The SSD Power Consumption Hoax (pics)

Message

morgandog · #1 Post by **morgandog** » Sat Jun 28, 2008 3:22 am

The subject above is the title of an interesting article on Tom's Hardware.

Check it out:

http://www.tomshardware.com/reviews/ssd ... ,1955.html

So is the presence of a SSD in the x300 is a good idea if it is not saving you battery life (and may not actually be a major performance boost).

After seeing the post on Tom's I am even more convinced that Lenovo needs to release an x300 with a decent high rpm HDD option...

EDIT: I thought I'd come back and add a bit of a summary of the Tom's article -

They tested a bunch of SSDs for actual power consumption in a real system and compared the consumption to a high performance 2.5" HDD (7200 rpm) and found that the HDD outperformed all SSD drives in terms of battery life (often by a wide margin).

The conclusion is that advertisements by all SSD manufacturers are fraudulent when they claim better power consumption. A lower performance HDD or any of the 1.8" drives should do even better and perform much better than SSDs.

Since they don't find a big performance advantage for SSD, they question the arguments for why one would ever want to use a current generation SSD in a laptop.

mod edit: added image warning to title

rbena · #2 Post by **rbena** » Sat Jun 28, 2008 5:22 am

How much of a notebook's total power does a hard drive use?

Of the total 13W my T42 uses on battery power, the 160GB-5400 hard drive averages 0.8W during use, and consumes less on idle. So a hard drive is certainly less than 10% of my notebook's power use.

It does seem questionable whether the SSD is more power-efficient than a hard drive. And with the x300 requiring a fan because of its other components, I don't see the SSD as making a major difference in a laptop running silent either.

IMHO the current generation SSD does not seem to offer a lot over a conventional hard drive, and offers a lot less storage capacity. A larger more cost-competitive SSD drive will have more attraction in the future, but I see it very much in the 'niche' category at the moment.... and also the x300 without a conventional hard drive option.

aaa · #3 Post by **aaa** » Sat Jun 28, 2008 8:34 am

rbena wrote:How much of a notebook's total power does a hard drive use?
Of the total 13W my T42 uses on battery power, the 160GB-5400 hard drive averages 0.8W during use, and consumes less on idle.

On my machine it's 12.5w when on, 10.1w when turned off. 2.4w difference.

And just because it has a fan doesn't mean it has to be on. I've coaxed the fan on my T40 to automatically turn off from time to time, and when the drive turns off as well it's totally silent.

Kinda sad that SSD manufacturers are apparently totally ignoring power consumption, though. I'm pretty sure there isn't a technical reason keeping them from having lower power consumption; they just seem to not care at the moment.

#4 Post by **erik** » Sat Jun 28, 2008 10:04 am

morgandog wrote:Since they don't find a big performance advantage for SSD, they question the arguments for why one would ever want to use a current generation SSD in a laptop.

they didn't test the samsung SSD in the X300 so you can only assume that the X300 doesn't benefit from having an SSD in terms of battery life. not all SSDs are created equal and the samsung is the best available right now.

but, even if its battery life didn't test as well as a hard drive, i wouldn't want a hard drive given the performance benefit of the SSD (see below).

rbena wrote:IMHO the current generation SSD does not seem to offer a lot over a conventional hard drive, and offers a lot less storage capacity.

the capacity argument is a personal preference and one that's difficult to debate given individual needs. the performance argument, however, is much easier to debate.

each of these disks were tested using the LSI SAS controller on my thinkstation to eliminate any variables in the thinkpads themselves.

200GB hitachi 7K200 (HTS722020K9SA00):

64GB samsung SSD (MCCOE64G8MPP):

FragrantHead · #5 Post by **FragrantHead** » Sat Jun 28, 2008 12:01 pm

Correct me if I'm wrong, but those benchmarks appear to be for access time and sequential read performance. It is well known that SSDs shine in that area. Their achilles heel however is write performance and in particular random write performance of small files, scattered across the drive. Such activity happens, for example, during software development when programs are compiled. Anecdotal evidence (e.g. a post in another forum) from as little as half a year ago suggests that some SSDs are up to 3 times slower than conventional hard disks in that scenario. Can you publish a comparative benchmark of small random writes?

#6 Post by **erik** » Sat Jun 28, 2008 12:15 pm

FragrantHead wrote:Can you publish a comparative benchmark of small random writes?

point me to a benchmark utility capable of small, random writes and i'll run the test.

aaa · #7 Post by **aaa** » Sat Jun 28, 2008 4:35 pm

erik wrote:
FragrantHead wrote:Can you publish a comparative benchmark of small random writes?
point me to a benchmark utility capable of small, random writes and i'll run the test.

It's called Windows Setup, they did it in the review. SSDs are often pretty slow at it.

akao · #8 Post by **akao** » Sat Jun 28, 2008 5:41 pm

Read the article and look at the raw graphs. The summary from the posters are misleading, so some of the statement from Tom's.

The graphs show that the different brands and models of the SSD have VERY different performance and power usage characteristics (by orders of magnitude).

As to the power usage, if you go to the last page, in fact only 2 of the SSD's tested advertised lower power usage that the HDD. So it's no surprise these power hungry SSD used more power in the benchmark tests.

As to the two SSD's with advertised lower power usage, one had a shorter runtime of 1 minute, over the course of 7 hours, which is definitely not within the precision of the test. The worst SSD had a shorter runtime by an hour. Again, this shows that the brand and model of the SSD is important.

The test is done with constant activity (MobileMark). Tom's hardware believes this is characteristic of their laptop usage, but I don't think it is true for me -- browsing the web requires very occassional disk usage.

Anyhow, this article is important in pointing out that SSD's are not all created equal, in performance or power usage. This isn't enough to write off all current SSD's (including the X300 SSD which wasn't tested).

And to the OP that says there isn't a big performance difference, look at the graphs from the article. (http://www.tomshardware.com/reviews/ssd ... 55-11.html) It shows the SSD is over 3x faster. I can tell from my own experince, that my desktop (7200 RPM, Quad Core, 4GB), is slower than my X300 in loading Vista and starting applications.

akao · #9 Post by **akao** » Sat Jun 28, 2008 5:53 pm

Sigh, I get too fired up when people post misleadingly (whether intentionally or not).

Anyhow, compare (http://www.tomshardware.com/reviews/ssd ... 55-14.html), which lists the advertised power consumption of the drives tested, to those of the SSD in the X300 (http://www.samsung.com/global/business/ ... asheet.pdf).

The SSD in the X300 uses 1/10 the power under idle, and 1/3 the power under load, compared to the best performing SSD in the test.

#10 Post by **erik** » Sat Jun 28, 2008 6:12 pm

aaa wrote:
erik wrote: point me to a benchmark utility capable of small, random writes and i'll run the test.
It's called Windows Setup, they did it in the review. SSDs are often pretty slow at it.

windows server 2008 installed slightly faster on my X300 than it did on my thinkstation, yet the thinkstation is much faster in real-world use. windows setup is a poor benchmark given the variables.

any other suggestions?

royhuang · #11 Post by **royhuang** » Sun Jun 29, 2008 7:05 am

i'm not going through the motions to provide benchmarks, but I used VMWare Workstation and installed Windows XP Pro and it seemed about 30% faster than using a 7200 RPM drive.

Put it this way: the Samsung SSD saves the X300 from being a sluggish laptop.

Also, I'm sure the underlying theme is whether the SSD is justifiable in cost, whether alone or for the X300. Economies of scale--and then the fact that it's a relatively new consumer technology--dictate that it won't vs. HDD. You do indeed pay a non-linear premium for speed and price.

So, I'm not sure how detailed this thread will go but I think AnandTech is a better place to bring this discussion up!

adrianaitken · #12 Post by **adrianaitken** » Sun Jun 29, 2008 8:32 am

Here is my unscientific results on WRITE speed only on a X60.
I was concerned that the usbstick might be a limiting factor but it isn't !!

usb -> ramdisk : 4.11 secs (FAT32 by Gavotte)
usb -> harddisk : 31.12 secs ST9120821AS (NTFS)
usb -> SSD : 7.73 secs 32Gig Samsung (NTFS)

862 jpeg total 370 Meg. File sizes 170k - 490k i.e. all small files

Why 862 files ? It's what I had in one directory !!!!
Now I am in no way saying the hd is the latest/greatest but I ran the test twice with similar results. I am happy staying with SSD and no, I have no idea how battery life is since I never run them for more than a few hours on battery then it's back on mains.

rbena · #13 Post by **rbena** » Mon Jun 30, 2008 6:23 pm

The hard disk test seemed a little slow. I ran a similar test on WRITE speed on a T42.

usb -> harddisk : 18 secs - Samsung HM160HC (NTFS)

866 jpeg total 370 Meg. File sizes 230k - 550k

Nonetheless, the WRITE time for the Samsung SSD 32GB is 2.5 times faster. Is something to look forward to with a higher capacity SSD, at that speed or possibly even faster as SSD technology improves.

adrianaitken · #14 Post by **adrianaitken** » Tue Jul 01, 2008 7:28 am

It's a 5400 speed one.

That's what happens when you have a fragmented hard disk !!!!!
Ramdisks and SSD's don't suffer the read/write penalties of fragmentation since they don't have some big old head to chuck around the casing

jketzetera · #15 Post by **jketzetera** » Wed Jul 09, 2008 8:54 pm

erik wrote:
aaa wrote:

ATTO Diskbench does the trick. It used to be a part of ATTOs HBA utilities that could be downloaded from ATTOs homepage (after registering).

royhuang · #16 Post by **royhuang** » Sun Jul 13, 2008 10:31 pm

FYI--check out Crave's blog entry "Samsung making 128GB solid-state drives--with a caveat":
http://news.cnet.com/8301-17938_105-998 ... subj=Crave

Just past midway in the article, it briefly mentions the power savings...

thomas565 · #17 Post by **thomas565** » Sun Jul 27, 2008 9:29 am

erik, I think iometer includes a random write test (www.iometer.org).

to those other's of you: try copying 500Byte sized files from some source onto the SSD and see how long this takes. I suspect maybe 5x longer than copying onto a HD. This is very relevant, since the windows pagefile, or linux swap deals with a lot of small random writes and reads, the demand in some way similar to RAM.

#18 Post by **erik** » Sun Jul 27, 2008 11:09 am

i finally found atto diskbench on their site and ran it on the samsung SSD. i no longer have the 7K200 and therefore can't compare them.

samsung SSD native in X300:

samsung SSD in thinkstation D10 on intel SATA controller:

samsung SSD in thinkstation D10 on LSI SAS controller:

from the above we can see that the bus controller makes a difference in the ultimate results.

i also have an X61 that i can test the SSD in if requested.

thomas565 · #19 Post by **thomas565** » Sun Jul 27, 2008 5:33 pm

but most importantly from erik's atto tests, the read/write is much slower than regular HDs when dealing with small files, 500Byte. Which becomes important when the drive is used for such files or for pagefile/swap.

T

Paul386 · #20 Post by **Paul386** » Sun Jul 27, 2008 6:00 pm

This article from Tom's Hardware is complete nonsense. It did not test using realistic scenarios. SSD's benefit from very low idle power consumption as compared to hard drives, so if you do a test that keeps the drives at full use the entire time (not realistic) then you never realize the advantage of an SSDs idle consumption.

Furthermore an SSD is generally faster than a HDD (at least the one found in the X300). Therefore an SSD will be able to preform a task faster than an HDD and go back into its idle state before the HDD can.

There are additional benefits to SSDs including the performance, silence, and immunity to shock (where an HDD can be damaged).

The shortfalls are the high price and potentially shorter lifespan. However, battery life is certainly not a shortfall.

akao · #21 Post by **akao** » Sun Jul 27, 2008 11:57 pm

I didn't think SSD's had such a large dip in speed for small files. The result is surprising. Where can I download atto diskbench? I want to see what the results are like for a normal HDD.

thomas565, I'm not sure there has been a benchmark showing that SSD's are slower in operation. Also, according to wikipedia, pages are usually at least 4k in size, 8 times larger than 500 bytes.

#22 Post by **erik** » Mon Jul 28, 2008 12:35 pm

akao wrote:Where can I download atto diskbench? I want to see what the results are like for a normal HDD.

https://www.attotech.com/software/files ... il_350.exe

this will save you from having to create a fake login like i did.

akao · #23 Post by **akao** » Mon Jul 28, 2008 6:31 pm

Thanks. I ran the benchmarks on my desktop, and I learned that I have no idea what I'm doing. Here are my benchmarks on two separate runs:

and with anti-virus turned off

The hard drive being tested is the system drive. The results make little sense to me. Writes being faster than reads? Does ntfs have journalling? Why the dip in the middle?

#24 Post by **erik** » Mon Jul 28, 2008 8:25 pm

akao wrote:The results make little sense to me. Writes being faster than reads? Does ntfs have journalling? Why the dip in the middle?

i have no idea. imagine my confusion when i ran this on my 300GB 15k RPM SAS RAID 1 array...

write speeds have never seemed as slow as what atto suggests. read speeds, however, are uncanny.

yesterday i bought a 73.4GB 15k SAS drive to use for the pagefile and swap space. it'll get formatted with the smallest possible cluster size then tested in atto to see if it's any faster with small files.

needless to say, we're getting off-topic here and should probably get back to discussing how SSDs are on battery life.

wizatdiz · #25 Post by **wizatdiz** » Tue Jul 29, 2008 12:13 pm

Results from a Hitachi 7K320:

The SSD Power Consumption Hoax (pics)

The SSD Power Consumption Hoax (pics)

Who is online