[0x4243]

Hi!

Today I did some more tests (already posted the result on "the other" forum ) with a comparison of a 300MB/s UHS II card and a 170MB/s UHS I card. Both tested in both slots. There is not much difference at all. The UHS II card in the UHS II slot has the worst performance though.

The cards are a SanDisk Extreme Pro 170MB/s V30 UHS I card with 128GB.

and a SanDisk Extreme Pro 300MB/s V90 UHS II card with 64GB.


In the PC I get a write speed of 86MB/s and a read speed of 96MB/s in an internal UHS I USB 3 reader for the 170MB/s card. That's pretty much what the interface can handle.
The 300MB/s card does 220-250MB/s writing on the PC (UHS II USB C card reader from SanDisk).



I measured both cards in both slots and get a data rate of around 60MB/s in each of the slots. (a bit less in Slot 2).


The speed of the 170MB/s UHS I card, compared to my 300MB/s card which makes 56MB/s in Slot 1 and 63MB/s in Slot 2 it is almost the same speed at one-sixth of the cost.


Here are my masurements from top to bottom:
170MB/s in Slot 1:
37 Pics for first burst, total of 65 pics during first 15s, 13,6s for clearing buffer, 61MB/s


170MB/s in Slot 2:
36 Pics for first burst, total of 62 pics durinfg first 15s, 12,8s for clearing buffer, 60MB/s


300MB/s UHS II in Slot 1:
36 Pics for first burst, total of 60 pics during first 15s, 13,8s for clearing buffer, 56MB/s


300MB/s UHS II in Slot 2:
36 Pics for first burst, total of 63 pics during first 15s, 11,9s for clearing buffer, 63MB/s


Note that the time for clearing buffer differs slightly due to my reaction time because here I watch the LED with my eyes and then pressed the shutter again.


The camera settings were all the same:
Manual focus, manual mode, 1/8000s, open aperture, ISO 100, lens cap closed (gets all black pictures), DNG only output, SR off, ISO noise reduction off


So I'd say, instead of buying expensive UHS II cards, maybe save your money for the next (Pentax??) camera because I see really no benefit of using UHS II cards.


Btw: The price is
180 € for the 300MB/s V90, 128 GB vs
28 € for the 170MB/s 128GB

Attached Images

 

[Kunzite]

Disappointing. I hope it's just a software bug/deficiency.

[stevebrot]

That's pretty much what the interface can handle.

By "interface", I assume you mean the reader. The card can handle almost twice that.* How does the UHS I card do in the Sandisk reader?


Steve

* Of course there is all the overhead of dealing doing filesystem stuff. A straight serial dump of the card would be much faster.

[0x4243]

By "interface", I assume you mean the reader. The card can handle almost twice that.* How does the UHS I card do in the Sandisk reader?


Steve

* Of course there is all the overhead of dealing doing filesystem stuff. A straight serial dump of the card would be much faster.

With interface I meant the UHS I card bus interface. As far as I know, this bus can handle up to 104 MB/s * (312MB/s for UHS II Half Duplex).

Using my SanDisk Extrem PRO SD UHS II USB C Reader (the one with the short USB cable) I get
89 MB/s reading and 84 MB/s writing with the 170MB/s card and
308 MB/s reading and 262 MB/s writing with the 300MB/s card


* I read that the 170MB/s UHS I card really can handle the 170MB/s in at least one SanDisk card reader. I don't know how this works and if they use a proprietary protocol. Would like to hear from someone who knows It seems that even the 170MB/s card is of course less than 100MB/s in "standard" UHS I readers... But let me know if I'm wrong here!


Christian

[Rico]

Here are the official SD Card Organization developers standard overview to the bus speed transfer rates for all the bus interfaces.

UHS II has headroom for 156mbs Full duplex i.e. both upstream and downstream data transfers at the same time or Half duplex of 312mbs in one direction at a time. So at the very least the K-3III's UHS-II slot 1 there should be 156mbs of room with 312mbs optimal.

[c.a.m]

Today I did some more tests

I have been following your posts on this subject. Thanks!

300MB/s UHS II in Slot 1:36 Pics for first burst, total of 60 pics during first 15s, 13,8s for clearing buffer, 56MB/s

From the strip plot for the third case, I see the following:

First burst: 36 frames in 3 seconds, followed by 29 frames in a 14-second period (2,1 fps). Total frames: 65

Buffer clears at 30,5 second mark.

Total data recorded: 65 frames RAW (DNG?) x 33 MB/frame average = 2145 MB. Average data speed to SD card over the recording period: 2145 / 30,5 = 70 MB/s

I am not familiar with the architecture of the image-file recording system. It's possible that during the slower image rate after the first burst, the buffer operates in read/write mode, i.e., writing some of the buffered data to the card. It might be useful to examine a sample run in which the shutter is stopped just at the end of the first high burst, then see how long the buffer takes to clear. I don't recall whether you had tested this scenario already.


FWIW, I looked up some performance times at various reviews done by Imaging Resource.

Sony A7III (ca. 2018): 24 MP, 14-bit uncompressed RAW. UHS-II. Continuous Hi+ (RAW only), 33 frames at 9,9 fps, then slows to 1,21 fps. Time to clear 33 frames after a high-speed burst: 26 s.

Fujifilm X-T3 (ca. 2018): 26 MP, 14-bit RAW lossless compressed, Mechanical shutter, UHS-II, 36 frames at 11 fps, then 2,7 fps. Time to clear 36 frames: 12,3 s

Buffer clearing times were measured with a 64GB Lexar Pro 2000x UHS-II SDXC card.

- Craig

[stevebrot]

Here are the official SD Card Organization developers standard overview to the bus speed transfer rates for all the bus interfaces.

UHS II has headroom for 156mbs Full duplex i.e. both upstream and downstream data transfers at the same time or Half duplex of 312mbs in one direction at a time. So at the very least the K-3III's UHS-II slot 1 there should be 156mbs of room with 312mbs optimal.

Thanks for pointing the discussion to "the source" for all things SD. The discussion for that page is helpful as is the full content of the SD Association Web site's section on "SD Standard Overview".

SD Standard Overview | SD Association

Nestled in, here and there, are discussions of read/write speed for serial transfer (bus rate) vs. read/write of file data. The latter may be hobbled by factors such as number of files already on the card, sectors written, and degree of segmentation. I am probably talking talking through my hat, but it seems that the best performance for write speeds might be attained with a freshly formatted (rewrite flavor) card and a single file with size approaching card capacity. That lacking, a single file at the maximum size the camera is capable of might be a good metric, say something like 25 minutes of high quality video (~2GB). Edit: It just occurred to me that video is streamed directly to file, so not a good example. How about a full-resolution Pixel Shift to DNG?


Steve

[0x4243]

I have been following your posts on this subject. Thanks!


From the strip plot for the third case, I see the following:

First burst: 36 frames in 3 seconds, followed by 29 frames in a 14-second period (2,1 fps). Total frames: 65

Buffer clears at 30,5 second mark.

Total data recorded: 65 frames RAW (DNG?) x 33 MB/frame average = 2145 MB. Average data speed to SD card over the recording period: 2145 / 30,5 = 70 MB/s

Yes, all test were done with DNG. Note that I didn't shoot exactly 15 seconds but a few seconds more. I only counted the images during the first burst plus those that were taken up to the 15s mark. The buffer clear time then was measured after releasing the shutter. But anyway, doesnt matter if 70MB/s or 60MB/s to me. I just expected a lot more with UHS II.

FWIW, I looked up some performance times at various reviews done by Imaging Resource.

Sony A7III (ca. 2018): 24 MP, 14-bit uncompressed RAW. UHS-II. Continuous Hi+ (RAW only), 33 frames at 9,9 fps, then slows to 1,21 fps. Time to clear 33 frames after a high-speed burst: 26 s.

Fujifilm X-T3 (ca. 2018): 26 MP, 14-bit RAW lossless compressed, Mechanical shutter, UHS-II, 36 frames at 11 fps, then 2,7 fps. Time to clear 36 frames: 12,3 s

- Craig

Thats in fact very interesting. This means that those cameras don't write faster too. Just looking at UHS II speed, I'd expect more fps after the buffer is full. Maybe this isn't possible at all.
I'll take a look at tests of some other "fast" cameras. Maybe we're not as bad as I thought...

[stevebrot]

I'll take a look at tests of some other "fast" cameras. Maybe we're not as bad as I thought...

Yes, they all "under perform". The research I have done seems to indicate that stills cameras don't benefit nearly as much from the faster cards as dedicated streaming of high quality video to file.


Steve

[c.a.m]

But anyway, doesnt matter if 70MB/s or 60MB/s to me. I just expected a lot more with UHS II.

I agree. Sorry, I didn't mean to suggest your analysis was incorrect. 60 or 70 is far from what we would expect with UHS-II.

The specifications for the K-3 Mark III simply indicate that the camera is "UHS-I/UHS-II compatible." As we know, there is no specification for ultimate buffer-clearing times. As you have mentioned, a specification for UHS-II could lead users to believe that it has a faster overall speed for writing data to the card in slot 1. In my case, I decided not to buy a UHS-II card until I really need it. I don't usually shoot long, rapid bursts.

I'll take a look at tests of some other "fast" cameras. Maybe we're not as bad as I thought...

Other cameras:

Sony A9 (2017), 24 MP, UHS-II, RAW 14-bit uncompressed, Cont Hi (RAW) 5 fps, 241 frames, 38 s to clear buffer.

A7R IV, 61 MP, UHS-II x 2, RAW compressed, 74 frames, 26 s to clear buffer.

Canon 90D (2019): 32 MP; Continuous High RAW, 10.0 fps, 27 frames then 3.9 fps; 5.6 seconds to clear after a high burst.


- Craig

[FantasticMrFox]

Why Pentax puts a UHS-II slot into its 2021 APS-C flagship, just for it to have a maximum write speed of around 90 MB/s, is beyond me ... The D500 at least gets to around 160 MB/s ...

[Adam]

Why Pentax puts a UHS-II slot into its 2021 APS-C flagship, just for it to have a maximum write speed of around 90 MB/s, is beyond me ... The D500 at least gets to around 160 MB/s ...

I performed a test with this card:
Lexar 128GB Professional 1667x UHS-II SDXC Memory LSD128CBNA1667

With all corrections and AF disabled, shooting until the buffer filled up, the measured write speed averaged out to 91 Mb/s in raw in 84 Mb/s in raw+. Perhaps not cutting edge, but that is a 2x speedup over the K-3 II (and other recent APS-C models) and a 3x speedup over the K-1. I'd consider that a job well done by Pentax!

[Kunzite]

Is cameramemoryspeed.com dead? The last results appears to be from 2018...
Anyway, perhaps it's worth testing using their methodology - so the results would be (hopefully) comparable?

[FantasticMrFox]

I performed a test with this card:
Lexar 128GB Professional 1667x UHS-II SDXC Memory LSD128CBNA1667

With all corrections and AF disabled, shooting until the buffer filled up, the measured write speed averaged out to 91 Mb/s in raw in 84 Mb/s in raw+. Perhaps not cutting edge, but that is a 2x speedup over the K-3 II (and other recent APS-C models) and a 3x speedup over the K-1. I'd consider that a job well done by Pentax!

First of all the situations in which one would shoot at a high continuous speed and care about the buffer clearing quickly are probably not situations in which one would disable the AF

Second of all, the five years old Nikon D500 manages 163 MB/s, the X-T4 manages 179 MB/s, the 90D manages 173 MB/s. So the K3 III certainly doesn't offer twice the write speed of 'other recent APS-C models', in fact it manages just a bit over half compared to those with a UHS II slot

Which warrants repetition of my question: why, Pentax, why?!

[Adam]

First of all the situations in which one would shoot at a high continuous speed and care about the buffer clearing quickly are probably not situations in which one would disable the AF

Fair point, but I'm just measuring the write speed here, lol.

Which warrants repetition of my question: why, Pentax, why?!

Probably fair to say that this is not their priority at the moment. Personally, I'd take the new massive viewfinder over a larger buffer. Hopefully they aren't losing too many customers over this, but I'm sure there are folks who have jumped ship due to these limitations.