Home · Register · Join Upload & Sell

Moderated by: Fred Miranda
Username  

  New fredmiranda.com Mobile Site
  New Feature: SMS Notification alert
  New Feature: Buy & Sell Watchlist
  

FM Forums | Nikon Forum | Join Upload & Sell

  

Z9 (and Z8) buffer / framerate in backup mode

  
 
maratus
Offline

Upload & Sell: Off
p.1 #1 · p.1 #1 · Z9 (and Z8) buffer / framerate in backup mode


Hi,

Iím trying to find solid information about this, but for some reason I canít. And especially for the latest firmware, which matters since Z9 has been out for a while.

The question for Z9 owners is if thereís any change in the effective buffer size and sustained framerate (once the buffer is full) between single card mode and dual card backup mode?

To be more specific, Iím looking for identical high-end 1400+ MB/s cards (i.e. Delkin Black or similar) in both slots, and Lossless RAW / HE* RAW written to both cards. Itíd make sense for Z9 to be capable of writing to both cards at the same exact speed. But, knowing how Z8 performance degrades even if tiny JPEGs are being written to its SD slot, Iím no longer confident that Z9 can maintain its write speeds to both cards simultaneously despite both of them being CFExpress.

As a bonus question for the Z8 users, what buffer size and sustained framerate are you getting when backup up to V90 (300 MB/s) cards?



Feb 22, 2024 at 03:55 AM
snapsy
Online
• • • • • •
Upload & Sell: On
p.1 #2 · p.1 #2 · Z9 (and Z8) buffer / framerate in backup mode


Dual-card tests start at 6:30. Results start at 11:25




Feb 22, 2024 at 04:13 AM
maratus
Offline

Upload & Sell: Off
p.1 #3 · p.1 #3 · Z9 (and Z8) buffer / framerate in backup mode




snapsy wrote:
Dual-card tests start at 6:30. Results start at 11:25



Thanks, I saw that video before and completely forgot the Z9 measurements he did! I thought he only did single card tests for it.

According to his results, Z9ís effective buffer size for Lossless + Lossless to dual cards is 92, and Lossless + JPEG fine is 193. If the camera was able to maintain the same write speed per card in backup vs single card mode, the numbers wouldíve been the same 193 in both cases. The discrepancy means that buffer + memory controller is bottlenecking the combined write bandwidth. The numbers are still incredible though.



Feb 22, 2024 at 10:28 AM
RoamingScott
Online
• • • • • •
Upload & Sell: Off
p.1 #4 · p.1 #4 · Z9 (and Z8) buffer / framerate in backup mode


maratus wrote:
Thanks, I saw that video before and completely forgot the Z9 measurements he did! I thought he only did single card tests for it.

According to his results, Z9ís effective buffer size for Lossless + Lossless to dual cards is 92, and Lossless + JPEG fine is 193. If the camera was able to maintain the same write speed per card in backup vs single card mode, the numbers wouldíve been the same 193 in both cases. The discrepancy means that buffer + memory controller is bottlenecking the combined write bandwidth. The numbers are still incredible though.


Note that if absolute speed is necessary, you can shoot to one card only and then copy the contents of that card to the backup card in one click in the camera menu.



Feb 22, 2024 at 10:33 AM
snapsy
Online
• • • • • •
Upload & Sell: On
p.1 #5 · p.1 #5 · Z9 (and Z8) buffer / framerate in backup mode


maratus wrote:
Thanks, I saw that video before and completely forgot the Z9 measurements he did! I thought he only did single card tests for it.

According to his results, Z9ís effective buffer size for Lossless + Lossless to dual cards is 92, and Lossless + JPEG fine is 193. If the camera was able to maintain the same write speed per card in backup vs single card mode, the numbers wouldíve been the same 193 in both cases. The discrepancy means that buffer + memory controller is bottlenecking the combined write bandwidth. The numbers are still incredible though.


Based on experiments I've done in the past the camera may serialize the writes between cards so that only one card is written to at a time. That would explain the 50% reduction he saw with dual cards. The experiment I did was to use cards with intentionally-mismatched performance, one very fast and another slow (I think it was on a Z8 with a Delkin Black CFE and a 10-year SD card that does 20 MB/s). If I ejected the SD card during a burst the camera spins waiting for the card to become ready. If I then pulled the battery and examined the still-inserted fast CFE it only contained the same number of images as the SD card, meaning the camera at the very least serializes on a per image basis across the two cards, if not also on a per-card basis.



Feb 22, 2024 at 10:52 AM
maratus
Offline

Upload & Sell: Off
p.1 #6 · p.1 #6 · Z9 (and Z8) buffer / framerate in backup mode


snapsy wrote:
Based on experiments I've done in the past the camera may serialize the writes between cards so that only one card is written to at a time. That would explain the 50% reduction he saw with dual cards. The experiment I did was to use cards with intentionally-mismatched performance, one very fast and another slow (I think it was on a Z8 with a Delkin Black CFE and a 10-year SD card that does 20 MB/s). If I ejected the SD card during a burst the camera spins waiting for the card to become ready. If I then pulled the
...Show more

There's something much worse than that. Z8 can have unlimited buffer @ 20 fps with lossless RAW and even lossless RAW + JPEG fine being written to single CFE as long as the card is fast enough. But even adding normal JPEG (8.5 MB per file, 170 MB/s effective bandwidth) backup onto a 300 MB/s capable SD drops the buffer to ~54 frames (2.7 seconds). Tiny 1.7 MB jpeg basic files still result in the buffer choking after 134 frames (6.7 seconds), and the impact of the SD card should be non-existent in this extreme case. Both cards are more than capable of handling the last configuration at 20 fps indefinitely, even if the camera can write only 1 file from the buffer to 1 card at a time.



Feb 22, 2024 at 03:46 PM
 


Search in Used Dept. 

maratus
Offline

Upload & Sell: Off
p.1 #7 · p.1 #7 · Z9 (and Z8) buffer / framerate in backup mode


RoamingScott wrote:
Note that if absolute speed is necessary, you can shoot to one card only and then copy the contents of that card to the backup card in one click in the camera menu.


That's a great idea. Does it do incremental copies? i.e. is it possible to copy only new files every hour?



Feb 22, 2024 at 03:47 PM
RoamingScott
Online
• • • • • •
Upload & Sell: Off
p.1 #8 · p.1 #8 · Z9 (and Z8) buffer / framerate in backup mode


Sadly it does not do incremental copies, I have bugged a few people who have Nikons ear about that.

maratus wrote:
That's a great idea. Does it do incremental copies? i.e. is it possible to copy only new files every hour?




Feb 22, 2024 at 03:50 PM
snapsy
Online
• • • • • •
Upload & Sell: On
p.1 #9 · p.1 #9 · Z9 (and Z8) buffer / framerate in backup mode


maratus wrote:
There's something much worse than that. Z8 can have unlimited buffer @ 20 fps with lossless RAW and even lossless RAW + JPEG fine being written to single CFE as long as the card is fast enough. But even adding normal JPEG (8.5 MB per file, 170 MB/s effective bandwidth) backup onto a 300 MB/s capable SD drops the buffer to ~54 frames (2.7 seconds). Tiny 1.7 MB jpeg basic files still result in the buffer choking after 134 frames (6.7 seconds), and the impact of the SD card should be non-existent in this extreme case. Both cards are more
...Show more

What you're describing is caused by the same issue I described - the serialization of writes between cards. When the camera is writing to the SD it's not writing to the CFE, which means the effective throughput to the CFE will be capped to that of the SD card's for all time periods SD writes are active. So then we have to calculate what the SD utilization/throughput would be for this workload, to know what % of time the CFE will be throttled to SD speeds, so that the effective CFE throughput can be calculated. The 300 MB/s rating for the SD is only for large sequential writes. A single image file involves at least two metadata writes plus data, and that's assuming the data is written as a single I/O by the camera rather than via scattered writes for the disjointed areas of the file, including the metadata and embedded jpg. Each of those I/Os incurs overhead in both the camera's I/O stack and in the SD firmware. For example, this review of a Sandisk 260 MB/s-write SD card shows the 4K QD-1 random write performance of the same card at less than 2.5 MB/s.



Feb 22, 2024 at 04:19 PM
eskimochaos
Offline
• • •
Upload & Sell: On
p.1 #10 · p.1 #10 · Z9 (and Z8) buffer / framerate in backup mode


I got ~200 with dual 512GB Pro Grade Gold 4.0 cards and HE* file size. I have not tried running one card to see if the buffer increases. I have yet to experience the Ďbottomlessí buffer that others speak of, even with a card rated to a min sustained write speed of 2400MB/S.


Feb 22, 2024 at 08:22 PM
RoamingScott
Online
• • • • • •
Upload & Sell: Off
p.1 #11 · p.1 #11 · Z9 (and Z8) buffer / framerate in backup mode


eskimochaos wrote:
I got ~200 with dual 512GB Pro Grade Gold 4.0 cards and HE* file size. I have not tried running one card to see if the buffer increases. I have yet to experience the Ďbottomlessí buffer that others speak of, even with a card rated to a min sustained write speed of 2400MB/S.


You'll never get "bottomless buffer" in backup mode. Literally all you have to do to get it is 1) shoot a fast enough card and 2) shoot HE* to it by itself.



Feb 22, 2024 at 08:24 PM
maratus
Offline

Upload & Sell: Off
p.1 #12 · p.1 #12 · Z9 (and Z8) buffer / framerate in backup mode


snapsy wrote:
What you're describing is caused by the same issue I described - the serialization of writes between cards. When the camera is writing to the SD it's not writing to the CFE, which means the effective throughput to the CFE will be capped to that of the SD card's for all time periods SD writes are active. So then we have to calculate what the SD utilization/throughput would be for this workload, to know what % of time the CFE will be throttled to SD speeds, so that the effective CFE throughput can be calculated. The 300 MB/s rating for
...Show more

I decided to calculate the exact slowdown with full serialization and you were completely right! That alone is enough to bring the Lossless + Fine* buffer down to 33 frames and less than 7 fps sustained. Which, in fact, isnít too bad at all.

I donít think that writing 1.5-2 MB files would cause a lot of IOPS to kill the write speeds so much for good SD cards, in fact it should be close to true sequential values. After all, thatís 500x larger than default 4K benchmark. And usually even 128K sequential writes are done faster than 90% of the max speed. We wonít have more tha 20 files per second either. What could be the case though is ridiculous overhead per file caused by the camera buffer hardware. I suspect that thereís a significant delay before next frame is ready to be written, and the negative effect of such delay is more obvious with tiny files.

But yeah, overall itís frustrating. Even more frustrating than using UHS-II in 2024



Feb 23, 2024 at 01:20 AM







FM Forums | Nikon Forum | Join Upload & Sell

    
 

You are not logged in. Login or Register

Username       Or Reset password



This site is protected by reCAPTCHA and the Google Privacy Policy and Terms of Service apply.