p.2 #1 · A Technical Look at CF Express Power Consumption and Why the Sony Alpha 1 May Be a Power Miracle
So I haven't found any specs for CF Express Type A cards but I did find a very interesting piece of relevant information looking at an AnandTech detailed review of the Samsung 970 EVO Plus NVMe drive. Since CFExpress is based on NVMe, it has a lot of the same characteristics.
In the Samsung 970 EVO Plus review, it lists the manufacturer specs for the different power states of the drive, similar to what we got from the Delkin CFExpress B card specs:
PS 0 - 7.8 W
PS 1 - 6.0 W
PS 2 - 3.4 W
PS 3 - 70 mW
PS 4 - 10 mW
The PS 3 and PS 4 are the standby and sleep states and the power draws are notably higher than what we saw in the Delkin. Interestingly, the Samsung 970 EVO Plus uses 4 lanes of PCIe 3.0 whereas CFExpress B drives only use 2 lanes but the Samsung PS 0, 1, and 2 power states are somewhere in between the Delkin CFExpress B power draw and ProGrade. This seems to indicate that active operational power draw is much more dependent on how efficiently a manufacturer designs the chips rather than the number of PCIe lanes used.
But that's not really the interesting part. What perked my mind is the real world measured Active Idle Power Consumption vs the Idle Power Consumption charts:
The Active Idle Power Consumption are the power draw numbers when the card is operational but not actively reading/writing AND with the APST/LPM disabled. The APST/LPM is the Autonomous Power State Transition feature from NVMe 1.1 and is where the operating system instructs the card on when it can go into the idle/sleep states. It is the primary power saving feature of NVMe drives. With it disabled, we see that an NVMe drive takes between 0.2 - 0.7 watts or 1 - 1.5 watts of power depending on what manufacturer/model of NVMe you have because it is not able to go into the PS 3 and PS 4 power states. (Ignore the Intel Optane drive, that is a fundamentally different technology than typical flash NAND)
The Idle Power Consumption chart is the power draw when the card is has the APST/LPM feature enabled so it properly goes into PS 3 and PS 4 idle/sleep states. This results in an order or magnitude or more of power savings. The power draw now is between 0.009-0.087 watts! Although there are two drives listed in that chart still using 0.322 and 0.732 watts even with power saving features enabled (Bad design likely).
We also see from the tables and charts that going into and out of the PS 3 and PS4 sleep states takes on the order of single digit milliseconds or less.
What does this all mean for CFExpress in our cameras? I think this builds a case that Canon did not properly implement APST/LPM in the Canon R5 or purposely chose to disable it leading to the excessively high power usage and heat in the CFExpress slot/card in the camera because a CFExpress card going into the the PS 3 and PS 4 states would only consume very small amounts of power. It would not be heating up to high temperatures like we see in the Canon R5 doing nothing.
The APST/LPM is an optional power saving feature so it is not inconceivable that Canon left this out of the Canon R5 firmware as a differentiating feature compared with the Canon 1DXiii that does use dual CFExpress B and where we have not heard of excessive heat or battery consumption. It is something that would force a professional to buy the upcoming Canon R1 over the R5. There is no reason to leave APST/LPM out or in a borked state since going into and out of true idle/sleep modes take single digit milliseconds or less. The buffer in the camera can easily store your photos for a few milliseconds while the CFExpress card gets into the P3 operational state. This is what enables CFExpress to use less power than UHS-ii for the same amount of data processed.
I think Sony will have properly implemented APST/LPM in the Alpha 1 firmware for the CFExpress A card slots and won't use this as a cripple hammer that Canon appears to be using to ensure future sales of pro bodies. It will ensure that Alpha 1 runs much cooler than the Canon R5 resulting in much less thermal noise and performance issues.
This also very likely means that our laptops and desktops are not putting the CFExpress B cards into the idle/sleep states because the card drivers of the card readers are not properly implementing APST/LPM or got lazy and left it out. A CFExpress B card doing nothing in a card reader should be room temperature, not hot, if the drivers are using the proper power saving features of CFExpress. Every spec sheet I have seen for a CFExpress B card mentions the power management features, there is no acceptable reason why it is not being used unless you are putting CFExpress cards into a desktop or server that runs 24/7.
p.2 #2 · A Technical Look at CF Express Power Consumption and Why the Sony Alpha 1 May Be a Power Miracle
I think since we are at the beginning of CFExpress usage, it makes sense to me that they are not implementing APST/LPM.
Since CFExpress cards are intended to be used, and cameras are typically not left idle for long periods of time without having photos taken (at least mirrorless), while the screen is on and the shutter can be pressed, entering LPM is impractical.
The wake-up time from LPM is, in the best case, hundreds of milliseconds and in most modern NVMe drives it is in the several seconds range.
For example, from the article you linked you can see that it takes nearly 14 seconds for the 970 EVO to come out of LPM
p.2 #3 · A Technical Look at CF Express Power Consumption and Why the Sony Alpha 1 May Be a Power Miracle
Jesse Evans wrote:
I think since we are at the beginning of CFExpress usage, it makes sense to me that they are not implementing APST/LPM.
Since CFExpress cards are intended to be used, and cameras are typically not left idle for long periods of time without having photos taken (at least mirrorless), while the screen is on and the shutter can be pressed, entering LPM is impractical.
The wake-up time from LPM is, in the best case, hundreds of milliseconds and in most modern NVMe drives it is in the several seconds range.
For example, from the article you linked you can see that it takes nearly 14 seconds for the 970 EVO to come out of LPM
Seeing these numbers I do not believe that Canon has cripple hammered much of anything....Show more →
You are mistaken, that latency chart is measured in microseconds. It takes less than 0.014 seconds in the real world for that Samsung NVMe drive to go from the idle to P3. This is why I say Canon cripple hammered the R5 by not properly implementing the power save features of CFExpress.
And on paper specifications for all the NVMe drives I have checked out show single digit millisecond wake up times for a modern drive. Multiple seconds is more for spinning rust. APST/LPM has been in the NVMe standard for a long time, it should be old news for CFExpress as the new kid on the block.
p.2 #5 · A Technical Look at CF Express Power Consumption and Why the Sony Alpha 1 May Be a Power Miracle
buffalowolff wrote:
So..... what are video recording times looking like on the a7sIII using cfe-a vs sd? Its not like we're waiting on the first use of this equipment.
Not sure if anyone compared that yet on the A7Siii but interestingly, Sony lists the power usage of the A7Siii as using 4.3 Watts when taking photos with the kit lens. The same test Sony did for the Alpha 1 says 4.6 Watts. Since the Alpha 1 shares the same processor, EVF and card slots with the A7Siii, I think this means that the new 50 megapixel sensor and AF on the Alpha 1 only uses 0.3 Watts more than the 12 megapixel sensor of the A7Siii. I think that's a very good number considering what we are getting.
p.2 #6 · A Technical Look at CF Express Power Consumption and Why the Sony Alpha 1 May Be a Power Miracle
TopPixel wrote:
You are mistaken, that latency chart is measured in microseconds. It takes less than 0.014 seconds in the real world for that Samsung NVMe drive to go from the idle to P3. This is why I say Canon cripple hammered the R5 by not properly implementing the power save features of CFExpress.
And on paper specifications for all the NVMe drives I have checked out show single digit millisecond wake up times for a modern drive. Multiple seconds is more for spinning rust. APST/LPM has been in the NVMe standard for a long time, it should be old news for CFExpress as the new kid on the block....Show more →
I am mistaken, I misread the chart. Thanks for pointing that out!
p.2 #7 · A Technical Look at CF Express Power Consumption and Why the Sony Alpha 1 May Be a Power Miracle
TopPixel wrote:
Not sure if anyone compared that yet on the A7Siii but interestingly, Sony lists the power usage of the A7Siii as using 4.3 Watts when taking photos with the kit lens. The same test Sony did for the Alpha 1 says 4.6 Watts. Since the Alpha 1 shares the same processor, EVF and card slots with the A7Siii, I think this means that the new 50 megapixel sensor and AF on the Alpha 1 only uses 0.3 Watts more than the 12 megapixel sensor of the A7Siii. I think that's a very good number considering what we are getting.
p.2 #9 · A Technical Look at CF Express Power Consumption and Why the Sony Alpha 1 May Be a Power Miracle
Hi..the expense is altogether because of Sony being the main producer of those cards. It's another tech, give it time and you will see costs going down.
Is there any valid reason why they wouldn't go that course? That produces more benefit for them.
A large portion of the size => you addressed your own inquiry. The cards being so little implies that they can fit 2 card openings in their cameras, which is a greatly improved arrangement than 1x CF-b + 1x UHS-ii space.
