I have heard a lot of hype about how much performance serial ATA was suppose to give you and I have been a little skeptical. Is it not true that an ATA-133 Drive is suppose to move 133Mb/sec, well it's more like 80Mb/sec because physically the heads cannot pull that much data off the platters so my way of thinking is why another controller upgrade?  Why don't we start working on making solid state drives a reality instead?  I had to be optimistic about this and see if adapting an old ATA-133 hard drive to a serial ATA controller would gain any performance.

For the test setup I will be using an IBM 75GXP because of a defect in the platter density that IBM doesn't want to live up to. These drives only last for about 6 months to a year so it makes a great test drive, But don't trust important data on them. The IBM drive will be using the HighPoint RocketHead 100 serial ATA adaptor connected to the onboard Silicon Image Sil 3112A serial ATA controller on my ASUS A7N8X Deluxe motherboard. This Serial ATA adaptor claims it will only work with HighPoint controllers but it works perfectly on the Silicon Image controller.  It however will not work with the Promise serial ATA controller on the ASUS A7V8X.

• ASUS A7N8X Deluxe
• AMD Athlon XP 2000+
• 512Mb PC3200 (2X256Mb)
• IBM 75GXP 20Gb ATA100 7200RPM
• HighPoint RocketHead 100 serial ATA adaptor (supplied by: SouthPacificSystems.com)
• Silicon Image Sil 3112A onboard serial ATA controller

Here are some photos of the adaptor connected to the hard drive to show you how low profile the HighPoint adaptor really is. I have seen most adaptors stick out to far.

Performance gains are not looking promising in the Sandra benchmark even though there is an improvement, it's not much.  In fact I highly doubt that you would even notice a real improvement in real time performance according to this test.

I had close to the same results in HDTach other then a boost in access time.  The first graph shows the access time of the Serial ATA controller is 12.1ms and the ATA-133 controller at 13.5ms.  Some people may feel a difference in some hard drive intensive applications with a 1.4ms improvement, but not much.  Before I wrapped up testing, I noticed that there is one improvement that may make the difference in upgrading to serial ATA, and that is CPU Utilization.  Check out the graph below.

The serial ATA controller only used 7.8 percent of the CPU to run the test while the ATA-133 controller used 43.7 percent of the CPU.  This shows that using serial ATA frees up 35.9 percent of the CPU for more important things, like blasting your buddy at your next Unreal Tournament match while his CPU is trying to keep up.

After finishing this testing I have a new respect for serial ATA.  I still wouldn't recommend upgrading yet because the technology is so new.  If you have the controller onboard already, the adapter doesn't cost a lot and it may be worth the money in the long run to extend the life of that ATA-133 hard drive and also give you bragging rights at the next LAN party.  I see a bright future for serial ATA and as the technology matures, we may see a lot more performance in the near future.  Get the HighPoint adaptor here.

I had a funny story to share with you guys before I get completely flamed for my comments about the IBM 75GXP. I had a 20Gb drive that I had gotten back from RMA a few months ago that has been in the box under my desk that I decided to use for this review.  After almost finishing all my tests, the drive died with the infamous clicking which in turn made me start the benchmarking process all over again with another one of my 4 IBM 75GXP drives that have had a total of 7 RMA's.  Keep in mind that this drive was still in its static wrap before I started testing and now holds the record for the quickest dieing IBM drive at just under two hours. I told IBM that I feel like I am being taken advantage of and if they didn't replace this drive with an upgraded drive that wouldn't die prematurely I would never stand behind any of there drives again.  The UPS guy showed up a few days ago and handed me another IBM 75GXP. I guess IBM doesn't feel they need to take care of their customers. I urge everyone that is using an IBM 75GXP to consider upgrading there hard drive to a non IBM drive or at least keep regular backups of your data because in the long run IBM doesn't care if all your business's financial data was on that drive. Data integrity is the most important performance trait. I don't know about you but 7 RMA's in two years with five different computers and operating systems ranging from Linux, Windows ME, Windows 2000, and Windows XP, tells me that there's a problem with these drives.

If anyone is interested in the reason IBM 75GXP drives die, this is the best explanation I have heard to date. Apparently the 75GXP has a really dense platter to accommodate 75GB on only two platters.  Normally as hard drives warm up there platters expand bringing the head closer to the platters surface so the head has to adjust itself away from the platter. In the IBM 75GXP's case the platter doesn't warm up evenly because of the platter density so the heads have to constantly adjust themselves as the platters spin around. This is where the clicking noise comes from. What eventually kills the drives, is the heads constant adjusting makes errors and places data in the wrong place on the drive and it gets stuck in a continues loop trying to find the data. I have been able to extend the life of the drive by a few weeks by just deleting the partition and reformatting the drive, however, the only true fix is to send the drive in on warranty.