Western Digital sells three models of hard drives with similar model numbers: WD10EADS (32 MiB cache), WD10EACS (16 MiB cache), and WD10EAVS (8 MiB cache). The last of these is intended to be sold to OEM manufacturers, but you can buy an individual unit from various sources as long as you don't mind the lack of a warranty. I picked one up for $70.
I had expected the performance of all three of these models to be basicly the same, but it turns out that that's not quite right. The following table shows the sequential read speeds (in millions of bytes per second) at the beginning, middle, and end of each disk.
The WD10EAVS is slightly slower than my WD10EACS. As you would expect, both are significantly slower than my 7200 RPM Seagate ST3500410AS drive.
Sequential read speed is not a particularly good indicator of how a drive will perform in actual usage, so I decided to use a search of the Linux kernel source as a benchmark. I created a copy of my /usr/src partition on each drive and ran the following command on the Linux kernel source:
On POSIX systems, when a file is read the access time of the file is updated, but Linux allows the update to be turned off in order to increase performance. The following table shows the run time of the command, in seconds. The second column gives times with updates of access times turned off:
On this benchmark, the WD10EAVS takes about 32% longer than the WD10EACS. The latter performs quite well, basicly matching the performance of the Seagate drive. The Seagate had a slight advantage because the file system was closer to the start of the drive on the Seagate disk than it was on the other disks. The last line shows the performance with all three disks combined into a RAID 5 configuration.
The WD10EADS/WD10EACS/WD10EAVS disks are "green" disks, so you probably aren't looking at these if performance is paramount. Still, the performance numbers suggest that the WD10EAVS should be avoided if you can get one of the other models for a similar price.
(Notes: To obtain the numbers in the above tables, I ran each test 9 times and selected the median. For the second set of tests, I unmounted the file system and did a raw disk copy. Copying the files rather than the disk could result in the files being laid out on disk differently, resulting in a meaningless comparison. The tests were run using a Gigabyte GA-MA790XT-UD4P motherboard, with an AMD Phenom II X3 710 processor overclocked to 3250 Mhz, and 2 GiB of DDR3-1333 memory.)
Last year, I picked up 2 WD10EADS drives to go with my WD640 Black.
The other major difference between the EACS and EADS models is the number of drive platters. The EACS uses four. The EADS uses three. I suspect that contributes more to the throughput than the size of the caches.
Thanks for sharing! Though given their listed cache sizes, I think people will tend to go for the highest they can see (32MiB), don't you think?
I doubt that cache size actually a significant effect on disk performance. Back 1999, Storage Review compared two drives which differed only in cache size and concluded that there was no significant difference between a 512K cache and a 1M cache. If they had been able to test the effects of smaller cache sizes, I'm sure they would have found that if the cache size is small enough, increasing the cache size improves performance. The point is that increasing the cache size only helps up to a certain point, and the 512K cache was past that point, so increasing the cache size to 1M didn't help.
Today disks have larger track sizes than 1999 vintage disks, so they may benefit from larger cache sizes. (In particular, I expect that the best measure of disk cache size is the number of tracks it can hold, rather than the number of bytes.) But it would seem that and 8M cache would be plenty for any drive on the market today.
The point of my posting was to warn people that the EAVS is not just an EACS with a smaller cache size. Jsc points out that the EACS is not just an EADS with a smaller cache size, either. So people may tend to go for the disk with the largest cache, but what they really should look at is performance measurements.
I am thoroughly confused.
In the initial posting you are talking about three drives:
WD10EADS (32 MiB cache)
WD10EACS (16 MiB cache)
WD10EAVS (8 MiB cache)
but you are presenting the resuts of your tests for only two of them:
WD10EAVS (8 MiB cache)
WD10EACS (16 MiB cache)
"On this benchmark, the WD10EAVS takes about 32% longer than the WD10EACS "
The WD10EACS has the bigger cache (16 MiB cache). This could suggest that the larger cache size is beneficial, hence the comments from the other readers. Suggestion that you questioned in your reply to r_manic.
Fine so far, but the strange thing is that you have not tested the EADS with the 32Mib cache, so we do not know how well (or badly) it would perform.
jsc also points out that the EADS is phisically different (three platters not four), his comment being
"I suspect that contributes more to the throughput than the size of the caches. ".
Fine... but how well (or badly) does the EADS perform? You did not test it. You showed that two supposedly identical drives (except for the size of the cache) perform quite differently. The third drive has a bigger cache AND a different platter architecture. Can we speculate how well the EADS could perform? I guess not, it is too different, and this is what you were saying all along.
The bottom line is.. how good is the EADS, with its 3 platter and a 32Mib cache?
Many thanks for your testing anyway, much appreciated!
p.s. btw, it seems that WD is not showing the EAVS anymore, only the EAVS and EADS (I am sure that I saw the 8Mb cache options few days ago, but I am always wrong anyhow)
I have a pc with Win XP Pro Installed on a Maxtor Hard Disk, controlled by a SATA Raid Controller Sil3114 by Silicon Image.
When I Connect My WD10EAVS To my controller and I turn off the Pc No operating system loads and the screen say: Press A Key To Reboot.
Maybe I must set the jumpers on the disk?
This is ofcourse for the WD10EAVS 8mb cache version. Got one OEM from Fry's $54 bucks Sun. sale - died the NEXT day (controller no - nothing, boot, read, even show up in ANY OS) - brought it back - exchanged (no refund because it is an OEM drive, even from Frys' of all things) - Next disk had a serial number roughly +1500 from the first one - installed it - copied about 250GB of video to it - it died, same exact thing as the first one. headed back to Fry's - got another one - this one was -17000ish serial numbers (so from an earlier batch if the serials are sequential from low to high) - copied data to it and whammo next boot - it died. Brought that one back and got another - this time from a diff Fry's (being in the south bay there are like 4 within 10 min. of each other - and you guessed it - DEAD as a mackeral in a salmon feeding frenzy the next day. - sooooooo - after emailing WD support they want me to send the drive back to them and they will check it out and swap me ANOTHER one.
Well to make a looooong story short and in barking at the tele support people I come to find out this series of drives was actually created for USB 5v systems and has a controller that switches (supposedly) the power from 5v / 12v depending on the input voltage receieved. Well come to find out aslo that they are very aware that the controller malfunctions if used as an internal normal 12v SATA connection - but due to marketing they don't want to tell people this, instead having them return the drives and WD then updates the controller board -
So, IMHO stay FAR clear of this product unless you plan on using it soley as a USB driven 5v setup.