junk data in order to test this, which was probably the most insane logistical challenge in putting this piece together (and the reason this is being posted on a Sunday rather than in DF's traditional Saturday slot).

This exercise offers up some interesting comparisons, but what we can't test is "wear and tear" on the drive - gradually filling a disk but deleting old data inevitably leads to fragmentation issues, which is almost certainly what happened in our Mass Effect 2 installation analysis last month, where we saw so much variance in our install times.

First up, let's look at results of our installs on freshly formatted drives. Also note that this is a comparison of three brand new drives up against a used (albeit restored) SSD along with a four-year-old 60GB PS3 drive.

Drive Gran Turismo 5 Devil May Cry 4 Mass Effect 2 BioShock
60GB Stock PS3 (5400rpm) 43:03 21:24 27:12 10:49
500GB WD5000BEVT (5400rpm) 40:14 21:09 26:45 10:47
500GB WD5000BEKT (7200rpm) 31:01 21:17 26:54 10:43
500GB Momentus XT (Hybrid) 30:02 21:56 27:25 10:43
128GB Samsung PM800 (SSD) 26:36 22:18 26:57 10:43

The results are intriguing in that even with an SSD in the equation there is no clear winner. With Mass Effect 2 and BioShock, we see that the drives are operating with near-identical levels of performance, suggesting very much that the Blu-ray drive's data throughput is the limiting factor.

However, on the flipside we see an extraordinary range of install times with Gran Turismo 5, with our SSD clocking in at 26 minutes at one end of the spectrum while the slowest drive - the 60GB launch HDD - achieves the same task in 43 minutes. Clearly, the drive is the differentiating factor over and above the Blu-ray unit, and in certain situations the 7200rpm HDDs clearly offer a significant performance boost.

Devil May Cry 4 has a small level of differentiation between the results, the biggest surprise being that the SSD is the slowest performer in this case. Quite why is something we cannot fathom, and we ran this test twice to ensure the accuracy of the result, but there it is: DMC4 installs faster on a traditional hard disk for reasons which defy rational explanation.

Next up, all of these tests are repeated with each drive approaching capacity. The 60GB PS3 drive acquits itself well here, but it is important to note that we had to leave enough room for all of our installs to fit in the remaining space, so allocating 20GB of spare space isn't exactly the same kind of stress test as leaving 50GB free on a 500GB drive.

The SSD was not included in this test as flash RAM is not subject to the same data throughput limitations as a mechanical drive. However, for the sake of comparison, the existing results are factored into the table.

Drive Gran Turismo 5 Devil May Cry 4 Mass Effect 2 BioShock
60GB Stock PS3 (5400rpm) 43:16 21:20 27:02 10:53
500GB WD5000BEVT (5400rpm) 71:50 21:30 27:14 10:44
500GB WD5000BEKT (7200rpm) 37:58 20:52 26:56 10:42
500GB Momentus XT (Hybrid) 33:09 21:58 27:30 10:43
128GB Samsung PM800 (SSD) 26:36 22:18 26:57 10:43

Interestingly we see in some cases that performance is actually faster compared to an empty drive, which is very surprising. This suggests that throughput from the Blu-ray drive on our unit may well be quite inconsistent in certain situations.

We see that once again Mass Effect 2 and BioShock post very consistent results effectively identical to the drive when empty, which really does suggest that in these cases the Blu-ray drive is the limiting factor. However, on GT5 and DMC4 the benchmarks show that drive performance has degraded. The real surprise is our cheapo 5400rpm Western Digital Drive, which collapses horribly on GT5 install times, taking well over an hour to complete, but otherwise still competitive.

It's interesting that an almost full 7200rpm drive can still outperform the 5400rpm drives (even when empty) but the biggest takeaway we get from this is the uniform benchmarks from the ME2 and BioShock tests. It suggests that there is an installation mechanism developers can adopt which provides level performance regardless of drive.

Why are install times so variable?

To better understand what these results actually tell us it would help to find out what the PS3 is actually doing during these installs. To do this, we compared what was installed on the hard disk with what is found on the original Blu-ray disc for each game.

We already know what's going with Gran Turismo 5 thanks to our initial GT5 installation analysis. The game is dipping into a single 12GB archive on the Blu-ray, selecting and extracting 40,000 files and dumping them individually onto the hard drive. It is a staggeringly inefficient way of going about things, but at least Polyphony didn't simply transfer 40,000 individual files from the Blu-ray disc itself. That could have been seriously nasty in terms of wear and tear on the BD drive mechanism.

On the flipside, the secret to BioShock's speedy install - fast in relation to the amount of data being copied, at least - appears to be entirely down to the amount of files being transferred: just 370 in all for 5GB of data. Again, like Devil May Cry 4, the vast majority of the game is simply being copied directly off the BD, but it's clear that both the Blu-ray drive and the HDD operate best when copying a small number of large files. The more continuous reads and writes, the faster the performance.

Devil May Cry 4 and Mass Effect 2 dump files directly across from the BD just like BioShock. ME2 at around 4.3GB takes three times as long to complete as BioShock, even though it is transferring a smaller amount of data. Again, it's all down to the number of files being transferred.

DMC4's 5GB install comprises 1035 different files of varying sizes: every major in-game asset appears to be incorporated into individual .arc files, each of which is stored on the HDD. Mass Effect 2 simply dumps across around 50 per cent of the game files stored on the BD: 2238 files in total make their way across.

All installs are not equal, clearly, but generally speaking, the fewer files to read and write, the better it is for both your patience and the wellbeing of the console hardware.

It is worth pointing out that developers do have an option available in avoiding the seek times of the Blu-ray drive by bundling up all the files into a single archive. Sony delivers all of its demos, DLC and PSN games in what's known as a .PKG file - think of it as the PS3's version of a ZIP archive. Installation files can be packaged into .PKG files too: Resident Evil 5 does this, for example.