URSA 12K SSD vs CFAST 2.0 Dual Record

Thought users here might be interested in some of these results.

Firstly we've been testing the Samsung 983 U.2 SSD @ 1.92TB of storage since getting a replacement USBC cable from BMD.

After a short test. Here are our results at the following specs.

12K - 5:1 - 40fps - 17:9 DCI = equates to roughly 963MB/s. We've gotten 16 minutes of record time and stopped due to power. We'll be doing more testing tomorrow to really push things.

Now that things are working as they should, we're completely blown away by this performance!

If anyone with Cfast 2.0 cards could do some tests for us in dual record that would be great! Gives us as well as others an idea of record capability. As the Cfast 2.0 recommended card list from BMD only indicates a single cards performance. Not sure why they did that... If we're wrong please feel free to point us in the right direction.

Don't forget to mention the details of the cards you use.

Please let's keep this on topic. Keyboard warriors stay out of it!

Thought users here might be interested in some of these results.

Firstly we've been testing the Samsung 983 U.2 SSD @ 1.92TB of storage since getting a replacement USBC cable from BMD.

It was a cable in the end ?

JB

Thought users here might be interested in some of these results.

Firstly we've been testing the Samsung 983 U.2 SSD @ 1.92TB of storage since getting a replacement USBC cable from BMD.

It was a cable in the end ?

JB

Yeah. Initially they had sent out a replacement. That was also a failure apparently. I was told that the supplied cables were faulty and that a new supplier was being sought. Touch wood this new cable lasts. It's worked flawlessly since!

Might you have time JB to try a few tests?

As the Cfast 2.0 recommended card list from BMD only indicates a single cards performance. Not sure why they did that...

The impression that I have is you need to be able to fill a single card at a sustained 400 mb/s to be on the approved list for the 12K. I'm presuming for dual card 800 mb/s would be the benchmark. Maybe that can be clarified / corrected if not accurate?

I don't have any tests to contribute here, generally I am well within single card range as I'm using more 8K 17:9 DCI 24fps 5:1 - 18:1.

I'm really trying hard to see what the drawbacks are to sticking with 18:1 compression because you can fit 2 hours per 512gb vs 1 hour in 8K DCI 24fps 5:1.

Travis, was the 40 fps test a torture test? Realistically would you be shooting in the 24-30 fps range. Even the best CFast 2 cards aren’t likely going to sustain single card rates the equate to over 480 MB/s. Pretty sure the Wise CFast2 cards can’t do that. Perhaps the Angelbird will.

If you’re recording 24 fps, the rate will be significantly less and should be feasible on Wise and Angelbird and others.

As the Cfast 2.0 recommended card list from BMD only indicates a single cards performance. Not sure why they did that...

The impression that I have is you need to be able to fill a single card at a sustained 400 mb/s to be on the approved list for the 12K. I'm presuming for dual card 800 mb/s would be the benchmark. Maybe that can be clarified / corrected if not accurate?

I don't have any tests to contribute here, generally I am well within single card range as I'm using more 8K 17:9 DCI 24fps 5:1 - 18:1.

I'm really trying hard to see what the drawbacks are to sticking with 18:1 compression because you can fit 2 hours per 512gb vs 1 hour in 8K DCI 24fps 5:1.

That’s kinda what we were thinking too Ryan.
That perhaps 800MB/s is what one can expect.

Maybe BMD haven’t specified maximum recording capabilities because they found varying results amongst the different CFast options? Just a guess.

Do you see much difference when it comes to quality between those two ratios? Or is it simply not noticeable?

Travis, was the 40 fps test a torture test? Realistically would you be shooting in the 24-30 fps range. Even the best CFast 2 cards aren’t likely going to sustain single card rates the equate to over 480 MB/s. Pretty sure the Wise CFast2 cards can’t do that. Perhaps the Angelbird will.

If you’re recording 24 fps, the rate will be significantly less and should be feasible on Wise and Angelbird and others.

Yeah we definitely shoot 24fps 99% of the time.

This endeavour was mostly just to see what real world data rates everyone was getting.

We continued with our test today. We had heavy rain, a stand in model and about 6 Apurture MC lights going off in colour variations in quick succession. Not sure that our real world requirements ever get that hectic but we figured it would at least provide our sensor with a lot of moving parts so to speak

Alas. We recorded with the settings above and filled our SSD completely. No over heating. Not dropped frames. Just pure capture! We we’re pumped to say the least.

Tomorrow we will do some more high frame rate testing in 8K. Pushing those 1000+ MB/s data rates.

Fingers crossed

We’re guessing this card maxes out at around 975 ish. But that’s just a hunch.

Will post results tomorrow gents.

Was hoping to see some results from CFAST users on this post.

Anyways we've done our testing and found rather interesting results.
in 8K 2.4:1 8:1 160p with a max data rate of 846MB/s we get a consistent 10 seconds before frame drops.

However, @ 4k 2.4:1 5:1 160p with the same max data rate of 846MB/s we got continuous card fills.

Seems with the 8K resolution at 160p the quality is simply too high for the SSD to process. However at @ 16:9 aspect ratio and a frame rate of 120p we get full drive recordings without issue.

In short it seems the MD (Max Data Rate) of 846MB/s in 8K 8:1 is simply too demanding for the SSD. So if 8K 160p is required we simply have to move over to 12:1 which has a MD of 564MB/s.

Hope this helps anyone looking at getting the Samsung drive we've mentioned. For the time being, we're going to stay with the drive. It's been performing flawlessly.

Was hoping to see some results from CFAST users on this post.

Anyways we've done our testing and found rather interesting results.
in 8K 2.4:1 8:1 160p with a max data rate of 846MB/s we get a consistent 10 seconds before frame drops.

However, @ 4k 2.4:1 5:1 160p with the same max data rate of 846MB/s we got continuous card fills.

Looks like there is something wrong here:
8K 2.4:1 8:1 160p should have 809 MB/s and
4k 2.4:1 5:1 160p should have 328 MB/s
(Byte conversion based on 1024)

They can't have similar data rates because 4k has 1/4 of data and 5:1 is only 1.6 times more than 8:1

8K 16:9 8:1 120p has 720 MB/s
8K DCI 8:1 120p has 768 MB/s

8K 2.4:1 12:1 160p has 480 MB/s

Alternatively you could use
8K 2.4:1 8:1 144p which has 728 MB/s and gives a nice 6 times slowmo

Was hoping to see some results from CFAST users on this post.

Anyways we've done our testing and found rather interesting results.
in 8K 2.4:1 8:1 160p with a max data rate of 846MB/s we get a consistent 10 seconds before frame drops.

However, @ 4k 2.4:1 5:1 160p with the same max data rate of 846MB/s we got continuous card fills.

Looks like there is something wrong here:
8K 2.4:1 8:1 160p should have 809 MB/s and
4k 2.4:1 5:1 160p should have 328 MB/s
(Byte conversion based on 1024)

They can't have similar data rates because 4k has 1/4 of data and 5:1 is only 1.6 times more than 8:1

8K 16:9 8:1 120p has 720 MB/s
8K DCI 8:1 120p has 768 MB/s

8K 2.4:1 12:1 160p has 480 MB/s

Alternatively you could use
8K 2.4:1 8:1 144p which has 728 MB/s and gives a nice 6 times slowmo

The data rates are supplied by BMD...
Tried uploading a photo but 2mb is too big of a file size apparently.

Take a look at the user manual at the max data rates for the URSA. These figures come from that exact page.

Was hoping to see some results from CFAST users on this post.

Anyways we've done our testing and found rather interesting results.
in 8K 2.4:1 8:1 160p with a max data rate of 846MB/s we get a consistent 10 seconds before frame drops.

However, @ 4k 2.4:1 5:1 160p with the same max data rate of 846MB/s we got continuous card fills.

Looks like there is something wrong here:
8K 2.4:1 8:1 160p should have 809 MB/s and
4k 2.4:1 5:1 160p should have 328 MB/s
(Byte conversion based on 1024)

They can't have similar data rates because 4k has 1/4 of data and 5:1 is only 1.6 times more than 8:1

8K 16:9 8:1 120p has 720 MB/s
8K DCI 8:1 120p has 768 MB/s

8K 2.4:1 12:1 160p has 480 MB/s

Alternatively you could use
8K 2.4:1 8:1 144p which has 728 MB/s and gives a nice 6 times slowmo

Not sure how you got those numbers, but they aren’t the Max Data rates for either resolution. Just thought you should know.

Was hoping to see some results from CFAST users on this post.

Anyways we've done our testing and found rather interesting results.
in 8K 2.4:1 8:1 160p with a max data rate of 846MB/s we get a consistent 10 seconds before frame drops.

However, @ 4k 2.4:1 5:1 160p with the same max data rate of 846MB/s we got continuous card fills.

Looks like there is something wrong here:
8K 2.4:1 8:1 160p should have 809 MB/s and
4k 2.4:1 5:1 160p should have 328 MB/s
(Byte conversion based on 1024)

They can't have similar data rates because 4k has 1/4 of data and 5:1 is only 1.6 times more than 8:1

8K 16:9 8:1 120p has 720 MB/s
8K DCI 8:1 120p has 768 MB/s

8K 2.4:1 12:1 160p has 480 MB/s

Alternatively you could use
8K 2.4:1 8:1 144p which has 728 MB/s and gives a nice 6 times slowmo

The data rates are supplied by BMD...
Tried uploading a photo but 2mb is too big of a file size apparently.

Take a look at the user manual at the max data rates for the URSA. These figures come from that exact page.

It seems that someone made a mistake in the manual because those data rates are not plausible.

As I have already explained - just think it through yourself:
Constant Bitrate means the compression ratio stays constant compared to uncompressed.
Uncompressed RAW data rate is calculated from the square amount of pixels and its bit depth.
Sensor uses 16 extra pixels on each side for calibration.

8K 2.4:1 has (8192+2*16) x (3408+2*16) = 28,290,560 px
4K 2.4:1 has (4096+2*16) x (1704+2*16) = 7,166,208 px

That is 3.948 times the amount of pixels.

And 8:1 compresses 1.6 times stronger than 5:1

So 4k 2.4:1 5:1 must have 0.2533*1.6 = 0.405 or 40,5% of 8K 2.4:1 8:1

So the data in the BMD manual table can't be correct.

Not sure how you got those numbers, but they aren’t the Max Data rates for either resolution. Just thought you should know.

They are from the calc sheet I made with information from Captain Hook and provided here:
viewtopic.php?f=2&t=118151

Well then that's really bad on BMD's part on supplying misleading information.

I imagine all the numbers are incorrect then.

We're able to record in the highest 12K compression at 40 frames with no snags whatsoever. According to the misleading information, it's 963MB/s as a MD. We were able to fill the card. Whereas according to the data supplied this Samsung SSD should technically only allow up to 752MB/s.

Thanks for taking the time to do the math and for pointing out your reason.

Travis I'm literally in the middle of a series and I don't have th ability to set up a camera with CFAST cards to test.

I know that CFAST has always been reliable to me. I haven't lost shots.

The only time I have it stop recording on the 12K is sometimes when I am using the Q1 setting on a single card and using a very wide lens at deep stops. The Q series vary compression by HOW MUCH IS IN FOCUS. In theory Q0 will max at a higher data rate than constant, however this rarely actually happens, and if it does is usually momentary.

There's no setting I can't shoot once I go to dual cards in Q series in my usual on set work, without me doing the specific work of studiously testing media. I did this early on but it's meaningless to compare because my numbers are with pre-production units and different iterations of BRAW.

I know also that you CANT just take the manufactures specs on ANY cards or any media as being the (whole) truth.

There's a lot of stuff going on underneath the hood that affects the performance of media. The sustained media numbers are simple MARKETING numbers for the media, and are typically generated but the manufacturer in different, usually more idea conditions. No media maker "test" their numbers with a 12K camera recording continuously. It's not JUST the data rate. It's the way the data is moved around. It's only an INDICATOR of performance in general conditions.

(EDIT even this post proves this viewtopic.php?f=2&t=137309&start=50#p748529)

I also know that my friends at BMD have told me the fastest card that tested WITH THE 12K when I was doing early testing was the Angelbird 256GB. Other Angelbird are also very good too.

A CFAST card is not that expensive in the context of a 12K. The best thing for you to do is to TEST the cards in your use scenarios. I've mentioned this to many threads. Go buy a couple of cards and use them for back up if you need. Just do it. Don't listen to me, or what others say. Go test it for yourself.

I've even used my Sony SD cards at Q5 successfully when I needed to.

I want to also say that many of us here did say that the problem with USB media is the EXACT problem you had. Here we are with all of that only strongly confirmed.

I understand you WANT to use the USB based media, I know the cost is an issue for you, and I understand you're looking at the specs and trying to make sense of it all.

CFAST media makes all of this go away. Take away the angst...

JB

Well then that's really bad on BMD's part on supplying misleading information.

I imagine all the numbers are incorrect then.

We're able to record in the highest 12K compression at 40 frames with no snags whatsoever. According to the misleading information, it's 963MB/s as a MD. We were able to fill the card. Whereas according to the data supplied this Samsung SSD should technically only allow up to 752MB/s.

Thanks for taking the time to do the math and for pointing out your reason.

Bitrate in Metadata

Codec_Bitrate.JPG (10.97 KiB) Viewed 1844 times

Do you have the clips saved that you made on your computer? Resolve will have the "Codec Bitrate" in the metadata and the corresponding compression saved that you can convert to mb/s to get a better idea of the actual data rate. I think it's averaged from a single frame(?).

For example, I have a 30 sec Pocket 6K Pro 16:9 3:1 24 fps clip that came out to 257 mb/s and a 12:1 clip at 64 mb/s both recorded to CFAST. That matches up pretty well with Robert's spreadsheet.

Blackmagic is using 1000 Bytes = 1 kB in their storage rate tables, which gives slightly higher numbers than the 1024 Bytes = 1 kB default in my spreadsheet.