HDD waterblock, BB style ;)

[BladeRunner]

Butcher

That Seagate link is good.

Not sure they all come with internal monitoring though, a couple of Maxtors that I had from 2000 / 2001 weren't supported as are quite a few according to this list.

http://www.siguardian.com/products/h...ture/hdds.html

I'm sure most new ones today come with it. I suppose its of only overall guidance if you don't know where the diode is mounted, could be next to a hot running IC :shrug:


bigben2k

I understand what you are saying there will technically be an optimum way to cool a particular drive, what I saying from my experience is just a thin copper block with one drive underneath, one on top both just connected by contact through the lid label is enough to significantly reduce the running temps. I could probably make some elaborate and complex design to get the dive 5c cooler overall but it's really not required. Like said I had intended to fix the block to the drives with a copper side plate, and probably still will at some stage, but it's not an important requirement for cooling them

Mine also don't touch the metal case sides as this leads to vibration induced seek noise.

[Butcher]

That list is definitely not definitive. It lists my IBM drive as not supporting monitoring, but I was able to monitor it with a tool from IBMs website (which sadly only runs in dos). The issue is there's 2 places in the SMART data that can contain temperature data, I'm guessing that tool only reads one of them and so claims not to support HDDs.

As for temp sensor placement, I have no idea, but it should be placed somewhere where it can reliably detect overheat as that's the point of having SMART on hdds at all. Most HDDs seem to list a temp of 65C on their internal sensor as maximum btw.

The most reliable IDE monitor I've seen so far is this tool: http://private.peterlink.ru/tochinov/DTemp/DTemp.zip
Sadly no SCSI though.

[Arcturius]

According to Hitachi, 50C is what they consider to be the nominal drive temp, and 65C is the final trigger point for overheating.
Interestingly enough, they figure that running a drive at 35C should extend its life 35-65% (so it can be obsolete even longer...). Overheating to 65C decreases it by about 22-33%.

Also, the document claims that they are measuring the base casting temperature.

Hitachi link

[Butcher]

Well I have a program cobbled together. It reads both my scsi and IDE HDD temps, including the one that was listed as not supported on your link Bladerunner.

I'll whack a GUI on it tomorrow and get it looking reasonable.

The current prog if anyone wants to test whether their drive has temp sensors.

Oh yeah, no win9x support for IDE disks atm (I'll sort that tm) and if you want SCSI disks to read you'll need an ASPI driver.

[Butcher]

Seems latest MBM has HDD monitoring, so I'm just going to use that.

[BladeRunner]

Originally Posted by Butcher

Seems latest MBM has HDD monitoring, so I'm just going to use that.

Yeah that's what I've been using as its been a feature in MBM5 for well over a year now. Before that I used the HDD Temp Pro app, but it was flaky to say the least. I did ask the author of MBM a few questions about HDD Temp Pro program and polling, (I'm sure I wasn't the only one), and a little while later he made HDD temp monitoring an included feature. He also included the Asus smart Doctor GPU & Ram temps as well which was great when I ran Asus 3D cards because my temps were reported in MBM. Here's an old screenshot when I had all temps reported.



Not sure if MBM supports the latest Asus cards based on ATI's Chipsets?. One thing I had to do to get the SMART HDD temp reporting working with Abit NF7-S r2.0 was ditch the Nvidia IDE controller that was default installed with the nforce driver package. It works fine using the default windows controller.... current HDD temp 22C

[j3rkoff]

^22c!!! :o :o


I guess that's why you should watercool your hdd's? My antec case has an 80mm fan in front of the hdd's, and temps still reach 38-40c. I like your design bigben, it looks like a clean, professional style design, that would perform well.

Also, can anyone tell me what i could expect performancewise from the copper tube/plate design on pages 2-3? Just wondering, because I can make one like that, and it doesn't seem restrictive to flow.

[bigben2k]

Thanks.

Volenti's solution is a really good one, because it doesn't interfere with the drive cage. The only drawback is the thermal resistance of the steel, which is going to reduce performance a bit, but hey, it's a hard drive, not a CPU. I'd still expect temps to be in the twenties though.

[BladeRunner]

Originally Posted by j3rkoff

^22c!!! :o :o

I guess that's why you should water-cool your hdd's?

I don't think my block on top cooling is that efficient.... some explanation required..

You need to take that figure I got in the context of my particular system, that is completely fanless with coolant at about 14c to 16C when that screen shot was taken in 2002. 22C Hdd temp is nice but anything under 40C all thats' required really. To me it's 45C and over that's unacceptable...

I'm not using the underground tank cooling now due to a house move but still have 16C coolant atm.



This is using a large tractor Rad & fan (at 4v) remoted along with the pump, in the much cooler loft space of my house. This allows a silent PC environment but of course is only a winter solution....... in the spring I'll be doing a new underground system.

[Yo-DUH_87]

Originally Posted by BladeRunner

I'm not using the underground tank cooling now due to a house move but still have 16C coolant atm.

Did you leave the pipes in the wall? Would be interesting to see the looks on the faces of the new homeowners if they happened to be doing remodeling

[nicozeg]

There's a lot of theory about the best place to watercool a Hard drive, but no real data. While I was in the process of making an HD block, I discovered that a friend had an IR thermometer buried in a tool box, and I asked for it.



This is a must have toy for chasing hot spots!
(Metal surfaces don't work, need to be painted or covered with tape.)


I wrote an article about this in my friend's page, You can read it here

If don't understand the languaje, just see the pics, are self explanatory.

[BladeRunner]

Yes I got one too, mine was very cheap, as in FREE we use them at work and one somehow ended up at my house To be honest they are not expensive and although a lot of people say they are not very accurate I've found them pretty good in my testing. Not very happy if I point it at shiny copper tho

was super usefull in finding hot spots in my PSU

=========================================

no pipes in the wall.... but the bomb is still in the ground, probably think it's a unexploded WW2 bomb if they ever dig it up

[aBo]

Originally Posted by nicozeg

There's a lot of theory about the best place to watercool a Hard drive, but no real data. While I was in the process of making an HD block, I discovered that a friend had an IR thermometer buried in a tool box, and I asked for it.

The temperature differences between those measure points were pretty much what I had gathered on numerous forums about this very same subject. However it is always great to have evidence, great job nicozeg!

IMO cooling the base of the HDD is waste of time because even the hottest spot is only 1.4C hotter than the sides, not worth the extra effort required to construct a fitting waterblock there.

So: If you're planning to keep the HDD's in the 3.5" rack, then I would go for the top cooling as it can fit the rack and is also more than efficient enough to cool down the HDD and if you're using 5.25" spots or making a sound dampening box for the HDD, I'd go for side cooling as then the required space isn't an issue.

$.2

[bigben2k]

Originally Posted by nicozeg

There's a lot of theory about the best place to watercool a Hard drive, but no real data. While I was in the process of making an HD block, I discovered that a friend had an IR thermometer buried in a tool box, and I asked for it.

This is a must have toy for chasing hot spots!
(Metal surfaces don't work, need to be painted or covered with tape.)

I wrote an article about this in my friend's page, You can read it here

If don't understand the languaje, just see the pics, are self explanatory.

Nice work Nico!

This demonstrates what I've been talking about.

The backside cover doesn't touch the spindle motor (air gap), and that's where the heat comes from, so cooling the cover isn't going to do much. It's pretty obvious from the pics that water cooling the spindle motor directly is going to be pretty inconvenient, hence my decision to just cool the sides, and stick a few miniature heatsinks to cool any hot electronics, with the help of a small fan.


BTW, I picked up this neat tapping tool yesterday: it's a screwdriver, capable of tapping three size holes: 6-32, 8-32 and 10-32. Check out "Klein tools" for more details.

[nicozeg]

Originally Posted by bigben2k

The backside cover doesn't touch the spindle motor (air gap), and that's where the heat comes from, so cooling the cover isn't going to do much. It's pretty obvious from the pics that water cooling the spindle motor directly is going to be pretty inconvenient, hence my decision to just cool the sides, and stick a few miniature heatsinks to cool any hot electronics, with the help of a small fan.

Remember that this is valid in a seagate barracuda, some drives have the cover screwed to the motor.

About the electronics: The smaller square chip was at more than 60ºC, but seems it works happy at that temp.

[BladeRunner]

If the link highlights anything its more that the whole drive is a pretty similar temp all over, backing up the belief its mainly general heatsoak over time, rather than high point heat. it's hard to be definite because all drives will differ by there design and components used.

Not wishing to shun that persons results at all, but it was measured with the drive on its edge. How long was it like that for? because we all know heat rises.

With those infra red sensors to get a close comparison, all surfaces should be the similar colour too.

It's easy to say in theory cooling the cover wont do much but it seems to in practise, true it may not be as good as cooling the motor or sides, but to infer it doesn't work well is incorrect in my experience. It really depends on how efficiently you need to cool them, balanced with other important factors like space block design & layout.

[nicozeg]

For the test I placed the disk on the edgeoutside the case, away of air currents. Then run different HDD intensive apps for about two hours. At that time dtemp had stabilized, reporting 42º-43º, then I measured.

Color don't make a noticeable difference, the black parts showed the same when covered with masking tape. Only metal surfaces read bad, so i covered the motor's center with tape, and read the cover on the white sticker.

Meassured temps were the following:

Motor = 40.2º
Base = 39.2º
Side =38.8º
Cover =36.4º

So the total delta between the coldest and hottest spot on the drive, excluding chips, was 3.8ºC. Not a big thing to me, considering that the goal is to keep it reasonably cool. I'm not going to overclock my hdd, just want to lower the airflow need.

I'm prety sure that all you need to watercool a drive is just a spit on any surface, but now i know that my spit is going to be 3-4ºC more efficient on the underside.

[fhorst]

Originally Posted by nicozeg

I'm prety sure that all you need to watercool a drive is just a spit on any surface, but now i know that my spit is going to be 3-4ºC more efficient on the underside.

I should agree with you for 100%, as it is the same way I cool my HDD's but.....
It's not only from cooling the hottest spot, contact area should matter also! And the contact area fron the sides is larger.

So I'm a bit confused here It seems to be that the HDD transfers the heat generated at the hot spot (motor) fairly well (good design).
So should I focus on the hotspot with a smaller area or the sides with a bigger area to get the best cooling :shrug:
Or both?

[bigben2k]

Originally Posted by nicozeg

Remember that this is valid in a seagate barracuda, some drives have the cover screwed to the motor.

About the electronics: The smaller square chip was at more than 60ºC, but seems it works happy at that temp.

That would be the difference that validates top cooling.

BTW, I read somewhere that a temp decrease of 10 deg will double the life of an electronic component, and I happen to know that 45 deg C is on the hot side of comfort for an IC, so I'd cool that 60 deg C chip if I was you!



Fhorst: do what you think is best. It's a balance between efficiency, and the amount of work you're willing to put into it.

[Butcher]

Most ICs are fine as long as they're below 100C. 45C is certainly not "on the hot side".

[Gooserider]

Fhorst :
I should agree with you for 100%, as it is the same way I cool my HDD's but.....
It's not only from cooling the hottest spot, contact area should matter also! And the contact area fron the sides is larger.

How do you figure? On a standard 1/2 height, 3.5" HDD, I get about 6" X .75", or approx 9"^2 for both sides. The top is about 6" X 4" or 24"^2, looks to me like the top has about 2.5 X the surface area of the sides. On top of that, the useful side surface area is reduced by the fact that the sides on many drives are milled to recess the casting everywhere but at the mounting screw holes.

I'm holding a Seagate Cheetah 10KRPM drive (ST39102LW) in my lap as I type this to get my measurments. I would estimate the actual contact area for a flat plate bolted to the sides is actually in 3 strips about 1/4" X 3/4" on each side plus a thin ridge ~1/16" along the base casting / top seam. Certainly less than 2"^2 total. IMNSHO, this is NOT consistent with an intent by the manufacturer to dissipate much heat from the sides, indeed it looks to me like they are trying to MINIMIZE the contact area!

The bottom (PCB side) is even worse, there are 6 little islands (4 threaded mounting holes, two smaller balance pads) coming off the base casting, each about the size of a 6/32 screw head. The drive motor is well below the plane of the contact pads. So is the rest of the circuit board, and all the electronic components. However, it does appear that the PCB is either touching or is in VERY close proximity to the base casting, I could not see a gap between them. It looks like the PCB is intended to be cooled by conduction into the base casting.

OTOH, the top has an irregularly shaped high area, most of which is covered by the drive label, however I would estimate that the useable cooling surface is about 75-80% of the drive top.

I would add that according to at least some of the material I've read (including a big article on hard drive mechanics on the big hard drive review site) the PRIMARY source of heat in an HDD is air friction and turbulence from the disks churning the air inside the drive. This air turbulence also means that heat will be transferred very efficiently from any hot spots to wherever the cooling solution is being applied.

Gooserider

[MMZ_TimeLord]

IMPORTANT NOTE:

I just learned a valuable lesson! IF you screw a block across the HDD motor on each side... like this...



MAKE SURE the block does not press the motor (i.e. - make sure you have about 2 thousandths clearance on the motor with the block tightned down.

Mine was pressing on the motor and eventually caused the drive to have physical errors... I'm still trying to get an accurate Ghost image off of the RAID pair.

This apparently warped the casing of the drive, throwing the platters out of alignment with the heads... USE CAUTION WHEN PUTTING ON THIS KIND OF SOLUTION!!!

I may change over to a drive cage made with copper and water channels. :shrug:

[BladeRunner]

Definitely worth warning about but you can't possibly be sure your block killed it? We take HDD's for granted now but they are still intricate complex items and occasionally they can just die for no obvious reasons....... :shrug:

In trying to get a ghost image have you tried the drive in all positions? If you don't get any joy and your data is important take it to a data recovery place they might be able to help...

[Butcher]

Also note IDE hard drives have much higher failure rates and are much less tolerant of vibration and heat than SCSI. It's possible a block bolted across the disk like that altered the vibration characteristics enough to throw the heads off. Additionally a lot of IDE disks only have a bearing at one end of the spindle making it relatively easy to put the spindle out of alignment.
Lots of nice data about disk design here.

Back on HDD cooling, I'm thinking about setting up watercooling for 3 HDDs. I was thinking about a pair of side blocks with the HDDs bolted in and then the bolts mounted on a frame with rubber grommets for vibration isolation. However reading that article has me concerned that the drives would vibrate each other a lot and cause poor performance or errors. Any idea how much more noise I'll get from just securely bolting the blocks to the frame (thus giving the whole assembly a lot more rigidity to reduce vibrations)?

[BladeRunner]

I find that its mostly seek noise that dramatically increases if you hard mount the drive to the case. Motors are virtually silent on cuda IV (sp) but this increased a little with rigid mounting. Different drives and different cases will of course give different results.

I'm not sure that the vibration issue is anything to be worried about, mine have been "non mounted" with bottom drive sitting on some open cell soft foam, motor side down block just sitting loose on top. The other drive flipped over with motor side upwards on top of the w/b. No case connection at all except for thin grounding wires, data and power... two years now no obvious issues, although one seems perhaps a little "nosier" in heavy seek..