Bigger Flash or external drive for editing?

We're getting a nMP and I'm not sure if the best would be to spend money on bigger FLASH internally or an external drive.

The internal FLASH drive of the nMP seems SUPERfast but external storage can be much bigger but then also more expensive.

Is there anything suggesting we should not use the internal FLASH drive for editing on nMP ?

Any disadvantages I should know?

Hey Kenneth,
Congratulations on the new nMP.
Despite the very fast flash memory, it is still good practice to separate the boot drive from the media storage drive.
Remember the boot/system drive has a lot of overhead I/O handling all apps and system operations while you are editing.
Some folks are concerned about the number of writes Nand flash can handle reliably, IMHO this is overblown with newer technology, but video is a stress factor.
If you can swing it, get a multi-bay ThunderBolt or even a USB3 enclosure that supports UASP for storing media files. How much throughput you'll need is dependent on what you edit, codecs, #of cams, effects, titles, etc…
External hdds do store lots of data and are cheaper TB/$$$. Easy to eat up space with every project, let alone multiple projects at the same time.
Having said this, I do sometimes use an external ssd(s) for single projects, erasing between jobs just for the speed. My internal is forever full and constant fight to slim down.
Just my 2cents worth, I'm looking for a 4 or 5 bay JBOD unit myself.
Hope this helps, Greg

Thank you

Regarding throughput, I'm looking at it this way:

When we're editing, the throughput is limited. We record in UHD (BlackMagic Production 4K recording Prores 422 LT) and for us, a typical production is fairly simple: one main cam on storyline with voice over and background music + additional stuff like extra clips, station IDs, lower thirds etc. We don't do many realtime effects, but color correction and sound plugins are added to all clips.

As we haven't tried the nMP yet but currently work on a hackintosh, I'm not sure how fast we would need the disk to be, to be snappy, but the hackintosh is constantly lagging - click play and wait 0.6 sec. for it to start, move the playhead and wait 0.4 sec for it to react, open a compound clip and wait 1.1 sec. for it to be ready.

Split seconds....... it's maybe no much, but we having this lagging feeling all the time and it slows our workflow down.

In this case, MAYBE the Pegasus Thunderbolt 2 R4 would be enough. It's still a lot of money, but as I understand it, it can deliver read/write that surpasses our needs - on the other hand, not faster than one, simple SSD at 1/4 of the price.

When exporting, my belief is, that the more you can save, the faster the encoding will be. In this case it seems to me, that the R8 will hammer through encoding much, much faster, because of the higher throughput, right? ..... on the other hand, if an export takes 18, 32 or 51 minutes almost never matters to us. It's just a background process.

Either way, even the R4, is quite expensive for our budget and I'm more and more in favor of not paying a lot of money for old, spinning technology.

In this case, for the same money, I can either go for:

1 ) MacPro 6 Core, d500, 256mb flash, Pegasus R4
2 ) MacPro 8 Core, d700, 1tb flash

To me, the 2. seems to be the fastest in total by far, even considering the "stressful" life of the flash and the time we'd spend on moving finished projects away from the flash drive?

Any insights on this is greatly appreciated

I use the 8 core with 1TB system drive and D700s. All my Libraries are on a Pegasus T2 RAID 5. I got the empty 4 bay, populated with enterprise level 4TB drives. I edit Blackmagic 4K native ProRes HQ files almost daily and have no performance issues at all.

I'd go with your number 2 solution. The difference between the D500 and D700 GPUs is very noticeable. I also packed mine with 32GB RAM, and am glad I did, as I can be running all four of the pro apps at the same time, plus other apps like Modo, Pixelmator, Studio Artist, Audacity, etc, at the same time.

The 4 bay P2 is in fact better "real world" performance than a single SSD, we have both. Remember that SSDs are still compressing and decompressing data during reads and writes. They throughput on paper is impressive, but it often doesn't pan out in real world use.

Exporting will depend on what codec you're going from/to. The nMP is actually slower than other Macs with H.264 encoding, as the CPUs don't have the same H.264 encoder built in to them as other CPUs.

Here's a performance test I helped out with.
www.macprovideo.com/hub/final-cut/final-...o-x-performance-test

Thank you

We're encoding almost 100% of the time to h.264 for youtube and podcast

It seems to me that the top end iMac 27" would actually be enough to handle our projects, cheaper AND faster for our h.264 encoding - but then again, the iMac is not THAT much faster than the rMBP 15" we're also using..... I can't figure out why Apple makes it so hard to figure out where to get the best performance / price ratio. Steve would NEVER have allowed that......

Have you checked out BareFeats?
They have some well done comparisons on Macs with different setups.
www.barefeats.com/tube05.html
Quad core comparison…
www.barefeats.com/quads.html
Another site I like for good testing is…
macperformanceguide.com
There are real differences between i7 and the xeons in the MPs. MP are still Ivy Bridge based and the i7 is Haswell based. Apple really had to cobble the Ivy Bridge set up for things like USB3 and HDMI. Not natively supported.
Also the GPUs really make huge differences in performance. If I was buying a nMP, I would go for the d700's as they are good value$$/performance. If you decide to get a new iMac, go for the 27" as it allows you to upgrade some of the components including ram, the smaller one does not.

I really have a problem with the link Ben provided as it does not equalize the testing, and does not isolate/ identify the various processes being run. Very hard to make a fair comparison between systems when some are using 4bay ThunderBolt in a RAID vs 2 5400 drives using 1 USB3 shared bus. IMHO I would not use this testing for selecting a system unless I was interested in an exact system match and wanted to know speeds. Why didn't they compare the systems using the same storage device? Then you could begin to see the differences in performance.

Thought you were already getting nMP?
I still would use a separate external enclosure for media. (wooden nickel advice)

Hope this helps, Greg

Greg,
Our purpose was to show tests of real world systems, not do a high level benchmark across the board. We were not replicating Barefeats, but doing something from a very different perspective. It does in fact show the difference between several real world systems. I don't think I'd ever dish out the money for a P2 16TB RAID to use with a MBP. Thus, the reason for testing drives we actually used with machines we actually use for real world editing jobs in various studios. And yes, we do specify what exact processes we were running.

As for you statement that our tests are only helpful if you are purchasing the exact same setup, well, any benchmark can only relate that exact same hardware setup. So that's just a given in any comparison article.

Well, I actually found the test in BenB's link more useful than the Barefeats ones.

I've looked at barefeats countless times and every time I think "Okay, but what about REAL LIFE FCPX work?"

I don't care about game performance or benchmarks that doesn't reflect how the combination of Apple hardware and Apple software rocks (especially due to the dual GPU of nMP) and to my best knowledge, Barefeats fails to answer that question

Unfortunately, I think Barefeats - most of the times - just end up with a explanation or interpretation of the data, that doesn't lead me anywhere in terms of "how will THIS test compare to my everyday workflow in FCPX?".

Just a quick test here:

Hackintosh : 4 core i7, 32gb ram, 4 disk raid 0 internal 3.5", nvidia 780ti gpu
MacBook Pro 15" retina early 2013 (650m gpu, 16gb ram, i7), 3 disk raid 0 thunderbolt disk.

The macBook exports to h.264 MORE than TWICE as fast as the much more powerful Hackintosh.

Living Smart TV wrote: ...We're encoding almost 100% of the time to h.264 for youtube and podcast

It seems to me that the top end iMac 27" would actually be enough to handle our projects, cheaper AND faster for our h.264 encoding - but then again, the iMac is not THAT much faster than the rMBP 15" we're also using...

I edit video professionally using FCP X on a top-end iMac 27 with 8TB Thunderbolt Pegasus R4. It works very well. That said, there's a good argument for the 6-core and higher nMP.

iMac exporting to *single pass* H.264 is much faster than a nMP because of Intel's Quick Sync. However multi-pass or high quality H.264 export cannot use Quick Sync, so the nMP is faster for that. Detailed discussion:
www.fcp.co/forum/imac/18907-mac-pro-vs-i...eo-compression#45368

FCP X is very render-oriented -- you are constantly staring at red render bars. Frequently these are CPU-bound, not disk-bound or GPU-bound. You can see this by examining core utilization using iStat Menus -- all available cores are usually harnessed. It's very good at background rendering but it's constantly doing it.

There is also lots of GPU activity, although the relative % between GPU and CPU fluctuates. Apple will probably further harness the GPU in later FCP X updates; that's the general trend.

I would personally not get a base 4-core nMP. However there's a good argument for the 6 and 8-core nMP with D700 GPUs. It partially depends on your workflow and work style. For contemplative, creative work you aren't going that fast from a UI standpoint. OTOH for rapid production work you are pushing the UI fast and hard, frequently waiting on the machine. A nMP is better for that, despite the single-pass H.264 issue.

Eventually H.265 or other codecs will replace H.264. It is unclear if Quick Sync will work for that. It's possible that advantage could disappear.

Re bigger flash drive vs external drive, more I/O is good but only to a point. Once I/O is not the bottleneck, further I/O performance doesn't help. Codecs strongly affect this. 1080p H.264 is fairly compressed so you're commonly more CPU-limited than I/O limited. OTOH editing raw video is more I/O-limited.

This assumes single-stream editing. For editing multi-cam streams, the I/O rate increases rapidly.

Okay, thank you very much for the advice.

I ended up buying the entry level 4 core, 12gb ram, d300 and we already had a 3 disk Caldigit thunderbolt raid.

I'm amazed to see how "snappy" it is in general use but I'm also a frustrated to see how slow the h264 encoding is.

It's remarkably slower in encoding the same files as our macBook Pro 15" Retina flies through and even though I heard and read about this prior to shopping, I'm still surprised to see how much slower it actually is.

It's like buying a sportscar and then driving it with the handbrakes on

The base nMP is a nice machine and I'm sure you'll be happy with it. It is very quiet. Normally the iMac is quiet but under very sustained heavy rendering, occasionally the fan is audible. Despite the slower single-pass H.264 export, the nMP is still faster than a top-end iMac on various things. I personally tested several FCP X tasks, and the base nMP was faster for stabilization motion analysis, optical reflow and "render all" (not export).

That said if you are still within the nMP return period you have the option of exchanging it for a maxed-out iMac. There are pros and cons either way.

Hey Ben, if I may ask how big of the difference you notice between D500 and D700?

A bit, when it comes to GPU intensive processes. More and smoother playback when render lines are present. I rarely render anything. But the D500 blows away anything else found in any other Mac, also.

I'd suggest to anyone, put a test Library on a USB 3.0 drive, go to your local Apple store, and plug it in to a nMP. These machines are all supposed to have FCPX on them.