another "is this enough?" thread lol

[thizisweird]

I'm well aware that I can get away with the hardware specs I list out below.... I'm currently editing video with FCPX with zero issues, Ableton works just fine, and I've had zero DAW problems overall while running this lean machine. Two questions

MP 5,1 (2009 w/EFI update)
E5520 Quad core 2.26GHz x2 CPU
12 GB memory
500GB HDD (single, will update this asap)
AMD 4870 (512MB)
Nvidia GT 120 (512MB) [PCIe slot 2]

My real question here: If I were to snag the Xeon X5675 upgrade CPUs, 32+GB memory, and eventually full SSD storage (M.2 for primary, and 4x 850pro RAID is the aim for now).... As a video editor who wants to branch into 3D work (including 3D printing, blah blah), someone who managed to throw FCP into an endless loop of freeze/crash from rendering) for a few days before realising the problem, whoops lol.... With the aforementioned upgrades, should I expect everything to be about as smooth as it will get within FCP? Whatever that is, I'll make do with it. It's currently the best computer I've had for the task to date. I'm sure I could get caught up in GPU specs and whatnot, and crashing Ableton is an acquired skill (mastered on my end), but I really have no idea how well the W700 would actually hold up with real-time performance; let alone do I really know if two of them would benefit me over the single in any way? Like, only in compressor? I don't think anything else would be as noticeable of a hardware choice.


Secondary question: Would there be much of a difference in stability between running 4 SSDs and 4 HDDs in a RAID 0 environment? That's the two big "upgrade" paths I've seen for this machine, and I'm fine with both, as the quad M.2 PCIe card I want is a few hundo on its own. Regardless, I'll be capping the transfer rates, but the SSDs just 'seem' (for lack of better words) like they'd be the more stable, reliable, long-term choice... not to mention the speed boost if I put the drives in another machine. So basically- any room for improvement, or am I thinking too much?

Hopefully the answers aren't too "it varies" because I forgot something dumb. Sounds like me! Fingers crossed, and thanks for tolerating my inane questions the learning curve for Mac OS in general has been nice and easy, bringing my Linux background and ancient Mac/PowerPC memories to my mind every so often. It's nice to be back! haha

 

[orph]

i dont know what software you want to use but might be worth checking there support pages or asking them in a email.
also why not just try software you want to use and see if it works now.

ps if you want to know how well 'X' software runs you do relay need to name 'X' software

edit
ops my bad missed you use FCX, must have been tired when i replayed

 

[thizisweird]

i dont know what software you want to use but might be worth checking there support pages or asking them in a email....

ps if you want to know how well 'X' software runs you do relay need to name 'X' software

Well I did mention Ableton and FCPX, that I've been known to crash Ableton just from using it lol (I love my plugins), and I did make my current setup crash with Final Cut trying to render something it couldn't handle. So while it runs, and I'm now forced to hand edit all my transitions (fun for learning, sucks for production time), I'm also extremely limited with what I can do. I'm flexible with what I need to do to stay within my computer's limits, but I don't know how to quantify performance of something like a W7000 (compared to my AMD 4870) for the exact type of potential stress I'll put on it in the next 6 months as I start getting parts in, etc.. Would it even matter that I'd prefer to stop using proxy media just to conserve space for raw data and samples? Lastly, on software being used: if I have any others I'll want to use heavily, I'll be sure to search and ask

I'm used to changing my work habits to match my machine's performance (never owned high performance for personal use), but it's difficult right now to say exactly what I'm going to be doing when it will be for clients I have yet to make contracts with.......... The more I can offer to do with my hardware, the better, though. I mean, the W7000 sounds like it would be a great card, and it's cheap..... but since I don't know exactly what is sent to the GPU when I'm using FCPX, it's hard to know a quantifiable amount of performance I'm looking for..... not to mention architectures efficiency, OpenCL, and all that stuff I ignore until it's shopping time. I know FCPX is way more stable with a second GPU in it, though I have zero reason for why, other than background rendering for something I'm doing. RAM may be low but that didn't change when I added the second GPU, so the stability change appears related.... That's all I know.

also why not just try software you want to use and see if it works now.

The computer doesn't even have 16GB of memory at the moment, so even OSX isn't the most stable at times. There's also a serious lack of storage at the moment (only 500GB, which will be remedied), so I don't have much luxury installing whatever I want when I'm working with 4K media. Not to mention, as soon as I start using more programs, storage consumption goes up regardless lol, and I'm sure we all know what that's like with no money lol. It's like the delete game on your old Nokia.

Bonus nachos: what upgrades would improve Compressor rendering time, or would a second, fairly dated 1GB-ish AMD card plugged in render-farm style be a better option? If it's cost efficient to have a couple slave towers for heavy Compressor rendering (like mass-rendering projects for a deadline), that might be worth looking into.

 

[kschendel]

Upgrade the CPUs to a pair of X5680 or X5690, put in a couple of SSDs in the drive bays and RAID 0 (stripe) them, and I think you can expect a rough doubling of perceived performance. Maybe not quite 2x but close. You have the 2009 dual so you'll need de-lidded CPU's which makes it a bit more of a hassle. I'm not sure what effect memory will have, you might want to put in 3 x 8 Gb sticks for 24 Gb total. I'm not entirely sure why OSX would be unhappy with 12 Gb but whatever.

I don't do video / photo editing myself so I'm not sure what impact the GPU would have on your particular workflow.

If you don't want to fool around with software RAID, just replace your 500 Gb HDD with J Random SATA SSD and you'll see a very nice improvement just from that.

M.2 is a form factor. I assume you mean M.2 PCIE not M.2 SATA because there's little point in the latter. If you are doing PCIE SSD on OSX you'll need an AHCI SSD, *not* NVMe, which will cost more and be harder to find. My experience with a 2009 cMP with both SATA SSD in the drive bays and NVMe SSD in PCIe (I'm running linux) is that you need to really be beating the bejabers out of the I/O to see the difference. HDD -> SSD in a drive bay is huge. SATA II SSD -> NVMe SSD is, subjectively for most things, meh.

 

[thizisweird]

While I appreciate the input, kschendel, you kind of are stating blanket "performance upgrades" without me even knowing how it will change my actual working performance while editing. Sure, better playback and overall stability goes up with better storage (part of the reason I want the quad SSD raid setup mentioned earlier); however, I'm well aware upgrading the CPUs and memory will only do so much... to a point.... and even less of a degree for some users with the GPU upgrade. Since I'm not entirely sure where the main focus will be in the near future, it would be nice to know what I want to look for in each graphics card (or what kind of GPU upgrade for different hardware demands) before I hit eBay and enjoy a price before I even know the workflow performance. I rarely see much info explaining how to arrive at an answer, and mostly people wanting to know "this GPU is what I need to buy." If I build a rendering farm or two, which I might need for other projects, I won't be as inclined to ask the same question again lol. Just sayin, mate.

I think a good example of what I'm looking for would be a workstation GPU version of explaining things like the JayzTwoCents 'How to choose a Power Supply' video. It basically breaks it down into something easy, helps a novice avoid a bad purchase from asking the "wrong" question, provides the core things to look for, and kind of leaves you holding the bag with all decisions. Big props to Jay for that video too, for those that need the reference material in there. But that's off topic. I will need to get fairly comfy here, I'm afraid.

As for "performance increases," I'm more focused on GPU acceleration, and figuring out just exactly how much performance is going to be affected after the diminishing returns kick in, and how one could figure out a more numbers-based approach to guiding my GPU purchases in the right direction. Although, consumer card feedback is welcome, I'd only want workstation cards (as this will be for professional editing). Funny how we can basically choose a gaming GPU with minimal effort but workstation cards are so foreign and barely covered for people not in the industries :/

Also well aware that FCPX doesn't use the graphics card a whole lot, and GPU upgrades are better for stuff like Compressor, but I believe the GPU still handles some stuff (like 3D rendering?).

 

[h9826790]

Also well aware that FCPX doesn't use the graphics card a whole lot, and GPU upgrades are better for stuff like Compressor, but I believe the GPU still handles some stuff (like 3D rendering?).

I think it's the other way around. FCPX use GPU a lot (for compute and rendering), but Compressor can hardly use the the GPU to encode.

AFAIK, Compressor can only use little bit GPU power if the source is from FCPX, otherwise, it won't use GPU at all.

 

[devon807]

I have heard of some people that use the W7000 in their Mac Pro's here is a video of someone who paired it with a GTX 980:

They say it performs the same as a single D300 found in the 2013 Mac Pro. In terms of compatibility, it has prototype drivers in MacOS, but i'm not sure if the video is still relevant. If you can confirm that the W7000 still works with Sierra then I would say go for it , as it's only around ~$150 USD. If the performance is good, then I would try using two of them in a "crossfire" of sorts because your Mac Pro could easily power two of them.

I believe when dealing with Compressor, it would benefit from the X5675's as Compressor is very CPU bound as h9826790 said. I haven't used Compressor for a while, but while I find it's optimizations with Final Cut useful, it's slow when rendering straight from FCP and I only found it useful when working with high bit rate stuff.

 

[thizisweird]

I think it's the other way around. FCPX use GPU a lot (for compute and rendering), but Compressor can hardly use the the GPU to encode.

AFAIK, Compressor can only use little bit GPU power if the source is from FCPX, otherwise, it won't use GPU at all.

I did find a nice little home test that put some of my worries at rest: https://www.studio1productions.com/Articles/Premiere-Benchmark.htm

While that may not be the best guideline, it seems to dish out the proper information, even if for the wrong software. It would also explain some of the things I've seen reported in forums (like using GTX 6xx and 7xx series with no real "need" to upgrade within the last few months). Unless you really need the time saved or you perform much more intensive editing than most of us would think about (mostly commercial applications), it seems like workstation cards are just not a real benefit any more. I guess that's cool? Kinda sucks, but I'll live with it. Also, the majority of what I've read has said "GPU acceleration is pretty much for FX and rendering, so buy a GPU according to those needs," and that a GPU will really only be a huge benefit for those using DaVinci Resolve (one of the next steps in software for me). I think I'll chat with Apple tonight and see if I can nail a system spec with them for FCPX, and go from there.

Concerning Compressor, it definitely can use hardware acceleration for video compression. Can it use it on the MP? Doesn't seem to be the case.... and that really sucks. If you look at Larry Jordan's article on Faster Video Compression, the MP seems to be the worst product for compressing web content simply due to a lack of hardware acceleration. Sure, the faster cores will help out, but that doesn't explain the results in the article. The iMac used in that article could use the GPU to speed things up; the MP cheesegrater just doesn't support that fancy pants tech in Compressor. I guess I can live with that, and I'll probably keep a few towers lying around for a Compressor 'cluster' I think. Might be the next best move in the short term if it actually speeds it up dramatically.

I have heard of some people that use the W7000 in their Mac Pro's here is a video of someone who paired it with a GTX 980

They say it performs the same as a single D300 found in the 2013 Mac Pro. In terms of compatibility, it has prototype drivers in MacOS, but i'm not sure if the video is still relevant. If you can confirm that the W7000 still works with Sierra then I would say go for it , as it's only around ~$150 USD. If the performance is good, then I would try using two of them in a "crossfire" of sorts because your Mac Pro could easily power two of them.

I believe when dealing with Compressor, it would benefit from the X5675's as Compressor is very CPU bound as h9826790 said. I haven't used Compressor for a while, but while I find it's optimizations with Final Cut useful, it's slow when rendering straight from FCP and I only found it useful when working with high bit rate stuff.

I actually watched that video the first day I heard of the W7000. I must say it didn't answer a single question I had. If he had run just the W7000 and compared the two cards somehow, I'd have thought it was worth watching. As it is, though, I don't see much help in that video. I could be overlooking something, but I want individual performance to show me even a little of what I'd get when I slap that card in my tower. Also, I doubt I'll gain anything from xfire of two W7000 cards, but the cost does make it tempting to stash a spare tower to play with this kind of guess work. Unfortunately time does not permit that, even if money were no object.


After doing a few hours of digging around last night, it just seemed that I was in a huge loop going nowhere fast. My performance does seem to be more or less limited by the CPU/memory (from what I've read), and given the AMD Vega support with the new iMac Pro I would expect the Vega 64 to be compatible with Sierra by the time I really need to look at a new GPU to increase my workflow. Not holding my breathe on it being listed as supported, but I doubt we'll see much holding that card back from OSX use. Definitely a strong contender at $500 MSRP. Not to mention, I have enough GPU power for my current uses, so I guess a W7000 upgrade when I feel a need for something faster is a good backup plan for now, and I'll keep an eye out for Vega 64 drivers and spring for a newer consumer Nvidia card worst case.

At least I learned a lot from this

 

[h9826790]

I did find a nice little home test that put some of my worries at rest: https://www.studio1productions.com/Articles/Premiere-Benchmark.htm

While that may not be the best guideline, it seems to dish out the proper information, even if for the wrong software. It would also explain some of the things I've seen reported in forums (like using GTX 6xx and 7xx series with no real "need" to upgrade within the last few months). Unless you really need the time saved or you perform much more intensive editing than most of us would think about (mostly commercial applications), it seems like workstation cards are just not a real benefit any more. I guess that's cool? Kinda sucks, but I'll live with it. Also, the majority of what I've read has said "GPU acceleration is pretty much for FX and rendering, so buy a GPU according to those needs," and that a GPU will really only be a huge benefit for those using DaVinci Resolve (one of the next steps in software for me). I think I'll chat with Apple tonight and see if I can nail a system spec with them for FCPX, and go from there.

Concerning Compressor, it definitely can use hardware acceleration for video compression. Can it use it on the MP? Doesn't seem to be the case.... and that really sucks. If you look at Larry Jordan's article on Faster Video Compression, the MP seems to be the worst product for compressing web content simply due to a lack of hardware acceleration. Sure, the faster cores will help out, but that doesn't explain the results in the article. The iMac used in that article could use the GPU to speed things up; the MP cheesegrater just doesn't support that fancy pants tech in Compressor. I guess I can live with that, and I'll probably keep a few towers lying around for a Compressor 'cluster' I think. Might be the next best move in the short term if it actually speeds it up dramatically.


I actually watched that video the first day I heard of the W7000. I must say it didn't answer a single question I had. If he had run just the W7000 and compared the two cards somehow, I'd have thought it was worth watching. As it is, though, I don't see much help in that video. I could be overlooking something, but I want individual performance to show me even a little of what I'd get when I slap that card in my tower. Also, I doubt I'll gain anything from xfire of two W7000 cards, but the cost does make it tempting to stash a spare tower to play with this kind of guess work. Unfortunately time does not permit that, even if money were no object.


After doing a few hours of digging around last night, it just seemed that I was in a huge loop going nowhere fast. My performance does seem to be more or less limited by the CPU/memory (from what I've read), and given the AMD Vega support with the new iMac Pro I would expect the Vega 64 to be compatible with Sierra by the time I really need to look at a new GPU to increase my workflow. Not holding my breathe on it being listed as supported, but I doubt we'll see much holding that card back from OSX use. Definitely a strong contender at $500 MSRP. Not to mention, I have enough GPU power for my current uses, so I guess a W7000 upgrade when I feel a need for something faster is a good backup plan for now, and I'll keep an eye out for Vega 64 drivers and spring for a newer consumer Nvidia card worst case.

At least I learned a lot from this

Work station card is just an expensive option in MacOS, no real benefit (apart from more single slot card option if you really need it).

I didn't read your link yet. But I believe that you mixed up QuickSync with GPU accalaeration. All modern i7 has QuickSync which speed up h264 encoding a lot. However, you have no control over it. If you need specific encoding options (e.g. 2 pass), then QuickSync won't be available.

So far, only very little process can utilise OpenCL (real GPU acceleration) for video encoding. Apart from that, the whole process is totally CPU limiting.

 

[orph]

i think final export is mostly CPU limited, the in app rendering in FCX is vary GPU accelerated and your GPU's are a tad old (two GPU's dont make final cut more stable but two newer GPU's can make it faster, matched GPU's are preferred i think).

but as your only working on one old HD then that might be also a massive slowdown for you. ram is only a problem if you see high memory pressure in activity monitor, if you do see that then a ram upgrade is worth it.

if you can get say a SSD for you OS (a small one is fine 128GB or 250GB) then something like a 2TB WDblue for all your files may be the biggest speedup for you at the mo.

a RX560 4GB may be worth a look to as the cheapest/easiest option.

 

[devon807]

I wonder what two of those RX 560's will do. I wonder if anyone has tried it yet.

 

[orph]

well it will work, it's whats in the imacs.
it's more of how fast will it be, as the OP has a AMD 4870 which is fairly old with only 512mb ram the RX 560 with 4GB ram will be faster http://gpuboss.com/gpus/Radeon-RX-560-vs-Radeon-HD-4870
as a cheep new option it's valid and half the price of the rx 580 (in the UK)

the rx580 will be faster just bugs me that price seems a tad high here thanks to demand from miners
there's also used ATI cards to depends what you want & there both not a huge cost so if you can swing it a rx580 is faster.

 

[bopajuice]

Id like to know how you are working with a single 500gb hard drive. One project at a time? External drives?

 

[h9826790]

I wonder what two of those RX 560's will do. I wonder if anyone has tried it yet.

Depends on your usage, for graphics stuff, that will be sucks. And MacOS can only utilise one GPU for graphics (lack of Crossfire support).

But for FCPX, I expect it will deliver pretty good performance. Especially if you pick the 4GB version (2GB may be a bit too limiting).

 

[thizisweird]

Thanks for all the response! You guys are great.

Work station card is just an expensive option in MacOS, no real benefit (apart from more single slot card option if you really need it).

I didn't read your link yet. But I believe that you mixed up QuickSync with GPU accalaeration. All modern i7 has QuickSync which speed up h264 encoding a lot. However, you have no control over it. If you need specific encoding options (e.g. 2 pass), then QuickSync won't be available.

So far, only very little process can utilise OpenCL (real GPU acceleration) for video encoding. Apart from that, the whole process is totally CPU limiting.

I had a feeling the workstation cards would be little more than the expensive option, but considering that I will be doing some fairly high profile work for a friend (he owns multiple businesses, and through him I'll likely be able to start chasing my TV broadcast editor dream lol), so I'd like as much stability as possible for all-around reasons I guess. I'd be willing to part with the cash for a business expense down the road. Worth looking into (I think), and having the horsepower at home for such projects would be a great benefit for getting used to the workflow of the equipment I'd use elsewhere. But hey, I love to overanalyse and be extra prepared for crap like this lol. I used to produce with Ableton 6 with a 2Ghz Core2 Duo and 4GB of memory....... the mixdown was the most brutal task ever, and I managed to produce a nice track with it that gained a little love from indie labels...... so I mean, I'm fairly realistic with expectations at the least lol. Forcing slow/old machines to work for me is what I'm used to!

I think you're right about QuickSync. I took a peak at the concept, and that sounds identical to what was portrayed in the link I posted (it was about compressing for web use, like youtube). Very good to know, though! I think that'll help me make some decisions for other hardware in the future, so a double thank you.

i think final export is mostly CPU limited, the in app rendering in FCX is vary GPU accelerated and your GPU's are a tad old (two GPU's dont make final cut more stable but two newer GPU's can make it faster, matched GPU's are preferred i think).

but as your only working on one old HD then that might be also a massive slowdown for you. ram is only a problem if you see high memory pressure in activity monitor, if you do see that then a ram upgrade is worth it.

if you can get say a SSD for you OS (a small one is fine 128GB or 250GB) then something like a 2TB WDblue for all your files may be the biggest speedup for you at the mo.

a RX560 4GB may be worth a look to as the cheapest/easiest option.

well it will work, it's whats in the imacs.
it's more of how fast will it be, as the OP has a AMD 4870 which is fairly old with only 512mb ram the RX 560 with 4GB ram will be faster http://gpuboss.com/gpus/Radeon-RX-560-vs-Radeon-HD-4870
as a cheep new option it's valid and half the price of the rx 580 (in the UK)

the rx580 will be faster just bugs me that price seems a tad high here thanks to demand from miners
there's also used ATI cards to depends what you want & there both not a huge cost so if you can swing it a rx580 is faster.

I've been letting all the info swim around in my head, and I've been trying to figure out just exactly what I can do for the bare minimum to get things running smoother in general lol. I think the 2x GPU "improvement" I've seen with these ancient cards is more placebo effect than not. After a little more editing, and getting my project to the point I have to keep hitting "Render All' to keep things going smoothly with only keyframe/manual transitioning/etc..... in hindsight I'd say nothing has really improved lol. I think I'm just trying to make it crash unintentionally, seeing how far I can push it lol.

I was thinking maybe a 960 Pro via PCIe, 1-2x 512GB 850 Pro drives for commonly needed data (ie: current project files/etc.), and a 1-2TB HDD (WD Black is ~$70, which is reasonable to me) for storage would be the best starting point for both current and future desired uses. I do want to upgrade to 32GB of memory asap, but mainly because that's ~$100 for used server DIMMs on eBay, and I'd pretty much never need to worry about memory again until I upgrade my whole tower.

As for the RX 560, I'm curious if the memory bandwidth will be anything to worry about? Still trying to grasp all the intricacies here. I know the 560 has half the bus width of the W7000, (128 v 256 bit bus, respectively), slightly lower bandwidth, but it also has a much higher clock speed. Almost makes me wonder if that's something similar to the ridiculous memory amounts on lower-end cards, and if so would it even matter for stability with my uses? Might as well ask while it's in my head. Both the 560 and W7000 are pretty much the same price on eBay, and while I'd love to get the W7000 so I can use slot 2 for a quad M.2 card (mostly for fun and bragging rights, but I'd love having overkill since I'd like to venture into 3D work with RealSet and more as I get more contracts), and to avoid potentially overheating a consumer card (which I understand may be a nonissue), I could always stick with 2.5" SSDs for the short term. It's all going to be a business expense for taxes, so I'm not too put off buying more/better upgrades within the next year to scale up.

The 580 does sound like a good option, but I'd be happy with a 560 to start with, and consider the RX 580/Vega 64 option when I start to see some performance issues with a heavier workload? Anyone have thoughts on that?

Id like to know how you are working with a single 500gb hard drive. One project at a time? External drives?

I'm glad someone is interested, because I didn't think this would be an enjoyable way to edit lol. The drive came with the tower (ST500DM002-1BD142, which could use an update). I did a complete wipe of all data, and it's been a very reliable drive for my uses so far. I can switch between my two current projects with minimal waiting time (maybe 30 seconds to fully load, plus rendering times). No external drives just yet, as both my external drives (Seagate) have corrupt filesystems, and I've nowhere to back up my 5TB backup drive so I can get it switched over from NTFS. Both externals are USB 3.0 as well, so when I can get them stable again I'll get a PCIe card for that too.

I have plenty of complaints with my current system, but it's all I have to work with... and this is the only way for me to get anything into my portfolio right now..... and given my history of low power machine abuse, I'm holding up far better than when I ran Windows. FCPX is the smoothest software I've used to date, with Ableton 6 (last version I used) being a very close second. Both have been super flexible with what I've done (I averaged 3-4 VSTs per track, plus a few built in mastering tools, at 30+ tracks in the mixdown w/Ableton on that ancient laptop), and hardware has been my only real bottleneck whenever using either of them.


Side note:
I attempted some light gaming for grins, and I'm quite impressed at how well the 4870 can handle things. Being a tweaker of micro-quadcopters, I decided to give the DRL simulator a try with an xbox 360 controller. If I have the graphics cranked down, I notice a low frame rate over water, trees have a short draw distance, and flying through crowded areas can be interesting; however, I don't really have much of an issue. That's pretty impressive for an ancient card! I didn't expect that much for a card well below the minimum hardware requirements. Even with the old A1082 display, my dad (very inexperienced) managed to do a decent flight without real complaints. I also tried the demo of Sole (indie game on itch.io that's in demo phase), and it performed quite well! Very impressed, and I may be giving these old AMD cards less credit than they deserve. Just a thought

 

[devon807]

Thanks for all the response! You guys are great.

As for the RX 560, I'm curious if the memory bandwidth will be anything to worry about? Still trying to grasp all the intricacies here. I know the 560 has half the bus width of the W7000, (128 v 256 bit bus, respectively), slightly lower bandwidth, but it also has a much higher clock speed. Almost makes me wonder if that's something similar to the ridiculous memory amounts on lower-end cards, and if so would it even matter for stability with my uses? Might as well ask while it's in my head. Both the 560 and W7000 are pretty much the same price on eBay, and while I'd love to get the W7000 so I can use slot 2 for a quad M.2 card (mostly for fun and bragging rights, but I'd love having overkill since I'd like to venture into 3D work with RealSet and more as I get more contracts), and to avoid potentially overheating a consumer card (which I understand may be a nonissue), I could always stick with 2.5" SSDs for the short term. It's all going to be a business expense for taxes, so I'm not too put off buying more/better upgrades within the next year to scale up.

The 580 does sound like a good option, but I'd be happy with a 560 to start with, and consider the RX 580/Vega 64 option when I start to see some performance issues with a heavier workload? Anyone have thoughts on that?

Im looking at the same GPU's too. Im torn between dual RX 560's or dual W7000's I would go for a RX 580/70 but they are impossible to find for a decent price these days. I was going to preorder the RX 580 Pulse from AmazonUK but I need the GPU's for a project pretty soon. The only thing shying me away from the W7000's is that the drivers may be non existent/half baked in Sierra. The drivers for the RX 560 is baked in to MacOS and two RX 560's are still cheaper than the price of one RX 570/580 with current pricing. I was also considering dual R9 380's but Im picky and prefer MSI cards and I cant seem to find those at a good price either. Another thing is the power consumption from the dual 380's is pretty high and would have to either undervolt them or do the Pixlas mod.

 

[h9826790]

Thanks for all the response! You guys are great.

I had a feeling the workstation cards would be little more than the expensive option, but considering that I will be doing some fairly high profile work for a friend (he owns multiple businesses, and through him I'll likely be able to start chasing my TV broadcast editor dream lol), so I'd like as much stability as possible for all-around reasons I guess. I'd be willing to part with the cash for a business expense down the road. Worth looking into (I think), and having the horsepower at home for such projects would be a great benefit for getting used to the workflow of the equipment I'd use elsewhere. But hey, I love to overanalyse and be extra prepared for crap like this lol. I used to produce with Ableton 6 with a 2Ghz Core2 Duo and 4GB of memory....... the mixdown was the most brutal task ever, and I managed to produce a nice track with it that gained a little love from indie labels...... so I mean, I'm fairly realistic with expectations at the least lol. Forcing slow/old machines to work for me is what I'm used to!

I think you're right about QuickSync. I took a peak at the concept, and that sounds identical to what was portrayed in the link I posted (it was about compressing for web use, like youtube). Very good to know, though! I think that'll help me make some decisions for other hardware in the future, so a double thank you.


I've been letting all the info swim around in my head, and I've been trying to figure out just exactly what I can do for the bare minimum to get things running smoother in general lol. I think the 2x GPU "improvement" I've seen with these ancient cards is more placebo effect than not. After a little more editing, and getting my project to the point I have to keep hitting "Render All' to keep things going smoothly with only keyframe/manual transitioning/etc..... in hindsight I'd say nothing has really improved lol. I think I'm just trying to make it crash unintentionally, seeing how far I can push it lol.

I was thinking maybe a 960 Pro via PCIe, 1-2x 512GB 850 Pro drives for commonly needed data (ie: current project files/etc.), and a 1-2TB HDD (WD Black is ~$70, which is reasonable to me) for storage would be the best starting point for both current and future desired uses. I do want to upgrade to 32GB of memory asap, but mainly because that's ~$100 for used server DIMMs on eBay, and I'd pretty much never need to worry about memory again until I upgrade my whole tower.

As for the RX 560, I'm curious if the memory bandwidth will be anything to worry about? Still trying to grasp all the intricacies here. I know the 560 has half the bus width of the W7000, (128 v 256 bit bus, respectively), slightly lower bandwidth, but it also has a much higher clock speed. Almost makes me wonder if that's something similar to the ridiculous memory amounts on lower-end cards, and if so would it even matter for stability with my uses? Might as well ask while it's in my head. Both the 560 and W7000 are pretty much the same price on eBay, and while I'd love to get the W7000 so I can use slot 2 for a quad M.2 card (mostly for fun and bragging rights, but I'd love having overkill since I'd like to venture into 3D work with RealSet and more as I get more contracts), and to avoid potentially overheating a consumer card (which I understand may be a nonissue), I could always stick with 2.5" SSDs for the short term. It's all going to be a business expense for taxes, so I'm not too put off buying more/better upgrades within the next year to scale up.

The 580 does sound like a good option, but I'd be happy with a 560 to start with, and consider the RX 580/Vega 64 option when I start to see some performance issues with a heavier workload? Anyone have thoughts on that?

I'm glad someone is interested, because I didn't think this would be an enjoyable way to edit lol. The drive came with the tower (ST500DM002-1BD142, which could use an update). I did a complete wipe of all data, and it's been a very reliable drive for my uses so far. I can switch between my two current projects with minimal waiting time (maybe 30 seconds to fully load, plus rendering times). No external drives just yet, as both my external drives (Seagate) have corrupt filesystems, and I've nowhere to back up my 5TB backup drive so I can get it switched over from NTFS. Both externals are USB 3.0 as well, so when I can get them stable again I'll get a PCIe card for that too.

I have plenty of complaints with my current system, but it's all I have to work with... and this is the only way for me to get anything into my portfolio right now..... and given my history of low power machine abuse, I'm holding up far better than when I ran Windows. FCPX is the smoothest software I've used to date, with Ableton 6 (last version I used) being a very close second. Both have been super flexible with what I've done (I averaged 3-4 VSTs per track, plus a few built in mastering tools, at 30+ tracks in the mixdown w/Ableton on that ancient laptop), and hardware has been my only real bottleneck whenever using either of them.


Side note:
I attempted some light gaming for grins, and I'm quite impressed at how well the 4870 can handle things. Being a tweaker of micro-quadcopters, I decided to give the DRL simulator a try with an xbox 360 controller. If I have the graphics cranked down, I notice a low frame rate over water, trees have a short draw distance, and flying through crowded areas can be interesting; however, I don't really have much of an issue. That's pretty impressive for an ancient card! I didn't expect that much for a card well below the minimum hardware requirements. Even with the old A1082 display, my dad (very inexperienced) managed to do a decent flight without real complaints. I also tried the demo of Sole (indie game on itch.io that's in demo phase), and it performed quite well! Very impressed, and I may be giving these old AMD cards less credit than they deserve. Just a thought

You better avoid 960 Pro (or any NVMe SSD) for production machine at this moment. They are not just non-bootable (most likely forever) but not natively supported in cMP yet.

I haven't check the RAM price closely, but 32GB (4x8GB) server pull RAM should not cost more than $75, in fact, it was $50 some time ago. $100 may be a bit too much.

No, memory bandwidth should be no big deal. In my own experience, video editing is not that VRAM intensive. Those much higher memory bandwidth usually can only improve at most few % performance overall. Of course, this very depends on your workflow. However, by considering both RX560 and W7000 is pretty low end now. I really don't think the memory bandwidth will make any difference.

RX580 is a very make sense choice, but if go for RX560, I personally may simply go for the R9 280, which can performance better than the RX560 (in term of video editing / compute in MacOS, I didn't check which card perform better in gaming) and flashable. A single R9 280 can replace the HD 4870 + RX560, so, you can have 3 PCIe slots for whatever you want. And the R9 280 is very cheap. You can even go for dual R9 280 without any extra PSU or Pixlas mod. But in this case, I definitely prefer a single RX580 now.

A reason of better stick to gaming card in MacOS is because of the driver. e.g. The top GPU D700 on Mac Pro 6,1 is just the R9 280X (they have identical Device ID). That means if you install a R9 280X into a cMP, MacOS guarantee has native driver support, and have better chance that the software (e.g. FCPX) is optimised for it. (P.S. The R9 280 I recommend has same Device ID as the HD7950 Mac Edition card, therefore, it should also be natively supported as long as cMP still officially supported by Apple). For any workstation card, they may or may not work properly. Apple has no obligation to support them. Nvidia provides its own web driver to fix this issue, but AMD do absolutely nothing on this matter now. If I were you, I won't let my production machine rely on a card that may lost support after any OS update. Also, all workstation card's features are disabled in MacOS due to lack of proper driver support (e.g. ECC VRAM). So, what's the point to pay more but get less? A R9 280X is definitely more powerful in MacOS than a W7000 (unless you are doing something that require >3GB VRAM), go for the guarantee natively supported faster cheaper card make more sense in general.

However, if you looking for a single slot card for whatever reason, then workstation card usually is the best option. AFAIK, none of the single slot RX460 work on cMP (most likely due to custom PCB design). However, W7000 really works fine natively in MacOS. That's pretty special among workstation cards. So, it is indeed a good choice if you need a single slot card.

 

[thizisweird]

Here's something I found today- Using Dual vs Single GPU for Video Editing In Premiere Pro CC and DaVinci Resolve 12.5, and while the article isn't too involved........ comment section usually draws a few people who have an opinion, and there's something that seems worthwhile this time. Oh yeah, and I just saw this in my YT sub feed...... Is AMD RX VEGA REALLY better for video encoding? We didn't expect this... (JayzTwoCents). I'd like to see more than just the rendering times, and more of a "benchmarking" from recording the GPU/CPU/RAM activity while he's editing, and at least explaining what he's using to cause GPU demand from both skimming and rendering. I know he's running a decent camera (I think a step or two up from a BlackMagic URSA?) May be close to it in production value? Either way, he has a nice camera and he seems to do some pretty involved post production for also running his channel/site/etc. for his primary income.

You better avoid 960 Pro (or any NVMe SSD) for production machine at this moment. They are not just non-bootable (most likely forever) but not natively supported in cMP yet.

I haven't check the RAM price closely, but 32GB (4x8GB) server pull RAM should not cost more than $75, in fact, it was $50 some time ago. $100 may be a bit too much.

No, memory bandwidth should be no big deal. In my own experience, video editing is not that VRAM intensive. Those much higher memory bandwidth usually can only improve at most few % performance overall. Of course, this very depends on your workflow. However, by considering both RX560 and W7000 is pretty low end now. I really don't think the memory bandwidth will make any difference.

RX580 is a very make sense choice, but if go for RX560, I personally may simply go for the R9 280, which can performance better than the RX560 (in term of video editing / compute in MacOS, I didn't check which card perform better in gaming) and flashable. A single R9 280 can replace the HD 4870 + RX560, so, you can have 3 PCIe slots for whatever you want. And the R9 280 is very cheap. You can even go for dual R9 280 without any extra PSU or Pixlas mod. But in this case, I definitely prefer a single RX580 now.

A reason of better stick to gaming card in MacOS is because of the driver. e.g. The top GPU D700 on Mac Pro 6,1 is just the R9 280X (they have identical Device ID). That means if you install a R9 280X into a cMP, MacOS guarantee has native driver support, and have better chance that the software (e.g. FCPX) is optimised for it. (P.S. The R9 280 I recommend has same Device ID as the HD7950 Mac Edition card, therefore, it should also be natively supported as long as cMP still officially supported by Apple). For any workstation card, they may or may not work properly. Apple has no obligation to support them. Nvidia provides its own web driver to fix this issue, but AMD do absolutely nothing on this matter now. If I were you, I won't let my production machine rely on a card that may lost support after any OS update. Also, all workstation card's features are disabled in MacOS due to lack of proper driver support (e.g. ECC VRAM). So, what's the point to pay more but get less? A R9 280X is definitely more powerful in MacOS than a W7000 (unless you are doing something that require >3GB VRAM), go for the guarantee natively supported faster cheaper card make more sense in general.

However, if you looking for a single slot card for whatever reason, then workstation card usually is the best option. AFAIK, none of the single slot RX460 work on cMP (most likely due to custom PCB design). However, W7000 really works fine natively in MacOS. That's pretty special among workstation cards. So, it is indeed a good choice if you need a single slot card.

This is the best response I could ever ask for. You're awesome. I'll use that to decide on my GPU when purchase time comes. I think I might just go with the 280X if that's the case. Although, I will keep an eye out for Vega drivers since that's going to be what's in the new iMac Pro. If it gets supported, I bet that will be the best bang for the buck with the cMP, and I wouldn't mind that card if/when I end up using Resolve. From what I've seen so far, it looks like just Resolve seems to benefit from having more and more GPU power (even getting benefit from SLI)

I've actually been watching the server pull memory ever since before these towers came into the house (a couple of months ago), and the cost has risen from ~$70 to $100, but unless I can definitely find it cheaper relatively soon, I can't really complain about the money. Here's what I have saved in my eBay for now, but I'll keep looking around until the purchase date comes: Samsung (4x8GB) 2Rx4 PC3-10600R (this is the average price I'm seeing)

You mentioned NVMe isn't bootable. Would AHCI be comparable in a 4 bay PCIe adapter? I guess I could probably stick with 2.5" drives if I need to. Although, if I could use quad NVMe to boot Windows as a second OS, and use it for overall storage as well....... ideas or thoughts?

If I'm skipping over anything, or if anyone has some ideas for me to mull over, it's much appreciated.

 

[h9826790]

Here's something I found today- Using Dual vs Single GPU for Video Editing In Premiere Pro CC and DaVinci Resolve 12.5, and while the article isn't too involved........ comment section usually draws a few people who have an opinion, and there's something that seems worthwhile this time. Oh yeah, and I just saw this in my YT sub feed...... Is AMD RX VEGA REALLY better for video encoding? We didn't expect this... (JayzTwoCents). I'd like to see more than just the rendering times, and more of a "benchmarking" from recording the GPU/CPU/RAM activity while he's editing, and at least explaining what he's using to cause GPU demand from both skimming and rendering. I know he's running a decent camera (I think a step or two up from a BlackMagic URSA?) May be close to it in production value? Either way, he has a nice camera and he seems to do some pretty involved post production for also running his channel/site/etc. for his primary income.


This is the best response I could ever ask for. You're awesome. I'll use that to decide on my GPU when purchase time comes. I think I might just go with the 280X if that's the case. Although, I will keep an eye out for Vega drivers since that's going to be what's in the new iMac Pro. If it gets supported, I bet that will be the best bang for the buck with the cMP, and I wouldn't mind that card if/when I end up using Resolve. From what I've seen so far, it looks like just Resolve seems to benefit from having more and more GPU power (even getting benefit from SLI)

I've actually been watching the server pull memory ever since before these towers came into the house (a couple of months ago), and the cost has risen from ~$70 to $100, but unless I can definitely find it cheaper relatively soon, I can't really complain about the money. Here's what I have saved in my eBay for now, but I'll keep looking around until the purchase date comes: Samsung (4x8GB) 2Rx4 PC3-10600R (this is the average price I'm seeing)

You mentioned NVMe isn't bootable. Would AHCI be comparable in a 4 bay PCIe adapter? I guess I could probably stick with 2.5" drives if I need to. Although, if I could use quad NVMe to boot Windows as a second OS, and use it for overall storage as well....... ideas or thoughts?

If I'm skipping over anything, or if anyone has some ideas for me to mull over, it's much appreciated.

Oh yeah, may be the price go up a bit, I expect it's just $50-60, but it actually cost exactly $75 now (4x8GB). This is the price that directly get from China taobao.com, the seller says the DIMMs are new, but I won't believe that anyway (Don't get me wrong. Their product usually works well, looks exactly like new when arrived. in fact I never get any faulty items from them, but I just don't believe it's new with that price)

ACHI is fine, in fact, IMO, that's the very best SSD that you can use in cMP with the current OS. However, it may be a bit hard to acquire. And NO, NVMe is not bootable regardless OS type (including Windows), it's a firmware limitation at this moment.

I personally still stick the 2.5" SATA SSD. For my usage, I can't justify the cost for a PCIe SSD. I prefer more capacity (for the same money) than high spec which that won't give me benefit most of the time.

I mainly use FCPX for video editing, it also utilise multi GPU very well.

 

[devon807]

I mainly use FCPX for video editing, it also utilise multi GPU very well.

Does MacOS see/use the dual 7950/ R9 280 as 6GB of VRAM or 3GB Mirrored?

 

[h9826790]

Does MacOS see/use the dual 7950/ R9 280 as 6GB of VRAM or 3GB Mirrored?

Not 100% sure, but since the VRAM usage in iStat is pretty symmetrical. I believe it's 3GB mirrored.

Edit:
I should emphasis that FCPX use 3GB mirrored. In normal desktop environment. They are obviously 6GB separated.

 

[devon807]

Not 100% sure, but since the VRAM usage in iStat is pretty symmetrical. I believe it's 3GB mirrored.
View attachment 715418

Ok, I wonder if 4k would eat up more VRAM than 3GB. With the GTX 960 monitoring the usage it currently uses 50-60% of the 4GB and only a quarter of the actual GPU. Do you think I may be safe with 3GB for 4k? Also how may one undervolt these cards. Unfortunately the 380's have went back up in price costing about $250-300. This narrows me down to the dual 560's again. Do you think that would be a viable option since it's so cheap?

 

[h9826790]

Ok, I wonder if 4k would eat up more VRAM than 3GB. With the GTX 960 monitoring the usage it currently uses 50-60% of the 4GB and only a quarter of the actual GPU. Do you think I may be safe with 3GB for 4k? Also how may one undervolt these cards. Unfortunately the 380's have went back up in price costing about $250-300. This narrows me down to the dual 560's again. Do you think that would be a viable option since it's so cheap?

In fact, if I have to buy a new card now (for dual GPU config), and confirmed the RX560 works OOTB. I will definitely go for the 560 4GB.

The VRAM usage in my last post was rendering 4K video in FCPX (about 3 filters with some special effects), So, it seems 3GB is actually enough for most of the time. And for that particular one, because I need to check something down to single pixel accuracy during editing, so I didn't use proxy. I am quite confident that in general (proxy available), 3GB VRAM is actually OK for 4K video editing (if you know how to do it correctly and use the resources effectively).

To undervolt the card, you have to do some test in Windows, find out the voltage and/or clock speed you want; dump the VBIOS; edit it with the correct editor (e.g. PolarisBiosEditor for RX560); then flash it back to the card (e.g. by ATIWinFlash).

Once flashed, the card will follow that voltage / frequency table to work under MacOS. This is how I undervolt my cards.

But TBH, I think for RX560, you don't need to down volt it for dual GPU setup. Each mini 6pin should able to power one card without any issue (standard clock, not OC). Undervolt just mark them run cooler, less chance in thermal throttling, and further save some power. I believe it's not necessary in cMP.

 

[devon807]

In fact, if I have to buy a new card now (for dual GPU config), and confirmed the RX560 works OOTB. I will definitely go for the 560 4GB.

The VRAM usage in my last post was rendering 4K video in FCPX (about 3 filters with some special effects), So, it seems 3GB is actually enough for most of the time. And for that particular one, because I need to check something down to single pixel accuracy during editing, so I didn't use proxy. I am quite confident that in general (proxy available), 3GB VRAM is actually OK for 4K video editing (if you know how to do it correctly and use the resources effectively).

To undervolt the card, you have to do some test in Windows, find out the voltage and/or clock speed you want; dump the VBIOS; edit it with the correct editor (e.g. PolarisBiosEditor for RX560); then flash it back to the card (e.g. by ATIWinFlash).

Once flashed, the card will follow that voltage / frequency table to work under MacOS. This is how I undervolt my cards.

But TBH, I think for RX560, you don't need to down volt it for dual GPU setup. Each mini 6pin should able to power one card without any issue (standard clock, not OC). Undervolt just mark them run cooler, less chance in thermal throttling, and further save some power. I believe it's not necessary in cMP.

Thank you for your response. And I agree, I don't think you need to undervolt the 560's but you would the 7950's/280/380's. I think I will be going with the 560's! I really tend not to use proxy media, I only use it for large client work (which is few and far between nowadays)

 

[thizisweird]

Oh yeah, may be the price go up a bit, I expect it's just $50-60, but it actually cost exactly $75 now (4x8GB). This is the price that directly get from China taobao.com, the seller says the DIMMs are new, but I won't believe that anyway (Don't get me wrong. Their product usually works well, looks exactly like new when arrived. in fact I never get any faulty items from them, but I just don't believe it's new with that price)

ACHI is fine, in fact, IMO, that's the very best SSD that you can use in cMP with the current OS. However, it may be a bit hard to acquire. And NO, NVMe is not bootable regardless OS type (including Windows), it's a firmware limitation at this moment.

I personally still stick the 2.5" SATA SSD. For my usage, I can't justify the cost for a PCIe SSD. I prefer more capacity (for the same money) than high spec which that won't give me benefit most of the time.

I mainly use FCPX for video editing, it also utilise multi GPU very well.

I don't touch chinese retailers, personally. Never bothered with Alibaba, Taobao, none of them. I'll stick with the typical US-based places for purchases, and if I find a Chinese seller that can get me parts within 2 week window or 100% money back, I'm not interested. I may take my time upgrading the GPU, but memory is a need as of last year with this tower. In fact, I just was able to play around with FCPX last night with 16GB, and it acted like my previous work was nothing in comparison to running with 12GB. Since that memory isn't mine, I was simply testing things out; however, it was genuinely a world of difference for what I'm doing. I also noticed it was balancing the load better (which I think many ignore/discredit). If I have to write off $100 as a business expense to get my tower that stable, then so be it lol. Especially if it means avoiding Chinese mail (which I've never had anything come from China faster than 1 month, no matter what it was or what the delivery promise was).

Then again, if someone has some extra memory they'd like to sell below eBay pricing, I'm listening! Same goes for the CPU upgrade, but I doubt anyone is going to have those handy haha.

I guess for my SSDs I'll just swap my original idea lol. 2.5" SSD array for OS, and M.2 down the road in a PCIe slot for fast storage. I won't complain about that. I may not need the crazy fast speeds with PCIe storage, but with transferring large files I'd like to avoid down time or pushing my transfers to the background while I edit. I've had too many bad experiences in the past where files didn't transfer properly (for a number of reasons), and I also would like to be able to dump data while someone waits (I know, a tall order with large files, but I am that kind of person with customer service lol). I also will be doing a lot of smaller work in my downtime for practice, and if I can dump an entire SD card onto my computer while the owner waits.... I think that's as good as it gets. I like those personal touches. This also wouldn't my first time doing business with that level of accommodation, so anything that takes a while will feel like an eternity for me... and it will only make me look even more average to clients, which I like to avoid

That all aside, it sounds like I might go for the RX 560 to start with! Much thanks for the info and SS, mate. It's also nice to know that 4K editing should be manageable with the 560, as I'm sure I will be handling plenty of 4k files.

This has really been a huge help. Endless thanks from me. I owe you a beer!