AMD Roadmaps published yesterday.

Hi.

Normally AMD make a number of announcements and roadmap reveals, togetter with breakdowns for financial analysts and investors. Here is a link with a webcast:

https://ir.amd.com/events/event-details ... t-day-2020

Regards Carsten.

Hi.

AMD held its Financial Analyst Day 2020 yesterday, to update the investment community on both the current state of the company and its future plans for the next 3-4 years. I watched the 3,5 hour long livestream. Here are first some main links to it:

https://www.tomshardware.com/news/amd-f ... a-big-navi

https://www.anandtech.com/show/15597/am ... 20-roundup

And then some sublinks covering some particular areas of it:

AMD have found out that it is difficult to optimize energy efficiently on a GPU to cover both Gaming and Compute. And will therefore introduce CDNA, that is AMD’s new designation for their compute focused GPU architecture and now will be a distinct family separate from AMD’s RDNA architecture and the consumer GPUs.

It was told that Zen 3 will arrive at the end of 2020. And later it was told, that the next gen EPYC CPU's the 7nm Milan will be coming at the end of this year. Personally do I expect the Zen3 based Ryzen CPU's first to arrive early in 2021, as AMD have for several years, promised to keep the AM4 socket until 2020.

The next next gen EPYC CPU's, the 5nm Genoa will be coming by the end of 2022.
Here is some more links:

'New AMD Ryzen and EPYC Roadmaps, Zen 4 and 5nm Genoa by 2022, Zen 3 and Milan in 2020'

https://www.tomshardware.com/news/new-a ... milan-2020

'Updated AMD Ryzen and EPYC CPU Roadmaps March 2020: Milan, Genoa, and Vermeer'

https://www.anandtech.com/show/15594/up ... nd-vermeer


'AMD Says its Upcoming RDNA 2 and Navi 2x Will Boost Performance per Watt by 50%'

https://www.tomshardware.com/news/amd-r ... 50-percent

'AMD's RDNA 2 Gets A Codename: “Navi 2X” Comes This Year With 50% Improved Perf-Per-Watt'

https://www.anandtech.com/show/15591/am ... erfperwatt


'AMD Announces X3D Chip Stacking and Infinity Architecture'

https://www.tomshardware.com/news/amd-a ... chitecture

'AMD Discusses ‘X3D’ Die Stacking and Packaging for Future Products: Hybrid 2.5D and 3D'

https://www.anandtech.com/show/15590/am ... 25d-and-3d


And finally some more links from Anandtech Only:

'AMD Moves From Infinity Fabric to Infinity Architecture: Connecting Everything to Everything'

https://www.anandtech.com/show/15596/am ... everything

'AMD Unveils CDNA GPU Architecture: A Dedicated GPU Architecture for Data Centers'

https://www.anandtech.com/show/15593/am ... ta-centers

'AMD's 2020-2022 Client GPU Roadmap: RDNA 3 & Navi 3X On the Horizon With More Perf & Efficiency'

https://www.anandtech.com/show/15592/am ... he-horizon

Regards Carsten.

Gaming GPU? Compute CPU? Data centre GPU? To which family does the unreleased Radeon Pro W5700X for the Mac Pro 2019 belong? Nice to see the roadmaps for GPUs, but hope the W5700X still makes sense to order for non-gaming use in Resolve and its various page tabs.

To which family does the unreleased Radeon Pro W5700X for the Mac Pro 2019 belong?

The W5700X isn't "unreleased" from AMD's perspective, just from Apple's. It's the same GPU as the existing 5700XT PC-based cards that are out right now, but with twice the VRAM. So: whatever family the current 5700XT belongs to is the same one the W5700X will.

If we ever see it.

Gaming GPU? Compute CPU? Data centre GPU? To which family does the unreleased Radeon Pro W5700X for the Mac Pro 2019 belong?

Hi.

The the unreleased Radeon Pro W5700X is a Gaming Graphics Card.

And now I will give you all the information I have for the coming Compute GPU's.
It is not a single introduction, but a class of Graphics Cards that will be improved from generation to generation.

I expect the first will be the successor of the Radeon Instinct MI50. The rumors call it Radeon Instinct MI100 and is expected in the 2nd half of this year. You can try to google 'Radeon Instinct MI100'.

But QUOTE: 'AMD's plan over the next two generations of products is for the IF ( Infinity Fabric ) to turn into its own architectural design, no longer just between CPU-to-CPU or GPU-to-GPU, and future products will see a near all-to-all integration.' for CPU-to-GPU connectivity.

From: https://www.anandtech.com/show/15596/am ... everything

And AMD's plans go futher. They just don't want to connect CPU-to-GPU with a very fast connection, but also plan to let the CPU and GPU to share the memory and its cache.

In this link is a 'Representation of AMD's diagram':

https://www.anandtech.com/show/15590/am ... 25d-and-3d

QUOTE: ' AMD’s diagrams show four main compute chiplets, arranged in a 2x2 pattern, and then 4-high stacked die with one per chiplet ....... it seems that the ‘die stacking’ element points to HBM or some form of memory, while the compute chiplets in the middle are only one high, but all connected through the interposer. AMD is claiming that this level of integration offers a 10x increase in bandwidth'

Here is a link to a 3:02:28 long YouTube with
'AMD Financial Analyst Day 2020' Recording from the Live stream.

From 2:58:20 to 3:00:12 in the Q&A part of the AMD event. AMD's David Wang, Senior Vice President Engineering explain the benefit of this unified memory.

As I see it, is this or a competing systems the future of Resolve.

Regards Carsten.

My thanks for clarifying the W5700X and this roadmap.

So this improved (Infinity) architecture sharing memory and linking CPU-GPU sounds like it will all depend upon Apple adopting AMD CPU processors and will be the end of life for Intel W processors in the Mac Pro. So maybe a next generation Mac Pro in only two or three years with 10x current bandwidths and on PCIe 4 and TB4 presumably handling 8K video everywhere whether or not that’s really useful and demanded by clients.

Coming back to 2020 reality, I am hoping for some advice or criticism of my intention to delay purchasing the Mac Pro 2019 until the W5700X (with 16GB DDR6 VRAM) can be included. I believe that’s the minimum I need to handle 4K video and it should be significantly less expensive than the older Vega II.

The W5700X includes hardware for processing h.264 and HEVC as I understand it and will match or outperform the older Vega II (32GB HBM2) that costs more but performs better for renders in the Timeline. However I read recently that the T2 chip in the Mac Pro also includes hardware acceleration related to HEVC I believe (but don’t know if it accelerates h.264 too) so maybe I don’t need to rely on the W5700X after all.

I honestly think I’ll be dead or mentally incapacitated before my clients demand 8K deliverables. Setting aside the cost differences for the moment, which card would you recommend for 4K now in the Mac Pro 2019 if there is only one GPU?

Much appreciate the advice as I can’t afford to make a mistake on this given the other demands for my limited budget in 2020. I understand by 2022 or 2023, this current Mac Pro will be sadly outdated, but still capable of managing 4K video.

which card would you recommend for 4K now in the Mac Pro 2019 if there is only one GPU?

I'd go with any Vega card. Radeon VII would be the best.

In the light of the CDNA announcement, I really can't see Navi being any more than mediocre in comparison. Would love to be proven wrong, of course

Sulo Kokki, thanks. That W5700X does have that advantage of compressed video data being sent to the new Pro HDR 6K monitor that then permits the addition TB/USB connections on the back of the monitor to function as intended. The Radeon Pro Vega II doesn’t allow that.

I will take a look at the roadmap re CDNA this weekend.

It's what you need for yourself that counts.

But Navi 5700XT's track record against Vega 10/20 (in Resolve) isn't the best. More on the nose, Vega makes it look like a Polaris refresh. Can't see W5700X being that much better - it will undoubtedly have new features that you personally may need. In that case, sure, get one.

But, be advised, at that point you'll be looking at a trade-off between features and computing power. On the 5700XT, the gap has so far looked formidable. AMD's insistence to segregate their gaming and computing cards does little to alleviate my worries for that gap to be closed any time soon.

These are tough choices in uncertain times, so it's wise to consider one's options.

Thanks again. The recent Barefeats benchmarks of GPU cards compared Vega with the (non-MPX) 5700XT using Resolve and the 5700XT was within a few percent of the single Vega II.

I think I’m just going to wait until Barefeats compared the actual W5700X card which might be better than the 5700XT in the Mac Pro 2019.

https://barefeats.com/mac-pro-2019-gpu-options.html

After that I’ll try to judge if the Radeon Pro Vega II is the better overall solution.

Barefeats ran their Resolve FX tests in HD.

Here's Puget with 4K (Windows system, but should give you an idea).

In my experience, if it quacks like a duck... Of course, one good bench with the W5700X and I stand corrected.

Yes, thanks so much for showing the Puget 4K test is not encouraging for the 5700XT, but with twice the memory and more features and I hope better tuned drivers in the Mac Pro 2019, all these benchmarks need to be updated when the W5700X is available.

Again, I don't mean to bum you down, but...

Navi cards use GDDR5 (or 6) memory. All Vega cards use High bandwidth memory (HBM2). Now, in case you're unaware, the CliffsNotes is that HBM is great for bigger data on wider bus (Resolve), whereas GDDR is roughly the exact opposite (great for gaming).

HBM is industrial catnip for a software like Resolve. Doubling the GDDR mem stack alone won't save the W5700X.

Sulo, thanks for your patience. Appreciate your additional comments. I’ll see if I can manage the budget for a single Vega II card.

Praying you don’t write back telling me I need the Vega II Duo card to get real-time 4K.

Originally that’s what I wanted but several folks have assured me that Resolve won’t use more than the 32GB HBM2 on a single card even though Resolve detects the presence of 2 GPUs. Makes no logical sense that the Duo card can’t use all 64GB of GPU memory, but I am old and have a hard time understanding why 1 + 1 = 1 in the new math; all my life I’ve thought 1 + 1 = 2 at least for this planet and classical mechanics.

No worries, mate. Buying a new GPU specifically for Resolve is a hairy proposition. You have to weight in all you'll need, have to look at what's available, make educated guesses on where the market's going, look out for the best offers... It can get daunting, because as a software, Resolve's a particular animal, with a lot of traffic between the CPU & GPU, heavy reliance on vRAM, etc.

There are variables involved that can be hard to make out at first glance. Add in the fact that not all OS' are alike - Radeon VII works well in a two-card config in macOS and Linux, but not Win10. macOS uses Apple's own Metal API, which allows it to utilize the Vega chips better than other systems. So, a Vega bench on Win10 is lesser than the same bench (with the same specs) on macOS.

As for Vega II's big memory... I, too, think that currently the 16gb Radeon VII is just as good for Resolve as a single 32gb Vega II. As the years pass and resolutions scale up, bigger GPU memory (or even, a shared RAM pool with the CPU) will make a difference. But not today.

Personally, if you're strapped for cash and need an affordable & solid GPU now, I'd recommend a Vega 56. It draws in significantly less power than the 64 and is more cost-effective than the VII. You should be able to pick one at $200-300, and later upgrade to a bigger Vega (or Navi) if you need to.

But again, up to you. I find this whole Vega II business to be muchly hype, and would welcome an Arcturus (Vega 30) at $1000. But AMD seems to have other ideas...

Yes, it’s a little like AMD has drawn some lines in the sand to differentiate their devices going forward and they’re not interested in what they’re done before now. Sounds like a plan for several years and that’s good; but hard right now when things are being disrupted. Same may be said of the Mac Pro 2013 and Mac Pro 2019 and Mac Pro 2022? The 2013 version is still usable even with the obvious advantages of the 2019 version, but my gut tells me the box (all the physical structure except the non-locking wheels) will endure for at least a decade, but the innards are a bridge to a better solution and AMD wants to be that solution in a few years.

With AMD potentially supplying all Mac Pro CPU and GPU including memory and with Apple moving to their own (iDevice) processors for all other Macs, Intel will be gone.

Praying you don’t write back telling me I need the Vega II Duo card to get real-time 4K.

For little it may be worth:

The cost of the W5700X will likely be around half of the Vega II upgrade. So assume it'll be ~$1200 or so. Give or take a few pennies. The GDDR6 and only half of the VRAM of the Vega II will be big contributors to that cut in price, but it is an Apple product, so it will cost more than a stand-alone PC GPU will. Just because: Apple.

The single Vega II processes 6K/60 footage just fine from the Canon EOS C500 Mk II. Playback is smooth and easy-peazy on both a 6K timeline and a scaled 4K one. It's got a lot of grunt for an "old" card; the only place it'll likely fall down in comparison with the W5700X is the encode/decode. It's real-time 4K/60 encoding, where the Navi's encoder should be able to do it in 75% of that time.

Given Apple's tripping around with releasing the MPX modules to purchase after the fact, and because the W5700X is still "Coming Soon", I've lost interest in it. I'll truck along with my Vega II until the bigger/badder/better compute card is produced by Apple. Next year. Or maybe in two years.

Jason, I appreciate the path through the jungle that you blazed so it’s easy for me to follow. On my iMac my h.264 renders (with Intel Quick Sync acceleration) can take 12 hours or more), I ‘know’ deliverables with the W5700X on the Mac Pro are going to be a small fraction of that. We learned a scary less when the sexy Mac Pro 2013 was introduced and h.264 took dozens of times longer than the iMac. I just I’m cautious that the Vega II without h.264 acceleration is going to lag badly.

Do you know how many fps you can create your 2K/HD deliverables with your Vega II on your system? My iMac can sometimes do 10 fps or 5 fps without overheating in the winter. In the summer I’m doing 5 to 2 fps h.264. Using 4K on the Resolve Timeline.

the bigger/badder/better compute card is produced by AMD for Apple. Next year. Or maybe in two years.

Fixed that for you. But why would Apple want to refurbish their MP GPU with an Arcturus semicustom chip, when they can sell you multiple Vega IIs for the next five years?

Originally that’s what I wanted but several folks have assured me that Resolve won’t use more than the 32GB HBM2 on a single card even though Resolve detects the presence of 2 GPUs.

It's because it's 2 GPUs on one card. Resolve uses them in parallel, and is optimized around the being the same. AFAIK unless something's changed recently it farms out frames to each card, so they need to be symmetrical for it so send them efficiently.

Rick -

Do you know how many fps you can create your 2K/HD deliverables with your Vega II on your system? My iMac can sometimes do 10 fps or 5 fps without overheating in the winter. In the summer I’m doing 5 to 2 fps h.264. Using 4K on the Resolve Timeline.

I just ran through a quick test using a 12:19.56 4K/59.94 timeline that I already had up. I used h.265 output, scaled it to 1080p/59.94, and it exported in 00:04:39. It stuttered during the export due to a my scene transitions having to render. Otherwise, smooth as silk and pretty quick. If I did the math right, that looks like 430.4 FPS? I think?

Fixed that for you.

No ya didn't. I specifically said card not GPU. Apple makes the cards using AMD's GPUs. Read carefully.

But why would Apple want to refurbish their MP GPU with an Arcturus semicustom chip, when they can sell you multiple Vega IIs for the next five years?

That's a good question, and they may very well not. It's possible they haven't learned their lesson with respect to keeping the Pro line up to date, even though they do manage with the laptops and iMacs. We'll see.

I specifically said card not GPU. Apple makes the cards using AMD's GPUs. Read carefully.

If you equal Apple Store graphics products with 3rd party Radeons (Sapphire, Gigabyte, etc), I see what you mean, then.

It's possible they haven't learned their lesson with respect to keeping the Pro line up to date, even though they do manage with the laptops and iMacs. We'll see.

You see my point, and I doubt they have. Apple's consistent in its habits, and has zero sense of its own history, characterized by Jobs' refusal to acknowledge the Apple II team's contributions.

That's why, with AMD's announced CDNA, and the fact that AMD has no real market for the Very Big Instinct card to begin with, the quadruple-Vega II setup seems like Apple's shot for stability.

I just ran through a quick test using a 12:19.56 4K/59.94 timeline that I already had up. I used h.265 output, scaled it to 1080p/59.94, and it exported in 00:04:39... If I did the math right, that looks like 430.4 FPS? I think?

Just using round numbers for simplicity:
60 fps Timeline 12’ 20” or 740” or 44,400 frames.
Rendered in 4’ 40” or 280”.
Frame rate to deliver:
44400 frames / 280 seconds ~ 160 fps

For my typical video:
24 fps Timeline 02 hours and 20 minutes or 8,400 seconds for 201,600 frames.
At 160 fps, video delivered in 21 minutes!

That kind of speed will allow me to add more effects so sure it might be an hour or two to deliver better quality. Fine.

My current iMac, set to render at 5 fps when it is summer, is 11 hours 12 minutes in theory, but, it’s actually slower than at around 12 hours—not much fun if I need to make a change as each run means a day lost.

Even my longest days that could be 10 hours of footage in June may render in an afternoon! Really a big relief if I go with that old Vega II 32GB HBM2. Thanks, Jason.

Hope I’ve done the math correctly.

Hope I’ve done the math correctly.

You did. I must have mis-keyed the calculator somehow to get that unrealistic frame rate. When I do it again, I end up with the same numbers you did.

Math is hard. Let's go shopping!

Sounds plausible. Vega 56 exports 15mins of similar 25fps timeline at about 4:30mins. So Vega II is about twice as fast. Nice card

...
Math is hard. Let's go shopping!

Great idea! I’ll need to defer memory upgrades beyond 48GB in the Mac Pro... might need to sacrifice the Apple internal 8TB SSD... maybe plan for 2022 upgrades with third-party SSD and memory.

Great idea! I’ll need to defer memory upgrades beyond 48GB in the Mac Pro... might need to sacrifice the Apple internal 8TB SSD... maybe plan for 2022 upgrades with third-party SSD and memory.

All kidding aside, buy the default 32GB and upgrade it yourself for pennies compared to what Apple charges. That's what I did. Buying RAM from Apple is basically like flushing hundreds (either CA or US) down the toilet. Don't do it.

As for the system drive, I wouldn't go with the monster 8TB, personally. The 1TB version, which is two 512GB drives in a hardware stipe, is far more than enough for most OS and app installations. Have other drives for your important stuff. Don't put that on the system drive.

Now back to kidding... I can understand "deferring". After renting that camera last weekend, I'm trying to figure out how to cough up $16K for one of my own. And I'll never make any money off it either!

If I don’t buy the Mac Pro, I can get you that for Christmas!

If I don’t buy the Mac Pro, I can get you that for Christmas!

In that case: here's to hoping you don't buy a new Mac Pro this year!

I finally went through some of the articles explaining AMD's roadmap, and I'm coming to the conclusion that Navi (2X, 3X, et al) will be the path going forward for all desktop, workstation, and gaming GPUs. RDNA looks like it's what they're going to make actual video displaying GPUs out of; CDNA looks like it's more akin to NVidia's V100, T100, et al cards. Ones that don't actually have any display ports on the back of them because they exist purely for compute. If I read that correctly, it means CDNA isn't coming to the Mac Pro or any other desktop rig.

If I read that correctly, it means CDNA isn't coming to the Mac Pro or any other desktop rig.

Hi.

Dual Radeon Pro Vega II is just four Repacked Radeon Instinct MI50's with added display outputs.
I wonder, why you don't think AMD can Repack coming CDNA Cards for workstation use again?

But the Radeon Instinct MI50 can run on CentOS 7. The same as Resolve. Here is a link:

https://www.phoronix.com/scan.php?page= ... MI60-Linux

I can't se any reason that the coming Radeon Instinct MI100 can't run in a Workstation either.
You just have to add an extra Graphics card or Decklink for display.

Regards Carsten.

Please read my posts carefully. I choose specific words for a reason.

I wonder, why you don't think AMD can Repack coming CDNA Cards for workstation use again?

I didn't say they couldn't. I said it appears like they're trying to make data center compute cards like the very expensive NVidia Tesla cards (those cost on the order of $20K US, give or take). No one said you can't put those into a desktop machine and buy another card for display output. But that's not what they're intended for.

If AMD follows that sort of path, there's no way Apple would try to package a compute-only GPU into an MPX module for the Mac.

Apologies to Jason, but I think the time is right for me to go with the Radeon Vega II 32GB card in the MacPro. I looked over the roadmap summaries at Anandtech and really the emphasis is heavily on RDNA gaming in the short term at AMD. The architecture differences for CDNA computing better suited to DaVince Resolve are further down the line.

I don’t want to go through anymore pain and suffering with the iMac 2015. I’ll still use the iMac 5K screen for now as my display for the Mac Pro as Apple assures me the Thunderbolt 3 to Thunderbolt 2 cable will interface correctly with the iMac’s TB2 connectors on the back for Displayport purposes. I’ll also use another TB3->TB2 to connect to my Pegasus2 R6 RAID. Promise chose to completely ignore an email I sent them weeks ago inquiring about Pegasus32, so I guess I’ll ignore them for the foreseeable future and use what I have.

If anybody considers this akin to a lemming jumping off a cliff, in other words a complete disaster, send a message very soon as I’m putting in the order shortly.

Apologies to Jason, but I think the time is right for me to go with the Radeon Vega II 32GB card in the MacPro.

Well congrats. But dammit, I wanted that camera!

I’ll also use another TB3->TB2 to connect to my Pegasus2 R6 RAID. Promise chose to completely ignore an email I sent them weeks ago inquiring about Pegasus32, so I guess I’ll ignore them for the foreseeable future and use what I have.

Are you running any of Promise's utilities or drivers with that RAID box? Or just using Apple's built-in? The reason is that Premiere Pro (if you use it at all) is colliding hard with Promise's latest round of drivers. Meaning the NLE just up and crashes; I don't know if Promise has identified why that is, but both companies are quite aware of it. If you don't use Pr, then cool. Carry on and install the drivers. Elsewise I'd try to find some older Promise drivers/utilities to make sure Pr doesn't crash.

Order submitted. I should be fine with my Mac Pro connecting to the Pegasus2. Big relief to get it over with. Bigger relief when it all works!

RDNA looks like it's what they're going to make actual video displaying GPUs out of; CDNA looks like it's more akin to NVidia's V100, T100, et al cards... If I read that correctly, it means CDNA isn't coming to the Mac Pro or any other desktop rig.

They certainly give that impression.

The pruning's been going on for a while now. Vega 10 had RX, Pro and Instinct models. Vega 20 had RX and Instinct. "Vega 30", only Instinct confirmed so far. And that's how it's likely to be, because it's easier for AMD to round the RX & Pro models up with Navi. We may get another big computing card, maybe in the Pro lineup. But that could take years.

Don't think Apple really needs CDNA as such to target content creators. Vega II suffices neatly, as it just happens to get everything done from video work down. Lucky for Apple to be high and dry while AMD goes after the Nvidia gamers.

I'm sure the ensuing cards will have nice ray-tracing, but the price of their Resolve performance is set to go up. Lukewarm is to be the new steep, apparently.

Don't think Apple really needs CDNA as such to target content creators. Vega II suffices neatly, as it just happens to get everything done from video work down. Lucky for Apple to be high and dry while AMD goes after the Nvidia gamers.

I don't see that as the only path. Remember that NVidia basically has three lines of product when it comes to cards. The super-expensive Teslas that are packed full of HBM2, have no display ports on them, and are meant to do AI, compute, path finding, etc. In other words, data center cards.

Then there are the Quadro and "GeForce" (for lack of a better word) lines. With the Titans bridging the gap between the GeForce and Quadros. The thing is: they're made using the same actual GPU chip. They have different amounts of shaders available, and different memory setups (HBM in some, GDDR6 in others). But they're ultimately the same chips.

Basically the way I'm reading AMD's stuff:
CDNA --> NVidia's Tesla cards
RDNA --> NVidia's Quadro and GeForce cards

I expect the next major "compute" card for the Mac Pro will be a Navi RDNA, but loaded down with HBM2(e). But we probably won't see that for a while.

Yes, I got that.

Doesn't change the big picture much, tho. AMD's next big computing card would go against Quadro, as you said. It could a Navi, or a Vega. But I doubt it will be either cheap or soon. And its Resolve performance as opposed to Vega II / Radeon VII could be a hurdle in justifying the cost.

Thanks for all your guidance on this, folks. I’m really a neophyte when it comes to the internal architecture of the GPU cards. I appreciate AMD revealing their Roadmap, but when someone needs to do work in 2020, what’s available now is very good in context.

I think I’ll be happy for many years with the configuration I ordered as long as I stay with 4K video primarily BRAW Q0.

I think I’ll be happy for many years with the configuration I ordered as long as I stay with 4K video primarily BRAW Q0.

Well good for you. I guess I'll have to work on selling my second kidney for that camera now.

I think I’ll be happy for many years with the configuration I ordered as long as I stay with 4K video primarily BRAW Q0.

I did some "fun" experimenting yesterday with some 4K/60 footage from my iPhone 11 Pro. As nice as they are, their sensors are noisy as hell in shadows. I ran a minute clip through Resolve's NR using both temporal and spacial NR; both luma only as the noise didn't seem to show color.

It took the rig 7:40 to export a 56 second clip in 4K/60 with that NR applied. The CPU was basically sleeping, but the GPU was running along at around 95% or so.

I'm sure someone who knew what they were doing (ie: not me) could have made that work a lot faster and produced identical if not better results. But the point is that if you do add a ton of NR to your workflow Rick, the rig will grind to a halt.

Thanks for all your guidance on this, folks. I’m really a neophyte when it comes to the internal architecture of the GPU cards. I appreciate AMD revealing their Roadmap, but when someone needs to do work in 2020, what’s available now is very good in context.

I think I’ll be happy for many years with the configuration I ordered as long as I stay with 4K video primarily BRAW Q0.

Congrats Rick! What configuration did you order? Mine is below in my signature.

... if you do add a ton of NR to your workflow Rick, the rig will grind to a halt.

Good to know and if I had known, there’s still nothing I could have changed yesterday as my budget was already exceeding what I should spend to add the Vega II. I’ve no intention to add another GPU until a few years down the road when something much better is available. The only thing I will increase when I can afford to will be more RDIMM memory.

Good to know and if I had known, there’s still nothing I could have changed yesterday as my budget was already exceeding what I should spend to add the Vega II.

I'm sure selective and far more intelligent applications of NR would have made things a lot easier on my experiment yesterday. I basically NR'd the entire 56 second clip, and I bet I could have gotten away with less. Breaking it up into smaller clips and adding NR to the ones that needed it, etc.

Oh, and I think I've figured out another way to get the camera that I don't need and will never make any money on. ;-P

Congrats Rick! What configuration did you order? Mine is below in my signature.

Mac Pro late 2019 16-core, 48GB DDR4 (6x8GB), Radeon Pro Vega II 32GB HBM2, 8TB SSD, wheels...

The lack of memory will slow my moving into more fusion work, but generally should be fine for Resolve 4K. 2021 I’ll add another 96GB (6x16GB) or 192GB (6x32GB) when I need to do more Fusion or apply more effects and perhaps add a monitor. 2022/2023 I’ll add another GPU and a monitor if not done in 2021.

I’ll be sharing my iMac screen as a monitor for the Mac Pro. Just don’t have over $9K CDN for that now.

If you think the 8TB SSD is bordering on insane, you may be right. But it’s not much more than the 4TB configuration that is fairly popular. I’ve been using 1TB SSD on my iMac since 2017 and that’s woefully inadequate for anything but programs and stuff that mysteriously must reside on the boot drive. So the 8TB SSD will be my home for my current projects with completed projects residing on the external pegasus2 R6 24TB RAID. When I do a feature film or documentary, I’ll have to expand, but only for feature clients. Until then I expect it will be a joy to have all current work on the SSD.

I'm sure selective and far more intelligent applications of NR would have made things a lot easier on my experiment yesterday. I basically NR'd the entire 56 second clip, and I bet I could have gotten away with less. Breaking it up into smaller clips and adding NR to the ones that needed it, etc.

I use Resolve NR, primarily temporal, but very sparingly and often none at all! All depends upon the lighting situation which isn’t an issue most times. So if the occasional deliverable takes another hour on the rough spots, I’ll live with it. Anything I do is gong to be better than my current 4-12 hour renders I think.

Oh, and I think I've figured out another way to get the camera that I don't need and will never make any money on. ;-P

Does this plan include a getaway car driver? If so, you can count on me.

Does this plan include a getaway car driver? If so, you can count on me.

I'll not discuss those plans on a public forum. For... reasons.