Ideas for SSD Power Consumption Testing

[W1zzard]

Just received my Quarch for SSD power consumption measurements (https://quarch.com/products/hd-programmable-power-module/)

Any ideas, requests, suggestions what power-related tests I should add to my SSD reviews? Gimme a wish list, let me know what you'd love to see, don't worry if it can be done

 

[Toothless]

Read, write, read/write, idle.

 

[Selaya]

total power consumption for the current benches? (like, booting windows et cetera)

 

[W1zzard]

Read, write, read/write, idle.

Random? Sequential? QD?

 

[Toothless]

Random? Sequential? QD?

I'd say whichever is worst case.

 

[nomdeplume]

Simulate heavy C/P-state settings v. unrestricted delivery?

To determine where a modicum of performance can still be realized inside power conservation. Firmware programming is rather intelligent these days. Almost wonder what you might uncover if you metaphorically lift the right set of rocks.

Tinkering with low power delivery might have consequences best saved until late in your testing.

 

[Wirko]

How fast it can enter and wake up from power saving states.

 

[Kapone33]

Sustained read and write but with real world applications. One of the uses for NVME is Epic. I have had to re-install Windows 11 several times. Just today I was getting blue screens and did a fresh install. This is where Epic comes in. Unlike Steam you can't just find your Game files with Epic. In many cases (Even on their Website) they will tell you to create an Epic folder on a different drive start downloading your Game, pause the download, Copy/Cut and paste your Data from the previous drive to the new drive. I know I am not the only person with this issue. I would like to see PCie 3/4 vs SSD and even RAID 0 SSD or NVME used in this scenario.

 

[Shrek]

Just received my Quarch for SSD power consumption measurements (https://quarch.com/products/hd-programmable-power-module/)

Any ideas, requests, suggestions what power-related tests I should add to my SSD reviews? Gimme a wish list, let me know what you'd love to see, don't worry if it can be done

Do solid state drives have power spikes like video cards? I imagine not, but if so they would be of significance.

 

[W1zzard]

Great feedback so far!

Do solid state drives have power spikes like video cards? I imagine not, but if so they would be of significance.

Their max power is around 5 W, so spikes are kinda irrelevant, because too small

Copy/Cut and paste your Data from the previous drive to the new drive.

So the test is basically "Copy 75 GB of game data to target drive" and measure the power + time it takes, then divide power by time to calculate energy efficiency, so that faster drives that finish the copy in less time dont get penalized for using more power?

Simulate heavy C/P-state settings v. unrestricted delivery?

Hmmm .. so 512K sequential write capped to 100/200/400/600/1000/1200/1500/.../unlimited MB/s and plot power? Or rather read? or mixed? or all 3 ?

 

[Flaky]

I think it would be good to check how drives handle power instability.

That would be:
- What is the minimum voltage it requires to run properly
- How long can power instability/voltage drop be without experiencing data loss or drive disappearing from OS (let's call that "internal hold-up time")

Why?
It's the PSU that generates 5V/3.3V rails SSDs run on. With certain mix of devices, PSU and power loads voltage may be impacted.
At least that's what I suspect may be a possible cause of rather rare problem of people experiencing drives disappearing.

It would be useful to see how picky are certain drives, and how far outside ATX spec can they really go.

 

[nguyen]

power measurement when SSD is constantly streaming textures in-game, like in PS5.

 

[W1zzard]

power measurement when SSD is constantly streaming textures in-game, like in PS5.

Any PC game recommendation?

 

[nguyen]

Any PC game recommendation?

Spiderman Remastered

 

[Wirko]

Are you going to measure the total motherboard or system power consumption at the same time as that of the SSD? It may turn out that the processor and chipset consumption delta is more than SSD's, compared to idle.

 

[W1zzard]

Spiderman Remastered

https://imgur.com/TcZyBrM

This is me zipping along the streets, only a few MB/s, with peaks at 20-30 MB/s .. not sure if this is what we understand as "streaming"

Are you going to measure the total motherboard or system power consumption at the same time as that of the SSD? It may turn out that the processor and chipset consumption delta is more than SSD's, compared to idle.

Drive only

Just spitballing, re-bar/sam vs traditionally access. Does it affect power usage.

On an SSD? Why would it? Edit: to elaborate, ReBAR is a technology for graphics cards specifically that changes the data transfer "window" size between CPU and GPU from 256 MB to the full VRAM size, so things no longer have to get funneled through a bottleneck.

 

[Blue4130]

Just spitballing, re-bar/sam vs traditionally access. Does it affect power usage.

 

[Ferrum Master]

It should be a casual task. Next thing should it be OS agnostic? I always thing do nvme(the really differ controller to controller) + pcie power states really work in Linux and how to trigger them.

Would like to to suggest creating a windows ISO from a downloaded ESD file, maybe add some custom jobs, as the process covers compression, DISM and dumping the result into ISO.

 

[Mussels]

Don't forget Gen 1/2/3/4 testing to see if the power differs under the worst case test (R+W?), as it may matter for cooling the drives

 

[Ferrum Master]

Don't forget Gen 1/2/3/4 testing to see if the power differs under the worst case test (R+W?), as it may matter for cooling the drives

How about only Gen 3/4/5?

 

[lowrider_05]

Maybe try a File Integrity check on a big Steam Game or a full Virus scan on the system SSD

 

[Frick]

J per file copied, ranging from 1KB files to 1TB files.

 

[nomdeplume]

Hmmm .. so 512K sequential write capped to 100/200/400/600/1000/1200/1500/.../unlimited MB/s and plot power? Or rather read? or mixed? or all 3 ?

I'm not confident of directly stating means of testing or delivering the results inside your well honed review template. With the latter entry being comprehensive enough to provide meaningful analysis that can be quickly compared across multiple drives. Defining the premise between segregating with power load and data load weighted heavily towards use of programmable power module in my head but ??? Hey, if this leads down a path that refines your data collection so be it.

I think it would be good to check how drives handle power instability.

That would be:
- What is the minimum voltage it requires to run properly
- How long can power instability/voltage drop be without experiencing data loss or drive disappearing from OS (let's call that "internal hold-up time")

Instability is one of multiple elements in the near area of my very loosely stated 'metric' to be developed. Trick here is zeroing in on a behavior that couldn't be eliminated through use of, in this case, UPS people will instantly seize on. Without expanding beyond his testbench into project status. We have similar ideas on exploring retarded timing and other tricks that might form a low power routine.

If I could state one definable worth to be attained here it would be charting the path of changes inflicted upon manufacturers to improve performance of their components (SSD in this case). It's not entirely beyond conception a properly developed schema could make the leap to testing of other devices and components. Beyond developing social relevance to be capitalized upon. W1zzard appears to be looking for new and exciting ways to play with his new toy that justify it's existence for business purposes. Eventually any resulting ideas need to become relevant to the audience of his reviews

 

[docnorth]

Also idle consumption and comparison with HDD(s), if possible. Thanks

 

[Wirko]

If you're going to plot graphs of power vs. time, chances are that they will look very hairy, due to large variations that look like noise. In such cases it would be good to add a smoothed line (short-term average).

I took a graph from one of TPU reviews and added a line to demonstrate this. The write speed vs. gigabytes graphs could use a smoothed line as well.