Polaris vs Pascal for Video Editing/Adobe Premiere Pro CC

[PhonakV30]

We already know Polaris is a terrible miner card due to low compute performance.
http://cryptomining-*********/tag/rx-480-hashrate/

Note something? Yep, lack of memory bandwidth is making it hard for both.

then GTX 1070 is a terrible miner card due to low compute 🙂

 

[kraatus77]

Kraatus77 -

You need to educate yourself a bit better on the history of CUDA/Premiere Pro/Adobe in general and maybe perhaps read my entire OP. CUDA is very relevant to this topic, my topic, as Premiere Pro is heavily accelerated by CUDA. You're just stating AMD cards are better because of OCL 2.0, you're missing the point entirely. Also, this has nothing to do with Mining. General compute performance does not translate/scale perfectly across the board for every app, especially when Premiere is as heavily tuned for CUDA as it is.

By no means I am saying you're wrong, but just to dismiss something based on not having the newest spec is silly. We all have the proof and results in front of us from Linus's video but he only used 3 examples/benchmarks and they were all exporting/encoding. Most of the speedup benefit you get is not having to pre-render everything and having a fluid and smooth scrolling timeline playback while you're applying multiple effects. I don't think these are things you can easily benchmark.

I think Linus's tests were a very limited sample of what you actually do within Adobe Premiere, which is more than just exporting. Most of the higher end visual effects are 32bit and cuda accelerated, which he completely ignores in this video. I was hoping to get more insight beyond what's out there.

linus 's video plus what sdk posted already showed almost no difference between all those cards, and since nv on cuda vs amd on ocl doesn't show much difference, i think 480 should be a better option since half the cost.

i said cuda is irrelevant because amd has ocl to fight it, and if you use anything that only supports cuda there's no point even making a thread like this since for cuda only nv is the option.

people still think amd on ocl vs nv on cuda is favorable to nv. those benchmark already shows thats not entirely true. plus ocl 2.0 ensures better future for hardware. what's the point of buying something out dated ? ( it's not about 2>1.2 but better features, we already know on 1.2 amd is better.

 

[thesmokingman]

Mac's are on AMD now and that's Adobe bread and butter. The industry is moving away from proprietary, get with the program.

 

[2blzd]

I will repeat what I already said.

Exporting videos =/= Video Editing. You can't sum up performance of an app by benchmarking one of its 1,000s of features.

The main benefit of hardware accelerated video in the case of Premiere Pro is "real-time editing" - Something that Linus didn't benchmark. The Puget Systems 1070/1080 article briefly touches on this with Render Previews but not compared to any AMD cards.

 

[Thala]

RX 480 and 1070 got the same TDP

Are paying your electricity bill based on TDP or based on actual consumption?

 

[kraatus77]

I will repeat what I already said.

Exporting videos =/= Video Editing. You can't sum up performance of an app by benchmarking one of its 1,000s of features.

The main benefit of hardware accelerated video in the case of Premiere Pro is "real-time editing" - Something that Linus didn't benchmark. The Puget Systems 1070/1080 shows this benchmark but not compared to any AMD cards.

For that you need to connect the dots and make an educated guess, or better just wait for launch/reviews. but i have a strong feeling it will be pretty much same.

 

[Bacon1]

I will repeat what I already said.

Exporting videos =/= Video Editing. You can't sum up performance of an app by benchmarking one of its 1,000s of features.

The main benefit of hardware accelerated video in the case of Premiere Pro is "real-time editing" - Something that Linus didn't benchmark. The Puget Systems 1070/1080 article briefly touches on this with Render Previews but not compared to any AMD cards.

And the Puget review showed that the new cards aren't better than the old ones really, and Linus's video showed that old cards are comparable to new ones, so why spend a lot of money on new pascal card? Why not go with older 960/970 and see how those compare to 1070?

http://www.dslrfilmnoob.com/2014/04/26/opencl-vs-cuda-adobe-premiere-cc-rendering-test/

He says that opencl and cuda are the same, do you have any sources that show CUDA is superior?

 

[Thala]

We already know Polaris is a terrible miner card due to low compute performance.

100W for 24MHS? That would be the best miner card by far!
You have a very strange view about what is a terrible miner card...like declaring the best terrible.
If i am getting 24MHS for 100W, i am going to buy 4 of these "terrible" miner cards for mining...

 

[sirmo]

100W for 24MHS? That would be the best miner card by far!
You have a very strange view about what is a terrible miner card...like declaring the best terrible.
If i am getting 24MHS for 100W, i am going to buy 4 of these "terrible" miner cards for mining...

Actually under-clocked Fury Nanos are better, but that's because of the efficiency of HBM memory. You can get them to use just 90watts, and the wide memory bus really helps the mining. Which really makes them hard to beat when it comes to MegasHashes/watt.

rx480 has a high frequency GDDR5 VRAM which makes them consume more power and the 256-bit bus is a bottleneck for mining since compression isn't leveraged.

rx480 @ 24Mh/s are still pretty solid but not better than already amazing Fury cards due to HBM.. Vega with HBM2 will be huge with miners though I suspect.

But to get back to the topic, other than perhaps some specific cases rx480 is easily the best card for the OP. It's affordable, relatively power efficient and plenty powerful for what he needs.

 

[Thala]

Actually under-clocked Fury Nanos are better, but that's because of the efficiency of HBM memory. You can get them to use just 90watts, and the wide memory bus really helps the mining. Which really makes them hard to beat when it comes to MegasHashes/watt.

Thats interesting because i currently have Nanos running and i get them nowhere close to 90W with undervolting. How much MHS are you expecting from a Nano/FuryX assuming you get them down to 90W? I currently cannot see my Nanos beating 24MHS/100W, i would like to be proven wrong though 🙂

 

[sirmo]

Thats interesting because i currently have Nanos running and i get them nowhere close to 90W with undervolting. How much MHS are you expecting from a Nano/FuryX assuming you get them down to 90W?

I don't have one, I just remember reading about it, I think it's about the same ~25mh/s

 

[Thala]

I don't have one, I just remember reading about it, I think it's about the same ~25mh/s

Thats a pitty, because i was experimenting quite a bit with underclocking/volting and i am pondering about getting few RX480 for mining. Other than that Nano is a great little card which easily reaches FuryX performance.

ps. Sorry for beeing off-topic here.

 

[2blzd]

Mac's are on AMD now and that's Adobe bread and butter. The industry is moving away from proprietary, get with the program.

Yea, no. Apple has nothing to do with this. Especially when they flipflop GPU brands every other generation. As someone who is in this industry, I can tell you Apple systems are in the minority.

For that you need to connect the dots and make an educated guess, or better just wait for launch/reviews. but i have a strong feeling it will be pretty much same.

And the Puget review showed that the new cards aren't better than the old ones really, and Linus's video showed that old cards are comparable to new ones, so why spend a lot of money on new pascal card? Why not go with older 960/970 and see how those compare to 1070?

http://www.dslrfilmnoob.com/2014/04/26/opencl-vs-cuda-adobe-premiere-cc-rendering-test/

He says that opencl and cuda are the same, do you have any sources that show CUDA is superior?

That was the whole point of this thread. Since the Mercury Playback Engine came out in Premiere, Adobe's official word has been to go with Nvidia cards because of CUDA performance. When OpenCL support was added a few years later, nothing changed. Adobe people, people in community, professionals all said the same thing. Get cuda. But I guess OpenCL support has dramatically improved since it first came out because there were early tests showing it lagging behind CUDA but I can't find them.

Like I said in my OP, it just seems to be what everyone says when in fact I too would like some hard proof. Linus's video was nice but definitely doesn't give the whole picture. I was hoping someone with a deeper knowledge of the subject could chime in.

As far as spending money on an older card, I don't buy old tech. I'm building a brand system from scratch and tend to make them last 4-5 years minimum. A 1070 would be a little bit more future proof than a RX480 as far as Im concerned. And if I want to game, it'd be nice to play them without compromising.

 

[Cookie Monster]

That was the whole point of this thread. Since the Mercury Playback Engine came out in Premiere, Adobe's official word has been to go with Nvidia cards because of CUDA performance. When OpenCL support was added a few years later, nothing changed. Adobe people, people in community, professionals all said the same thing. Get cuda. But I guess OpenCL support has dramatically improved since it first came out because there were early tests showing it lagging behind CUDA but I can't find them.

Like I said in my OP, it just seems to be what everyone says when in fact I too would like some hard proof. Linus's video was nice but definitely doesn't give the whole picture. I was hoping someone with a deeper knowledge of the subject could chime in.

As far as spending money on an older card, I don't buy old tech. I'm building a brand system from scratch and tend to make them last 4-5 years minimum. A 1070 would be a little bit more future proof than a RX480 as far as Im concerned. And if I want to game, it'd be nice to play them without compromising.

It would actually be nice if reviewers could do a deep dive into non-gaming application performances as video cards are becoming much more than just 3D-accelerators. But the hard part is how to actually benchmark these features (such as ones from Adobe Premiere Pro) in a meaningful way..

 

[trane]

I would wait and see. The Pascal cards are unstable with Premiere Pro right now, especially RED media in unusable with constant black frames. In terms of performance, anything R9 290 to 980 Ti pretty much offered the same performance. Heck, even R9 380 was close enough. It is only with very heavy, high effects 4K+ footage that we saw some variation, and in which case Fury X was the top card. Otherwise, CPU and I/O are the most common bottlenecks - not GPU.

I'm willing to bet RX 480 ends up offering pretty much the same experience as 1070. Drivers could be the deciding factor, which is why I recommend wait. Of course, go for the 8 GB variant - will come in handy if you are doing 4K work.

 

[2blzd]

I would wait and see. The Pascal cards are unstable with Premiere Pro right now, especially RED media in unusable with constant black frames. In terms of performance, anything R9 290 to 980 Ti pretty much offered the same performance. Heck, even R9 380 was close enough. It is only with very heavy, high effects 4K+ footage that we saw some variation, and in which case Fury X was the top card. Otherwise, CPU and I/O are the most common bottlenecks - not GPU.

I'm willing to bet RX 480 ends up offering pretty much the same experience as 1070. Drivers could be the deciding factor, which is why I recommend wait. Of course, go for the 8 GB variant - will come in handy if you are doing 4K work.

Yea Ive been reading on the Adobe Hardware forums about Pascal and Premiere problems with RED stuff. Its to be expected with brand new hardware though so I'm not too concerned with that. I don't edit too much RED footage currently anyway, maybe 1 out 5 jobs.

Anyway, thanks for the great insight. That's something to think about. My target was to build a new rig before the free Windows 10 expires at the end of July, so maybe in about 4 weeks. Hopefully AIB cards for both RX 480 and 1070 are out by then as I really don't want to purchase a reference card. I know there are plenty of non-FE 1070s on the way (thats why I was targeting it, $379 isn't so bad) is there any news about non-ref RX480s?

 

[Bacon1]

As far as spending money on an older card, I don't buy old tech. I'm building a brand system from scratch and tend to make them last 4-5 years minimum. A 1070 would be a little bit more future proof than a RX480 as far as Im concerned. And if I want to game, it'd be nice to play them without compromising.

Uhh ok

Getting to the point I was pretty set on getting a 1070 for the price/performance and amount of cuda cores and vram. But since I saw Linus's video, I'm starting to wonder if a RX 480 for $200ish less would be a better buy?

I'm not a heavy gamer. Thoughts?

So what exactly is the point of this thread if you already have your mind up on buying a 1070? It won't have better performance for Premiere than older cards or 480, yet you aren't a heavy gamer and at the same time want to spend a ton on a gaming card... :whiste:

 

[Janooo]

Yea, no. Apple has nothing to do with this. Especially when they flipflop GPU brands every other generation. As someone who is in this industry, I can tell you Apple systems are in the minority.


That was the whole point of this thread. Since the Mercury Playback Engine came out in Premiere, Adobe's official word has been to go with Nvidia cards because of CUDA performance. When OpenCL support was added a few years later, nothing changed. Adobe people, people in community, professionals all said the same thing. Get cuda. But I guess OpenCL support has dramatically improved since it first came out because there were early tests showing it lagging behind CUDA but I can't find them.

Like I said in my OP, it just seems to be what everyone says when in fact I too would like some hard proof. Linus's video was nice but definitely doesn't give the whole picture. I was hoping someone with a deeper knowledge of the subject could chime in.

As far as spending money on an older card, I don't buy old tech. I'm building a brand system from scratch and tend to make them last 4-5 years minimum. A 1070 would be a little bit more future proof than a RX480 as far as Im concerned. And if I want to game, it'd be nice to play them without compromising.

It appears Apple is the reason why OpenCL support is in better shape now than it used to be.
It would not surprise me if somebody tested both and OpenCL came on top.

 

[coercitiv]

Anyway, thanks for the great insight. That's something to think about. My target was to build a new rig before the free Windows 10 expires at the end of July, so maybe in about 4 weeks.

Offtopic but might be useful for the OP:

The free Win 10 expires in the sense that they will replace the available build with one that no longer offers the upgrade option, but I'm confident that using an earlier build will get you upgraded. AFAIK they have no way to do that for offline installs (install original media > activate online > install Win 10 offline with a build prior to expiry date) Not only that, but I have just successfully "moved" a Win 10 upgraded installation from one system to a new one with no shared components, by installing the original OS and performing the upgrade a second time.

Whatever your choice of time and product is, Win 10 upgrade path is not a limiting factor.

 

[2blzd]

I don't have my mind made up. Linus's video and some newer articles has given me some doubt, hence this thread. Trane had some good insight and was hoping to get a discussion going about things I wasn't aware of. That right there is worth it alone for me.

I don't have to be a heavy gamer but to want to game without compromise when I do. They aren't mutually exclusive. For example. I don't need ultra settings but I would enjoy playing at native resolution @ 60FPS, which I currently can't do (2560x1600).

Why do you care so much ? If you don't like this thread move on.

 

[HiroThreading]

It sounds like you're already pretty set on the CUDA environment. Many users here have presented fairly reasonable arguments as to the benefits of OpenCL, but you've swatted them away without much consideration.

To be frank, the current state of the situation is in favour of CUDA. Adobe recommends it, many of Adobe's users/developers recommend it and CUDA support is a lot better than OpenCL support. No doubt about that. But this is all fairly natural when a vendor with as much clout as Nvidia comes out in support of its own in-house API.

However, there is no argument when it comes to OpenCL's momentum in the industry. Obviously, AMD is making a comeback -- it looks like Apple, which is a major supporter of OpenCL and open standards, is firmly sticking to Radeons and FirePros. But most importantly, Intel (remember those guys?) is making a push with its iGPUs and OpenCL support and compute performance.

If the programs you work with are better supported under CUDA today, then by all means invest in a GeForce/Quadro card. In a couple of years however (or maybe even next year?) expect to move towards OpenCL.

 

[renderstate]

Apple is a major supporter of open standards?!? How do you come up with this stuff? Even rocks know Apple is killing OpenCL in favor of compute on Metal and they are transitioning away from pretty much any open standard to their own APIs for gfx and compute. Bye bye OpenGL, OpenCL, etc.

 

[flopper]

480 is the better choice vs any Pascal card.

 

[renderstate]

480 is the better choice vs any Pascal card.

You won! Time to close this thread and move on, nothing left to discuss 🙂

 

[HiroThreading]

Apple is a major supporter of open standards?!? How do you come up with this stuff? Even rocks know Apple is killing OpenCL in favor of compute on Metal and they are transitioning away from pretty much any open standard to their own APIs for gfx and compute. Bye bye OpenGL, OpenCL, etc.

...huh? It's understandable if you hate Apple because... they're Apple. 🙄 But your hatred for the company shouldn't get in the way of facts: Apple has been a huge supporter of open standards in the industry.

On the topic of Metal, Apple is simply responding to recent developments -- DirectX12/Mantle/Vulcan -- in order to improve performance on all of its platforms. Apple's hardware is quite standardised, so it makes sense to adopt APIs that are "closer to metal". Besides, Metal is pretty much based on OpenCL and OpenGL.

Also, Apple has always supported open standards such as OpenCL, OpenGL, WebKit, Darwin, HTML and PCI Express, and will continue to do so in the future. Furthermore, as declared at WWDC last year, Swift (its replacement for Objective-C) will be open.