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."
It's about damned time. We have more than enough reviews on speed and performance, but there is a serious dearth of information on noise.
If you follow discussions at other forums for ReplayTV and TiVo owners, you already know that in that situation you don't really care about performance. A 5400rpm drive can easily handle the job. However, noise is critical, and hence, some of these systems don't have fans, making heat also critical--if you upgrade with a drive that runs hotter than the original, you're likely to have random failures.
So this sort of review is wonderful, both for the information it provides, and for encouraging manufacturers to pay attention to these factors so that they will look good in the future.
I personally care a lot about noise. I pick out my CPU fans, Hard drives and power supply/case fans based on "Which, under X decibals, provides the very best performance?". The difference between 99th percentile of cooling performance and 80th percentile usually means about 50% less noise with CPU fans.
Has anyone built a case that wasn't made out of thin sheets of metal? What if you made a case filled with sound insulation such as styrofoam or eggshell foam, leaving only the air intake/exhaust vents exposed?
Seems to me that so much money is being spent on making PC components quiet, presumably so we can nuzzle our faces next to the motherboard and take a nap, but why can't we just isolate the sound inside the box? It's my -novice- understanding of airflow design that little or no heat is dissipated by using a metal box; the heat is transfered through the moving air. Well, keep the air moving and soundproof the case.
Is this a stupid idea? Maybe it's like my idea to make a solar powered, weather balloon lifted, permanently high altitude platform for launching space missions - a fool proof and economical plan for capitalist conquest! I just need to develop an attention span and find some fundin-- HEY! Something shiny!
I wouldn't lump the entire line of drives from a single company into a category of "Bad Drives." Western Digital has many flavors of drives for different applications. I've been using their top-of-the-line JB and BB series drives for the last few years under very strenious conditions in database servers and IDE RAID configurations and they have fared exceptionally well.
One thing that could be improved on many of these quantitative reviews is if they quit relying upon surface temperature probes (which is HORRIBLY unreliable. A slightly grainy texture would make the drive appear much cooler because of reduced heat transfer), and instead go right to the source: Power consumption. Is it so hard to measure the current on the 5 and 12V inputs, and deriving an actual power consumption metric for the drive? Not only is this valuable as it absolutely directly relates to heat, but it additionally is useful for those building low power rigs.
Anyways, just a thought.
I see a couple of problems here. Not to say it can't be done, of course, but it's not going to be that easy:
- Bit densities are getting so high that minor fluctuations in drive speed could cause all sorts of read/write errors.
- All operating systems that I've ever used seem to hit the disk periodically during idle time. In order to get the benefit here, you'd have to have some threshold for activity before spinning up. However, that could be catastrophic for any kind of real-time app, like video encoding. Allowing the drive to spin down would limit the kinds of apps you could use.
It's definitely an interesting idea, but I think it may fall into the causes-more-problems-than-it-solves category.
Cheers
-b
Why not develop a speedy drive that can slow itself down if it starts to generate too much heat or if it's not being used (as opposed to shutting it completely off)? I assume it's probably much easier to create a single speed motor than a variable speed one, but what would the disadvantages be?
The heads in your hard drive fly above the surface of the media due to the circulating air inside of the harddrive. Typically the heads are also engineered to fly at a specified height at all times given a specified amount of air moving through the drive. If you slowed the drive down the head would most likely fly lower or not fly at all, this would be quite bad for your data and the head.
Hard drive spindle motors are variable speed and are typically servo controlled for speed. Adjusting the speed wouldn't be a big deal except for the above.
Honestly - I do not feel the need of anything faster than the slowest 5400 RPM HDD for now - my PC has gobloads of RAM and I can make use 128MB-per-HDD as buffer. Easily.
What worries me is, the faster you spin, the more catastrophic a failure is.
What I DO care about, is reliability and shock resistance. If anyone sells a reliable HDD that would survive 5-inch drop and still operate with all my data intact, I'll buy it in a heartbeat.
Compared to my data, the HDD and the theoretical time that can be saved with higher speed worth REALLY little. Almost nothing.
``the CPU is able to keep processing while waiting for data from the hard drive.''
IDE drives can do this. It's called DMA.
Please correct me if I got my facts wrong.
I haven't had a single WD drive that didn't turn into worthless slag within six months.
I've two IBM Deathstars that have been performing flawlessly for quite some time now. (And with mad performance, I might add.)
On the other hand, I've known people who swear by WD drives, and have had nothing but problems from their Deathstars.
Why is this? None of the people I know like hurling their drives against a wall or anything.
Simple. Production runs. A machine fscks up, you get crap. Luck of the draw, and all.
It's interesting how anecdotal (but informed) evidence like this always seems to run from absolute bad to absolute good.
Back in the Good Old Days, I worked for an engineering firm that had a Northstar mini with a CP/M console. The console talked to the Northstar and made paper tapes that would be fed into CNC machines to make big steel widgets. It took 8-inch floppies. Other support boxes took 5.25-inch floppies.
At one point we started to have constant data failures on the Verbatim floppies we were using. Disks would just die, or would not take a write the first time. We kept throwing them out, and went through cases of them. Eventually we threw out the rest of the boxes of verbatim and switched to Dysan and the problems just went away.
Of course, I never used Verbatim again. I won't even use their optical media.
I now assume that this was just one of those truly random, "Estimated Mean Time Between Failures" issues, and not everyone was experiencing such a high failure rate.
I mean, I've been using the same two Western Digital drives for years now. Right now, one of my years-old 2Gb drives is in my firewall/webserver where it gets daily constant abuse. We've also used them at work, and they don't seem to fail more than the other brands we have.
-- clvrmnky
The CPU will still be interrupted by disk I/O IRQ's unless you use SCSI. Lots of discrete disk accesses means lots of interrupting, which is why SCSI is used nearly exclusively in server environments.
"Gold still represents the ultimate form of payment in the world." - Alan Greenspan, 1999
The problem is that everybody's sample set is too small to really make good reccomendations on this. Even people with computer labs or white box guys don't have anything close to a single percent of the number of drives at any time. For instance, I have two WD drives, an 800MB and a 2.1GB drive. Both still work great despite being abused for years. The Quantum drives however....
I read the internet for the articles.
Forgive me if I'm missing the point, but I can't see that choosing a hard disk on its noise production is in any way sensible.
Noise is damned important in many applications. TiVo, for instance, needs lots of capacity, but speed is not a critical issue. Any modern 5,400RPM drive is more than sufficient. Who wants to watch a movie and listen to a loud whine from a disc drive? Also, since TiVos tend to live in "entertainment centers" and have limited cooling, heat is a big concern.
Another good example is my firewall machine. It runs my mail server, FTP server, and web server. It performs NAT for my network. The a-number-1 thing that I want from that machine (outside of reliability) is quiet. None of the applications on that machine get much action. My web server is a private page that lets me look at my system temperatures and voltages -- so it does not generate a lot of hits. The mail server serves me and a few freinds. But the machine is in my office running 24/7. I don't want to hear a loud hard drive, nor do I want to put six fans in the case to extract heat. So I run a slow, low wattage Duron (650mhz) and a 20GB, 5,400RPM hard drive.
It's all additive. The machine on which I work is loud enough because of my "need-for-speed." It's got multiple fans, hard drives, etc. And it sounds like it. The quieter I can make the other machines, the better off I will be.
ok... is every one on crack because $1000 for what is basicly RAM is bs. Thier QuicCache card that is basicly a place for the page file to go is just dumb, buy more sytem memory and then get rid of needing a page file. While thier other products are cool they are way to expensive for what you get, plus in situations where you need faster access a raid set up can be way faster than what you are going to pull through the pci bus. I can't belive anyone would seriously buy these. is there something I am missing here?