Dave Haynie
Forum Replies Created
-
Yeah… drivers make a big difference. Both good and bad, sometimes.
I’m playing around with two GPUs right now: an AMD HD 6970 and a nVidia GeForce GTX570.
When I first got the GTX570 installed, the performance was disturbingly lower than the AMD on nearly everything. After the latest upgrade, things are better, though the AMD is still faster. AND one extra present… accelerated Main Concept CODECs seem to always crash.
The nVidia also seems to have a questionable video quality… hard to figure out how, this being a digital interface and all (DVI on the PC, HDMI on the monitors). But it looks like nothing much different than an analog display with a bit of impedance mis-match, so you see echos at high contrast transitions. That’s probably going to be the deciding factor. I had an actual change to put Vegas 11 to work on a real job, and I think the nVidia display has been giving me a headache!
-Dave
-
No idea… seens a bit nuts to require a GPU.
Even if they do, there is a possible loophole — AMD has an OpenCL driver for CPUs. You must gave a CPU with SSE3.
-Dave
-
Basically, I woudn’t worry about the PCIe speed. The speed of the interface isn’t the only issue, or even the most important one. The on-board memory speed, number of stream processors, etc. has more of an impact on performance. Chances are, newer GPUs support 2.x speeds, so you may find the device you want based on other concerns has the 2.x interface. You get no benefit from that feature, so just ignore it, and consider the other issues.
-Dave
-
Pretty good. The system I’m running was a quickie replacement. Last November, my Q9550 main board self-destructed. I needed a replacement that day (it was Friday, I had lots of weekend work looming on video projects), so I ran to Microcenter.. about the only place in the South Jersey/Philly area with New Egg or Frys competitive prices.
Anyway, the 1090T (fastest at the time from AMD), main board, and 8GB DDR3 DRAM ran me about $425. Not too shabby, but it looks like you get another few hundred MHz today (is that a 1100T?) for a bit less. Of course, the “Bulldozer” CPUs are on the way… mixed reviews so far. But unless you’re spending big, it’s all about performance for the price, not absolute performance.
-Dave
-
Bus speed is a potential. Think of it like a monitor. I have a monitor here that displays 1920×1200 (actually, two), but a graphics card that can do 2560×1440 or more. When you plug them together, they negotiate capabilities (standard stuff over DVI, HDMI, DisplayPort, etc) and agree to a resolution (aka, data rate) they can both support.
A PCI Express 2.x card is capable of supporting 500MB/s per link. Doesn’t mean it has to ONLY support 500MB/s per link. So when you plug it into a PCIe 1.x board, the controller can support it, but only at the PCIe 1.x speed of 250MB/s per link.
Similarly, a PCIe 2.x main board will accept a PCI 1.x graphics card, but it will only use the lower rate for communications.
Same idea as your wireless network.. modern networks negotiate 802.11 a, b, g, or n connections, and from there choose one of a dozen or so rates at which to communicate, in this case dynamically, based on the quality of the current link.
Of course, I like to say “should work”… a correctly implemented PCIe board works in a PCIe slot of the proper width, regardless of the generation of the PCIe bus. But there’s no telling if someone’s made a design error that might get in the way here. Unlikely, but it doesn’t hurt to check return policies. Or the support site for your particular main board or PC. Just sayin’…
-Dave
-
Ok… you’re mixing specs here. PCI express is always a serial bus that runs between two devices, a card and a motherboard in this case. A slot can support 1, 2, 4, 8, 16, or 32 of these links; PCs usually limit the links to x16 for video cards, x4 or x1 for regular peripherals. The original 1.x PCIe supports 250MB/s per link; thus, one link is not quite twice as fast as regular PCI. The 2.x specs change this to 500MB/s per link.
All 1.x motherboards are supposed to work with 2.x cards; all 2.x mothetboards are supposed to work with 1.x cards. But you can’t upgrade a device to the next spec.
BTW, the word is “bus” (a collection of signals) not “buss” (archaic synonym of “kiss”).
-Dave
-
I claim that 8GB is plenty… depending. You’ll use a bit more for each core you add, but I also have an AMD x6, and 8GB was enough for any video project I threw at it. I did upgrade to 16GB last summer, but that, oddly enough, was for photography. Wicked hard-core photography.
Intel i7s with hyperthreading really don’t consume as much on the “virtual” processors as one would have with a real extra core (I’ll skip the tech talk, unless anyone’s seriously interested). They can also make things go slow, they’re not also a performance boost, it depends on the specific work being done.
Stephen’s also right about the CPU — buy the bad-assest CPU you can afford. That will be your primary bottleneck in just about everything. The GPU becomes important with Vegas 11, but this is still being characterized by the community (and hopefully, Sony), so it’s not as easy to say just how much they’ll help. I also suspect that if you have more CPU power, you benefit a bit less from more GPU power than someone with a weaker CPU.
-Dave
-
You ought to KNOW the HD6870 works in Vegas 11… Sony’s own published benchmarks use this card. If you’re not seeing it, you probably don’t have OpenGL installed.
Sadly, that might be common. Over ten years ago I had an ATi graphics card, a decent one. I didn’t care about GPUs or anything, but ATi was one of the companies actually producing clear analog video. But their installer was a mess… it failed, repeatedly, to properly install the drivers.
Jump forward to the HD6970 I bought last week… same problem. I found it’s reporting errors on really stupid things, too… in my case, a version of a Microsoft support module didn’t install… because I have a more recent version. But the installer reports an error, and doesn’t tell you what failed.
And, with everything installed, the main ATi configuration program won’t load. In fact, it tends to crash my whole system.
I naturally grabbed all the latest stuff online. Still failed. So grabbed the individual pieces, including the OpenGL drivers, which are separate. It works.
My concern right now is that, yeah, the GPU stuff is real; sometimes moderate, but they’re real. Worth it, and Sony’s doing right by supporting OpenCL. But there are still companies doing plug-ins supporting only CUDA. So I’m concerned that CUDA might be more important than I was thinking. All current CUDA implementations also support OpenGL and the Microsoft thing no one uses, DirectCompute, but only nVidia supports CUDA. I’m getting in a nVidia card to evaluate, to see if it compares to the AMD/ATi or not on Vegas performance.
-Dave
-
What was the problem with the Radeon? An HD 6870 should be a fine GPU for Vegas, at least in theory. What are you seeing with it?
-Dave
-
I used a pretty cool TBC years ago, a plug-in for my Amiga 2000. It didn’t just provide time base correction, but you could run special effects, or just things like luma level, color adjustments, etc. So in 1994, in a weird way, I had video “plug-ins”, even though I was editing tape to tape. Trick was, the edit was automated by computer.
What a TBC does is rebuffer your video. Tape is stretchy, and successive generation of tape to tape editing will have a nasty effect on the timing of your video signals. Most digital TBCs work by digitizing your video, storing it to a frame buffer, then generating video back out again.
Thing is, a good digitizer is doing that first part anyway — resynching the video. The big difference is likely to be tolerance: a TBC is designed with the idea that crappy video is coming in; a cheap digitizer may lose its mind if the video is too far off. But in general, you don’t need TBC in front of a digitizer.
-Dave