I don't disagree that the Geforce 900 series is indeed less power-hungry than the Radeon 200 series, but to AMD's credit their GPUs are a year older than Nvidia's currently.
You are not logged in. Please login or register.
SmoothVideo Project → Posts by Nintendo Maniac 64
I don't disagree that the Geforce 900 series is indeed less power-hungry than the Radeon 200 series, but to AMD's credit their GPUs are a year older than Nvidia's currently.
Did you have a model with the reference air cooler? They look like this:
http://cdn5.thinkcomputers.org/wp-conte … -290-1.jpg
FYI, AMD's been basically using the same stock cooler since the 5000 series, which is no wonder it's as crappy as it is - it'd be like using a reference Nvidia non-Titan cooler on the Titan X. This is very likely why there are literally only 2 SKUs available that use said cooler.
Go with the 980, and I am saying that based on my personal experience with AMD, I found they have a tendency to run too hot when pushed too hard. That may be fine for an AMD card, but all that heat transfers into the case where other components have to deal with it.
I must disagree with this for a few reasons...
1. AMD does not have any more tendency to run hot than Nvidia does. AMD's 3000, 4000, and 5000, and 6000 series GPUs all ran cooler and were less power-hungry than Nvidia's flagship GPUs at the time (9000 series up until the 500 series); note that I'm excluding dual-GPU models for obvious reasons.
2. Temperature and heat output are not the same thing - more heat output can give higher temperatures, but higher temperature doesn't necessarily mean higher heat output. This is important because the current Radeon 200 series has quite high transistor density, which gives higher temperatures but doesn't mean higher heat output - this is the same case with Ivy Bridge and Haswell CPUs.
3. The Radeon 200 series reference air cooler was crappy. Models with the stock cooler are still available, but you'd be an idiot to buy them since there's literally only 2 models of for example the 290 on Newegg compared to like 70 models with non-reference coolers, and those 2 reference models are not the cheapest models either. Nevertheless, most reviews used the stock cooler since the non-reference models weren't available back in the day, so for most people the temperatures in said reviews aren't going to be accurate unless the review specifically uses a card with an after-market cooler.
4. Radeon GPUs have a user-selectable maximum temperature setting in the drivers that is set to 95c by default for whatever reason, and users that don't want it as hot merely have to turn the setting down.
5. It would be silly to buy any GPU right now - we know the 300 series is coming this quarter. Even if you buy an Nvidia GPU, you would still be able to take advantage of the according price-drops.
DISCLAIMER: My current PC uses an Intel integrated GPU and my (currently unplugged) discrete GPU is an old-ish Geforce 8800GS.
CTRL+RIGHT moves forward frame by frame.
Alright, that'll be easy to remember since it's the same as MPC-HC.
CTRL+LEFT does something weird... not frame by frame but it moves
I guess I'll investigate that as well once I get around to this whole thing...
Yes it does.
Okaaaayyyy... may I ask how?
I still have not looked into this whole MPDN thing yet, but I would like to have a question answered before I eventually do dive in.
Can MPDN to single-frame advancing while paused like MPC-HC can? In MPC-HC you can hold Ctrl and then use left/right arrow to navigate foward/back by a single frame.
I tried to install SVP in a separate directory but it still uses shared settings =[
You could probably do it with Sandboxie, but just note you'd also have to run the media player in Sandboxie as well.
So I was fiddling with my CPU overclock today and in the process of said fiddling I decided to try my hand at RAM overclocking. Well it turns out that, as long as the memory clock is higher, you can keep the timings super-loose and RAM voltage down and yet you'll still get better SVP performance, even if you're not using GPU-acceleration. Typically in programs, once your timings get pretty loose, you start to have worse performance...
For reference, I took 1333MHz 9-9-9 RAM up to 2133MHz 14-14-14 without any voltage increase (I tested 1600 and 1866 as well, but 2133 performed the best and my PC couldn't boot with anything higher)
My CPU is a Pentium G3258 if it makes any difference...
you may consider upgrade to 960 which has most advanced NVIDIA video engine today.
https://en.wikipedia.org/wiki/Nvidia_Pu … reVideo_HD
Considering the rest of his PC, it's very likely that the pricepoint of a 960 would be quite out of his league ($200 USD) - he could buy a whole processor, motherboard, and RAM for that price.
Honestly, GPU-wise I would look more towards HTPC-focused cards like the Radeon R7 240, but even then it's not that cheap. ($70 USD)
Besides, it wouldn't be the wisest to buy a new GPU now because AMD will be launching their 300 series this quarter (Q2). Even if one doesn't buy a 300 series GPU, it should still mean price-drops on both older-gen AMD products and on competing Nvidia GPUs.
Also it's not good to just assume that the likes of video game footage will always to 30fps - I myself have come across game footage in both 24fps and 25fps even though the games themselves did not run at those framerates not multiples of said framerates.
But by that logic it shouldn't be limited to specifically that but instead be user-customizable - in particular, some people hate high framerates and interpolation for movie content but are perfectly fine with it for video game footage and content.
So, SVP cannot use DXVA right?
The only DXVA it can use with SVP is copyback - if it is set to DXVA native, it will simply not be used when using SVP.
You can alternatively set the LAVfilters decoder to QuickSync, but this may increase the load on your integrated GPU - you're welcome to try it if you'd like, especially since it may also decrease your CPU utilization.
I have integrated one, but can it be used together with the discrete one?
I do know that, in SVP, you can select which GPU to use; don't know about madVR.
How to do it?
I don't have any first hand-experience, but at least for SVP it should be an option below "Interface type" when you right-click on the SVP tray icon.
Just beware that, if your integrated GPU is too weak, you may actually have reduced performance.
I'm using DXVA native
Uhhh, SVP doesn't even work with native...
Also note, copyback has considerably better CPU utilization than it used to in the past in LAVfilters v0.64 and/or MPC-HC v1.7.8.
Lastly, you didn't say anything about your iGP. Do you not have one? I know that SVP in particular can be set to use your integrated GPU rather than your discrete GPU.
You know you can use SVP v3.1.5 with MPC-HC v1.7.8... it can decode HEVC, VP9, etc.
Then the obvious answer is to turn down your madVR settings.
Alternatively, assuming that you aren't using an LGA2011 or Nehalem-based CPU, you could try to set either SVP or madVR to use your iGPU (probably which ever uses the least amount of GPU utilization).
If you do the above, also try setting your LAVfilter decoder settings to DXVA2 copyback (assuming it isn't already set to that) which may help your GPU utilization a bit.
I thought the default renderer was EVR custom?
I'd say try the following...
- Enable "alternate vsync"
If that doesn't work...
- Disable "accurate vsync"
If that doesn't work...
- Disable both "alternate vsync" and "accurate vsync", then enable "D3D Fullscreen"
This would be awesome, but I can't think of any practical way that would make thing possible yet convenient to use without some sort of AI programming.
I actually find SVP to be strangly good in terms of stress-testing - I was able to pass running both prime95 (with priority 8) at any fft setting and furmark at the exact same time for hours, but I'd get a BSoD with SVP after 10-20 minutes until I increased my CPU's vcore.
However, I'm using integrated graphics rather than discrete graphics, so stressing the CPU and iGP at the same time is even more important.
If your CPU has hyper-threading it's possible may in fact be hitting a CPU performance bottleneck - hyperthreading doubles the thread amount, so if your physical cores are maxed out but your virtual threads have no load, it would show as only 50% CPU utilization.
I think it's arrow left/right
It's actually Ctrl + left/right arrow; just left/right arrow skip forward or backwards by 5 seconds.
Also, where would you get 4:4:4 content from? I’d love it if it was standard, but unfortunately almost everything is in 4:2:0 even.
I made my own using said test image.
EDIT:
Because SVP is inactive initially when skipping to a bookmarked timestamp, it was hard for me to capture a frame both with SVP active and without.
I find that I can navigate frame-by-frame in MPC-HC with SVP activated... (I realize that you're using PotPlayer instead).
Uhh... I find that SVP already supports 4K resolutions.
Also the source codec does not matter since SVP does not interface with the video file directly, rather the video player outputs a raw video stream that SVP can use.
Wait, SVP can't handle 4:4:4 chroma as the input? I thought it was able to do so...
EDIT: Just tested it using this chroma test pattern and sure enough SVP cannot. Huh, not sure why I thought otherwise...
Is it at least able to output at 4:4:4 chroma, like when downsizing a 1080p video to a 768p screen?
EDIT 2: It would seem that SVP does indeed process (or at least resize) and output at 4:4:4 chroma.
Well first off, are you sure those are the exact same frames? Because the position of the eye doesn't quite line up 100% (it's off by literally a couple pixels).
Other than that, I personally can't really help you since it can't be what I initially assumed since you specifically said it's an 8bit clip.
Edit: They were much smaller, but still too big to attach. So I've added 3 more links to my initial post.
Most of the time people just use external image hosting sites like Imgur or Minus (I prefer the latter since the former sometimes has a tendency to re-encode PNGs above 1MB into JPEGs)
Also there's not really a point to having BMP links at all if you have the PNG links...
SmoothVideo Project → Posts by Nintendo Maniac 64
Powered by PunBB, supported by Informer Technologies, Inc.