Slashdot Mirror
Hard Drives Evaluated for Noise, Heat and Performance
Sander Sassen writes "Ever wondered what harddisks offer the best combination of performance and low noise? Hardware Analysis evaluates all recent 5400 and 7200-rpm harddisks and focuses on noise, heat production and overall performance. Their results show that 7200-rpm spindle speed is no guarantee for high-performance and that low-noise and high-performance is not an impossible combination with some harddisks."
Seagate's Barracuda IV drives are great! Exceptionally quiet (the CPU cooling fan generates more noise) and I've not run across a single failure in ~100 sold.
So rise up, all ye lost ones, as one, we'll claw the clouds.
If you're looking for a good 7200-rpm harddisk then look no further than the Western Digital WD800BB, with 2MB cache, just a tad bit slower than the WD800JB which features 8MB of cache. The surprising newcomer is the Samsung SP8004H that scores well on all fronts and certainly deserves your attention too.
Equally surprising was the performance of Western Digital's 400AB and 800AB, both 5400-rpm harddisks showed exceptional performance on par with all but the fastest 7200-rpm harddisks. If you're looking for an affordable, high-performance and yet silent 5400-rpm harddisk either of these will fit your needs exactly.
If you're however looking for a harddisk that offers an impressive combination of performance and low noise then look no further than Seagate's ST380021A Barracuda IV, it really is an engineering marvel that combines the best of both worlds. No match for the IBM or Western Digital but a fair trade-off between performance and noise level.
Sig for sale or rent. One previous user. Inquire within.
This is a timely article, what with hard drive warranties having just been bumped from three years to one in a few of the leading brands (including Maxtor). Word is the WD w/ 8 meg cache still has 3 years.
The problem you describe will happen when the processor needs the data to continue processing and it isn't in memory yet. The solution to this would be more memory and programming to take advantage of it, in other words, transfering the data from the fixed disk to the memory early enough that the process doesn't wait for IO. That increases the memory footprint. People will complain about the footprint and they'll complain about waiting for IO, so pick a middle road and stick to it.
I would sooner transcribe all of my work to punch-cards than trust Western Digital with a single BIT of my data. I've owned several WD drives and they are ALL TRASH. They almost always fail (read: TOTAL HEAD CRASH) within 2 months of usage. We use WD drives at work and they are always failing, failing, failing. JUNK. Do NOT trust Western Digital. Do NOT get suckered in by their high storage capacity--you'll just lose all your data all the quicker.
IBM on the other hand, made (sadly, past tense now) some damn fine hard drives. I own several IBM SCSI drives and they work, period. They are on and spinning 24x7, and have been doing so for about 3 years without a hitch. There isn't a Western Digital drive in the universe that can come close to reliability like that.
Amen to that. You'll notice that they give high marks to the WD800BB (7200RPM, 80GB, 2MB cache). I bought this drive last March, and it served me well: fast and reasonably quiet. However, in July, it died. Corrupted boot sector, various bad sectors, etc.
Unfortunately, this was not a one-off manufacturing error. A friend purchased the same drive about a month later, and his died the same crash-and-burn death as mine about two weeks later. Sounds like these drives just suck. I'm using the replacement they sent me, but only until I can get the scratch together to replace it with a Seagate.
Caveat emptor on these suckers.
storagereview.com
Huge database of very indepth reviews on hard drives. Scsi, ide, 5400-15000rpm.. Basically everything, with noise, temperature, and a few different benchmarks for different usage conditions.
Definatly the best resource I've found for hard drive tests. I always consult this site before a hard drive purchase.
Blessed are the pessimists, for they have made backups.
Remember, if he were to be believed the Intarweb would no longer exist due to non-superuser raw sockets in Windows XP. Oh, and he still has the "next generation" DDoS attack article up, detailing the next generation attack and his wonderous solution. Even though the problem had been encountered and solved many moons ago, the solution being far more comprehensive and elegant.
My father bought the identical drive you speak of one month ago, and it *refuses* to be the main partition to run an OS. If you do, it just utterly fails at random points during installation/usage. When you make it a slave, it seems to work as anticipated.
In another light, I've purchased quite a few IBM Deskstar's (80gig, 120gig) and haven't had a problem yet -- and they are doing hardcore 1-5mb/sec transfer 24/7.
k.
--even a broken watch is correct twice a day.
Storage Review has been doing noise and measurement tests in thier reviews for several months now.
IntroductionBy: Sander Sassen
A modern harddisk is not that different, mechanically, from the first generation of harddisks that debuted with the IBM PC in the '80s. Today's harddisks are also mechanical parts that use spinning platters and read/write heads to store or read information from them. That also explains why harddisks haven't seen the rapid pace of innovation as for example CPUs have; simply because the mechanics are holding the harddisk back from making similar leaps in performance. But to be honest that's not entirely accurate, modern harddisks could be substantially faster, but not without either driving up the price significantly or introducing unwanted side effects.
One of these side effects is excessive noise; because a harddisk has a number of spinning and moving parts it is virtually impossible to make a harddisk noiseless. Anything you'll do to counter the noise will either influence the performance, drive up the price, or make the harddisk physically larger. For example one way to reduce the noise would be to reduce the rpm of the platters which would mean we'd end up with a slower harddisk overall. And vice versa, by increasing the rpm of the platters we'll get a better performing harddisk but the noise level will also increase.
Naturally we could counter the noise production by adding sound insulation. Unfortunately insulation is not the preferred way of tackling the noise production as it'll make the harddisk physically larger, and thus leave less room for storage capacity. Furthermore it also works as an insulator for the heat produced by the harddisk, which would then cut into the harddisk's MTBF, Mean Time Before Failure. That actually brings us to the second unwanted side effect of high-performance harddisks and that's excessive heat production.
Modern IDE harddisks feature platters that revolve at either 5400 or 7200-rpm and thus revolve at about half the speed of the fastest SCSI harddisks that top out at 15.000-rpm. The main difference is that these SCSI harddisks are used in professional applications such as database servers where noise- and heat production are second to performance, and thus these levels are substantially higher than consumer level products. In consumer level PCs however heat production is becoming an increasingly important issue. PCs continue to get smaller and CPUs on average dissipate more than 50-watts of heat, so the system temperature will rise significantly if a harddisk is mounted with excessive heat production. In the following pages we'll take a look at all recent 5400 and 7200-rpm harddisks from IBM, Maxtor, Western Digital, Seagate and Samsung with a focus on noise and heat production as well as overall performance.
Next >>
Noise ProductionBy: Sander Sassen
As mentioned a modern harddisk still is a mechanical device, and as with most mechanical devices it is hard to completely cancel out all noise. One part that contributes to the noise significantly is the spindle, which drives the platters. The spindle is usually direct-driven and has an rpm of 5400 or 7200-rpm, which translates itself into a high-pitched whine. Harddisks that use high-quality bearings or even fluid-bearings are less noisy in this respect than others that use conventional bearings. Then there are the harddisk platters; due to the high speed at which they revolve they generate both noise and heat. And naturally the more platters, higher capacity harddisks, the more heat and noise are generated. Due to space constraints modern harddisks don't use more than four platters which does limit the heat and noise production to a certain level.
Fortunately a manufacturer has a number of options at hand to reduce the noise level of the spindle and platters, such as using high-quality bearings or decoupling the spindle and spindle-motor from the harddisk casing by using a sound and vibration dampening gasket. Of course another way to reduce the noise and heat production would be to use less platters of a higher density, as this will not influence the storage capacity of the disk.
Fig 1. The interior of a typical harddisk whilst operating. This movie requires the Windows Media player to be installed and requires a broadband connection for streaming playback.
But there's other moving parts that also contribute to a harddisk's noise production; the read/write heads are notorious for causing the 'rattling sound' many harddisks make when accessing your data. And again the manufacturer has a number of options to reduce this noise which usually mean that the heads are operated less abruptly when sweeping across the platters. For example by smoothing out read/write operations by slowing the heads down when they need to reverse direction or by intelligently combining read and write operations and thus reducing head movement. Unfortunately many of these noise reducing measures do affect the harddisk' performance and some can even be set arbitrarily through software to optimize for a specific application.
Heat ProductionBy: Sander Sassen
Any part that is operated above its maximum temperature rating over a prolonged period of time will see its MTBF drastically lowered. All parts that make up a modern PC have such a maximum temperature rating, most of which are around 50...65 degrees Celsius, or 122...149 degrees Fahrenheit. An average CPU dissipates about 50-watts of heat inside a PC case and is one of the major contributors to a rise in system temperature which could easily be as high as 45 C/ 113 F. If the case temperature is already that high, some of the parts of the PC could already be operating at or above their maximum temperature rating.
Especially 7200-rpm harddisks are suspect as they are known to get substantially hotter than their 5400-rpm counterparts. Naturally 7200-rpm and multiple platters all contribute to more heat being produced, as the spinning platters get hot due to the friction with the surrounding air, the more platters and the higher the rpm, the more heat. Unfortunately we can't simply say that 7200-rpm harddisks get hotter than 5400-rpm harddisks by default as our measurements will clearly show, but we'll get to that in the next few pages.
Harddisk Interior
Fig 2. The interior of a typical 80GB 7200-rpm harddisk, in this case a Maxtor D740X. This harddisk features two platters with 40GB/platter density.
What is important to consider though is the question whether the harddisk doesn't get too hot when mounted inside the PC. Most modern harddisks have a maximum operating temperature ranging from 55 to 65 C or 131 to 149 F, and if operated above that temperature you're really putting yourself at risk of losing valuable data or a harddisk malfunction. If the system temperature is already at 45 C/ 113 F the harddisk could very well be operating over it's maximum temperature, which will cut into the disk's MTBF and reliability significantly.
Platypus Technology does make something similar. They have both internal PCI and external enclosures to just hold sticks of ram; some models have stadard hard drives for times of power loss. Unfortunately under linux it requires a kernel module (and at the time I was using them, if I upgraded the kernel the company had to compile a new module to match). They fly though, they're sooo fast. Really nice for my mail queues.
Probably because hard drives (ide, anyway) cost about a buck a gigabyte. SDRAM costs about a buck a megabyte; maybe a little more once you add a power supply and an interface. But, look here and here . The first is pretty much what you're looking for, I think, and the second is a bit more cost effective.
See what I've been reading.
Storagereview.com has had noise and heat statistics for years.
Actually, it is a better reference than this quoted article because you can tell SR.com to compare all the drives you are interested in purchasing and get good* benchmarks, heat/noise, and can sort by specific benchmark.
Go to the website, click "database" (near the top) and choose your criteria. In ten seconds you can find out the noise/heat/speed of every drive SR has ever reviewed, with a rather nice labelled bar graph for clarity.
You can also visit the forums and get advice from some of the most knowledgeable people in the IT industry, and get information that is difficult to come by anywhere else--for example, that Samsung makes the most reliable (albeit close to the slowest) IDE hard drive. SR was also the first to discover that Seagate planned to reduce their warranty and that there are terrible SCSI performance bugs in Windows XP, among others.
A very good resource, and it's been slashdotted without the server being brought to its knees. (It runs Linux/Apache/PHP)
Computer Science is no more about computers than astronomy is about telescopes. --E. W. Dijkstra
The most quiet drives, the seagate Barracuda IV atas have a problem in Raid configurations. When used in a Raid configuration, the performance is less than a single drive by itself. Raid is not officially supported by this drive. More here.
Take the cheese to sickbay, the doctor should see it as soon as possible - B'Elanna Torres, "Learning Curve"
Since things seem to be getting bogged down on Hardware Analysis's end, here are two mirrors:
1. Earlham College
2. UW-Madison
These are in PDF format, which I converted from the printable HTML provided on the website. It is missing one eye-candy picture of a hard-drive's interior.
That's something that's always bugged me. Why is it that the heads are allowed enough flexibility to touch the heads in the first place? Why not make them rigid enough so they stay in a fixed location relative to the platter regardless of air movement?
Like other people have replied already, tolerances are the issue. The gap between the head and disk during operation of the drive is crazy small. The wobble in the spindle bearing is easily more than this gap, if your head was in a fixed position at one point it would be too far away from the surface to operate, at another it would be digging into the surface. The only real solution is to have it fly above the surface. If it flies too high then there won't be enough lift and it will fly lower, if it flies too low then there will be extra lift so it flies higher. It is a continuously adjusting system based on the flow of air in the drive and it certainly works quite well. I don't think any harddrive today could be built with a fixed height head, there just isn't enough precision in a device as large as a harddrive.
no. the speed on the head flying across the disks to grab and write data are precisely calculated with lots of painful calculations. to get a 6000 rpm drive they would have to go thru all that again, find a vendor selling 6000 rpm drive motors & recreate the ASIC where those calculations are stored instead of buying it off the shelf.
Although it would take longer to accelerate the platter up to speed with 7200rpm, it would not nesisarrily require more power (amps). The bearing losses would be higher, but possibly not as much as you would expect.
I'd wager that the real world energy usage of a 7200rpm is significantly higher than a 5400rpm (which meshes with the fact that 7200rpm drives are generally quite a bit hotter, again correlating with energy draw). Indeed, one of the big conclusions of this article is that the extra power and heat of a 7200 might be unncessary.
The better approach is to actually measure the heat gain in a controlled environment over time for a variety of different usages. That is a little more complicated than just using a DC ammeter...
I'm not quite sure if you're dismissing the idea of measuring current, however in reality truly measuring the current over time is far from the trivial task that you make it out to be. One would have to actually measure many samples (the tighter the interval the more accurate the total draw) as surely no drive is going to be consistent in its energy draw: I would imagine it would constantly fluctuate by 100% or more as the drive does different access patterns, etc, not just hooking up a radio shack multimeter and eyeballing the draw. The net result though would be an extremely accurate gauge of the true power draw, and hence energy discharge, of the hard drive (easily as accurate as putting the draw in a controlled environment: It isn't generating heat magically).
Most drives get noisier over time. I've had many drives that were nearly silent when I purchased them, but after not much time they start to get noticeably louder, until they're unbearable. Especially Maxtor. :> Anyway, since this article (and all the others that I've seen) don't address noise levels after a month or six, they're really not all that helpful.
I've given up trying to find a quiet enough drive for my living room and just put the living room system in another room (the basement) with long cables. It's a bit awkward (though will be better once I acquire an external DVD-RW) but a much simpler solution.