[Dave Haynie]

Just for grins, try relocating your images to a different drive, even the C: drive … shouldn’t take long, and you can point Vegas at one of the moved files and it’ll find the rest. Render out to yet another drive… even a USB dongle or SD Card is fine. Now check your rendering time.

Chroma keying can certainly slow things down in a render, but it’s CPU intensive, so that’s not your immediate problem. When you don’t see the CPU hitting close to 100%, that’s a clear indication the CPU isn’t your bottleneck. It should be, since it’s usually pretty easy to ensure you do.

A modern HDD is pretty fast, of course, and most folks don’t run into bandwidth issues. But most native video is around 25Mb/s… Cineform can go 50-100Mb/s depending on settings. So that’s going to consume more HDD performance. Then, add in enough graphics, and you may find your HDD is thrashing… its performance can become dominated by seek time, not linear performance.

-Dave

[Dave Haynie]

Oh, I see your problem… only 32MB on PC#2 🙂 Assuming you mean 32GB… well, sure, an image resize is an image resize. I have the same processor, and can’t really imagine any project I’ve done taking 12 hours on it.

First thing I’d check is my performance. Fire up Task Manager when you’re rending and look carefully at the CPU performance. If you’re not seeing at least 90-95% (or maybe 80-85% with GPU rendering enabled), you have other problems. Make sure your data drive isn’t thrashing… it may be fast, but too many things from the same drive and you’re just seeking all the time. And it can actually be worse with a RAID.

It’s quite possible to slow down based on still graphics. Not what I’d expect if that’s limited to like a lower-thirds graphic or two per frame, but sure… for something I use all the time, couldn’t hurt to pre-render it to an output-sized uncompressed PNG — naturally, with its own alpha (easy for a still), no chroma keying or any crazy stuff.

-Dave

[Dave Haynie]

I totally agree. Microsoft has a terrible steering problem. They’re going along their chosen road, drive right past something (office tools, The Internet, PC-based entertainment, Gaming, Mobile), then think they can correct course if only they steer hard enough toward that missed thing. Which, as you might expect, has more of a tendency to run them off the road than copy that thing well.

Windows 8 is the perfect example. They have a touch interface now, as John pointed out, basically the one from Windows Phone, which itself was inspired by the one from the Zune (you remember the Zune, of course). I have used just about every touch interface… Microsoft’s for me is the worst, on a touch device. It’s just not intuitive to the same degree as iOS, Android, Ubuntu for Phones, Blackberry 10, WebOS, and … there’s got to be another in there somewhere. Like many things Microsoft “embraces and extends”, they’re not the inventors of these ideas… they’re someone else’s. Whoever’s designing these seems to have both missed some fundamental aspect of touch UI harmony the others pretty much get (I had some issues with Android before 4.x, too, but I think it’s mostly fixed). And at the same time, they’ve forgotten the even more fundamental rule of any GUI — it’s a user interface that uses graphics in order to make things intuitive. The Windows touch interface is chock full of “you just have to know it” things that no amount of intuition will reveal.

Then they made the mistake of doing this everywhere. Sure, touch is critical on smartphones and tablets. Even there, touch is a blunt thing… no precision. If you’re doing to do much creative work (eg, input) with a tablet, you probably want some other kind of input device. And that’s also where Microsoft fails… you pretty much need the keyboard for Windows 8 anyway. When I was messing around with Windows 8, in a VM on my desktop machine, I was annoyed at all the things basically missing or impossible to find that I had in Windows 7. The stock answer: there’s a hot key for it. On a touch UI. Sure.

Despite moving the world from pen/stylus to touch, Apple’s been far smarter here. They’ve actually been looking at the things from the iPhone/iPad that might work on the desktop, and rejecting the things that won’t. Like touch screens. Apple did a study a year or two ago, and found exactly what had already been learned and apparently forgotten back in the 80s… vertically oriented touch screens you have to reach for are a terrible idea. The CAD industry did this in the 80s.. there were all sorts of dedicated CAD systems, before PCs could really do the job, and most of them used touch screens, light pens, or some other system for input to the screen. And in a few years, these companies were all selling users larger digitizers, adding keyboard commands, anything to get them off the screens. If you think RSI is a problem with keyboard and mice (and it can be), that’s nothing compared to a few months of reaching out to your big-screen monitor. And today, pretty much everyone is using monitors folks in the 80s, even with $20K-$50K CAD system budgets, could only dream about.

In short, desktop touch screens will NEVER work. And still, we’re forced into this on Windows 8, even before Microsoft has touchscreens or any other HID device to make Windows 8, well, maybe still a trip to the dentists’ office, but less of a root canal.

And perhaps the most annoying thing about this: none of it was necessary. The Windows 8 UI is just a “graphical shell”, after all. The UI shell has always been replaceable in Windows, just as you can use different command-line shells (PowerShell, “Command Prompt”, bash+cygwin, etc). They could easily have kept the Windows 7 shell available for desktop users, if they didn’t want to include the full compatibility, as they have in the past (eg, Windows XP could give you pretty much the Windows 2000/98 experience if you wanted it). And there are certainly 3rd party tools to fix this up, though you’ll need many to fix everything. And I’d be concerned about reliability at that point, particularly if you’re mixing a bunch of random fix-up tools from different companies.

And that’s before you get to Microsoft’s WinRT… the whole tablet OS that’s there along with the tablet UI. Full-screen only (well, ok, now offering a couple non-overlapping panels, just like Windows 1.0 and some Android home screen shells) is apparently also Microsoft’s vision (Apple and Ubuntu are a little into that, too, unfortunately, but not locked in the way WinRT is).

-Dave

[Dave Haynie]

If you’re looking for most kinds of apps, they’re on Android as readily as on iOS. I use my tablets for camera control, video preview (my recently-damaged Asus tablet has a clip-on keyboard with extra battery and full SD card and USB interfaces — also allows backing up SD cards to NTFS or exFAT formatted HDDs in the field), etc.

If you’d like to do any sort of music development on them, you need the iPad. Android’s audio system is using older Linux components which, like most of the audio APIs throughout the history of Linux and UNIX, have weird latency problems. That’s been fixed with newer audio subsystems in Linux itself (a buddy of mine does all of his music production on Linux using free tools… kind of a masochist, but it’s possible), but not in Android. Playback is just dandy, but it’s not a good DAW or instrument controller.

Samsung specifically has the Note series, which adds another dimension if you’re into photo editing on a mobile. They include a Wacom digitizer pen, which is dramatically more accurate than your fingertips on the capacitive touchscreen (touchscreen is in there, too). Microsoft’s been including a similar thing on their Surface Pro (of course, that’s a full fledged PC, more of a convertible Ultrabook than a tablet), to give you the accuracy you need to control standard Windows applications without a mouse.

In short, yeah, they’re useful for lots of little things. Nothing you can’t do with a laptop and a smartphone, though some of the more useful aspects of both in a single 1-1.5lbs device with all-day battery life and pretty stellar screen quality these days (the latest Samsung Galaxy Note 10.1 has a 2560×1600 pixel IPS screen). Of course, you’re paying for the premium features here, as you are in the iPad.

But I’d watch out for $100 tablets; they work, but they’re using older technology, and this has been a very fast moving market so far. Samsung’s already shipping tablets with 3GB RAM, the practical limit until they move to 64-bit chips next year. Apple’s got less RAM, but just released the next iPhone with a 64-bit ARM SOC of their own design, and you can bet that shows up in a new iPad in a month or so. An cheap tablet may not run today’s apps; same goes with an older iPad model.

-Dave

[Dave Haynie]

Consider for a minute what you’re trying to do here. You have 60fps video (actually 59.94fps) and you’re trying to stuff that into a 30fps (29.97fps) output… that’s a 2 pound chicken in a 1 pound pot.

So something’s got to give. When you way “blend frames” or “interpolate frames”, what you’re asking Vegas to do in each case is to take those two 60p frames that occur in 1/30th of a second and stuff them both into one 30p frame. That’s your “ghosting”; they both do the same kind of thing, just using different algorithms.

Your only other alternative — and presumably what your other software does — is to drop every other frame. That’s what Vegas does when you don’t ask it to interpolate/blend. Then you have only one 60p frame dropped into each 30p box. Works fine, no ghosting, but of course you’re tossing out half of your information. You have to decide which is right for your project.

-Dave

[Dave Haynie]

Yup. I used Win7 Pro on my build last month as well. The main reason you need Pro for video work is memory; Windows 7 Home only allows 16GB to be used. Yeah, I did say “only 16GB”.. and while that’s probably enough for most single-application video editors, getting more doesn’t cost that much, and it could be pretty useful down the road.

-Dave

[Dave Haynie]

[Joe Mantaratz] “Can you explain why using a dedicated SSD/HDD doe rendering would not result in shorter render times? With no heads and spinning discs one would think it would be a faster process. No doubt I do not have a good grasp on the meaning spec data of either. As mentioned before, I have always used dedicated drives for separate tasks and enjoyed efficiency and thought the same would apply using SSD’S.”

Any render is a pipeline… think of it as a bucket brigade if that helps. The first bucket is any number of disc to buffer transfers, then those decoded to raw video, then FX applied, then compositing, then rendering to your output format, then than buffer dumped back to disc.

Each of those buckets consumes some resources: memory, disc, CPU, etc. If each of these were unlimited, you’d have a finished product seconds after you pressed your “Render” button. Which of course doesn’t happen. The speed of the render is determined by what’s essentially the weakest link your system.

Given the work that needs to be done, the weakest link should be computation. When you look at your system’s CPU and see every CPU core in the system going at close to 100%, you know your CPU is at that point the weakest link in the system. Even with infinitely fast discs, you’re not going to see much faster rendering.

Why? Well, it’s simple.. most of the time, your “write buffer to output file” bucket is empty, waiting for more data from the rest of the pipeline. If you’re rendering to AVC or MPEG-2, you’re probably not going much more than 25-50Mb/s. You can write this to an SDHC card (10MB/s = 80Mb/s), a single HDD (100MB/s = 800Mb/s), a RAID (250MB/s = 2Gb/s), an SSD (500MB/s = 4Gb/s), it doesn’t matter… that media will never stall that part of the pipeline. And on a modern PC, the CPU has very little work to do in that write… most is managed by DMA channels in the system hardware.

It’s more possible to run into stalls on input media, IF you have lots of input media (multiple AVC/MPEG/DNxHD/raw files, large still photos, etc). This is mostly the product of hard drives thrashing… as you read more separate streams from the same HDD, your throughput tends toward seek time rather than peak transfer time. On the other hand, if you’re reading and decoding N AVC streams, you’d be hard pressed to make the HDD the weakest link even at that, because AVC streams are relatively small, so it’s pretty easy to suck up many frames of each stream at once… and then they’re a big CPU hog to decode. I’ve never had a problem with too many AVC streams from a single HDD. I have run into it intentionally trying to make the HDD the bottleneck with multiple 144Mb/s DNxHD streams, and ran into it ages ago with a couple of HDV streams and hundreds of 6 and 8 MPixel JPEG photos in a project.

There was also a concern about using the same drive for programs and data. Today, I look at that as more an SSD vs. HDD issue, but it was a valid concern, especially for music (I was doing DAW stuff for years before video). Again, with many audio tracks, you have lots of seeking. You’d like to have 8, 16, 64 of these, depending on the project, and they need to be mixed in realtime, particularly if you’re overdubbing. In short, a more difficult problem than video editing.

So the system drive… when a program starts up, it can simply load into memory. Some do that. Some load DLLs (dynamic libraries), and make those on-demand… so only features you’re using get loaded. Another way to do this is via “speculative loads”… the program is marked in virtual memory as having been loaded but isn’t actually loaded, so that chunk of the program doesn’t get loaded until run, then it’s trapped and loaded from disc via the VM system). And in the past at least, you might not have enough memory to run all aspects of the program at once. Code could be kicked out when not used, reloaded, etc. That results in disc activity; even more if you have to much data in memory and some starts to get swapped out to your swap file. So a system disc can be active, which means a few more seeks, and something that could throw a twig in your realtime mix, or technically, slow down a video render, at least a bit.

Thus, we started using secondary drives, to keep this (and any other disc-related system or app activity we leave running while running that DAW or NLE) from glitching our primary work. I have 64GB of RAM on my system… I’m pretty such these are not rational fears about the main drive versus the data drive. Nothing’s getting tossed out of memory. There could be speculative or DLL loads, but they’re only an issue once… not a big deal.

[Joe Mantaratz] “Understand about the defined write cycles of SSDS but don’t have a comparison of what the life would be vs HDD. Not really concerned about the SSD dying as all the media would be archived and only the Vegas project file would be updated. I’ve always made sure to save the project to other drives with all the media in one folder. Nothing of mine is ever spread across multiple drives.”

There are different measures of lifespan. In terms of just sitting there, an SSD should outlast an HDD. The systems are often rated in MTBF… Mean Time Between Failure…. basically the average time a device of any given generation will run before it fails. HDDs are of course mechanical systems, and current ones should deliver an MTBF of 500,000 hours or more. SSDs, being all electronics, should deliver a MTBF of better than 2,000,000 hours.

But that’s not the whole story, because of the difference in write cycle life. As mentioned, a typical MLC Flash cell has a life of only a few thousand write-erase cycles. All SSDs use wear leveling logic, so you can write the same location 100,000+ times without problem — the drive will logically map that write to different areas of the flash chip from cycle to cycle. Consumer drives should be good for better over 100,000 write-erase cycles randomly across the drive. Of course, if you decided to re-write the entire drive, you’d get a couple thousand cycles of that before it failed. Enterprise-class drives use SLC flash, which can have 100,000 cycles per cell or more. So these last millions of cycles in practical use.

HDDs don’t usually have any specific limit to the number of write-erase cycles. But they will wear out, mechanically… and of course, at the worst possible time. Drive heat can also cause premature of the disc.

In the do as a I do department, I have an SSD as my system drive and a RAID10 as my data drive. The RAID10 configuration delivers 6TB for less cash than most 1TB SSDs, it’s pretty fast (350-375MB/s for reads, typically) and it’ll ensure the abuse of constantly changing data much better than an SSD would. I also need the space. Sure, it won’t be as efficient dealing with a huge number of assets, due to seek time, but that’s only a problem in 60 layer animations using raw video files, so far. Intermediate files for a project like that can get well into the hundreds of GB, so for a serious project there’s always a project drive as well, with some of the assets living there. Either way, FX and compositing are often still the bottleneck on this kind of project…

-Dave

[Dave Haynie]

[Joe Mantaratz] “I am researching using two 960 GB SSD drives that sell for about $1,000 each and smaller maybe 520 GB for rendering. Not totally convinced but I would prefer to have Vegas, OS on one and the media on the other and a third to render plus externals for archival. Not interested in RAID drives at this point. I’m not shooting 4K or Red footage (yet) so not a must have. “

Keep in mind, unless you render in raw HD (and I sometimes do, for animation), you’re getting absolutely no advantage out of an SSD as a rendering drive. That one 25Mb/s or 50Mb/s stream is not going to need 350-500MB/s (2.8-4.0Gb/s)… rendering won’t go a hair faster than if you dumped it to a USB stick. Not true, of course, if you’re using just the one drive for assets and rendering… SSD will outperform single drive and even a small magnetic RAID there, due to seek times.

I like a RAID for the main drive because, once your software is in place, it’s mostly used for reading. The OS does writes, sure, but small ones — another thing that magnetic drives don’t use. But SSDs still have a much shorter life cycle than magnetics, due entirely to the limited erase-write cycles. Used as a system drive, it should be a fairly moot point, since you’ll probably want to replace it with a larger, faster drive, before it’s dead. But writing many, many gigabytes to the drive on a regular basis? Not ideal. Particularly if you’re using an MLC-flash drive rather than a very expensive SLC-flash drive.

I should mention I also have two project drives… I mount SATA drives and use them as dedicated project discs, for larger projects.

[Joe Mantaratz] “When is the latest and greatest from Apple being released? No info I have seen from Apple or what the specs are. “

I’m not John, but the only Mac I’d consider viable for video work is the new Mac Pro. And it’s a wonky system — looks like a trash can, no internal expansion, no HDD, no optical. It’s kind of cute — about 1/8 the size of the current “looks like a desktop PC” Mac Pro.

You have dual ATi/AMD FirePro GPUs, on custom boards, so you’ll have these for the life of the PC. Which is probably not a problem, though keep in mind, very few OpenCL programs can use dual GPUs right now, though OpenGL can. These share a fairly interesting cross-point switch with some Thunderbolt ports… which is of course a new Intel thing that only Apple is currently using. Of course, there’s FirePro and there’s FirePro — today’s top of the line W9000 cards cost $3500 each, low end are under $300. This is one of the reasons no can but guess what these things are going to cost.

The high speed expansion is Thunderbolt, and you have six ports. But as mentioned, there’s a cross point switch, which flips the port to either Thunderbolt or DisplayPort, depending on how it’s connected. So if you have two monitors connected, you only have four TB ports available.

One Thunderbolt port runs two 10Gb/s channels, a bit faster than a single PCI Express 3.0 link. They also support Thunderbolt 2.0, which lets you aggregate the two links into a single 20Gb/s channel, much as PCI Express aggregates links, like the 16 that usually feed a GPU card. Great for external storage… you’ll need that. Not usable for external video cards or some of the other things pundits immediately thought about when this machine was debuted. And of course, four USB ports… gotta connect that keyboard, still.

This is going to be pricey, when it debuts later this year… they’re waiting on Intel’s 12-core Xeon; again, a part that’s not out yet, and no specific version mentioned. Given the retail price of the high-end 6-core i7 is over $1000, this, too, could add some weight to the price. Same memory type as my far cheaper machine, not sure what they’ll be shipping with, but it’s an LGA2011 Xeon, so four memory lanes of DDR3-1866.

For storage, it’s got PCI Express based Flash, which they claim is 1250MB/s… that’s damn fast. Faster than SATA 3.0… presumably they’re using a 2x PCIe link. The problem is, you won’t get much of that storage… going to make dual-booting Mac and PC an issue.

This will be a kick-ass machine. Then again, assuming the typical Apple pricing, any machine in that range will be.. I’d be surprised if they sell a 6-core version for much under $4K… probably more. Though it was nice to see Apple getting back into the pro market — their current Mac Pro is ancient, the design dates back to late 2009.

More here: https://arstechnica.com/apple/2013/06/a-critical-look-at-the-new-mac-pro/

[Joe Mantaratz] “Xeon processors are designed for servers not workstations so I am wondering the application realism. Are there any dual processor motherboards using other Intel CPU’S? “

Xeon processors are usually just modifications of Intel desktop processors, with more cache and sometimes more core units. There’s nothing inherently “workstation” vs. “server” about them, except that they support server-class RAM (registered DDR, FDIMM, depending on generation). You optimize your processor selection for the job. Some Xeons add cores and lower the peak clock, which pushes them toward server optimization; others keep the higher clocks, perhaps with fewer cores. No idea about the 12-core chip, since that’s not out yet.

Intel doesn’t support dual-socket for anything but the Xeon, even though the LGA2011 i7s and LGA2011 Xeons are essentially the same chip in some versions. They could, of course, they choose not to… they get more money for Xeons. Ages ago, there were hacks for this with some of the chips, but it’s simply not work messing with.

-Dave

[Dave Haynie]

[Stephen Mann] “lso, I seem to recall that the MS Windows license is one per CPU, so that a dual-cpu board would require two licenses for Windows. Since I’ve never built a dual-processor rig before, I never looked into this.”

Not quite. Microsoft only licenses Windows Home for a single core. But you’ll need Windows Pro to get beyond 16GB anyway, so it’s not a significant issue.

Dual processor is a way to go, but you have to be careful about how that propagates pricing. Most dual socket motherboards are intended for servers, so they’re large, they require registered DIMMs or even FDIMMs, they may not have enough PCIe or USB ports. Some are intended for workstations… that’s what you’re looking for. Expect to pay about 3x as much for the main board, but for that level of performance, maybe not a problem. Asus has a line of workstation main boards like this (example: Z9PE-D8 WS), Supermicro does too still, I think (I had a dual-socket Supermicro system, ages ago).

Similarly with the CPU… the i7-39xx are essentially the same as the Xeon E5-16xx series. The Xeon E5-26xx are just like that, too, only allowing dual socket operation. What you want to watch out for is that most server work demands more cores, not faster cores. So some of the Xeons drop their speed considerably, but enables more cores. You want both, speed and cores, in a workstation, and shouldn’t take a hit on peak speed.

Also, consider drives and other I/O… LGA2011/X79 systems usually have 40 PCIe lanes. The consumery LGA1155 has only 16 PCIe lanes. These aren’t just used for adding multiple GPUs, but for more SATA ports (my main board as 12), more USB, perhaps other peripherals. As well as those sockets that can handle 3 or 4 GPU boards.

-Dave

[Dave Haynie]

[Joe Mantaratz] “However I would gladly like to hear what the experts are using and more importantly perhaps, not using. LIke you I want to edit and not spend my time finding new ways to work around problems related to hardware/software.”

I built out a new system last month:
Intel i7-3930K processor (Socket 2011, 6 cores)
64GB DDR3-1866 DRAM
Gigabyte main board, Intel X79 chipset
Crucial 960GB SSD main drive
Western Digital 3TB x 6 RAID10
AMD Radeon HD6970 GPU

Very happy with the result. The 64GB is for photo editing; I’ve rarely come close to the limits on my old 16GB system doing video editing. On the other hand, the Intel Socket 2011 chipset runs a four-lane memory bus, so if you’re going to buy four DIMMs anyway, might as well go for the 32GB total. It’s not that expensive.

I use my PC for video, music/audio, photography, and electronics CAD. So even with the data drive, I have more than 300GB in application installs… and I’m not done yet restoring old software to the new system. You can get a faster SSD in some of the smaller form factors, but I’m very happy with the speed. The RAID10 is using Intel’s C600 chipset RAID, which is “software” RAID with some hardware acceleration in the driver. I tried it RAID5 mode and it was unacceptable… the main reason for RAID in-the-box is speed. But RAID0 actually drops your reliability versus a single drive, and I have an external Drobo anyway. This is a good solution, the RAID10, fast and reliable, though of course, you’re halving your effective storage as the price for that reliability.

[Joe Mantaratz] “With great dismay I have been reading the plethora of issues with Vegas Pro 12 and can hear the exasperations of the many waiting on the next build to address the woes. Which versions after 8.0 have been found to be the least troublesome?”

I have actually found Vegas 12 to be better than Vegas 11, at least as good as Vegas 10. And I even use the GPU… which really does help, even with the new system, on some kind of work (for example, the Vegas GPU benchmark project… it’s at least 3x faster with GPU enabled).

[Joe Mantaratz] “I plan on also installing Boris Red, After Effects, Win 7 64 bit Pro, VASST products.”

You’ll also want a good GPU for most video effects. Even when they’re not using the GPU for compositing (OpenCL/CUDA stuff), there’s still a lot of OpenGL/3D stuff you’ll find within the packages. I don’t recommend “pro” series GPUs… I think they’re a waste of money. They’re no faster than the consumer/gaming sort, sometimes actually a generation behind. If you’re doing some really heavy duty 3D stuff, like electronics CAD, you’ll find they’re faster… but at least in some cases, that’s because the consumer cards are intentionally made slower. Seriously, there are some articles that prove this conclusively. Not that you can go wrong with a “pro” series GPU, just that you’re not likely to actually get any benefits, other than perhaps a little more driver stability (consumer cards get frequent tweaks to fix gaming issues, pro series drivers pretty much ignore the gaming issues and concentrate on OpenGL).

[Joe Mantaratz] “I currently own Sony EX cams and have had ZERO issues with MXF format. So no AVCHD on this end.
I dont film with DSLRS at all….”

If you’re doing AVC editing and compositing, you NEED a pretty high-end system to be happy. For MPEG-2, not so much. But some of that is because just decoding an HD AVC stream can take up the better part of a slower 4-core machine, since AVC decode itself is rarely GPU-accelerated or even well multithreaded in Vegas. Once you move to multiple layers, of course, you’re multithreading. But if you can afford it, if you’re doing lots of compositing or FX, you want the extra performance.

-Dave