SSD power consumption question

[Jay Ko]

I'm thinking about installing an SSD in a Thinkpad X60s. Given the well-known thermal problems with this machine (the palm-rest gets uncomfortably warm under many circumstances), I'm double-checking power dissipation before I make a decision.

My machine currently has a Hitachi 7K100 5400RPM 60GB drive that's rated at 2.3W read/write, 0.9-2.0W idle and 0.25W standby (all ratings are 'average', whatever that means). Other 5400RPM laptop drives I've looked at have similar specs.

For the SSD, I was looking at the Western Digital SiliconEdge 64GB SSD (http://www.wdc.com/en/products/products.asp?driveid=787). WD's website lists the maximum read power at 2.0W, max write at 3.5W and standby power at 0.6W. Average read and write power will obviously depend on usage, but 0.6W standby for an SSD seems pretty high. The Intel X25-M 80GB has an idle power of 75mW typ. Why 8x more power for the WD unit in standby? I read someplace the WDC has a 64MB DDR cache while the Intel has a smaller, slower cache... if true, this could account for much of the difference.

Intel doesn't spec peak power for their drive (I wonder why), but they do list active power during a particular benchmark. Unfortunately WDC doesn't provide comparable data about their drive.

Has anyone seen an SSD review that measures power under various conditions? It's hard to believe any SSD wouldn't be significantly better than a mechanical drive, but I'd really like to see some data.

- Jay

 

[sub.mesa]

0.6W idle at the wall socket (taking into account ~80% efficiency power supply) seems reasonable; that's what review sites like TechReport tell. That would be 0.5W DC usage (without the power supply inefficiency).

So generally you could say:

SSD : 0.5W idle
Notebook HDD : 0.7-1.5W idle
velociraptor idle: 4W
5400rpm 2TB green idle: 4.5W
7200rpm single-platter: 6W
7200rpm multi-platter: 7-9W

Exceptions are crappy SSDs like OCZ Apex; which use two crappy JMicron controllers to make a little bit less crappy RAIDed controller; still crap by my definition. But it does cause its idle power to be significantly higher.

You should know that unless your computer is running some background tasks daily, the only thing that really counts is the idle power consumption. Even when streaming a movie, that would only cause minor sequential reads which do not require much power at all. HDDs use most power when seeking; doing random I/O at below 1MB/s throughput; not when they read sequentially at 80MB/s+.

SSDs on the other hand, use more power as they use multiple flash channels in parallel, writing on multiple flash chips at the same time; normally this kind of I/O would only last very shortly. Only in benchmarking situations or server usage would this be something that really impacts on power consumption.

In other words, both SSDs and HDDs consume in most common scenarios extremely close to the idle power consumption; whether they are being used or not.

 

[Emulex]

so when you set the ssd drive to sleep after 1 minute it still draws 0.5w?

 

[Jay Ko]

0.6W is not at the wall socket - it's into the drive.

I don't quite understand how mechanical drive idle powers are specified - the data sheet for the 7K100 I have now lists 'performance idle' , 'active idle' and 'low power idle' - the range is from 2.0-0.9W, but who knows what mode the drive is operating in. I assume all of the idle modes keep the disc spinning, and if that's the case it's pretty sad that the WDC SSD uses nearly the same idle power as a spinning hard drive. It looks like 64MB of DDR2 uses about 150mW in standby (per Micron's data) and the flash should be close to zero when not selected, so probably WD's controller just does a poor job of internal power management. Intel's 75mW idle power is more what I'd expect from a well-engineering SSD.

 

[jchu14]

Intel SSD's 75mw is very optimistic (or just measured in some odd fashion). This review http://www.cmoullas.net/reviews/44-ssd/90-intel-ocz-ssd-power-consumption
pitted the G2 Intel ssd against a OCZ vertex and 5400rpm hdd.

The G2 uses 0.6w while idling and up to 2.4w while writing. The Vertex is more power efficient in most operations and the 5400rpm is worse in all conditions.

 

[Jay Ko]

Thanks for the link. A couple of things concern me about that test. First, what do those different states mean (i.e. Idle, Burst, Load, etc.), and how does he know what mode the drive is in while making his measurements? Second, he says "measurements represent the maximum observed readings". Again, if the drive were mostly in idle mode but periodically woke up to do something (or if some OS task accessed it occasionally), this could cause a short spike in the current that he would have presumably recorded. When we talk about idle mode what we really care about is the average power while the drive isn't doing anything. It would have been interesting to also see the minimum recorded current while in idle mode. If both the minimum and maximum observed current fell within a relatively narrow range close to the reported 120mA then I'd be more convinced that Intel's 75mW/15mA idle spec is bogus. The author mentions in one of the comments that he never saw the drive "anywhere close to what Intel said" (referring to minimum power I think), but there's no data. Also it's not clear the power management on his test system is configured as aggressively as one would like to see in a laptop (and if you care about a watt or two, you're probably running this in a laptop).

Really what I want to know is the long-term average (since this relates to heating), but you need a more complex setup to measure that.

Most SSDs have a few tens of MB of SDRAM cache (at most), and keeping that refreshed doesn't take a lot of power (well under 100mW). The flash basically draws nothing when deselected. And the micro/ASIC controlling the drive should put itself into a low-power state when it's not doing anything. 150mW sounds doable, which I doubt a mechanical drive could touch unless it's spun down.

- Jay

 

[IntelUser2000]

Intel SSD's 75mw is very optimistic (or just measured in some odd fashion). This review http://www.cmoullas.net/reviews/44-ssd/90-intel-ocz-ssd-power-consumption
pitted the G2 Intel ssd against a OCZ vertex and 5400rpm hdd.

The G2 uses 0.6w while idling and up to 2.4w while writing. The Vertex is more power efficient in most operations and the 5400rpm is worse in all conditions.

The system needs to support DIPM(Disk Initiated Power Management) in order to get 75mW of power for the X25-M. Only on laptops with supported chipsets you get DIPM, forget about it on desktops. That's why the idle power numbers are higher on that review.

http://www.anandtech.com/show/2614/16

 

[Jay Ko]

IntelUser,

Do you know if there is standard support for DIPM on Win 7? XP? Linux? Does it rely on the host chipset at all? I did a google search and it looks like a registry patch is sometimes required to enable it. Maybe this is something Intel's SSD toolbox takes care of.

- Jay

 

[IntelUser2000]

I think Windows 7 is fine, I'm not sure about earlier versions. You can check once you have the product using the Intel SSD Toolbox, on my system it tells me its not enabled. When I was reading about this certain brand of laptops needed driver(firmware?) update to get DIPM enabled. Or was that a Windows patch?

Anyway your Thinkpad is a Core Duo based one? I think its 50/50 on whether it'll work or not. DIPM is part of the mobile SATA spec and is supported on the latest drives. Even regular platter HDDs will benefit from this. On SSDs its better because spin down/up time is basically zero in terms of user experience.

 

[jchu14]

The system needs to support DIPM(Disk Initiated Power Management) in order to get 75mW of power for the X25-M. Only on laptops with supported chipsets you get DIPM, forget about it on desktops. That's why the idle power numbers are higher on that review.

http://www.anandtech.com/show/2614/16

Great info! Thanks.