PS3 Hard Drive Performance Test: 5400.3, Scorpio Black, X25-M G2

Introduction

Questions about hard drive upgrades regarding the PS3 have been asked many times.  I have asked this question on multiple occasions and have never gotten a concrete answer.  I have gotten some “out there” responses such as “All drives are automatically reduced to 5400RPMs” and “The PS3’s hardware is only compatible with 5400RPM drives.” I’ve also gotten some responses that have merit, like “The Blu-Ray drive will bottleneck the drive anyway” and “SATA1 will limit faster drives.”  So, I decided to find out for myself how hard drives compare to each other when installed on the PS3 and if the BD-ROM and/or SATA1 limits load times.

Hardware Info

Seagate Momentus 5400.3 80GB 2.5” HDD (PS3 Stock HDD)
Seagate Official Specifications

Configuration
Disks/Heads: 1/2
Cache: 8MB
Transfer Rate
Sustained Internal: 44MB/s
Maximum External: 150MB/s
Average Seek: 12.5ms
Average Latency: 5.6ms
Power Consumption
Read/Write: 1.9W/1.8W
Idle/Standby: 0.6W/0.2W

Western Digital Scorpio Black 7200RPM 160GB 2.5” HDD
Western Digital Official Specifications

Configuration
Disks/Heads: 1/2
Cache: 16MB
Transfer Rate
Host to/from Drive: 100MB/s (max sustained)
Average Latency: 4.2ms
Power Consumption
Read/Write: 2.1W/2.1W
Idle/Standby: 1.1W/0.2W

Intel X25-M G2 80GB 2.5” SSD
Intel Official Specifications

Configuration
10 Parallel Channel Architecture with 34nm MLC ONFI 1.0 NAND
Transfer Rate
Sustained Sequential Read: up to 250MB/s
Sustained Sequential Write: up to 70MB/s
Latency: 65µs/85µs
Power Consumption
Active: 150mW Typical
Idle: 75mW Typical

Methods

The tests were performed by using a stopwatch to record load times.  Since the start and stop times on the stopwatch were controlled by me and not a computer, there is a chance of “human error” in the form of precision.  All the tests were done on offline single player.  This is because when I tried multiplayer tests the results consistently varied too much to do any analysis.  Three load times were recorded for “OS to Menu” and “Menu to In-Game” per drive per game, for 180 samples.  Three load times were recorded for “Off to OS” per drive, for an additional 9 samples.  Those 3 times were averaged together to help eliminate some of the “human error.”

Off to OS = Load time between power button “beep” and user selection screen.

OS to Menu = Load time between selecting the Blu-Ray icon and game menu (spam “Start” and/or “X” while loading when needed).

Menu to In-Game = Load time between “the last button press to load a saved game” and being “in-game.”

Results

The majority of the time, the first load would take anywhere between 5s and 20s longer than subsequent loads. This is because the PS3 stores data in its RAM, and on the first load the game’s data isn’t in the RAM. To compensate for this the first load times were ignored in all calculations (4 loads were actually done, but the first was ignored to make 3).

Off to OS

OS to Menu and Menu to In-Game

In a few instances, load times actually increased when going to a “faster” drive.  This happened on Fallout 3, Call of Duty 4, Resistance, and Motorstorm when going from the 5400.3 to the Scorpio Black.  This was not a one-time fluke, times were consistently slower on the Scorpio Black with these games (usually less than 1s, but slower nonetheless).  I do not know why this happened, everything about the Scorpio Black looks faster than the 5400.3; more cache, higher density platters, and more RPMs.

Percentages can be misleading when the overall load time is relatively small in all drives. So, the average change in seconds between drives is shown along with the percent performance increase.

One would think that the size of the game data and save games on the drive would affect the load times of games. So, the two load times were added together to get an “OS to In-Game” load time, and the size of the game data and save games were added to get a “Total Game Data.” Then they were plotted with the x-axis being the increasing size of game data.

Conclusions

Is a Faster Drive Worthwhile?

Yes, upgrading to a 7200RPM 16MB Cache from stock gives added speed and storage for cheap.  For example, I paid $50 for the Scorpio Black and got more speed and double the storage.  If you need more storage anyway, as many people do, then getting a 7200RPM drive is a “no-brainer,” in my opinion.

Upgrading to a SSD for the PS3 shouldn’t even be a thought; I mean SSDs cost almost as much as the PS3 itself and in some cases more.  I cannot justify the price for a ~15% increase in performance, but if you have the cash then by all means go for it.

Is BD-ROM a Limiting Factor?

2x Blu-Ray Drive (72Mb/s or 9MB/s)

I used the amount of game data installed as a reference to how much the Blu-Ray would be used.  My assumption was that, less game data installed meant that more data would have to be read from the Blu-Ray, and more game data installed meant less reading from the Blu-Ray.  It seemed like a logical assumption to me, but the data from the “Data Size -vs- Load Time” graph shows that some of the games with the most (Oblivion) AND least (Call of Duty 4, Dynasty Warriors 6) game data both had some of the fastest load times.  Based on this, I would have to say the Blu-Ray is not limiting the load times of games.

Is SATA1 a Limiting Factor?

SATA 1.5Gb/s (178.81MB/s; w/ encoding overhead it is 1.2Gb/s or 150MB/s)

Average reads, 35MB/s (5400.3) vs 150MB/s (SATA1 limit), suggest a 428.5% improvement is possible, but from the numbers that is obviously not the case.  Since the X25-M G2 has average reads well over 150MB/s and load time improvement between the two is only ~18%, average reads do not give a good indication of a drive’s performance in the PS3.  So, it seems that the X25-M G2’s average reads in the PS3 could be estimated to be around (35×1.18) = 41MB/s.

Now burst speeds, if we take the 5400.3’s average burst speed and divide by SATA1’s limit we get:  124/150 = 0.826 or 82.6% of the SATA1 limit.  This means there’s room for around 17.4% performance gain on the SATA1 interface.  Both the Scorpio Black and X25-M G2 have burst speeds of at least 150MB/s.  Since the Scorpio Black and X25-M do not perform the same, burst speed doesn’t seem like a good indication of drive performance either.

After going through the results, it doesn’t seem like the SATA1 interface ever gets saturated, even though a couple of the drives are capable of 150MB/s.

Possible Improvements

• Test 5400RPM and 7200RPM drives that have the same amount of cache and same platter density.  This will find load time differences based on RPMs alone.
• Test two 5400RPM OR two 7200RPM drives with different cache sizes and same platter density.  This will find load time differences based on cache size alone.
• Test two 5400RPM OR two 7200RPM drives with same cache size, but different platter density.  This will find load time differences based on platter density alone.

– MattNo5ss