D850 Buffer Can Exceed Specs (under certain conditions) ..

[Tony Bonanno]

Use to shooting the D5 for action. Very pleased to get a D850 to work alongside the D5. Was curious about the buffer capacity. Generally, with most of the normal settings, Nikon's spec of 51 frames for 14 bit Lossless Compressed full size RAW files seems to be fairly accurate. There are several features and variables that come into play regarding buffer capacity. Generally, the higher the ISO (and noise that affects file size), the fewer frames before buffer slows down. Card speed is another variable, etc.

HOWEVER, I did find an easy way to pop that buffer capacity up to 80-90 14-bit lossless compressed full size RAW files before the buffer slows down.. Not that most of us need that kind of performance very often, but if you do, try the following:

1. Use ONLY an XQD card (remove the SD card)

2. Set Image Quality to 14 bit lossless compression full size RAW

3. Cut OFF Auto ISO

4. Use a low ISO... at 200 ISO I was getting over 80-90 frames before the buffer slowed down. At ISO 400 getting about 80 frames, etc. You get the picture..

Try it and see if it works for you..

Cheers..

Tony Bonanno Photography, Santa Fe, New Mexico
ASMP Professional Member, APA National Member
Instructor, Santa Fe Photographic Workshops
www.bonannophoto.com

 

[Tom Frerichs]

Use to shooting the D5 for action. Very pleased to get a D850 to work alongside the D5. Was curious about the buffer capacity. Generally, with most of the normal settings, Nikon's spec of 51 frames for 14 bit Lossless Compressed full size RAW files seems to be fairly accurate. There are several features and variables that come into play regarding buffer capacity. Generally, the higher the ISO (and noise that affects file size), the fewer frames before buffer slows down. Card speed is another variable, etc.

HOWEVER, I did find an easy way to pop that buffer capacity up to 80-90 14-bit lossless compressed full size RAW files before the buffer slows down.. Not that most of us need that kind of performance very often, but if you do, try the following:

1. Use ONLY an XQD card (remove the SD card)

2. Set Image Quality to 14 bit lossless compression full size RAW

3. Cut OFF Auto ISO

4. Use a low ISO... at 200 ISO I was getting over 80-90 frames before the buffer slowed down. At ISO 400 getting about 80 frames, etc. You get the picture..

Try it and see if it works for you..

Cheers..

Tony Bonanno Photography, Santa Fe, New Mexico
ASMP Professional Member, APA National Member
Instructor, Santa Fe Photographic Workshops
www.bonannophoto.com

Thank you for that very useful information, although as you pointed out it's not often that we need that much performance. On the other hand, several fairly short bursts can quickly fill a buffer.

Tom Frerichs

 

[Slapstick Noir]

That's an interesting test.
Does the crop mode increase the buffer?
Do you have to necessarily remove the second card or setting it to "Flow" would get you the same increase?
Cheers!

 

[Tony Bonanno]

That's an interesting test.
Does the crop mode increase the buffer?
Do you have to necessarily remove the second card or setting it to "Flow" would get you the same increase?
Cheers!

Good questions. Haven't tested those options yet .. I suspect that having ANY SD card in any configuration in the camera will slow things down again. The speed at which the buffer can flush out the files to the card will be dependent on the "slowest link"... which in the case of an 850 is the SD card. I'm using the Sony G series XQD.

 

[Slapstick Noir]

That's an interesting test.
Does the crop mode increase the buffer?
Do you have to necessarily remove the second card or setting it to "Flow" would get you the same increase?
Cheers!

Good questions. Haven't tested those options yet .. I suspect that having ANY SD card in any configuration in the camera will slow things down again. The speed at which the buffer can flush out the files to the card will be dependent on the "slowest link"... which in the case of an 850 is the SD card. I'm using the Sony G series XQD.

 

[Tony Bonanno]

Thank you!

I didn't know that the slower card would be the bottleneck regardless if data is being written on it or it's idle.

Hi Slapstick Noir,

Oops, I provided you with some erroneous information. My apologies.. I shouldn't have said "any configuration". The slower card is the bottleneck ONLY when it is also being written to. For example, if you have your camera set so that one card is a backup (files are written to both cards at the same time) or set so that one card is a RAW and the other is a JPEG, then, yes, the slower card will be the bottleneck. IF you have one card set as "overflow" so that the secondary card is idle until the primary card is filled, then there is no bottleneck (since both cards are not being written to at the same time). For maximum performance in the "overflow" mode, you want to set up your fastest card (XQD) as the "primary" card in the camera menu.

Cheers,

Tony

--
Tony Bonanno Photography, Santa Fe, New Mexico
ASMP Professional Member, APA National Member
Instructor, Santa Fe Photographic Workshops
www.bonannophoto.com

 

[Tom Frerichs]

Thank you!

I didn't know that the slower card would be the bottleneck regardless if data is being written on it or it's idle.

Hi Slapstick Noir,

Oops, I provided you with some erroneous information. My apologies.. I shouldn't have said "any configuration". The slower card is the bottleneck ONLY when it is also being written to. For example, if you have your camera set so that one card is a backup (files are written to both cards at the same time) or set so that one card is a RAW and the other is a JPEG, then, yes, the slower card will be the bottleneck. IF you have one card set as "overflow" so that the secondary card is idle until the primary card is filled, then there is no bottleneck (since both cards are not being written to at the same time). For maximum performance in the "overflow" mode, you want to set up your fastest card (XQD) as the "primary" card in the camera menu.

Cheers,

Tony

 

[Slapstick Noir]

Hi Slapstick Noir,

Oops, I provided you with some erroneous information. My apologies.. I shouldn't have said "any configuration". The slower card is the bottleneck ONLY when it is also being written to. For example, if you have your camera set so that one card is a backup (files are written to both cards at the same time) or set so that one card is a RAW and the other is a JPEG, then, yes, the slower card will be the bottleneck. IF you have one card set as "overflow" so that the secondary card is idle until the primary card is filled, then there is no bottleneck (since both cards are not being written to at the same time). For maximum performance in the "overflow" mode, you want to set up your fastest card (XQD) as the "primary" card in the camera menu.

Cheers,

Tony

 

[HSHN]

Thank you!

I didn't know that the slower card would be the bottleneck regardless if data is being written on it or it's idle.

Hi Slapstick Noir,

Oops, I provided you with some erroneous information. My apologies.. I shouldn't have said "any configuration". The slower card is the bottleneck ONLY when it is also being written to. For example, if you have your camera set so that one card is a backup (files are written to both cards at the same time) or set so that one card is a RAW and the other is a JPEG, then, yes, the slower card will be the bottleneck. IF you have one card set as "overflow" so that the secondary card is idle until the primary card is filled, then there is no bottleneck (since both cards are not being written to at the same time). For maximum performance in the "overflow" mode, you want to set up your fastest card (XQD) as the "primary" card in the camera menu.

Cheers,

Tony

 

[Tony Bonanno]

For what it's worth, Thom Hogan suggests that image review performance can still be limited by a "slower" card.

http://www.dslrbodies.com/technique/technique-articles/quick-and-dirty/qad-card-practices.html

Obviously this is his opinion based upon his observations. I wouldn't take it as gospel, but he does tend to be fairly careful about technical matters.

HI Tom,

Yes, I pretty much agree with Thom. That is a good article. My observations are pretty much the same. Thanks for sharing the link.

Tony

--
Tony Bonanno Photography, Santa Fe, New Mexico
ASMP Professional Member, APA National Member
Instructor, Santa Fe Photographic Workshops
www.bonannophoto.com

 

[jeanphilippe Goube]

The more noise on the image, the less efficient the compression is (it is based on encoding areas with same or close values). Thus keeping low iso allows a more efficient compression and more images in the buffer.

I'll use Lexar XQD, I think they are a bit more faster than sony, I'll add some figures next week whe, my D850 arrives

 

[Surselva]

The manual says, that the capacity of the buffer is only 29 without the battery grip. How can you get 51 images in the buffer? Or do you have the grip already?

 

[Tony Bonanno]

The manual says, that the capacity of the buffer is only 29 without the battery grip. How can you get 51 images in the buffer? Or do you have the grip already?

Hi Surselva,

29 is the capacity for 14 bit uncompressed RAW. 51 is the capacity for 14 bit Lossless Compressed RAW. Lossless Compressed 14 bit RAW is the camera default and for most users, the preferred setting. You do not lose any quality. However, the files are smaller and therefore the buffer capacity is increased.

All the best,

Tony

 

[Vince P]

The manual says, that the capacity of the buffer is only 29 without the battery grip. How can you get 51 images in the buffer? Or do you have the grip already?

I have the grip and I find the buffer is larger without. I can get in excess of the 51 frames without the grip 14 Bit lossless Raw to Lexar 2933 128GB XQD (The buffer display shows R20)

Just adding the grip the displayed buffer falls to R15 and I get 30-35 frames..

It does a great job of emptying the buffer to the cards looks almost the same rate as the D5 which is amazing given the files are much bigger. So frequent shorter bursts are less likely to fill the buffer than the D810 for example. Even with a full buffer you are getting a few FPS.

Instagram @vinnypimages

 

[nextlife]

The manual says, that the capacity of the buffer is only 29 without the battery grip. How can you get 51 images in the buffer? Or do you have the grip already?

I have the grip and I find the buffer is larger without. I can get in excess of the 51 frames without the grip 14 Bit lossless Raw to Lexar 2933 128GB XQD (The buffer display shows R20)

Just adding the grip the displayed buffer falls to R15 and I get 30-35 frames..

It does a great job of emptying the buffer to the cards looks almost the same rate as the D5 which is amazing given the files are much bigger. So frequent shorter bursts are less likely to fill the buffer than the D810 for example. Even with a full buffer you are getting a few FPS.

Instagram @vinnypimages

Is it typical for a buffer to decrease with an added grip? I have not experienced that before.

 

[Vince P]

The manual says, that the capacity of the buffer is only 29 without the battery grip. How can you get 51 images in the buffer? Or do you have the grip already?

I have the grip and I find the buffer is larger without. I can get in excess of the 51 frames without the grip 14 Bit lossless Raw to Lexar 2933 128GB XQD (The buffer display shows R20)

Just adding the grip the displayed buffer falls to R15 and I get 30-35 frames..

It does a great job of emptying the buffer to the cards looks almost the same rate as the D5 which is amazing given the files are much bigger. So frequent shorter bursts are less likely to fill the buffer than the D810 for example. Even with a full buffer you are getting a few FPS.

Instagram @vinnypimages

Is it typical for a buffer to decrease with an added grip? I have not experienced that before.

Neither have i; i guess it is because of the faster frame rate having to deal with more data in the same time. Or I guess the same data in less time

--
Instagram @vinnypimages

 

[ericbowles]

Thanks for the post. Are you using the Grip to achieve the faster results?

The actual buffer is around 1.25 GB for the D500 and D850. The write speed to XQD cards is around 220 MB/s for the D500 and I expect around the same for the D850. Files are around 54 MB for 14 bit compressed RAW, which means you get about 4 fps written to the card, and the balance to the buffer. The buffer holds approximately 23 files if they are 54 MB 14 bit lossless compressed. At 7 fps that means it takes about 7.2 seconds to fill the buffer - the same time as to shoot 51 images. With the grip and a frame rate of 9 fps, you are writing 5 fps to the buffer so it fills in about 4.5 seconds or after 39-40 frames. All these calculations are estimates as the actual Buffer size and write speed are not published by Nikon.

Camera Memory Speed reports a faster write speed in the D500 (290 MB/s), but it does not match the camera write speed in filling the buffer (at that speed the buffer would be unused at 10 fps) so there is something else going on. I wonder if it is a power boost from the grip or an error in calculations. It is more likely to me that the D5 which has the same 290 MB/s reported speed is designed to boost power with the EN-EL18a/b. But I have seen no evidence of a power boost in the D850 - just a shorter maximum burst length. This might be why you are picking up some maximum burst length.

Nikon has made this more confusing because they use the term "buffer" to refer to "maximum burst length" as well as the "actual buffer size" where images are temporarily stored before writing to the card. The buffer is much smaller than the maximum burst which would include images in the buffer but also those written to the card during the burst.

--
Eric Bowles
BowlesImages.com

 

[Slapstick Noir]

That's strange indeed. May be the reason is the battery itself. Do you use the en-el15a or en-el18b in the grip?

 

[RoyC01]

The manual says, that the capacity of the buffer is only 29 without the battery grip. How can you get 51 images in the buffer? Or do you have the grip already?

I have the grip and I find the buffer is larger without. I can get in excess of the 51 frames without the grip 14 Bit lossless Raw to Lexar 2933 128GB XQD (The buffer display shows R20)

Just adding the grip the displayed buffer falls to R15 and I get 30-35 frames..

It does a great job of emptying the buffer to the cards looks almost the same rate as the D5 which is amazing given the files are much bigger. So frequent shorter bursts are less likely to fill the buffer than the D810 for example. Even with a full buffer you are getting a few FPS.

Instagram @vinnypimages

Is it typical for a buffer to decrease with an added grip? I have not experienced that before.

Adding a grip does not decrease the buffer, however adding a grip that increases the FPS does. Frames are going in faster at the higher FPS BUT they are written to the card at the same speed. Grip does not increase write speed to card.

Recent grips have not increased FX frames speed but if you go back to the D300/S and the D700 where FX frames went up, buffer filled up in fewer seconds just like the D850.

 

[Vince P]

Thanks for the post. Are you using the Grip to achieve the faster results?

The actual buffer is around 1.25 GB for the D500 and D850. The write speed to XQD cards is around 220 MB/s for the D500 and I expect around the same for the D850. Files are around 54 MB for 14 bit compressed RAW, which means you get about 4 fps written to the card, and the balance to the buffer. The buffer holds approximately 23 files if they are 54 MB 14 bit lossless compressed. At 7 fps that means it takes about 7.2 seconds to fill the buffer - the same time as to shoot 51 images. With the grip and a frame rate of 9 fps, you are writing 5 fps to the buffer so it fills in about 4.5 seconds or after 39-40 frames. All these calculations are estimates as the actual Buffer size and write speed are not published by Nikon.

Camera Memory Speed reports a faster write speed in the D500 (290 MB/s), but it does not match the camera write speed in filling the buffer (at that speed the buffer would be unused at 10 fps) so there is something else going on. I wonder if it is a power boost from the grip or an error in calculations. It is more likely to me that the D5 which has the same 290 MB/s reported speed is designed to boost power with the EN-EL18a/b. But I have seen no evidence of a power boost in the D850 - just a shorter maximum burst length. This might be why you are picking up some maximum burst length.

In terms of frames per sec written from buffer to card the D850 clears about as fast as the D5 (With or without grip) so I am concluding it's a faster data rate.