Guest Eugene

Testbed4 and Power Consumption

Recommended Posts

Guest Eugene

A long time go, in a thread far, far away ( http://forums.storagereview.net/index.php?showtopic=13469 ), I mentioned that an important metric for notebook drive tests would be the assessment of power consumption. Though we attempted a variey of solutions to the problem, none were satisfactory.

Eventually, however, we turned to long-time SR community member jtr1962, a participant known for his prowess in electronics. After laying out the requirements, jtr custom designed and built this handy device:

drive_tester.jpg

We were initially going to roll out a roundup of notebook drives in December of 2004 but decided to hold off since Testbed4 was, uh, right around the corner. Now, I'm happy to say that results for notebook drives are, uh, right around the corner...

In the meantime, however, we've decided to replace our standard 3.5" drive temperature measurements with power dissipation results. While measuring top plate temps as outlined here proved quite repeatable, we've never been solidly convinced that we were providing the best solution to thermal assessment- there simply remained too many confounding variables. Now, however, thanks to the laws of thermodynamics and the somewhat closed-system environment of an SR testbed, overall power dissipation of a drive can take the place of direct temperature readings and provide SR readers with an excellent idea of the total heat a drive generates.

Here's a look at how current desktop and server drives fare:

power_usage.png

"Measured Idle" is exactly that- the power dissipated in watts when a drive undergoes no activity. "Measured Seek" is the power drawn with the drive under a full-bore random seek test.

Again, these results with replace our top-plate thermal measurements as well as satisfying an important facet of SR's impending notebook drive tests. Enjoy!

Share this post


Link to post
Share on other sites

This is a terrific surprise. Great job jtr and Eugene!

Share this post


Link to post
Share on other sites
Wow. Nice. Very nice!

Any tests-- even just a few one-offs-- planned to show drive peak startup current on each rail?

212462[/snapback]

I'd like to see some too... it'll be useful in identifying how much power is required in a RAID setup on the rails. Some controllers don't stagger spinup so this would be great info!

Share this post


Link to post
Share on other sites

Agreed, would be great to know start-up power.

But this is great! You guys have outdone yourselves!

Share this post


Link to post
Share on other sites
Wow. Nice. Very nice!

Any tests-- even just a few one-offs-- planned to show drive peak startup current on each rail?

212462[/snapback]

I'd like to see some too... it'll be useful in identifying how much power is required in a RAID setup on the rails. Some controllers don't stagger spinup so this would be great info!

212485[/snapback]

The thing is, most power supplies will quite happily deliver significantly more than their rated power for a short period of time, so, knowing the spin-up consumption doesn't really help you. The equation is a little more complicated.

However, most companies provide spinup power in their datasheets. And not only do they provide watts, they often provide V x A ratings. I checked Fujitsu, just because I like them, and found this example.

12V +5% @2.5 A, @3.0 A for  < 100us

That's better data than SR could ever give you ;). What more could you want.

Share this post


Link to post
Share on other sites
Wow. Nice. Very nice!

Any tests-- even just a few one-offs-- planned to show drive peak startup current on each rail?

212462[/snapback]

I'd like to see some too... it'll be useful in identifying how much power is required in a RAID setup on the rails. Some controllers don't stagger spinup so this would be great info!

212485[/snapback]

The thing is, most power supplies will quite happily deliver significantly more than their rated power for a short period of time, so, knowing the spin-up consumption doesn't really help you. The equation is a little more complicated.

However, most companies provide spinup power in their datasheets. And not only do they provide watts, they often provide V x A ratings. I checked Fujitsu, just because I like them, and found this example.

12V +5% @2.5 A, @3.0 A for  < 100us

That's better data than SR could ever give you ;). What more could you want.

212490[/snapback]

Agree Gilbo manufacturers use expensive and accurate measurement means most of us do not have access to. But as we all know, some PS's are more robust (will handle strain better, hence last longer before crapping out!) in component design than others-something I think jtr and I will regularly agree on ;), if nothing else. As with all results you read, you can be sure that for some reason or another, those results maybe different in your particular application. Meaning these results need to be taken with the idea that some 'fudge' factor of - or + X% should be added into the analysis.

You know Gilbo, that 'just because I like them'; could be applied to just about every drive out there, someone's going to favor one drive over another; like we all have our preferences for beauty. Apples and Oranges comparison you say? Now how is it we should interpret this 'dissipation' data? Is it not more just a measure of power consumption???

Explain this to me :confused: ---we find the formerly 'hottie' drives from WD are now the lowest consumers of power in 3.5in size, and therefore theoretically are the coolest, go figure!

I'll go for the Samsung or Seagate laptop drives just based on the earlier results with TB4 for 3.5 drives, seems like a crap shot to me... unless you have ultimate actual measured performance needs like those FS was talking about, or STR's for HD video editing or multiple video streams (or uncompressed higher res. 2k or above 'big leagues' HD video); I'd just by the drive for quietness, price, warranty, reliability...before considering minor performance gains for a given rpm range. That's just me, I don't need the 'emperor's clothes' or bragging rights. I still can't wait to get my hands on a single laptop with 60+ or greater GB SSD in the next few years; and happily get more runtime (still always have one backup battery) off of a single charge, not having to rely on slow spin up of a sleeping hard drive. SSD's are going to be awesome, if and when they arrive ;)

This consumption under various types of load, will give someone a baseline empirical type of data comparison for which to guesstimate if a drive might get too hot as a replacement in a laptop, but it's just a guess. Different laptops, have different types of means to dissipate, and remove heat from the inside of the machine. One that has a particularly efficient design may allow for a hotter running 7.2k rpm drive, but you'll not know until you install it. See, an all metal shell Momentus, may dissipate heat more uniformly over a greater surface area, than a mostly plastic one. Then again, perhaps all that metal could increase overall internal temperatures with a hotter 'hot-spot'. Just engaging in pure speculation here, I have no real experience of insight into what would or would not factor into laptop cooling/thermal issues. I wouldn't know from using measurements in TB4, whether or not this data represent something I can accurately rely on to make a decision. Say for instance, a 7.2k drive is only slightly hotter than a 5.2k drive you have in the laptop now, but that you replaced the original slouch cooler still 4.2k drive. Throwing in just a tad more heat to the equation could be all that is necessary (now maybe if you replace the DVD burners with a newer generation that consumes less power, generates less heat that will help) to got beyond what the laptop ventilation scheme is capable, and you end up with constantly overheating CPU or cooked 'nuts'/lowered sperm count ;):o

Perhaps Eugene would like to address the criticism silentPCreview has levied against the former method of testing for heat in TB3 era, and whether or not he thinks their own analysis is equally flawed? Not sure how you could measure temperatures accurately... I haven't thought about how to design a meaningful/useful means of measurement. However, even within TB4 (are there no pictures of the entire setup, with close-ups?... I'm having trouble visualizing how jtr's setup integrates) I wonder if you could employ on of those Infrared beam temperature measuring meters that many service tech's now use in the HACV industry?

http://www.infrared-usa.com/

We would expect there to be less of a difference between 2.5in laptop drives across the board, whether or not they are single or dual platter; as compared to the difference btw server class SCSI and desktop 3.5in drives. The question that will remain is whether or not these smaller differences you measure in a relatively more 'open' desktop like enclosure, amount to significant differences in space constrained laptops. Particularly in the smaller sizes like the Apple PB 12in, which uses an inferior GPU to the 15in & 17in models, probably less to do with cost than heat generation in such confined spaces.

Share this post


Link to post
Share on other sites
You know Gilbo, that 'just because I like them'; could be applied to just about every drive out there, someone's going to favor one drive over another; like we all have our preferences for beauty. Apples and Oranges comparison you say? Now how is it we should interpret this 'dissipation' data? Is it not more just a measure of power consumption???

Explain this to me :confused: ---we find the formerly 'hottie' drives from WD are now the lowest consumers of power in 3.5in size, and therefore theoretically are the coolest, go figure!

212504[/snapback]

There's no linear relationship between power consumption and temperature (IMO).

There is between power consumption and heat generation, but the 'speed' with which that heat is transfered 'away' determines temperature.

So you can't derive temperature from power consumption without taking into account that transfer 'efficiency'.

Share this post


Link to post
Share on other sites
You know Gilbo, that 'just because I like them'; could be applied to just about every drive out there, someone's going to favor one drive over another; like we all have our preferences for beauty. Apples and Oranges comparison you say? Now how is it we should interpret this 'dissipation' data? Is it not more just a measure of power consumption???

212504[/snapback]

Every drive manufacturer publishes this data. I'm sorry if it that wasn't clear. My example document came from Fujitsu simply because I have their hard disk product page bookmarked --which, indeed, is because I like them. No other reason.

It is indeed apples vs oranges. The data I quoted was spin-up power consumption. I quoted it because several posters above requested that Eugene conduct tests noting the spin-up consumption of the drives. IMO, since the information is already available with more accuracy than we can generate, including duration of peak load, it would be a waste of Eugene's time to add such a benchmark.

I don't think measuring idle or seek draw is a waste of time however. It is useful to independently, empirically measure that data. It also, as Eugene notes, provides a most accurate method of measuring heat production. Spin-up power is accurately reported by manufacturers for the simple reason that people absolutely need to know that number when working with large arrays.

There's no linear relationship between power consumption and temperature (IMO).

There is between power consumption and heat generation, but the 'speed' with which that heat is transfered 'away' determines temperature.

So you can't derive temperature from power consumption without taking into account that transfer 'efficiency'.

I don't think hard drives differ consequentially in their heat transfer efficiency. They are all virtually identical structurally. Considering this, power consumption is the most accurate method of measuring heat dissipation for hard drives. Watts in is proportional to watts out as heat.

Share this post


Link to post
Share on other sites

Many users over at [H]ard|Forum, when planning large storage arrays, are extremely concerned about spin-up current draw. At first glance at the test gadget, there's a peak option, seems like it would be fairly easy to get a spin-up draw measurement. Even a spinup-draw measurement of a single 7200, 10K and 15K drive would clear up a lot of FUD out there on HDD spin-up current draws.

Share this post


Link to post
Share on other sites
I don't think hard drives differ consequentially in their heat transfer efficiency.  They are all virtually identical structurally.  Considering this, power consumption is the most accurate method of measuring heat dissipation for hard drives.  Watts in is proportional to watts out as heat.

212509[/snapback]

It's not just in proportion, it's equal. If you're concerned about the (thermal) effects on your system then it's indeed enough to know.

But udaman was talking about HDD temperature.

Share this post


Link to post
Share on other sites

JTR1962,

Please develop some type of LABEL for your boxes, either a dry transfer or a nice case etching would do, that has your ID on it. Or for that matter, take the jpg posted above and Photoshop it on. A great piece of work - you deserve to be recognzed for it by all that see it (long after this thread has faded).

FS

Share this post


Link to post
Share on other sites
Many users over at [H]ard|Forum, when planning large storage arrays, are extremely concerned about spin-up current draw. At first glance at the test gadget, there's a peak option, seems like it would be fairly easy to get a spin-up draw measurement. Even a spinup-draw measurement of a single 7200, 10K and 15K drive would clear up a lot of FUD out there on HDD spin-up current draws.

212529[/snapback]

Please note my post immediately above your own. Manufacturers provide more detailed information than SR can gather. For example they include a duration for the peak draw.

Share this post


Link to post
Share on other sites
Guest Eugene

Guys, thanks for the feedback.

Gilbo brings up a very good point- as nice as this device is, it can't hope to match the precision presented by the manufacturers on their spec sheets. All we'll get is an "instantaneous" reading, whatever that means... no duration or the like.

That said, I should note that in this thread e_dawg argued that measuring any kind of power draw at all would be a waste since manufacturers spec it. If one quickly compares any of the idle measurements we present with those found on a manufacturer's spec sheet, one will notice discrepancies.

Does the same gap exist bewteen claimed peak power-on draw and a potentially accurate third-party measurement? Perhaps... but its much easier to measure and present idle and seek power draw in comparison.

Preliminary experiments with the device and some drives show the repeatability is there. The 5V line and the 12V line, however, peak at different times. Further, some drives register their highest readings during spinup while others do so durinig actuator initialization. What kind of standardization would readers want to see here?

Share this post


Link to post
Share on other sites

I'm still going to say that I'm more concerned with heat generation, than actual current draw, because I keep an extra battery around. However, e_dawg's hypothetical 0.2w is only valid for consideration of sleep/standby, even low idle consumes 0.85w. There has to be some current draw waking from sleep or standby idle, no free lunch. These are compromises used to lengthen battery-life. However, if one is doing work that (or play, vid games) requires mostly constant disk reads and writes; power consumption does matter. Let's do the simplistic math.

Hitachi Travelstar 7K100 series data sheet

Hitachi 7K100 line (they only specify a 'general' data set for the entire line, one or two platters) read/write average current is 2.0 watts. CPU will use perhaps 15w at full bore. But think about it 2.0watts is not that small an amount over several hours of runtime.

I think jtr can appraise those who are not familiar, but battery manufacturers typically specify capacity and voltage under 1/10C loads (1/10th the rated capacity of the battery in question). That's fine for lower power uses. The most common 'capacity' for 18650 Li-Ion batteries now (there are some 2400ma LG's available, and Sony just announced a 2650ma battery that hasn't yet made it to market) is currently 2000ma. 1/10C on a 2000ma battery, that's 200ma. The 7k100 all by itself is drawing an average of 400ma (2w/5v = 0.4a). Now a laptop battery is comprised of a set of these 18650 batteries, so the actual rated capacities is higher than a single 2000ma. Never the less, a single 18650 of 2000ma capacity would lose 20% of it's rated capacity in one hour, if that was the only load placed on it by the drive itself.

Li-Ion of the consumer type, when subjected to higher current demands will drop in ability to supply that 1/10th load rated capacity. Usually not a lot until you hit 1-2C (in the case of a 2000ma battery, 2C would be a 4A or 200w load. But at lesser loads of 1/5C to 1/2C you do lose some capacity, maybe something only as small as 10% with a new battery. But it has been stated before on other forums Li-Ion are not designed for long life, and like other battery chemistries, the more you cycle them in deeper charge/discharges the faster the capacity drops over time.

Therefore, if you are using the laptop for uses that hit the disk often, or more likely constantly, a 10% drop in capacity- say for a 4000ma rated (more likely you would find a 4000ma rated battery, is delivering ~3750-3500ma in real world usage, or less as it gets older, or less if their is heavy demand placed on the battery when using the DVD/player burner as part of the total laptop system power requirements) is nothing to disregard as insignificant or even 'minimal', in comparison to the LCD screen, or other higher drain components. SSD's are not being considered just for the 'high-tech' of it all, there are real gains to be made in power saving by replacing a HD with a SSD if you do not need ultimate capacity that the HD will provide.

Note also, for the SATA spec, read/write average is fully 25% higher at 2.5W...wonder why the substantial difference?

Share this post


Link to post
Share on other sites

I'm not sure this tells us what we might otherwise be inclined to think it does.

Each drive design is going to dissipate the power it consumes differently, and it's that dissipation that determines how hot the drive ultimately runs. Put another way, the power the HDD consumes dictates how much of energy goes in, but it's the rate at which that energy comes out that determines how quickly the HDD heats up and how hot it gets. There are many factors that dictate the rate of dissipation, including the design of the HDD itself, the efficiency of elements like the Motors and ASICs, how the HDD is mounted, what the ambient temperature is, how much airflow is presented to the operating drive, and of course the workload.

But it's important to understand that the drive that consumes the most power may not in fact be the drive that runs the hottest.

When we design HDDs, we use Infrared photography to determine what gets hot (and how hot it gets), and how well airflow cools these things, in order to be sure that the drive won't self-heat itself in to oblivion *if* operating according to the manufacturer's specifications.

For desktop and server applications, the drive's power rating is really designed more to meet legacy-box power supply constraints than thermal constraints (though older boxes with weaker supplies tend to be the same ones that have poor ventilation). The HDD could consume much more power than it does, and get better performance as a result, but the drives would crowbar power supplies in decade-old systems (that still get their fair share of upgrades). For portable applications, power is dictated by constraints defined by the battery and related power system.

That being said, power dissipation has recently started to be more important in some specific markets related to more traditional desktop applications. Consumer electonics like PVRs and set-top boxes are designed to operate in "non-traditional" environments like bedrooms, in spite of the fact that they're really pretty much desktop computers in disguise. Manufacturers want their boxes to be as quiet as possible, and in some cases omit fans altogether. The drive has to be designed to minimize power consumption while maximizing dissipation in order to work reliably in these environments. Luckily, the workloads tend to be predominantly sequential, which minimizes seek power dissipation.

Desktop and Server drives are often rated for a specific minimum operating airflow (such as "100 LFM", or 100 linear feet per minute airflow). When operated in an environment that meets the spec, drive temperatures will "stabilize" at some (fairly constant) temperature offset above ambient.

For example, when operated in the specified airflow, an Atlas 15K II will run about 10º-15ºC above the temperature of the ambient air. When operated in still air, the drive will run 30º-40ºC or more hotter than the air temp. If you put the drive in an enclosed box without airflow, it'll heat up until it gets hot enough that something fails and it stops working.

High temperatures can cause problems long before they cause outright failures. Cooler is generally better. But the only way to meaningfully measure thermal performance is in the specific enclosure of interest.

Share this post


Link to post
Share on other sites

I've been collecting my thoughts before posting in this thread. Lots of issues to address.

First off, let's start with the tester itself. The tester can obviously measure steady state power quite well, and I checked and adjusted its accuracy to within 1% under nearly all conditions, and in most cases better than 0.5%. Ultimately, the accuracy is limited by the circuit design and components used. It is possible to design a device to measure power to 0.001% but in this case it's pointless. First off, random variations in power usage will probably be much more than that. Second, the meters are only 1999 counts, meaning that results can't be displayed with any more accuracy than 1 part in 2000 or less (i.e. 0.05%).

Peak power is another animal. The tester can measure peak power fairly well, but its response time is limited by how fast the circuit can capture transients, how fast the peak hold capacitor decays, and how fast the meters update their readings (2.5 times per second in this case). It's safe to say that 100 µsec transients cannot be captured. The peak hold circuit cannot respond that fast for starters. Even if it could, the transient reading would quickly lost when the peak hold capacitor voltage decayed before the meters could update their readings. Even if the meters updated very quickly, the user would need to catch a high reading visible for perhaps milliseconds. There exist more complex ways of capturing transients, but this was beyond the scope of this tester.

Despite the limitations, the peak hold function is useful. The duration of the peaks captured are precisely those that would affect a power supply. The filter capacitors in all power supplies are designed to deal with transients in the µsec and sometimes msec ranges. This means that they essentially act as small capacity batteries. Any power demands lasting longer must be supplied by the active circuitry of the power supply. It is this circuitry which if overloaded can either fail, or cause the supply to go into thermal shutdown. These longer term, more important peak power demands are exactly what the tester captures.

Second, some concern has been raised as to why bother to even measure steady state power when the manufacturers already supply this data. Well, they also supply seek time data so why measure that? The answer is that the very point of testing is to independent verify manufacturer's claims, including claims of power usage. After that, power usage is exactly proportional to heat production. This is important because physically all drives of a given size more or less dissipate heat equally well (at least on the outside). By measuring power you know if one drive will get physically hotter than another, on the outside anyway. How well the internal components are thermally coupled to the drive case is what will ultimately determine their temperature. This coupling is technically termed thermal resistance, and is measured in °C/W. The higher the thermal resistance, the hotter the component will get. Since there is really no way to know how well thermally coupled the drive's internal components are, all we really have is case surface temperature. In any given situation, case temperature rise above amibient will be directly proportional to power usage. In a well-ventilated case of course a drive will be cooler, but the same rule holds. A drive using twice the power will rise twice as much above ambient air temperature. This is basic thermodynamics. Temperature measurements can also given us this information, but they are situation dependent. Power usage isn't. If you have a drive in your machine and it uses 10 watts without becoming excessively hot then you can replace it with another drive using 10 watts and probably not be concerned (assuming similar internal geometries).

Third, there is another very valid reason for wanting to know power usage. This is in portable systems where even an extra half of a watt can be significant over the span of a few hours. Batteries are almost always rated in amp-hours but this is often a 5-hour rate, a 10-hour rate, or in the case of lead-acid batteries a 20-hour rate (termed C/5, C/10, and C/20, respectively). If you draw power faster than the rate at which the batteries were rated then you *don't* get the rated capacity. Extra power usage then will shorten run times much more than at first glance. Here again independently verifying manufacturer's claims is a good thing.

Fourth, yes, the box would look nice with some decals. :) I usually have those made when I make production quantities of something, and also have my initials on the PC board when I have it professionally made. Since this is a one-off project, I had to dispense with those things.

Edited by jtr1962

Share this post


Link to post
Share on other sites

I'd like to see the 12V and 5V current listed separately. It will be very helpful when building machines. I'm building a machine now with 15 disks, and Seagate's documentation of max current consumption on the 7200.8 is just plainly incorrect. Some science will be nice.

SCIENCE!

Share this post


Link to post
Share on other sites

LOL watch out Udaman might get upset, how dare you say a manufacturer gave incorrect information, they are supposed to be Gods at rating power consumption of their drives, we have no need to test them ourselves with our inaccurate devices. :lol: Joking of course, just glad to see a real world test that disproves a manufacturer spec on power consumption which in turn proves what SR is doing needs to be done, good work guys can't wait for those test results. :)

Share this post


Link to post
Share on other sites

An example of power consumtion data being incorrect would be just about any WD drive that has been around a bit... Since WD change their drive internals with each new platter generation, but dont always update their model numbers or spec sheets their data often appears grossly incorrect (might as well be looking at a completely different drive).

Share this post


Link to post
Share on other sites

Thank you for the detailed description of the theory behind your methodology, jtr1962; I'm glad to see that you seem to understand thermodynamics a lot better than most, and that kind of approach gives me more confidence in the validity of the Test Bench results as a whole.

A couple of additional comments, for those who have trouble equating the similarity between power dissipation and heat generated: First, keep in mind that ALL power that goes into the hard drive becomes heat. Technically, a very small amount becomes mechanical energy, but since that's dissipated frictionally anyway, in the end it is ALL heat.

Second, as pointed out, since hard drives are externally pretty much the same--the shape and materials used are close enough to be functionally identical--the only thing that could cause a difference in how much heat is effectively "in" the drive would be interal layout, which the SR crew have no way of telling anything meaningful about (even if they tore the thing apart, they still probably woudn't be able to learn anything meaningful).

You can do this set of thought experiments to get an idea for why top-plate measurements aren't particularly useful:

1) One can assume that it's usually actuators, motors, bearings, and chips that fail due to heat. All of these are located inside of the drive or on the bottom surface. Let's say that, hypothetically, a drive manufacturer figured some magical way to transfer all the heat generated by the drive to the top plate, where it's away from the electronic components that it might damage. The drive would, based on this measurement, "look" incredibly hot, despite being very cool where it counts.

2) Similarly, pretend you had two drives, one with perfect thermal transfer (pretend it's solid diamond) while the other completely isolates the guts from the outside of the case with a layer of styrofoam. The styrofoam-encased drive would appear incredibly cool on the outside, since almost all the heat generated would be held internally, even though it'd rapidly bake itself to death, while the cool drive would appear comparitively warm on the outside surface.

As such, it's quite possible for an apparently cooler drive to actually be running hotter where it counts--in fact, if you give me two drives with the same power dissipation, I'd wager the one that felt hotter is the better design, as it probaly does a better job of dumping the heat to the surface where it can be carried away.

None of this, however, takes into account how good a job the drive does internally of keeping the important components cool, which leads to...

3) The only really good way to experimentally measure how heat would affect a drive would be to take a thermal scan of the chips on the bottom and, ideally, also the internal components. That strikes me as well out of the range of any realistic test SR could do, and in any case few of these components are near the top plate where it's easy to take a reading.

Hey, I could be full of it here, but it jibes with what I remember from thermodynamics.

Edited by Makosuke

Share this post


Link to post
Share on other sites

I'd still like to see the temperature measurement for future drives in addition to the power consumption profile.

1) A drive maybe more efficient at getting heat away from it's surface.

2) Still comparable with old test data.

3) Provide future reference data for comparison of drive temperature and power consumption. Nice pretty graph showing high corelation (won't be perfect, but the outliers would be interesting).

What-goes-in-must-come-out.

The temperature of the drive will rise until it can get rid of the average amount of heat per second. The heat came from using input power that same power heats the space inside the case and goes out the case somehow (eg. system fan, PSU fan, ducted cooling, radiated through warm case metal).

So I agree with the fact that knowing the power consumption is a more reliable indicator as to how it will fit into a system (requirements of power supply and cooling).

Leave the simple power consumption figure in the SR database. Leave further analysis to focused articles about a group of HD's and investigate their thermal behaviour under varying test conditions. Don't repeat test for single drive at a time. Detailed testing takes to long, uses expensive equipment and uneconomical for your usuall 1-2 drive per month testing rate and deadlines.

----------------------------

A near perfect thermal conductor will have a low surface temperature rise per watt radiated (the heat source will also be cool). A poor thermal conductor will require a higher temperature rise to radiate the same amount of heat as before. Yes the outside temperature would be much lower than the heat source but the internal and external temperatures will rise to a point that the outside will radiate generated heat.

Different materials and designs can assist with moving heat away from the source and improve heat transfer to surrounding environment. Some designs will be better suited for low air movement while others rely on ducted air like in servers.

A good performer in still air may gain less of a cooling improvement when provided with higher airflow than a drive optimised for higher air flow.

For example:

Drive 1 (15W Average) - 40C above in still air, 30C above with 100LFM.

Drive 2 (15W Average) - 45C above in still air, 27C above with 100LFM.

* Thermal picutres to show hot spots and how well it can get rid of the heat.

* Test HD to death? Large temperature controlled space with HD inside. Let the temperature inside rise (slowly?) and see at what ambient the drive starts to fail. Or monitor hot spot for reaching manufactures limit (maybe silicon manufature spec not nesessarily drive OEM spec).

* It's starting to be more significant to test several models from the same family. The power consumption, nise and performance is different depending on data density and how many platters/heads.

* Is it worth including a short stroking metric. eg. Can random/sequential, read/write and/or synthetic/realworld performance with 50GB on a 74GB 10000rpm WD Raptor be worse than 50GB on a similar priced ~250GB 7200rpm SATA drive ???

Share this post


Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now