What would you like to see in hard drive reviews?

I know a lot of you are hard drive junkies, so I ask - what specifically would you like to see measured/analyzed in a hard drive review? Please provide a list if possible - any replies would be greatly appreciated.

 

Here's my short list for mechanical hard drives:

1) Amount of platters
2) Heat signature of hard drives at idle and at 100% load
3) Acoustics of hard drive at idle and at 100% load
4) Any associated software that goes with the hard drive (such as WD's Data Lifeguard Tools for example)
5) Full hard drive diagnostics/analysis with the Pro version of HD Tune

 

power drain at startup, idle, and 'active' i/o loads
warranty compared to other drives
known problems with other drives in the mfgrs related ranges
availability and effectiveness of on-board hardware encryption
use and compatibility with TrueCrypt and any performance hits

length of time internal technologies have been in large scale production
such as:
heads
arms
motor

 

4k random read/write...

 

The only thing I care about is timing a solid known configuration with real work (whatever 'your' real work is) and then showing how the new drive handles the exact same install/'work'.

Also, by 'exact same install', I do not mean a clone of the original HD; a clean install and timing of the installed O/S and programs is what I consider the 'norm' for me.

Would I like to see benchmarks? Sure, but only if the paragraph above was adhered too. Only then do the benchmarks make any sense. And, I will be able to see if there is any correlation between the benchmarks and the 'real work' timings provided (there usually isn't - in my experience).

I also like what garetjax wrote too, except for point #5 (see above) and points #2 and #3 would be pretty hard to compare to unless we happened to have the exact same notebook (and the exact same hearing...).

This is asking a lot (and I know it is much more work than running benchmarks, because I've done it in two or three threads here), but I also believe that this is the only way to really know if Drive A, is better/worse than Drive B.

Thanking you in advance!

Cheers!

 

Will you also be reviewing Solid State Disks?

 

4k random readwrites, sequencial ones and parallel ones.

and comparison of some workload (like installing adobe reader, or something) and another (compressing some stuff with winrar or such?), and then again both at the same time.

i do a lot of stuff in parallel (as you might guess), and i like to see how much parallelisation harms a hdd/ssd.

another simple test could be copying 10 iso files from one folder to another, and then copy all 10 at the same time from one folder to another, resulting in 10 copy dialogs. does it take the same speed, not?

oh, and noise.

 

This might seem odd, but how far can you drop it without it breaking/ corrupting files? - laptops always get dropped somewhen in their life.

 

As an extension to that, you might want to at least note whether a drive has any sort of free-fall sensor or other shock prevention.

 

problem is that distance is not a constant. one of my hdds died because of a touch to the desktop that shaked it for half a cm. and that desktop got transported to lanparties, moved around while running, etc. but one day, those 5mm or so that i moved it was too much.

 

idle/normal (run a script which visits a few websites, starts a few programs, ... for maybe half an hour and calculate the average power consumption)/stress power consumption
idle/normal/stress noise and hints to clicking noise or other noticeable problems
large file transfer copy (no move) (>4GB), small file transfer (a few thousand files), large file extraction with a mixed archive which contains many small files and some huge files
and a real world benchmark. Either use some PCMark or so, which seems to test more than just access times and max. files transfer or write an own program which simulates real world use, like one minute doing nothing, then opening several files, moving them, doing nothing, ... And measure the time each time.

 

Boot time, application loading time, file copying time.
As long as it's done on the same system with the same image so it can be compared apple's to apple's between different drive's you review

 

As Frank mentioned. If you guys at notebookreview can upload the same you know 4gb file consisting of thousands of small files. A timed test on how long it takes to copy is more important than hdtune. hdtune is synthetic only, timed tests are real life.

and please NO unboxing videos. They are lame

K-TRON

 

Quite right, and I agree 100%. The only way to really obtain an across the board, equal playing field among all hard drives is to use the same laptop/OS combination every time for HDD testing. This type of testing has been done for years when benchmarking CPUs, CPU coolers, thermal greases, and memory modules for example.

For acoustics, that could be a problem if you don't own an anechoic chamber, but from what I've researched you could build one on the cheap. However, the devices required to record HDD acoustics and interpret the data isn't going to be cheap I wager.

 

Temperatures.

 

read and write speeds... basically a HD tune benchmark and how long it takes to transfer a 4GB file and especially temps....

 

Great, thanks for all the feedback. I am definitely going to run PCMark, HDTune Pro, and I will be doing some file transfers to see how long it takes.

What programs do you suggest using to measure things like 4k read/write, and power consumption?

 

the new crystal disk mark beta maybe? it might not be the best, but it's simple, quick, and consistent.. which is what we care about in tests, not?

power consumption, good question..

 

For power consumption:

Idle: Computer is doing nothing but sitting there with HD not timing out and stopping itself from spinning. Time how long the computer sits there like that before it shuts itself down while on battery power.

Playing DVD video from HD: Computer plays DVD file from HD and timing it until it shuts down while on battery power.

Run PerfectDisk 10 in infinite loop: The HD is fragmented and then PD10 is run in infinite loop until the battery gives out. ( Loop = fragment HD, run PD, repeat). Also, keeping a note on how many Runs were completed in a given amount of time would also be very useful too.

Between these 3 states, we should be able to know which HD makes a meaningful difference to power consumption.


Specifically, I don't care if at idle the numbers are .9 watts or that during seek r/w the number jumps to 3 watts. What I would love to see is that at idle, Drive 'A' lasts 10 minutes longer than Drive 'B'. Or, constantly using the HD like in the PD scenario Drive 'A' lasts 30 minutes longer than Drive 'B'.

Along with the notes that " not only did Drive 'A' last 30 minutes longer looping frag/defrag runs, it also completed 11 cycles instead of the 7 cycles that Drive 'B' did" we would not only be able to see which drive is more power efficient, but we would also be able to see which drive does the most work on a single limited battery charge too.

Why are the above distinctions important? Because the firmware electronics of the drive make up it's power profile too and the above scenarios make sure that those electronic circuits are put to use and measured too.

Cheers!

 

- number of platters
- temperatures.
- power consumption idle and under load
- noise, idle and under load.

 

love #5, and i think #2,3 are going to have problems as some people may test it in the laptop, and laptops differ in their cooling... also room temp, if my laptop is left in a car at night, and i pull it out, i can get 10celcius!
maybe if they its externally connected, laying on a specific surface, and with a certain room temperature???

 

"This might seem odd, but how far can you drop it without it breaking/ corrupting files? - laptops always get dropped somewhen in their life."

hmm, is that dropping it in a laptop?
and it matters, cause a thinkpad might have better results than a 400 dollar toshiba from bestbuy

 

The test can also be done with an external enclosure. This would seem easier for a shock resistance test.

 

There is no such thing as a definitive 'drop' test.

The angle of the fall has as much to do with any resulting damage as the height of the fall does.

This is a non-sense test for me - if I actually drop, jar or otherwise shake a working HD while it's 'on', I replace it as a matter of course. Not a matter of if it will develop symptoms, but when.

And, unless you keep testing it till it fails, then who's to say the next drop won't kill it?

My vote? Drop the drop test.

Cheers!

 

If you are going to fall for the 4kb cluster read/write test, you may as well do it on (ntfs) filesystems that are 'normal' and then have encryption and/or compression enabled to see if there are any performance hits.

Ideally you'd be able to do the same tests with two or three of the more common linuxii filesystems.

It's not uncommon for video card firmware to be optimized to cheat common benchmark tests, may as well make sure that hard drive firmware hasn't been 'optimized' the same way.

A test bed system might be linux based and capable of driving the tests/benchmarks on ntfs, ext3/4, jfs, and xfs. That way you'd be running your tests using the same os, kernel, and device driver revisions minimizing variables. The major variable would then be the filesystem structure itself.

 

Guys synthetic benchmarks don't cut it. Here's some evidence: I have a Seagate 5400.6 here and a WD1600BEVT.

The Seagate beats the WD in WEI (5.8 vs. 5.4), HDTune, Crystal Mark (also the 4K reads and writes) and Atto.

In real life the WD beats the Seagate in everything I throw at it, booting, application launching, copying and installing.

Are you seeing the value of synthetic benchmarks?

Even PC Mark Vantage is a synthetic benchmark.

Real life benchmarks. Timed by hand.

For example:
-booting Windows 7
-Install Microsoft Office (or another large application)
-hibernating and resuming
-duplicating a 5GB folder with a variety of small and large files
-opening a 40GB tiff file in Photoshop
-launching Word
-scan a 10GB folder for virusses

The best hard drive review are made by Laptopmag in my opinion. Here's an example: http://www.laptopmag.com/review/storage/intel-x25-m-g2.aspx?page=6

The drives that I'd like to see included:
Intel G2 80GB (and/or 160GB)
Corsair P64 (and/or P128)
WD5000BEVT (500GB)
WD3200BEKT (320GB)
Hitachi 7K500 (500GB)

Why these drives? They're widely available, within most people's budget and offer good price/performance.

Ideally speaking the tests would have to be run on drives that have been used for 3-6 months.

 

synthetic benchmarks still cut. they still use the drive. they just focus on specific tasks and how well it handles that. every "real life test" is the same: a syntetic test of a specific task.

is boottime relevant to how fast i can copy files from one disk to another? no. but knowing it still helps to determine a fast drive.

i just hope there will be multitasking checks. they show if a disk really is fast.

a singular mtron is f.e. very fast, but not if you have more than two big disk related jobs running. the intel ssd is still fast even when 4-5 different jobs are running (or more).

and in a lot of cases, even just starting an app, for a very short period, there are tons of parallel jobs.

 

Tomshardware, who solely rely on synthetic benchmarks, still believe the Seagate 7200.4 is the fastest hard drive. Go figure.

Yes multi tasking would be nice to include. Laptopmag does that very nicely.

 

tomshw is a not accepted source of information for me anyways. when ever they do something wrong, notice "omg look at those results", they release an article about it. instead of thinking "they could be wrong".

edit: "synthetic or real" benchmarks, it's how you interpret and understand the result, that matters. if its repeatable, it allows comparisons. what those mean, then, that's what matters.


imaging doing a benchmark of the intel ssd, and only testing how fast you can copy big files onto it. it would be below most ordinary hdds. that this doesn't matter, and that it's still much faster for most uses, that is what the tester has to understand, and thus weight and interpret the test.