Welcome to our community

Be a part of something great, join today!

  • Hey all, just changed over the backend after 15 years I figured time to give it a bit of an update, its probably gonna be a bit weird for most of you and i am sure there is a few bugs to work out but it should kinda work the same as before... hopefully :)

New Mac Pro

Honestly it's not like a 6 Core machine is going to be dramatically slower or things will take longer to open. I have a second mac pro coming to split render loads, and having played with a 6, 8, and a 12 core, each has their pros and cons. Faster cores sometimes beats MORE slower cores.

Either way, you're not making a permanent investment. I'd go six and worst case you CAN upgrade later.
 
The problem is the miss conception that more cores add's speed in terms people think of it. That is not the case. More Cores adds a far higher ceiling to the workflow with codecs like Red that take advantage of it. That means an editor can have far more complex sequences that play realtime without requiring rendering to playback cleanly. The speed increase is in workflow. AE the speed increase can be in ram preview time depending on the complexity of the comp and if you have the ram preview included with the multiprocessing options. The exports can significantly be improved for complex compositions. An example a comp I have that is 1 min long with 67 layers take 1.5 hours to export on Dual Xeon 12 Cores and 2.5 to 3 Hours on a 6 Core at 4.5GHz.. That is a significant time improvement for complex comps. The problem I see is people suggesting what works best for their workflow as the best for everyone. That is simply not the case. Each workflow is different and some require far greater processing than others. My single 12 Core 2697 V2 plays back 4 layers of R3D 4K at 1/2 resolution with a 770GTX 4GB card in Premiere. That is significantly better than any 6 core can handle right now even at 4.5Ghz.
 
To that point the main take away is that no one solution solves everyone's problem. And deciding what kind of work you do, and the apps to do them with, will help advise you of the right purchase.

Also keep in mind RIGHT NOW the software still doesn't take full advantage of these machines, so any machine you guy will be a little behind the curve regarding how fast it feels vs how fast it's TRULY capable of becoming. Early adopter tax is in full effect.
 
Thanks again Isaac.
The work I finalized/ mastered today had up to 45 layers and something like a 12000x9000 plate with multiple multilayered 3D elements on it + lighting - ugh the render pain :)

with some stuff it's just an overnight render and then I don't really care if it's 3 or 6 hours :) now if that's being rendered in 1.5hrs - that's nice ;) but anyway, I guess 6 cores it is.
 
With the current generation of Xeons we have to make our CPU choice in a way that balances cores against speed. Makes it very difficult for people trying to purchase one system to do it all. That's where building or buying a dual CPU PC has advantages. It's not so much that we can really utilize all the cores offered by having two or more CPUs in there. Some things can use all that -- 3D rendering, for example. But generally speaking, even in the world of R3D and Resolve, Premiere, etc.. 12 cores or 24 threads is pushing beyond the threshold of diminishing returns. But with multiple processors, we can have both the speed and the cores via using 6 or 8 core CPUs to get a total of 12 or 16 cores in there at the faster GHz. It also allows for more RAM and more PCIe lanes.

Anyway, getting back on topic... If you have a way to buy the new Mac Pro with a company discount, you will do better to just get the RAM through Apple. Their RAM prices are not out of line at all -- actually cheaper than what Dell and HP are charging for the same stuff. I'm a small business, but the local Apple store gives me a 10% discount as a business customer. With that, my 64GB RAM came out cheaper than buying 64GB from OWC or Crucial. If you're outside the USA, then who knows, I can't really say how that works out.

The hard part about buying one of these still comes back to which CPU should one choose. I think the 6 and 8 core systems are going to be the sweet spot for most people. 12-core is great if you don't mind the slower CPU and you spend a lot of time in software that can use all those cores. If you spend most of your time in Resolve or rendering out of AE or Maya or other things like that which scale well to all those cores, then the 12-core may be beneficial. For Premiere, some Resolve and Speedgrade, more time doing 3D modeling, animation, compositing, etc.. you're going to do better with a 6 or 8 core system. The quad-core is only about 10~15% faster than the top-spec'd iMac and a lot more expensive. The reason to buy it over the iMac is if you can use the dual GPUs and extra connectivity.
 
Thanks again Isaac.
The work I finalized/ mastered today had up to 45 layers and something like a 12000x9000 plate with multiple multilayered 3D elements on it + lighting - ugh the render pain :)

with some stuff it's just an overnight render and then I don't really care if it's 3 or 6 hours :) now if that's being rendered in 1.5hrs - that's nice ;) but anyway, I guess 6 cores it is.

For projects like that I'm seeing more use by something like amazons cloud for being useful to bring a lot of CPUs online for tackling big renders I've had to deal with.

Either way way I don't think you'll be unhappy with a 6 core.

Also Jeff said smart stuff, I agree with him.
 
Last edited:
Where's the LIKE button when you need it? Thank you so much for the deep insight to both of you!
 
Someone should do a test with all CPUs/dual xeons and publish the render times. That would be really helpful.
 
Hi Jeff and all,

Jeff, you said in your recent post that Some Resolve is good for 6 or 8 Core but also listed it as a reason to go 12 Core. Would you care to elaborate - what are the tasks that take full advantage of all the cores and where don't they?

For me I am looking at the 8 Core but now considering dropping to the 6 Core for higher clock speeds - I will be mostly transcoding from various camera formats to edit friendly DNXHD or ProRes and applying basic LUT's or CDL corrections, possibly also creating H264 dailies for review. All of this I do using Resolve or Scratch Lab. Occasionally I do jobs where they need secondaries and some skin softening/beauty work on fashion films and they take my transcodes as an almost finished grade through the edit and show an edit to the clients on set. The other significant factor is the slowdown caused during checksum verification when backing up the files. I have yet to establish the key hardware factors affecting this - obviously Data IO on the media and destinations is key but there seems to be other overheads which cause the system to slow down. I imagine System RAM is significant and when using ShotPutPro it runs multiple copy threads at once so maybe the number of cores becomes a factor during verification? As I am often copying in the background whilst working and transcoding in Resolve or Scratch this can often be an important consideration. Down the road I am considering setting up two machines - one as a fileserver with fast RAID and client drive IO to manage data and the other to work on, transcode and review the footage off the shared high speed storage - but this is outside of my budget for now.

I look forward to hearing some informed opinions on the above.

Cheers,

Q.
 
Jeff, did you end up hooking up your friend's ASUS 32" successfully? I've just got my 12-core nMP and ASUS PQ321 yesterday and so far I can only get 4K 30hz running under SST. When I enable MST on the ASUS I get nothing.

Sorry missed this the other day. Almost no luck here with the ASUS. Apparently the Sharp is supposed to be the only 4K MST monitor that works at 60Hz on the nMP. However, we rounded one of those up as well and it does not work properly at 60Hz, either.

The Sharp and Asus 32" monitors and the Dell 24" monitor all perform flawlessly with 60Hz MST when running Windows on the new Mac Pro. It's an OSX driver issue. The drivers attempt to support MST and I've had some degree of success with the monitors at one point or another, but the troubles range from intermittent flickers and drop-outs on the secondary stream to just having a completely blank display.

IMO, the best UHD/4K monitor solution at the moment would be the 24" Dell as it's the one that's appropriately sized for running HiDPI mode. …And it works quite well other than being locked at 30Hz. My friend continues to use his ASUS 32" display on his HP workstation and I returned the Dell 24" some time ago. I'm sure Apple will get this sorted, but in typical Apple fashion they may not fix it until they have their own 4K display solution to offer.


Someone should do a test with all CPUs/dual xeons and publish the render times. That would be really helpful.

render times for what specifically and in which app? All makes a difference. Premiere does a decent job of using a lot of the CPUs. Resolve does a much better job. REDCINE-X hardly scratches the surface of multi-processing unless you have good management tools to batch-out REDLINE nodes on individual threads.

A little A.D.D. torture test would be great...

A.D.D. appears to be single-threaded with no Rocket or GPU acceleration at the moment. The quad-core Mac Pro is going to win that test. ;)

Hi Jeff and all,

Jeff, you said in your recent post that Some Resolve is good for 6 or 8 Core but also listed it as a reason to go 12 Core. Would you care to elaborate - what are the tasks that take full advantage of all the cores and where don't they?

It's an issue of balance. Resolve does rather well at using 12 cores and for decoding R3D footage this can be very helpful. However, not all Resolve operations benefit from the multiple cores -- many of your node functions are GPU-bound for acceleration and/ or single-threaded, so more CPU cores isn't necessarily going to help there. The slower clock speeds of the 12 core CPU can slow things down. So in addition to Resolve, we have to weigh the other things that will be done with this system. If you're spending 90% of your time in Premiere, SpeedGrade or Resolve, or any combination of those apps, then I would steer you toward the 12-core. If you only do Resolve and Premiere a combined total of 50% of your time and you spend the rest of your time in AE, Nuke, Photoshop, surfing the web, etc.. Then the 8-core is probably more to your liking. When it comes to transcoding various camera formats, the various encoder and decoder software will determine how many cores are optimal. Generally speaking, the demanding ones are CinemaDNG, R3D, H.26x/HEVC.

For me I am looking at the 8 Core but now considering dropping to the 6 Core for higher clock speeds - I will be mostly transcoding from various camera formats to edit friendly DNXHD or ProRes and applying basic LUT's or CDL corrections, possibly also creating H264 dailies for review. All of this I do using Resolve or Scratch Lab. Occasionally I do jobs where they need secondaries and some skin softening/beauty work on fashion films and they take my transcodes as an almost finished grade through the edit and show an edit to the clients on set.

Just based on this, I would say the 8-core, possibly consider the 12-core if this system will spend most of its time doing this sort of work day in and day out in Resolve or Scratch.

The other significant factor is the slowdown caused during checksum verification when backing up the files. I have yet to establish the key hardware factors affecting this - obviously Data IO on the media and destinations is key but there seems to be other overheads which cause the system to slow down. I imagine System RAM is significant and when using ShotPutPro it runs multiple copy threads at once so maybe the number of cores becomes a factor during verification?

Unfortunately, the number of cores here doesn't come into play. These are not multithreaded tasks, so faster CPU speeds are going to help a bit. But what it really comes down to is I/O speeds and potential bottlenecks. Checksums can be time-consuming and doing multiple copies and checksums with something like ShotPut just becomes demanding on your storage and it's slow. period.

As I am often copying in the background whilst working and transcoding in Resolve or Scratch this can often be an important consideration. Down the road I am considering setting up two machines - one as a fileserver with fast RAID and client drive IO to manage data and the other to work on, transcode and review the footage off the shared high speed storage - but this is outside of my budget for now.

Separate machines for ingest and checksum operations is definitely the way to go. This can be an argument for the 8 or 12 core machine as well. Even though some of these tasks are not multithreaded, they can still run on their own threads and you will have more cores/ threads still available. The down-side is that it takes up your I/O bandwidth, regardless and adding more cores won't solve that. You may need to run a separate RAID/ storage for these tasks and another for your work in Resolve/ Scratch.
 
Parallelism outside the box

Parallelism outside the box

Big fan of having at least 2 machines, usually 3. I've been saying this for 7 years on RU - many people spend big bucks to build out monster systems but end up with poor bang for the buck in typical use cases (there are exceptions of course). There are a number of things that may be required of you on/near set and being able to flip files into required formats quickly counts. That said, 90% of the time the process I'm waiting on at the end of the day is making and verifying copies.

All the cores, RAM and GPU muscle in the world won't make busses faster. In my experience the expensive baller system is no faster than a mid-level one if its simultaneously engaged in background operations. Perhaps once some of the new architecture is sorted that penalty will lessen, we'll see.

I know this is the MacPro thread, but on a macro level; if I have a $10K budget for computer hardware I'd rather have 2 low end $1,000 "copy machines" and one $8,000 bomber for Scratch/Resolve/whatever than a $10K hero machine with a couple more cores and a third Titan. YMMV.

Cheers - #19
 
Yep, I always advocate the same in terms of multiple systems. I'm not sure how so many people get by with just one box and trying to build a one-system-to-rule-them-all machine just never works out that well. Even in the circumstances where it is possible to build such a beast, it's inevitable that it will be tied up doing something while you need or want to do something else. Building THE ONE computer right now isn't possible anyway with the current selection of CPUs and how the number of cores has to be balanced with GHz speed.

I know this is the MacPro thread, but on a macro level; if I have a $10K budget for computer hardware I'd rather have 2 low end $1,000 "copy machines" and one $8,000 bomber for Scratch/Resolve/whatever than a $10K hero machine with a couple more cores and a third Titan. YMMV.

Exactly.
 
I'm not sure how people use multiple machines (and get the best out of them) without investing in decent shared storage with 10gig....(I use a Editshare server)..

What are other people using? Either in studio or out and about?

Very interested in thunderbolt connectivity for some interesting LW super fast networking? Jeff?. What do you think?

Thanks

M
 
I'm actually WAITING on my second 8-Core machine to ship. It will become my travel/DIT station machine, but at home i'll be using it to split the load on larger renders.

Also I just received my EIZO 277s. These are the ones that JUST released. Worth the wait. I know everyone here is all 4K all the time for their displays, but for color accuracy these things are awesome. Half the power consumption and longer lifespan. Also I couldn't resist setting all 4 up (2x 276s and 2x 277s) since I was sending the 276s back and setting the 277s up.

1959395_10152614057079186_1485641677_n.jpg
 
I'm actually WAITING on my second 8-Core machine to ship. It will become my travel/DIT station machine, but at home i'll be using it to split the load on larger renders.

Also I just received my EIZO 277s. These are the ones that JUST released. Worth the wait. I know everyone here is all 4K all the time for their displays, but for color accuracy these things are awesome. Half the power consumption and longer lifespan. Also I couldn't resist setting all 4 up (2x 276s and 2x 277s) since I was sending the 276s back and setting the 277s up.

1959395_10152614057079186_1485641677_n.jpg

Nice,
I can't wait to get my 12 core
 
Just got the ASUS 4K monitor (PQ321) working Jeff. After having no luck finding cables specifically rated for MST - I ordered a bunch of DP->miniDP cables in the hope one would work. In the end a cable advertising 'audio support' was the winner and I'm a proud owner of a 60hz 31.5" 4K display! Only just started using it but so far I'm finding the size of text & icons really usable and the screen real estate is just bonkers. In the attached pic that's a pixel-for-pixel 1080p video window in FCPx so you get a feel for how much space is left over. Reminded me of how it felt staring into my first 30" Cinema Display (which I foolishly dropped $6k when they were first released).

I already had SwitchResX installed from past troubleshooting and in Display Prefs it shows up as a SwitchResX monitor with no refresh rate listed. Then when you enter the SwitchResX app it shows the full res and 60hz - STOKED!! Also running a secondary 27" Apple Thunderbolt display for now and it's all playing super smooth.

Ultrastudio 4K is running well too - once they sort the stupid audio issues out in the 10.0 drivers. RedRocket in the Magma box also running great and will only get better once their TB2 upgrade is available. Next up is to test an Apple Fibre Channel card in there to hook up to my XSan. Preordering the Promise SANLink 2 as well which will supply 2 x 8gbps fibre connections over TB2.
 

Attachments

  • 4k01.jpg
    4k01.jpg
    95.8 KB · Views: 0
  • 4k02.jpg
    4k02.jpg
    98.3 KB · Views: 0
Big fan of having at least 2 machines, usually 3. I've been saying this for 7 years on RU - many people spend big bucks to build out monster systems but end up with poor bang for the buck in typical use cases (there are exceptions of course). There are a number of things that may be required of you on/near set and being able to flip files into required formats quickly counts. That said, 90% of the time the process I'm waiting on at the end of the day is making and verifying copies.

All the cores, RAM and GPU muscle in the world won't make busses faster. In my experience the expensive baller system is no faster than a mid-level one if its simultaneously engaged in background operations. Perhaps once some of the new architecture is sorted that penalty will lessen, we'll see.

I know this is the MacPro thread, but on a macro level; if I have a $10K budget for computer hardware I'd rather have 2 low end $1,000 "copy machines" and one $8,000 bomber for Scratch/Resolve/whatever than a $10K hero machine with a couple more cores and a third Titan. YMMV.

Cheers - #19

On the $1K copy machine front - is there a huge difference with more RAM and processing power as well or is it only-mostly a matter of I/O ?
 
Wow, thanks for the very full reply Jeff, you are very generous with your time, thank you. And everbody else who pitched in on the subject of multiple systems.

I am thinking of a rack mac mini server running as a SAN as my data manangement system - then I'll hook up the MacPro to it to work on the footage. Any thoughts on this? Would if be too tied up running the SAN to manage offloading and checksum verification at the same time? Would it be better to simply run a powerful raid on the mac mini through a PCI card in the Sonnet rack and make it a shared volume and maybe network via Thunderbolt2 between the machines so that possibly unnecessary SAN functions don't become an overhead. The world of SAN is foreign to me and I'm not sure if it's entirely necessary for my needs or what kind of overheads it creates - though there are occasional circumstances where I work with editors or VFX teams on set and it would be nice to have the ability to share directly at good working speeds without copying to shuttle drives. Peoples thoughts and experiences on this would be much appreciated.

I'll re-post this in a new topic if the feeling is that this is veering too off topic but I think it could be a relevant factor to others considering building a rig or facility around the new MacPro.

Cheers,

Q.
 
Dan do you have link for that cable that works with your Asus?

Thanks!
 
Back
Top