AMD Radeon VII

[09872738]

I don't actually care if this situation is nVidias fault or Apples. Even if nVidia were incompetent enough to screw driver development up - which I highly doubt - Apple WILL lose pretty much the whole AI research/development crowd, which is why it is in Apples best interest to come up with drivers themselves just as they did in the past.

Let's not forget that nVidia has been the market leader in years prior up to this date for a reason; and while Radeon VII / Vega / Navi seems interesting nVidia still has the edge. Ignoring and even fighting a feud with the market - and let's face it: technology - leader seems incredible myopic and certainly highly unprofessional.

Apple cannot have any interest in losing the whole AI/ML field; not only would this mean losing a large fraction of this highly innovative field but also the halo effect it causes.

Its just so ridiculously childish. They should be ashamed of themselves...

Apples behaviour on the matter actually makes me think. Not only do they lock their systems down; demand outrageous prices for stuff that's not worth it. They also behave like discussed, unprofessional, apparently equipped with the mental maturity of toddlers. Outright DISGUSTING.

Do I really want to buy stuff from people with that kind of mindset? No. Not really. Really not. Their bragging about how deeply they care about their products and customers has become nothing but a marketing lie - their deeds prove it. They like to appear like standing on moral high ground - yet they are on the verge of misusing their position. Its just disgusting what Apple became.

This is not the Apple - as a company - I switched to 20 years ago. It actually became quite the opposite

 

[cube]

Now Navi is rumored for consoles instead of Vega.

 

[Synchro3]

 

[Reindeer_Games]

This is a board for Mac Pro right? Pro app users on macOS? Good hope so )

So let's be pros and not gamers. Let's talk about pro use.

The Vega64 and Vega VII are marketed towards pro users. Geforce is a gaming card.

"THE WORLD’S FIRST 7nm GAMING GPU"

https://www.amd.com/en/products/graphics/radeon-vii

I'd encourage you to read your sponsor's website. Gamers built this entire industry.

 

[itdk92]

It is not dead. What the Apple toy company does now is irrelevant.

Yeah, right.

 

[DavideDigitalFilms]

I am getting 13 sec in BruceX with my 7970and Radeon VII combined in the 2.93 cMP have not go HW acceleration going yet but I don’t think it makes a difference on PR 422
Anyone else tried any tests

 

[itdk92]

I am getting 13 sec in BruceX with my 7970and Radeon VII combined in the 2.93 cMP have not go HW acceleration going yet but I don’t think it makes a difference on PR 422
Anyone else tried any tests

Have you turned off background render and deleted all generated files?

Your result seem a few seconds too fast.

 

[DavideDigitalFilms]

Have you turned off background render and deleted all generated files?

Your result seem a few seconds too fast.

yes i got 15 with the vega 64 the 7970 makes 1sec the dif

 

[itdk92]

yes i got 15 with the vega 64 the 7970 makes 1sec the dif

Do you export the project from a timeline or from the project in Fcpx browser?

 

[DavideDigitalFilms]

Do you export the project from a timeline or from the project in Fcpx browser?

Timeline why? Each time i do a bruce x test I run a title over it and move the title position so that i know it has to re-render.

 

[itdk92]

Timeline why? Each time i do a bruce x test I run a title over it and move the title position so that i know it has to re-render.

I mean that's not enough, you need to delete all generated render files, used and unused. Do you do that every time you test?

 

[DavideDigitalFilms]

I mean that's not enough, you need to delete all generated render files, used and unused. Do you do that every time you test?

Umm, if you create a bruceX project with a title on it how can the old files be of any use?

 

[itdk92]

Umm, if you create a bruceX project with a title on it how can the old files be of any use?

Not sure it's worth using time discussing. There is a very specific procedure, and it needs to be followed. Period.
Otherwise the results of a specific test setup are useless to the community.

 

[DavideDigitalFilms]

Not sure it's worth using time discussing. There is a very specific procedure, and it needs to be followed. Period.
Otherwise the results of a specific test setup are useless to the community.

Ok sure will do

 

[deconstruct60]

I don't actually care if this situation is nVidias fault or Apples. Even if nVidia were incompetent enough to screw driver development up - which I highly doubt - Apple WILL lose pretty much the whole AI research/development crowd, which is why it is in Apples best interest to come up with drivers themselves just as they did in the past.

Apple did the drivers in the past? ( like the past back to MacOS 1-8 ? ) In Mac OS X ( macOS) Apple did the upper layer of OpenGL (the top 'half'), but Apple didn't do the graphics driver stack all the way down to the lowest levels down the the silicon.

Getting Nvidia hardware up and running on various OS that their GPUs are participating/partnering with is in part a Nviida responsibility. Competency in the subcomponent/subcontractor role is getting done what they are asked to do. Not running a "embrace , extent, extinguish" exercise in bigger moat building. If look over on the Linux environment Nvidia doesn't do "open soruce" driver development. The notion that Nvidia is generously handing over every scrap of IP to Apple to have full optimization access all the way down to the silicon is a stretch.

OpenGL is a much "taller" , bigger graphics stack that something like Metal. Before there was a big kitchen and two groups of chefs could each work apart with relatively narrow interaction zone. Moving o Metal is effectively like moving to a kitchen that is dramatically smaller. That means more coordination and cooperation; not less.

As far as loose all of AI ... ML inference is actually where the value add is. Training something completely detached from inference isn't generally going to be useful. And the notion that can't do any significant ML without being hooked to KW of computational heat sink is way overblown. Data mining for big companies not necessarily the only application.

Real progress on the graphic card for macOS systems means Apple and the GPU vendor being deeply involved. It isn't a one or the other thing. Years ago when some folks who saying "screw Apple.... Mac Pro's 2009-2012 have PCI-e slots so hackery will provide GPUs no matter what apple does going forward on macOS ..." ... well reality isn't so one sided over time.

Apple can't run rogue solely by themselves for an long period of time. Neither can the other side.

 

[itdk92]

Any new tests with the newest 10.14.6 beta and Catalina’s beta?

 

[AlexMaximus]

Hi guys, here is my latest Heaven 4.0 benchmark. I just compared it to a view blogs over at the "mac mini with eGPU" forum.

For everyone, that wonders if it would be better to switch horses and go with a new mac mini and eGPU. Here is the Apple to Oranges comparison with the same settings, decide for yourself.

Mac Mini 2018 Hex core with Hyperthreading and Vega64 eGPU:



Mac Pro 2010 Hex core with Hyperthreading and PIxlas VEGA VII:



Unfortunately, I could not find a Mac Mini with Vega VII eGPU. But I think it shows clearly how strong pixlas propels the 5.1 forward at least in this benchmarks. The political argument.. oh the 5.1 is CPU bottlenecked is getting awfully thin here.

 

[goMac]

Hi guys, here is my latest Heaven 4.0 benchmark. I just compared it to a view blogs over at the "mac mini with eGPU" forum.

For everyone, that wonders if it would be better to switch horses and go with a new mac mini and eGPU. Here is the Apple to Oranges comparison with the same settings, decide for yourself.

Mac Mini 2018 Hex core with Hyperthreading and Vega64 eGPU:

View attachment 845294

Mac Pro 2010 Hex core with Hyperthreading and PIxlas VEGA VII:

View attachment 845295

Unfortunately, I could not find a Mac Mini with Vega VII eGPU. But I think it shows clearly how strong pixlas propels the 5.1 forward at least in this benchmarks. The political argument.. oh the 5.1 is CPU bottlenecked is getting awfully thin here.

This video shows the Radeon VII getting 77.5 FPS off a MacBook Pro + eGPU.

There's a lot of flux there, that video was done with older drivers and performance may have gotten even better. But I'd be careful trying to infer something from testing two completely different GPUs on two systems.

 

[mattspace]

I'd be interested to see the test comparison run in 4k, to get a good proxy for VR performance.

 

[H2SO4]

PRO SUB-FORUM. Mac Pro. Pro users. Pro apps.

Geforce doesn't do 10 bit. You talking like we don't have Geforce cards and don't know. Even the Nvidia sales shill said he only uses Windows has come to post to tell us we should upgrade to $3000 card (with no macos drivers) to get this basic modern feature. Guys....serious. Brand tribalism makes brains look very deficient. It's a very unevolved and unenlightened waste of energy.

Your above post starts......"This is a board for Mac Pro right? Pro app users on macOS? Good hope so )

So let's be pros and not gamers. Let's talk about pro use.

The Vega64 and Vega VII are marketed towards pro users. Geforce is a gaming card.".

Nope.