Tech-Hounds.com

80 and 160 GB SATA Hard Drives Round Up

SATA II hard disk drives
7200 rpm, 80 & 160 GB

You don't have to be a gamer to understand the value of hard disk drives. They have been the main stay for storage in PCs for a long time. Like processors, the biggest improvements in size, speed and quality have occured in recent years. The use of more, larger platters have made possible to reach capacities unheard of in desktops several years ago - you can easily find drives with 500 GBs of space nowadays. Storage interface also enjoyed a much needed boost with the introduction of first generation SATA drives and onboard controllers in motherboard chipsets. SATA II brought support for features usually reserved for enterprise use - NCQ, hot swap - to desktop drives.

The introduction of onboard SATA I and II controllers have also propelled the widespread use of RAID, particularly RAID 0 (striping) and 5 (striping with parity). Those who have experience hard drive failures know the value of making backups and of course RAID 1 and 5 fault tolerant capabilities. Most onboard storage controllers now support a variety of RAID levels RAID 0, 1, 0+1, 5 are RAID levels commonly supported by various onboard controllers

For the most part, desktop users are more concerned with performance in single drive configurations and RAID 0 (striping). RAID support do offer the capabilities to 'squeeze' more performance out of hard drives, but not without some compromise. Even if they don't use RAID capabilites, users can still use two (or more) drives in their system - one for the systems and applications and the other for data.

Out of the many SATA II hard drives available on the market, the drives we're going to look at today are the 80 and 160 GB Hitachi's  Deskstar (HDS721616PLA P220 and HDS728080PLA PF20), Seagate's Barracuda 7200.9 (ST3160211A and ST380211AS) and Western Digital's Caviar SE (WD800AAJS and WD1600AAJS).

Problems

At first, we planned to use the onboard ICH8R SATA controller for testing purposes. Everything ran just fine during our initial tests with the 80 GB Hitachi Deskstar. We followed Intel's advice to use Hitachi's own Ftool utility to enable SATA II mode on the drive as recommended. However, we started noticing very odd behavior when we started testing drives from a different manufacturer, the 80 GB Samsung Spinpoint. Fearing this is caused by defective samples, we return the drives to the local representative in our area. Its when we put both the Western Digital's Caviar SE and Seagate's Barracuda drives that we began to suspect that the problem lies on the controller, not the drive.

The problem we saw was quite odd. Any write operation on the drive will cause very high processor loads, something like running a hard drive or CD / DVD-ROM in Programmed I/O (PIO) mode instead of UltraDMA. This basically affects all drives except the Hitachi Deskstar and our own Maxtor DiamondMax 10 drives.. We've tried plugging the cable to different SATA ports, switching cables, falling back to IDE instead of ICH8R's AHCI and RAID mode, even enabling Legacy IDE mode. Nothing seems to work. Driver problems? We're already using the latest, up to date drivers from Intel and driver issues does not explain the flawless performance of the Hitachi and Maxtor drives.

The Seagate Barracuda 7200.9 drives are not as severely affected by this problem, unlike the Western Digital Caviar SE drives. Below is a snapshot of Intel Matrix Storage Console before and after a write operation (copying a file to the drive).

 

Results with SATA I modes are so irregular, particularly with any write operation, that they are useless for comparison. Obviously it would not represent what the Caviar and Barracuda drives can do. After looking around on the Internet for answers, all we could find was a reference to RAID problems with the ICH8R on some Gigabyte P965-DS3 motherboards. Gigabyte's representative we talked to didn't offer any solution, but suggested updating the BIOS to see if that could help. So, we tried updating the motherboard BIOS to see if that will fix the problem - it didn't. That leaves us with only one solution - use the Gigabyte relabeled J-Micron controller for testing. After some testing, we found no such problems with the J-Micron controller

Although it works, there are several nitpicks we have with the J-Micron controller. First, CPU utilization in RAID 0 are higher compared to Intel ICH8R. CPU utilization seems to be the main culprit for fluctuating transfer rates when testing sequential reads / writes with HD Tach. Second, we feel burst transfer rates testing results with the J-Micron controller are way too inflated to be of any use for comparison - reaching numbers as high as 2.1 GBps. Third, we have no way of seeing whether the drive is operating in SATA I or II mode. With Intel ICH8R, we can use Intel's Matrix Storage Console and confirm the mode by looking at burst transfer rates in HD Tach. The final nitpick is the unability to monitor drive temperature with any SMART monitoring utilities such as SpeedFan.

There are less annoying issues, like the IDE port would stop working in Windows XP when we enable RAID (though, interestingly it works fine with Vista). However, there are several plus points with this controller. The most important one - it does not require a Windows reinstall if you want to change back and forth between IDE, AHCI and RAID modes - unlike Intel ICH8R. With ICH8R, if you do want to change modes, Intel suggested using / installing RAID controller drivers instead of AHCI drivers since it also support AHCI and IDE modes. Or you could force an array to work in AHCI mode (but not recommended). Unfortunately, this is only possible if you install RAID drivers by pressing the 'F6' key when you first install Wndows XP. Motherboard manufacturers also seem to 'conveniently' failed to supply the needed F6 drivers on the bundled driver CD or floppy, so you'll have to download them from Intel. Great. We expect better from Intel, who is supposed to be the leading manufacturer of chipsets and processors. We think this is one area where even ATI / AMD, NVIDIA, VIA and SIS have the upper hand - ease of installation.

Preliminary Tests

Before starting our game related tests, we ran some preliminary tests using synthetic benchmarks. At the very least, it should give us an idea of what kind of performance we can expect from these drives. It's also easier to see any odd problems we talked about earlier. We're going to use two synthetic benchmarks today - HD Tach 3.0.1.0 and IOMeter 2004.07.30.

HD Tach

First off, let's look at HD Tach results. The graph below should give us an idea of what kind of data transfer we can expect from these hard drives. Since HD Tach's internal graph tool is limited to two hard drives / devices, we're not going to use a screen dump this time. Instead we're going to use the data in the .csv files to make our own graph.



(click the picture for a clearer graph)

	Deskstar 80 GB	Barracuda 80 GB	Caviar 80 GB
Random Access (ms)	12.8	17.3	14.7
CPU Utilization (%)	1	3	3
Average Read (MB/s)	48.9	59.8	66.5
Average Write (MB/s)	48.9	60.0	66.6

(click the picture for a clearer graph)

	Deskstar 160 GB	Barracuda 160 GB	Caviar 160 GB
Random Access (ms)	13.1	17.1	13.1
CPU Utilization (%)	1	2	1
Average Read (MB/s)	65.7	61.9	65.7
Average Write (MB/s)	66.3	61.9	67.0

Since burst transfer rate seems pretty inflated with J-Micron controllers, we're not going to display or discuss them here. From the graph, we can see both Seagate's Barracuda 7200.9 and Western Digital's Caviar SE HD Tach results are very different to Hitachi's Deskstar's results. On the 80 GB arena, the Caviar provides the best results in terms of average read and write speed, followed by the Barracuda then the Deskstar. The saving grace of Deskstar is the smaller random access time. Moving on to the 160 GB drive, we're seeing the 160 GB Deskstar giving us the same results as the Caviar. The Barracuda must settle for third place.

Now if you look at the graph for the 80 GB drives, you'll see the read / write graphs for both Barracuda and Caviar are bunched up together. We do have to note that the Barracuda provides a much smoother write graph than the Caviar - notice the dips with the Caviar's write graph. The Deskstar seem to display the same behavior, though not as drastic as the Caviar - the peaks and dips are about 5 MB/s worth on the Deskstar while the dips on the Caviar is closer to 10 MB/s. On 160 GB drives, the situation is quite the opposite. The Caviar's read and write graphs are definitely much 'smoother' - less peaks and dips - than Deskstar's and is definitely higher than Barracuda's..



(click the picture for a clearer graph)

	Deskstar 80 GB - Array	Barracuda 80 GB - Array	Caviar 80 GB - Array
Random Access (ms)	12.9	17.0	14
CPU Utilization (%)	1	4	4
Average Read (MB/s)	97.7	96.1	110.6
Average Write (MB/s)	78.2	98.0	87.4

To avoid confusion, we've decided to use to use two graphs to show read and write performance of these drives in RAID 0 (striping). If we look at the numbers on the table above, it's easy to say the Caviar is the better drive of the three (the Barracuda do offer a clear advantage in write). However, when we look at the graph, we see a very different picture. Of the three drives, the Deskstar provides the smoothest graph of all when reading data. The story with write performance is not so clear. The Barracuda's may seem to excel but that large dips do bother us a bit.



(click the picture for a clearer graph)

	Deskstar 160 GB - Array	Barracuda 160 GB - Array	Caviar 160 GB - Array
Random Access (ms)	13.5	16.9	13.5
CPU Utilization (%)	4	3	5
Average Read (MB/s)	112.0	92.2	111.8
Average Write (MB/s)	88.0	96.9	86.5

From the numbers on table above, it's easy to say the Deskstar is a match to the mighty Caviar. However, if we look at the graph, neither the two and the Barracuda can sustain read and write performance through out the entire array. Of course, this is only HD Tach results, we'll have to see if we experience a similar siuation in real world scenarios. The Barracuda seems to have much drastic dips when reading than both the Deskstar and Caviar, though.

From these results, it looks like that if you're looking at single drive performance, you'll get the most out of your money with Western Digital's Caviar drives.  The Hitachi Deskstar put up a fight on the 160 GB arena, but the peformance is not as constant as the Caviar's. With RAID 0 (striping), the two are evenly matched - the 80 GB Deskstar putting a much smoother performance when reading. With 160 GB drives, the situations is not as clear cut. Worth noting, the Barracuda seems to be consistently following the other two drives. The only strong point it has is write performance with RAID 0. 

IOMeter

Now it's time to look at IOMeter results. We split the results into four tables - one group for 80 GB drives and a second group for the 160 GB drives. In each group, you'll find both sequential and random results for both read and write operations. In case you're wondering, these tests results are taken with test patterns using 32 KB transfer request size. Since we're using IOMeter as a synthetic test, data patterns are set to 100 percent (100% Sequential Read, 100% Sequential Write, 100% Random Read and 100% Random Write). 

Sequential
	Deskstar 80 GB single	Barracuda 80 GB single	Caviar 80 GB single
Read (Iops)	1829.2	2322.56	2350.78
Read (Mbs)	57.17	72.58	73.46
Average Access Time (ms)	0.55	0.43	0.43
Maximum Access Time (ms)	5.2	8.87	13.59
CPU Utilization (%)	1.76	2.54	2.86
Write (Iops)	1828.97	2323.1	2290.5
Write (Mbs)	57.16	72.59	71.59
Average Access Time (ms)	0.55	0.43	0.44
Maximum Access Time (ms)	17.06	9.63	15.43
CPU Utilization (%)	1.65	2.86	2.67
	Deskstar 80 GB array	Barracuda 80 GB array	Caviar 80 GB array
Read (Iops)	3310.46	2879.57	3358.02
Read (Mbs)	103.45	89.96	104.94
Average Access Time (ms)	0.3	0.35	0.3
Maximum Access Time (ms)	15.03	15.69	10.33
CPU Utilization (%)	3.61	6.05	6.72
Write (Iops)	3178.24	2478.98	2680.93
Write (Mbs)	99.32	77.47	83.78
Average Access Time (ms)	0.31	0.4	0.37
Maximum Access Time (ms)	0.93	5.11	17.77
CPU Utilization (%)	3.14	3.93	4

IOMeter results from the 80 GB drives seem to confirm our HD Tach results - the Deskstar is slower on read and write than both the Barracuda and the Caviar. At least, when it's used in a single drive configuration. Put two Deskstars together into a RAID 0 stripe and you'll get performance close to two Caviars in a similar RAID 0 stripe. Even better sometimes, just look at the strong write performance in RAID 0 with two Deskstars. This is very different to what we're seeing with HD Tach. According to HD Tach results, the Barracuda should be leaving the other two drives in the dust in this test. Let's look at the Random test results.

Random
	Deskstar 80 GB single	Barracuda 80 GB single	Caviar 80 GB single
Read (Iops)	72.34	56.1	69.09
Read (Mbs)	2.26	1.75	2.16
Average Access Time (ms)	13.82	17.77	14.47
Maximum Access Time (ms)	24.96	37.33	27.48
CPU Utilization (%)	0.03	0.31	0.19
Write (Iops)	104.95	89.76	100.36
Write (Mbs)	3.28	2.63	3.14
Average Access Time (ms)	9.53	11.14	9.96
Maximum Access Time (ms)	643.81	31.82	32.35
CPU Utilization (%)	0.11	0.16	0.78
	Deskstar 80 GB array	Barracuda 80 GB array	Caviar 80 GB array
Read (Iops)	71.07	54.42	65.11
Read (Mbs)	2.22	1.66	2.04
Average Access Time (ms)	14	18.37	15.36
Maximum Access Time (ms)	24.48	34.76	27.75
CPU Utilization (%)	0.11	3.59	3.7
Write (Iops)	145.92	137.12	131.12
Write (Mbs)	4.55	4.26	4.1
Average Access Time (ms)	6.8	7.33	7.63
Maximum Access Time (ms)	713.98	50	43.95
CPU Utilization (%)	0.71	11.21	18.12

According to this, the 80 GB Deskstar offer better results in this 100 % synthetic random test both when used in a single drive setup or RAID 0. The 80 GB Caviar comes in at a very close second place and the Barracuda follows at third place. One thing that's really puzzling is CPU utilization on random write tests. Both the Barracuda and Caviar post numbers above 10 percent while the Deskstar practically don't put any burden on the processor. So according to these test results, we're slightly better offf with the Deskstar than the Caviar.

What about 160 GB drives? Let's take a look.

Sequential
	Deskstar 160 GB single	Barracuda 160 GB single	Caviar 160 GB single
ReadOps	2390.79	2305.64	2438.04
ReadMBs	74.71	72.05	76.19
Average Access Time	0.42	0.43	0.41
Maximum Access Time	11.53	7.69	9.98
CPU Utilization	2.87	2.83	2.77
WriteOps	2380.76	2305.48	2433.4
WriteMBs	74.4	72.05	76.04
Average Access Time	0.42	0.43	0.41
Maximum Access Time	18.43	9.44	14.01
CPU Utilization	3.31	2.93	2.81
	Deskstar 160 GB array	Barracuda 160 GB array	Caviar 160 GB array
ReadOps	2989.39	2845.8	3351.38
ReadMBs	93.42	88.93	104.73
Average Access Time	0.33	0.35	0.3
Maximum Access Time	16.05	12.9	10.45
CPU Utilization	6.96	6	6.22
WriteOps	2698.06	2444.57	2698.24
WriteMBs	84.32	76.39	84.32
Average Access Time	0.37	0.41	0.37
Maximum Access Time	1.7	7.62	16.24
CPU Utilization	4.24	3.76	4.31

HD Tach results showed us that the 160 GB Deskstar is on a par with the 160 GB Caviar. According to IOMeter, the Deskstar puts up performance close to the Caviar but nowhere near of passing it. Well, except for one test - sequential write with RAID 0 - the two drives are practically offer identical.level of performance. However, look at read performance with RAID 0 - the Caviars are definitely have the lead here. As expected, the Barracuda brings up the rear, though in all honesty it isn't that far behind the Deskstar. We still do not see the write performance advantage HD Tach showed with RAID 0.

Random
	Deskstar 160 GB single	Barracuda 160 GB single	Caviar 160 GB single
ReadOps	71.61	56.16	71.82
ReadMBs	2.24	1.76	2.24
Average Access Time	13.96	17.8	13.92
Maximum Access Time	26.55	36.54	27.34
CPU Utilization	0.1	0.13	0.1
WriteOps	103.57	87.82	120.04
WriteMBs	3.24	2.75	3.75
Average Access Time	9.63	11.39	8.33
Maximum Access Time	615.79	30.31	32.45
CPU Utilization	0.13	0.11	0.16
	Deskstar 160 GB array	Barracuda 160 GB array	Caviar 160 GB array
ReadOps	69.44	54.92	69.81
ReadMBs	2.17	1.72	2.18
Average Access Time	14.4	18.21	14.32
Maximum Access Time	29.67	40.61	26.51
CPU Utilization	3.72	3.67	3.96
WriteOps	115.91	133.3	183.17
WriteMBs	3.62	4.17	5.72
Average Access Time	8.57	7.5	5.46
Maximum Access Time	862.22	50.66	32.33
CPU Utilization	14.07	11.71	16.08

Finally, the random test results. Looks like we found a chink in Deskstar's armor. Though the Deskstar's random read performance, be it as a single drive and in RAID 0 stripe is on the same level as the Caviar's, it's not as strong when doing random writes. In RAID 0, the number of Write Operations per seconds is actually lower than Barracuda's - a sign that perhaps HD Tach's sequential results is not really that sequential at all? The difference is significant enough for us to say that from these results, the 160 GB Caviar is the better choice.

Intel recommends using 128 KB strip size for RAID 0 so we also conducted tests with 128 KB transfer request size on these drives, mainly to see whether or not CPU utilizations will be lower with larger transfer request size.  Here are the results for 80 GB drives.

Sequential
	Deskstar 80 GB single	Barracuda 80 GB single	Caviar 80 GB single
Read (Iops)	457.32	580.8	587.71
Read (Mbs)	57.16	72.6	73.46
Average Access Time (ms)	2.19	1.72	1.7
Maximum Access Time (ms)	4.97	7.24	11.95
CPU Utilization (%)	0.64	1.44	2.21
Write (Iops)	457.09	580.82	572.53
Write (Mbs)	57.14	72.6	71.57
Average Access Time (ms)	2.19	1.72	1.75
Maximum Access Time (ms)	16.06	5.91	17.28
CPU Utilization (%)	0.62	1.4	1.35
	Deskstar 80 GB array	Barracuda 80 GB array	Caviar 80 GB array
Read (Iops)	914.51	808.54	999.29
Read (Mbs)	114.31	101.07	124.91
Average Access Time (ms)	1.09	1.24	1
Maximum Access Time (ms)	10.88	19.85	18.16
CPU Utilization (%)	1.21	4.04	4.11
Write (Iops)	910.01	738.21	828.14
Write (Mbs)	113.75	92.28	103.52
Average Access Time (ms)	1.1	1.35	1.21
Maximum Access Time (ms)	17.91	7.93	23.11
CPU Utilization (%)	1.19	2.18	2.69

And 160 GB drives.

Sequential
	Deskstar 160 GB single	Barracuda 160 GB single	Caviar 160 GB single
Read (Iops)	597.94	576.43	608.78
Read (Mbs)	74.74	72.05	76.1
Average Access Time (ms)	1.67	1.73	1.64
Maximum Access Time (ms)	12.53	8.34	11.8
CPU Utilization (%)	1.48	1.59	1.78
Write (Iops)	515.92	576.26	608.9
Write (Mbs)	64.49	72.03	76.11
Average Access Time (ms)	1.94	1.73	1.64
Maximum Access Time (ms)	24.62	11.51	16.12
CPU Utilization (%)	1.23	1.54	1.42
	Deskstar 160 GB array	Barracuda 160 GB array	Caviar 160 GB array
Read (Iops)	795.01	798.28	988.58
Read (Mbs)	99.38	99.78	123.57
Average Access Time (ms)	1.26	1.25	1.01
Maximum Access Time (ms)	25.11	28.13	17.94
CPU Utilization (%)	4.27	3.91	4.28
Write (Iops)	852.44	733.46	838.16
Write (Mbs)	106.55	91.68	104.77
Average Access Time (ms)	1.17	1.36	1.19
Maximum Access Time (ms)	2.6	23.84	15.66
CPU Utilization (%)	2.56	2.23	2.34

Looks like they're right, larger transfer request size do help keep CPU utilization down. So, from these prelimnary test, we've learned that very little separates the Deskstar and Caviar in terms of performance. The 80 GB Caviar drives puts up a good balance between single drive and RAID 0 performance. The Deskstar is slightly slower as a single drive, but put a much smoother performance in RAID 0 according to HD Tach. Write performance is similar between the two drives, though according to IOMeter the 80 GB Deskstar have a slight advantage in write performance with RAID 0 higher performance. The 80 GB Deskstar also seem better in handling random operations.

On the 160 GB arena, the Caviar is without a doubt the better drive. HD Tach shows a much more smoother performance throughout the entire array with a 160 GB Caviar than a 160.Deskstar. This is confirmed by slightly higher performance numbers in IOMeter sequential tests in a single drive configuration and read performance in RAID 0. The clincher is of course single drive and RAID 0 random write performance.

Heat

For thermal testing, we monitor the hard drives' temperature by monitoring SMART output via SpeedFan. To push the drive, we perform a drive to drive copy of about 60 GBs worth of data with Western Digital's Data LIfeguard tool to the first drive and copy several 2 GB files to the second hard drive. The first and second drive are positioned as they would inside a case (one drive on top of another - separated with plastic CD case - but without the confines of a drive enclosure. The drives were actively cooled using a standard 80 mm, 0.22 Nidec Beta SL ampere fan configured to suck of air instead of blowing into it.

Here are the results.

	Deskstar 80 GB	Barracuda 80 GB	Caviar 80 GB
Low	44'°	40'°	33'°
High	47'°	44'°	42'°

	Deskstar 160 GB	Barracuda 160 GB	Caviar 160 GB
Low	40'°	38'°	32'°
High	45'°	42'°	40'°

(temperatures are in ° Celsius)

Looks like the trend is fairly constant - Caviars are the coolest drive, Barracudas the second and Deskstars are the hottest. Looking at these results, we think it would be much better to choose the 160 GB Deskstar intead of the 80 GB drives. Kudos to Western Digital engineers for not only providing high performing drives, but also the coolest as well. We think Barracuda's slightly smaller dimension have a role in this test - much less dissipation area than the other two drives, so they also deserve a thumbs up. You could probably fit three Barracuda in a space where you could only fit two Deskstars or Caviars

Performance

To measure real world performance, we choose to measure the time it takes for a game to load a level. We choose three games for this test - F.E.A.R, Quake 4 and Serious Sam II. Because of the way the games work (caching some elements in RAM and virtual memory), we ran test for both first time load and reloads. We restart the computer after each first time load, clearing the RAM and virtual memory from any game data. Reloads or consecutive loads are done consecutively - in F.E.A.R and Quake 4, we load another level before reloading the first chapter. In Serious Sam, we reload the game directly after finishing one test run.

We chose both F.E.A.R's and Quake 4's opening cinematic simply because these levels are generally much larger than the average levels for both games. For Serious Sam II, we choose the final Mental Institution level for the same reason. Measurements are taken with FRAPS, from the time the game starts loading the level to the start of the cinematic sequence. Since we're measuring load times, the performance metric is time (in msec).

We also perform another additional test for this round up. We made an ISO image of a Company of Heroes DVD, copying into the first and second drive in a single drive configuration and the RAID 0 stripe in RAID configuration. We mounted the ISO, installed the game and record the time taken for the game to install itselt to the hard drive. The tests were perform three times, from which we then compute an average. By placing the ISO in the same drive (and in the same RAID 0 stripe) as the system, we're testing the I/O performance of the drive, forcing the drive to always swtich between reading the ISO and writing the game data. By placing the ISO on the second drive, we basically testing file copy performance - one drive reads while the another (the system drive) writes. From this test, we can basically see the 'burden' of using RAID 0.

We'd like to thank Tagan for supplying the additional power supply and Gigabyte for supplying both the Gigabyte P965-DS3P and Radeon X1950 Pro for this article.

Our test setup
Intel Core 2 Duo E6300 socket LGA-775
4 x 512 MB A-DATA Vitesta 5-5-5-18 PC6400 DDR2-SDRAM
Gigabyte Radeon X1950 Pro 256 MB graphics card
Gigabyte P965-DS3P Intel P965 motherboard
LiteOn 1673S DVD-RW
Tagan TG530-U15 530 watts ATX/BTX power supply

Windows XP Professional with Service Pack 2 installed
ATI Catalyst 7.2 reference driver
Intel Chipset Software Installation Utility 8.1.0.1006
DirectX 9.0c
all respected games used for benchmarks have been updated to their latest,final builds.

The graphs are pretty self explanatory, but in case you can't see the text, they're arranged in groups of three colors: green for the first run,  blue for the second and red for the third (final) run. Results in black are the average of three runs. The purpose of including all three runs is to see how much variability these test have. Results are in milliseconds, sp 25821 means 25.821 seconds, so shorter bars means better performance. We'll take a look at 80 GB drives first, then 160 GB drives later on.

The results:

Single drive configuration first. It's easy to see, the Caviar SE 80 GB has the upper hand here. On average, we save 3 seconds with the Caviar than if we were to use the Deskstar and 11 seconds if we were to use the Barracuda 7200.9. However, the Deskstar is the stronger performer with reloads - around 3 seconds faster than both Barracuda 7200.9 and Caviar SE.

With RAID 0, the Deskstar and Caviar SE are neck to neck from the get go, the Deskstar is faster by 1.5 seconds on average.  The Caviar regain the lead when you reload - it's faster by 0.7 to 0.8 seconds, with much more consistent results. The Barracuda 7200.9 simply can't catch up to the competition, its 8 to 9 seconds slower than the other two drives on first loads and about 1.4 to 2.7 seconds on reloads. Not surprisingly really, since we saw earlier sequential read performance is not one of Barracuda's strong points.


In Quake 4, we're seeing the Deskstar put up slightly stronger performance than the Caviar SE, both as a single drive and in a RAID 0 configuration. The differences is about 2.4 seconds with a single drive and 2.8 seconds with a RAID 0 stripe. Quake 4 doesn't push the hard drive enough on reloads, so we don't see any differences between the drives, single or RAID. The Barracuda again brings up the rear.


Serious Sam II loads very quickly, so any differences related to hard drives are actually a moot point. There is a consistent trend here. The RAID 0 Deskstars are slightly faster than the Caviars. And of course, the Barracuda trails behind the two, thankfully only by a very small margin.


Moving on to 160 GB drives, we can see the Deskstar is actually a little slower than the Caviar in a single drive configuration. The Caviar do have the more consistent results from run to run, particularly for first loads. On reloads, the Deskstar is faster (0.5 seconds) than the Caviar - again a moot point. The picture with RAID 0 stripes is pretty much the same. The Deskstar is faster on first loads and reloads but only by a small margin, you wouldn't notice it in real life. Results are much more consistent with RAID 0 for both drives. As for the Barracuda 7200.9, well, we think the graph speaks for itself.


Like F.E.A.R, we're seeing basically the same graph. The Deskstar is marginally faster in both single and RAID 0 configuration than the Caviar. About 1.2 to 1.8 seconds on first run loads. The Barracuda 7200.9 is still, well, the Barracuda 7200.9. We basically run out of words to say. Looks like in real life situations where read and CPU load is the emphasis, the Deskstar 160 GB has a slight upper hand above the 160 GB Caviar SE.


Hmm, redemption for the 160 GB Caviar SE? On single drive configurations, the Deskstar still leads over the Caviar but use them in RAID 0 and the Caviar pulls slightly ahead. Too bad the load times are much too short to really make a difference.

Company of Heroes ISO Test

Finally, with this test we can test write performance of these drives in situations close to real life. Admit it, most of us prefer to install games with ISO images than discs. For one, its faster and it also mean we could keep the original disc in storage for safe keeping -away from moisture, fungus, heat and dust that can make those precious original discs unreadable.

On to the numbers.

	Deskstar 80 GB	Barracuda 80 GB	Caviar 80 GB
ISO on the second drive	125.67	167	104
ISO on the same drive	264	390	230
	Deskstar 80 GB - Array	Barracuda 80 GB - Array	Caviar 80 GB - Array
ISO on the same array	148	229	167

	Deskstar 160 GB	Barracuda 160 GB	Caviar 160GB
ISO on the second drive	103.33	186.33	99.33
ISO on the same drive	194.33	374	204
	Deskstar 160 GB - Array	Barracuda 160 GB - Array	Caviar 160 GB - Array
ISO on the same array	169.33	222.67	142

(results are in seconds - smaller is better)

With writes coming into play, the single drive Caviar SE leaves the Deskstar behind. By using two, identical 80 GB Caviar SE drive, you could save 20 to 34 seconds when installing a game than if you're using the Deskstars. The Deskstars manage to get the lead back if you're using them in a RAID 0 stripe. However, it's interesting to note that it's still nowhere close to the 104 seconds we got with two Caviar SE on a single drive configuration. RAID 0 penalty is much lower on the Deskstar than on the Caviars (43 seconds on average than 63 seconds).

The 160 GB is a win for the Caviars, though only by a slight margin. The difference is 4 seconds when we mount the ISO on the second drive and 27 seconds if we're using them in a RAID 0 stripe. The Deskstar seems to have the better I/O throughput - 10 seconds faster when we place the ISO on the same drive. What about RAID 0 penalty? The 160 GB Caviars SE experience less of a penalty (42 / 43 seconds on average) compared to the 160 GB Deskstars (66 seconds on average)  Looks like the supposed RAID 0 write advantage on the Barracudas we saw in HD Tach never materialize here. On all configurations, the 80 and 160 GB Barracudas are significantly slower than the Deskstars and Caviars. Oh, well.

Conclusion:

Hard drives are perhaps the most often overlooked component in a gamer's system. Most performance reviews and articles focus on processors, memory and graphics cards. That doesn't mean hard drives are less important - they have less direct influence on performance, like when the game is loading a new level or map, when we're installing the game and to some extent, every time we save a game. The game related test we conducted here today shows the Deskstars is slightly ahead of the Caviars. Slightly - the differences are so small, you will hardly notice the difference. For most people, the deciding factors will be numerous, but we think we already found two good reasons.



WD800AAJS
WD1600AAJS
HDS721616PLA P220

We recommend the 80 and 160 GB Western Digital Caviar SE drives whole heartedly for gamers concerned with performance and temperature. Without a doubt Western Digital engineers are experts in making hard drives. Reasons? One: read / copy performance which should be handy when you're installing games and Two: operating temperatures. The only nitpick we have with the Caviar SE is its erratic behaviour with our platform of choice, not necessarily Western Digital's fault.

We also recommend the 160 GB Hitachi Deskstar. It offer performance close to the 160 GB Caviar SE, with slightly lower operating temperatures than its smaller brother. It also helps these drives are the only drives able to work flawlessly as 2nd Generation SATA drives with our test platform. We've actually already reviewed the 160 GB Hitachi Deskstar before (the review is here), though with different firmware and a different platform. The odd HD Tach sequential read pattern is thankfully absent on these drives. We think the lower CPU utilization of the controller used helped here - the NForce 4 SATA controller we used before have about 13 percent utilization (give or take 2 percent). Performance level have also improved, again more likely related to the SATA / RAID controller used (and a dual core processor).

Go to top