Hard Drives Have Less Environmental Impact Than SSDs, Seagate Says
- Reference: 0177059691
- News link: https://hardware.slashdot.org/story/25/04/18/127212/hard-drives-have-less-environmental-impact-than-ssds-seagate-says
- Source link:
On a per-terabyte basis, hard drives generate less than 1 kg of CO2/TB versus 160 kg for SSDs. The power consumption difference is equally notable. Hard drives operate at 9.6 watts with 0.32 watts per terabyte efficiency, while SSDs consume 20 watts with 0.5 watts per terabyte. "Hard drives exhibit the least carbon footprint, both in total and on a per-TB basis, offering the most carbon-efficient sustainable storage solution," the report states.
[1] https://www.seagate.com/content/dam/seagate/assets/resources/decarbonizing-data-report/decarbonizing-data-report-040325.pdf
asterisks everywhere (Score:1)
> Hard drives operate at 9.6 watts with 0.32 watts per terabyte efficiency, while SSDs consume 20 watts with 0.5 watts per terabyte. * ** *** **** *****
> * HDD compared to the worst case scenario SSD made by the worst manufacturers worst product
> ** SSD performance rounded up to the nearest whatever we felt like
> *** Based on laboratory testing, not real real-world conditions
> **** Not reflective of overall energy consumption
> ***** We will say anything to make line go up, please buy our stock.
seems legit.
Re: (Score:3, Interesting)
Not so fast. If I need to install an Operating System or piece of software; the SSD completes copying of the file in about 5 minutes, whereas a spinning disk would take 30 to 60 minutes for the same operation.
The mechanical disk has to spend an order magnitude more time to complete the same operation. 5 minutes at 20 watts (100 watt-hours) is still less than 60 minutestimes 10 watts: 600 watt-hours.
But wait: There's more. The total amount of time I leave my whole computer including the CPU an
Re: (Score:2)
> The mechanical disk has to spend an order magnitude more time to complete the same operation.
It looks like they did account for this. The W / TB value is a little hard to parse, but what they really meant was: "Power efficiency, measured as the average power consumed in watts to write or read 1 TB of data." So it does account for the greater speed of SSDs... but doesn't give them much credit for it. The numbers they give are for an SSD which is only 33% faster than the HD.
Also, this is for datacenter drives so the computational overhead is probably going to be pretty small compared to the couple
Re: (Score:2)
> It looks like they did account for this. The W / TB value is a little hard to parse, but what they really meant was: "Power efficiency, measured as the average power consumed in watts to write or read 1 TB of data."
That's only the power consumed by the drive itself, not the entire system:
> total power consumed by a drive
Re: (Score:2)
Yes, I got that. There were two separate things there: the parent was saying that the greater speed of the SSD would imply lower total power consumption by the drive, even if the drive consumed more power while in operation. And I was saying that it seemed as though they had accounted for speed.
And then the parent was also talking about total system power consumption as a separate point.
Re: (Score:2)
Isn't all this the kind of thing that should be priced in with a proper carbon tax?
Just tax carbon, then let consumers decide how much money to spend on SSD vs. HDD, time spent copying, electricity consumption, etc. They are already making all those decisions now, it's just based on pollution being free.
I don't mind it (Score:2)
if SSD is really worse at carbon.
At least I'm compensating with running the latest Devuan Linux on perfectly good old hardware instead of forced upgraded overpowered hardware with Windows.
This seems hard to believe... (Score:2)
in fact, I don't believe it. Their competitor says "On average, SSDs consume around 2-3 watts during active use, while HDDs can consume up to 6-7 watts or more."
Re: (Score:2)
It is because it includes the "embodied carbon—or, carbon emitted during upstream extraction, production, transport, bill of material, manufacturing, packaging, and distribution stages of a product’s life cycle (Scope 3)." (TFA, page 8).
In an SSD, each data bit is silicon area from the foundry, which is known very energy intensive. For HDD, there is only the controller. The bits are on the glass platter. Say at constant technology, when an SSD manufacturer proposes 2x the capacity, it duplicates
Re: (Score:2)
I'm not arguing your initial thesis: pre-consumer resources dominate this calculation. However, improved areal density is far more common than just throwing more chips in the case. M.2 drives in particular have a fixed numbers of chips, so increases in capacity means more efficient, and smaller transistors. It feels like this was a sponsored study, where they used the most efficient HDD they could find alongside the least efficient SSD. I certainly don't have any SSDs using 20W. My 4TB NVME gen3 drives h
Re: (Score:2)
> "I certainly don't have any SSDs using 20W. My 4TB NVME gen3 drives have a listed max of 6W, and typically are using about 3.5W during major operations."
But they listed "for equivalently sized data center SSDs", and I doubt your drive is a "data center" drive. Even so, it still seems like their power metric is too high.
Re: This seems hard to believe... (Score:2)
And how many ssdâ(TM)s are all consuming that amount of power to give you 30TB of storage? vs one hard drive.
If your goal is data storage volume, hard drives are still the cheapest
We can't go back. (Score:2)
In this day and age it is simply impossible to go back to traditional hard drives. Simple reason. Windows does so much under the hood, in the background, that a PC with a traditional hard drive running Windows is completely unusable.
Re: (Score:2)
> "running Windows is completely unusable"
That part, intentionally cut out, I can certainly get behind :) It is why all my machines run Linux.
But seriously, of course we can't go back. Spinning rust is still king when you need tons of slower storage at a low price. Plus it is arguably more reliable in cases of lots of writing. But if you need the speed and/or the massive reduction in physical size and/or shock durability, and can pay the price, SSD is just a hands-down game changer. It has done more
Re: (Score:2)
Indeed, when the first SSDs appeared, I cynically observed at the time that the only reason they existed was to make Micros~1's shitty filesystem seem faster than it was. Thus, it was no surprise when Micros~1 started heavily promoting them -- especially those weird "hybrid" drives that bolted an SSD on the side of a traditional hard drive to use as an extended cache.
I happen to have a few SSDs and HDDs sitting here (Score:1)
Seagate 16TB spinning disk +5VDC 0.90A +12VDC 0.72A
Samsung 4TB SSD +5VDC 1.2A
Seagate 4TB spinning disk +5VDC 0.55A +12VDC 0.37A
Kingston 480GB SSD +5VDC 1.0A
Seagate 2TB spinning disk +5VDC 0.5A +12VDC 0.33A
Dell 256GB SSD +5VDC 2.0A
I'm not going to pull SSD's and drives from running systems, but there may be something to it.
Re: (Score:2)
Those are peak numbers though, unless you are in a data center (which admittedly Seagate is targeting with their own numbers), that SSD is mostly idle because it is just that much faster completing a task. That idle disk will always be spinning (yes it can sleep but modern systems simply write or read too often for that to occur) using close to that 12v number to spin spin spin.
I don't think anyone misses the noise of a spinning disk.
Some actual data (Score:2)
from a real, very simple live server.
Running Ubuntu Server 20.04.6
Jetway fanless industrial mainboard (industrial means single 12V power)
8GB RAM
2TB, 3½" Toshiba HDD
Because the mainboard is 12V only I can power this from my laboratory power supply which indicates a fairly steady 1A current draw i.e. just 12W for the entire server minus power supply losses.
As for the HDD itself,
=== START OF READ SMART DATA SECTION ===
SMART Attributes Data Structure revision number: 16
Vendor Specific SMART Attributes
Re: (Score:2)
NVME has a different set of stats that show the available power states. PS3 is idle (0.05W), PS1 is the typical write state (5.8W). In practice, I typically see about 3.5W power draw when the drive is doing sustained activity.
NVME Identify Controller:
vid : 0x1e4b
ssvid : 0x1e4b
sn : XXXX
mn : SPCC M.2 PCIe SSD
fr : SN022788
ps 0 : mp:6.50W operational enlat:0 exlat:0 rrt:0 rrl:0
rwt:0 rwl:0 idle_power:- active_power:-
active_power_workload:-
ps
What's the denominator? (Score:2)
Even as SDDs have become more and more common in different applications, the cost per byte ratio has stayed roughly the same. The reason is that HDDs care more about cost per byte, while SSDs care more about cost per latency and bandwidth. If the denominator for comparison were performance instead of capacity, the conclusion would be different.
SSDs use heatsinks, HDDs don't (Score:2)
The point they are making seems confusing at first, but then it just comes down to power consumption per terabyte. SSDs get hotter than HDDs because SSDs use more power.
Re: SSDs use heatsinks, HDDs don't (Score:2)
SSD run hot in use but their time of use is so small due to speed. I still want a 16TB SSD on par in cost with HDD but here we are.
Cooling! (Score:2)
Exponentially higher amounts of cooling are needed for spinning drives vs SSD. Major cabon/electricity use there.
A typically narrow view of just manufacturing costs....by a manufacturer.
They didn't factor in transfer rate (Score:2)
Which is certainly a metric of use.
Did they use chatgpt to come up with the numbers? (Score:1)
Coz they certainly feel oh the high side for SSDs. 5 tons of carbon per SSD. Really?
Re: (Score:2)
Those are mind blowing numbers, certainly, and I will treat them with extreme skepticism until I see some corroboration.
Re: Did they use chatgpt to come up with the numbe (Score:2)
Do you think microchips grow in the forest?
Re: (Score:2)
> Coz they certainly feel oh the high side for SSDs. 5 tons of carbon per SSD. Really?
No kidding, and it says it's on per-terabyte basis. So 5 tons of CO2 for one TB of SSD. It's clearly impossible. If we, for example, take the fossil fuel that creates the most CO2 per unit of mass: coal; and we assume the worst case scenario of 3.37 kg of CO2 per kg of coal, that's about 1.5 tons of coal. Coal is also one of the cheaper fossil fuels. Current coal prices per ton are $103.50 USD per ton. So that would be about $155 worth of coal per TB of SSD, not to mention alll of the other costs. Now, obvi
Re: Did they use chatgpt to come up with the numbe (Score:1)
... You need a lot of rare earth minerals for solid state storage as opposed to electro magnetic storage. The CO2 numbers include the mining and refining process of said elements. The platter drive has very little in terms of chips compared to the ssd.
Re: (Score:2)
Arrgh. Ok. I realize I misread the summary now. The per TB basis was with a different set of numbers. Apparently the 5 tons has to be for an SSD around 30 TB in size (32 maybe?). Ok. So that throws everything off. So, that makes things a little more credible. I'll redo that. That would be 43 kg of coal per TB of SSD. So that would be more like $4.50 per TB. For retail, I can find prices as low as $45 per TB with larger than 1 TB drives with a quick search. When you consider profit margin for the manufacture