Very happy with my Hynix P31. Power efficient, doesn’t get hot or need any heat management add-ons, no issues with boot or hibernate on my Framework (batch5 i5 DIY).
1 Like
I managed to snag a 1tb model for just under $100 on eBay although I realize that is not really a proper “retailer”
Also found this listing over on Newegg, a price that is technically under MSRP
Certainly the 2TB model seems to be…let’s say rare but while not as plentiful as other SSDs, they are still available
Let me check UK sites and see if I can dig up a listing for you
Edit: Dang, can’t find even out of stock listings-best advice is to get a DRAMless drive if power savings is what you are after, Framework supports HMB so the performance drop shouldn’t be too bad, just get the best one you can I guess
1 Like
I think on “power efficiency” (and money efficiency) terms SATA SSDs tend to fare very well. They don’t have as much a heat buildup issue (although it mostly depends on controller firmware & algorithm)
Technically, if both under full load a 2.5 inch 7200 (or 5400) rpm HDD and a m.2 2280 SSD can draw comparable amount of power (1100mA at 5V versus 1700mA at 3.3V), which is 5.5W and 5.6W respectively. However, the HDD will draw about 500mA constantly while I would imagine that an SSD at idle would draw no more than 100mA, so you can quite certainly say that SSD is winning here.
However, regarding to ratings I honestly don’t see a difference between SATA and NVMe, with drives for both often rated at the same 1.7A at 3.3V. However I would imagine that a SATA drive will use slightly less power as the slower transfer speed as well as the lack of a PCIe device on the bus allow components to work with more headroom.
Unfortunately, Samsung no longer offers SATA based drives, which is an oddity
I still rock the 860 evo 250GB, although I feel like that with my new laptop coming up, I might use a new drive. Depend on how many slots there are, I guess.
WD and Seagate seem to have a few (current) offers if you want SATA m.2 SSDs.
Ultimately the best way to get extra miles out of your laptop is to dim the display (maybe reduce resolution too) and turn off the fan.
1 Like
Framework doesn’t support SATA M.2 SSDs, so I wouldn’t recommend going that route for the Framework Laptop!
3 Likes
wait what
why
Have you tried that
having a M key slot doesnt mean it can’t take SATA sticks (which are often M + B keyed)
Like many things, being able to physically fit one doesn’t mean it’s compatible.
On Tiger Lake (mobile at least), you can’t have PCIE 4 NVMe on the same slot as SATA, so obviously Framework chose PCIE4 because that’s what most people are likely to want.
4 Likes
Ok. In that case I believe the most low power storage would be Intel Optanes.
Not like the Optane SSDs, those actual 32/48GB Optane sticks. Obviously you can’t get much out of it, but that would be your answer.
Other sticks should behave exactly the same.
Now, it is not impossible to have SATA on Tiger Lake. You would have a PCIe to SATA chip somewhere to handle the protocol. But if Framework didn’t put it there (or if Intel dropped support), then let it be. SATA is pretty deprecated by now anyway.
Thanks all, some good ideas to look into there. Options so far:
a) Try to source a Hynix P31 from…somewhere. It’s probably possible with some effort.
b) Investigate DRAM-less + HMB
c) Investigate Intel Optane – but probably I don’t want anything this exotic
I’m also planning to scour the benchmarks for anything that will be at least a bit cooler/less battery-draining than average.
It’s possible there’s not much difference between the WD Blacks listed on Framework’s site and anything else that’s easy to get. But let’s see.
I found this reddit discussion: https://www.reddit.com/r/framework/comments/qur1bf/is_the_sk_hynix_gold_p31_worth_it/
Summary:
- P31 is low power at idle and under load
- WD SN750 reportedly is not great at idle
- WD SN850 reportedly uses lower power at idle but might get warm under load, but no big deal, it will just throttle if it needs to
It looks like I can get a P31 on eBay from the USA for not too much money, so I might try that. But probably an SN850 is not a terrible choice for anyone who wants something easily available.
Out of interest, is the difference really going to make that much of a difference? Are we talking single digit minutes on a charge and 2-3 degrees, or something actually significant?
Honestly, I suspect it will be hard to notice. If a drive is going to pull 750mW when idle instead of 4mW, I’d expect it to make a small difference to battery life. If it’s going to pull an extra couple of W under load and load goes on for a long time, I’d expect it to get hot and make a noticeable but not catastrophic difference to battery life, and probably also lose some performance from throttling.
I guess the thing is, getting the Framework DIY is an opportunity to over-optimise and over-engineer everything. And I will know I have a nice, cool, un-stressed, efficient storage device.
It really does make a difference, I remember reading an article from Laptop Mag testing various SSDs and they squeezed an extra hour just by changing the SSD in one case
Personally nothing I do is truly that intensive so a decent gen3 drive satisfied my performance needs but better battery life is appreciated by everyone in a mobile device
1 Like
You WANT something with DRAM so the drive can cache when it flush the buffer because SSDs can write by bit but only erase by block (especially the new high density ones)
Not having a DRAM reduce both speed and increase wear.
Yes and yes. Intel Optane is basically a technology that can both write by bit as well as erase by bit, with drastically increased durability. This means it need extra wiring (and transistors), which is why the storage density is pretty low.
yeah. so it depends more on the controller and firmware version. I might look into samsungs’ controller differences. So far I knew that the pm941 (not a) is the one used on old Macbooks. But in reality it didnt make a difference because I cannot get a working Mac OS installer.
1 Like
DRAMless SSDs do have lower power consumption on average and HMB gives DRAMless designs the benefits of a DRAM cache albeit at a latency cost
I don’t think durability really should be that big of a worry unless heavy writing is involved
1 Like
I have been checking amazon UK everyday for a few weeks now and this morning the 1TB P31 was available for dispatch in 5 weeks so I placed an order. (hopefully this means stock is reserved for my order when a shipment arrives at the amazon US warehouse)
Just keep checking and hopefully you’ll be able to snag an order. I’m still going to be checking each day to see if 500GB becomes available but I’ll be happy with the 1TB if I can’t get a 500GB.
1 Like
Just checked amazon again for the 1TB and it’s available for dispatch in 4-5 weeks now so I’d place an order if you still want one.
Fwiw I found that there were P31s available on eBay from the USA and they looked legitimate. The price was marked up a bit and the shipping cost was a bit high, but to me it was worth a try.
I tried to find some decent charts comparing power consumption. Anandtech does not seem to have done an SSD review for a long time. Tom’s Hardware does them but more recently has not been including the P31 on charts. For idle power usage they are showing the usage with “ASPM/LPM disabled” which I imagine on a laptop is not going to be the case for very long. Other than that the Samsung 980 Pro seems to score best on most power charts.
In fact you can combine the charts from this:
and this:
…looking at which, the WD Black SN850 actually scores better than the P31 for efficiency of a 50GB write. But the ASPM/LPM disabled power consumption on the P31 is massively lower at 359mW compared to 1530mW for the SN850.
It looks like the Samsung 970 might perform slightly better with ASPM/LPM enabled. But the P31 looks like the best all-rounder for low power.
Acquiring Hardware would be an interesting problem.
However, there are quite a lot of SSDs out there. Instead of SK Hynix I might go with Samsung, or maybe try a stick of WD, Kingston, Sandisk or Toshiba.
As I said, there are lots of variables and even the same company may produce sticks with drastic different power consumption/ratings, let alone different ones.
Speaking of power …
WD Green SN350 have 3.5W max and 5-25mW sleep/idle.
WD Blue SN550 have 3.5W max and 5-30mW sleep/idle.
WD Blue SN750 have 4W max and 5-30mW sleep/idle.
WD PC SN730 have 6.27W max and 3.5-30mW sleep/idle.
WD Black SN750 SE have … hm. They don’t tell you.
WD Black SN750 have 9.24W max and 3.5-70mW sleep/idle.
WD Black SN850 they don’t tell you …
WD Black Call of Duty®: Black Ops Cold War Special Edition SN850 shuld just be a regular SN850. Again, they don’t tell you.
WD PC SN810 have no details.
WD IX SN530 have 5.5W max and 3.5W average.
WD CL SN520 have 5.8W peak, 2.5-25mW sleep/idle.
WD Red SN700 have 9.24W peak, 3.5-70mW sleep/idle.
WD PC SN530 have … hm. 75mW active, 5-20mW sleep/idle.
Samsung is … hm.
Listed by amount of power consumed, in increasing order:
(ugh my pentium 6405U is struggling to render the webpage)
980 PCIe 3 (4.3W average)
970 Pro (5.2W, 8.5W burst)
970 Evo Plus (5.8W, 9W burst)
970 Evo (5.7W, 10W burst)
no data on 980 pro. All drives are either 500GB version or 512GB version.
All the 970 have 30mW idle, 980 PCIe3 have 45mW.
A 860 SATA 500GB have 2.5W write, 4.0W burst and … no data on idle. So yeah. there you go.
Kingston’s DC1000B here have 1.90W idle, 1.81W read and 5.47W write.
Their NV1 here have 5mW Idle (wtf), 1.1W Read and 3.3W Write.
And the Fury here have 5mW Idle, 2.7W Read and 4.1W Write.
My thought is that the DC series are crunching numbers because they are self-encrypting (among other things)
Note that the “max” power is the absolute maximum (10us) and happens during, say, drive bootup/refresh (perhaps charging up internal capacitor/grid for clearing blocks). Normal write is about half of that, and normal read is about 1/4 of that.
Also note that WD have a “sleep” power consumption and an “idle” power consumption. These have to do with different sleep states (S3/S4 respectively). You can assume it’s the high number when the system is on and the low number when the system is sleeping.
Further note is that a faster drive (e.g. Samsung 970 Pro) will be able to finish read/write operations faster (or don’t have to work as hard) than a slower drive (e.g. WD Green SN350) and therefore enter sleep/idle early, affecting total power consumption.
1 Like
The problem with WD is that their “datasheet” for the Black drives don’t contain the “boring info chart” found on many other components (e.g. max tolerance, minimum guaranteed performance, shock tolerance, temperature, humidity, mean time between failure, different commands, baud rate, additional capabilities) and instead focus on vaguely brag about its performance (e.g. 7000MB/s (burst) read/write), so consumers are more likely to purchase it (because nobody care about shock tolerance).
I categorize that as a “flyer” rather than a datasheet. But, they did provide very detailed sheets for the other drives and I am pretty happy about it.
I went to the Crucial site to look for their SSD today. Nothing about power consumption.
They even labeled the link as “consumer flyer”, so its neither a datasheet nor will it contain any “industrial” regulatory info (such as power consumption).
I guess I would have known that since it mentioned " Designed specifically for intensive workloads, high-quality creative content, and hardcore gaming", which are very consumer things.
@Xavier_Jiang tomshardware.com has some data on the SN850, and they do efficiency testing (which I find to be more useful than pure w/h comparisons). Unsurprisingly, the P31 tops their charts, but the 980 Pro comes in second, which delights me as I bought one months ago.
1 Like