Sidebar Sidebar
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.

MP 1,1-5,1 MP 5,1 Video Card for CAD and 3D

kittonian

kittonian

macrumors regular
Original poster
Aug 4, 2008
178
53
Austin, TX
I work primarily in ArchiCAD and Cinema 4D and currently have a flashed NVidia GTX 780 (6gb) running perfectly on 10.13.6, and driving an LG Ultrawide monitor.

My machine has already been upgraded to a 12-core 3.46ghz with 64GB of 1333mhz DDR3, and a Samsung EVO 850 SSD.

The problems are:

1. ArchiCAD feels sluggish some of the time and I could use faster drawing of heavy lines

2. Cinema 4D in heavier scenes can get really slow

3. I am stuck on High Sierra and as we just purchased the new 16" MBP that runs Catalina, I would like to have all my systems on the same OS, but as NVidia web drivers for any OS past High Sierra have never been available I cannot.

I am looking for a flashed card with full boot support (Apple logo, boot options, etc.), and since AMD seems to be the reigning champ with Apple at this point, I was hoping someone could point me to some options to consider. I definitely do not want to step back in terms of power/performance.

Thanks.
 
kittonian said:
I work primarily in ArchiCAD and Cinema 4D and currently have a flashed NVidia GTX 780 (6gb) running perfectly on 10.13.6, and driving an LG Ultrawide monitor.

My machine has already been upgraded to a 12-core 3.46ghz with 64GB of 1333mhz DDR3, and a Samsung EVO 850 SSD.

The problems are:

1. ArchiCAD feels sluggish some of the time and I could use faster drawing of heavy lines

2. Cinema 4D in heavier scenes can get really slow

3. I am stuck on High Sierra and as we just purchased the new 16" MBP that runs Catalina, I would like to have all my systems on the same OS, but as NVidia web drivers for any OS past High Sierra have never been available I cannot.

I am looking for a flashed card with full boot support (Apple logo, boot options, etc.), and since AMD seems to be the reigning champ with Apple at this point, I was hoping someone could point me to some options to consider. I definitely do not want to step back in terms of power/performance.

Thanks.
Click to expand...

TBH, there is no choice for you but MVC RX580.

You can't downgrade your 16" MPB. So, Catalina is a must.

No high end flashed Nvidia card works in Catalina.

On the AMD side, only RX580 is stronger than your GTX780 6GB, and can be flashed at this moment. All other GPU that can show boot screen and supported in Catalina are weaker than your GTX780.

Anyway, I assume that you know you can run Catalina with your GTX780. You don't need to upgrade your GPU.

Of course, CUDA won't work, but CUDA won't work on AMD card anyway. So, that won't make any difference.

RX580 should compute better than GTX780, but you better find out if your workflow is really GPU limiting at this moment. I've seen many guys who keep looking for better GPU but they are actually CPU limiting. End up purchased a new GPU, but helps nothing to his workflow. Those "slow motion scenes" still slow.
 
  • Like
Reactions: Flint Ironstag
That's what I am finding as well. The Apple EFI RX580 (8gb) Pulse from MVC looks like the best option. Unfortunately, while I "can" run Catalina with the NVidia card, it's going to be severely handicapped as there are no true drivers available post High Sierra. That's why I am looking at other options. I not only need to run the OS, I need the video card to perform to its fullest potential so that my redraw rates increase (something that even with the latest NVidia web driver doesn't seem to be happening with my current setup).

I originally bought the NVidia card because of CUDA but it turned out that since VRay is all CPU based and they still haven't got the GPU feature-set right after all these years, I never really used CUDA anyway. Plus, with ProRender, the AMD card is being used, so there are options.

I doubt that with a 12-core machine I am being CPU limited. This thing is really fast. It is just the redraw rates when I have tons of lines in ArchiCAD or really complex scenes in Cinema 4D.

I e-mailed MVC about all of this as well, but my main question prior to making the purchase is whether or not the power supply in this MP 5,1 is going to allow for the full potential of this new card, without modifying and splicing the power.
 
An RX580 pulse will work off the two mini 6 pin connectors.
Why do you need boot screens?
 
kittonian said:
That's what I am finding as well. The Apple EFI RX580 (8gb) Pulse from MVC looks like the best option. Unfortunately, while I "can" run Catalina with the NVidia card, it's going to be severely handicapped as there are no true drivers available post High Sierra. That's why I am looking at other options. I not only need to run the OS, I need the video card to perform to its fullest potential so that my redraw rates increase (something that even with the latest NVidia web driver doesn't seem to be happening with my current setup).

I originally bought the NVidia card because of CUDA but it turned out that since VRay is all CPU based and they still haven't got the GPU feature-set right after all these years, I never really used CUDA anyway. Plus, with ProRender, the AMD card is being used, so there are options.

I doubt that with a 12-core machine I am being CPU limited. This thing is really fast. It is just the redraw rates when I have tons of lines in ArchiCAD or really complex scenes in Cinema 4D.

I e-mailed MVC about all of this as well, but my main question prior to making the purchase is whether or not the power supply in this MP 5,1 is going to allow for the full potential of this new card, without modifying and splicing the power.
Click to expand...

The Apple driver in Catalina for GTX780 is a true driver. The card should work as expected. Same as any AMD card. Web driver is just an alternative, not necessary make the GTX780 run better.

AFAIK, Radeon ProRender can use any GPU. AMD GPU isn't a requirement.
AMD-Radeon-ProRender-Autodesk-3ds-Max-2017.png


TBH, for cMP, even 12 cores, 99% software still CPU limiting. Your workflow no necessary CPU multi thread limiting, but highly likely CPU single thread limiting. All cMP suffer from this problem, our CPU simply too old. Even the GPU driver performance is CPU single thread limiting, which means, our GPU performance can also be CPU limiting.
 
Last edited:
Thanks. I know it "will work" off the two mini 6 pins joined together into a combined 8 pin, but as the other power port on the card (6 pin) isn't being connected, I am wondering whether or not I will achieve the full performance of what this card can offer.

As to why I want boot screens, this is a mission-critical office machine and if I need to access boot options to solve an issue, I need to be able to do so.
[automerge]1575737472[/automerge]
h9826790 said:
The Apple driver in Catalina for GTX780 is a true driver. The card should work as expected. Same as any AMD card.
Click to expand...

You have to install the proper web driver in order to gain full support of the video card. Otherwise it defaults back to the default mac os graphics driver. Even in El Capitan, Sierra, and High Sierra this is the case. There is no built-in driver for this video card, regardless of whether or not CUDA is being used.

https://devtalk.nvidia.com/default/...drivers-be-released-for-macos-mojave-10-14-/1
 
Btw, love the chart. What I really wanted was the Vega card but it seems for that one I do need to mod the power supply and there is no one providing an Apple EFI for it to work in a 5,1
 
Just to let you know some idea about "how GPU limiting by CPU"

This is how my 1080Ti perform on my cMP.
Native.JPG


And this is how my same 1080Ti perform on my Hackintosh.
Heaven.PNG


Same GPU, same OS, same benchmark, same settings. CPU never run into 100% (multi thread) in both case. However, on cMP, the CPU single thread performance is limiting the GPU's performance.
[automerge]1575738035[/automerge]
kittonian said:
You have to install the proper web driver in order to gain full support of the video card. Otherwise it defaults back to the default mac os graphics driver. Even in El Capitan, Sierra, and High Sierra this is the case. There is no built-in driver for this video card, regardless of whether or not CUDA is being used.

https://devtalk.nvidia.com/default/...drivers-be-released-for-macos-mojave-10-14-/1
Click to expand...

The build in driver is a full and proper driver for GTX780, got it?

It's a full driver, same as web driver. OpenCL works, OpenGL works, Metal works. It's been like that since the very first Kepler card supported in OSX.

You can try to run some benchmarks, remove the web driver, then run the same benchmarks again. Most likely it perform more or less the same.

The only function that need web driver is CUDA, and you never use it.
 
Last edited:
  • Like
Reactions: crjackson2134
Interesting. So maybe there's no point in this at all and I should just buy one of the new Mac Pro machines when they are finally released?
 
kittonian said:
As to why I want boot screens, this is a mission-critical office machine and if I need to access boot options to solve an issue, I need to be able to do so.
[automerge]1575737472[/automerge]
Click to expand...

Then you don't need this function daily. Just keep your current flashed GT780 safely in the drawer. And upgrade to Radeon VII or Vega as you wish.

Only on that "bad day" (if that really happen", then spend 5min to swap the GPU and fix it with the boot screen.
[automerge]1575738464[/automerge]
kittonian said:
Interesting. So maybe there's no point in this at all and I should just buy one of the new Mac Pro machines when they are finally released?
Click to expand...

If budget allow, of course the 7,1 is a better option. The cMP may have higher performance to cost ratio, however, the cMP has no way to match the 7,1's performance. Especially CPU single thread performance.
[automerge]1575738582[/automerge]
For multi thread performance, a cMP with dual X5690 can match the 8700K. However, the single thread performance is night and day. This is the problem. And more than 90% software out there still very very very CPU single thread limiting.
8700K single.png
 
  • Like
Reactions: crjackson2134
Anyway, this is what happen when CPU single thread limiting.
Screenshot 2019-12-08 at 1.15.58 AM.png


As you can see, the CPU load is low (overall), but only one thread run just a bit above 100%.

Some people confused that only one core will be use (when CPU single thread limiting). That's not the case in general, the CPU will balanced the load. Also, single thread limiting only means the "the next step can't be started until the last step is completed". However, there is no rule about "both steps must be done by the same CPU core".

The CPU can handle millions of calculations in a second. It can be core 1 do step 1, core 2 do step 2, core 3 do step 3.... And eventually, you can see all cores work to 20-30%, but you are already CPU single thread limiting (because core 2 cannot start to do step 2, until core 1 finish step 1).
 
Last edited:
  • Like
Reactions: LoopDotDrifter
kittonian said:
Well this is disheartening. Doesn't look like AMD is holding a candle to NVidia when it comes to ArchiCAD performance. The Quadro P6000 is the best performing card for large projects with high poly counts according to Graphisoft's recommendations and testing. https://helpcenter.graphisoft.com/knowledgebase/87338/

Even the RX580 is outperforming the Radeon Pro cards.
Click to expand...

That's expected. P6000 is a high end Nvidia GPU. However, that's a Pascal card, wont' work in Mojave or Catalina. So, if you don't mind to switch to Windows 10, that's not your option. But if you are happy to switch to Windows 10, then you don't need a Mac Pro.

For AMD, especially in macOS, workstation cards won't work better than gaming cards. Workstation card basically means "same GPU but lower clock speed". Therefore, it can never run faster than gaming cards.
 
kittonian said:
Well this is disheartening. Doesn't look like AMD is holding a candle to NVidia when it comes to ArchiCAD performance. The Quadro P6000 is the best performing card for large projects with high poly counts according to Graphisoft's recommendations and testing. https://helpcenter.graphisoft.com/knowledgebase/87338/

Even the RX580 is outperforming the Radeon Pro cards.
Click to expand...

The tested Radeon Pro cards are missing the WX9100 ( Vega 64 with 16GB VRAM) and the new W5700 ( Navi and 8GB VRAM). The W5700 has about twice the memory bandwidth of the WX7100 ( which pragmatically replaces) and the 16GB VRAM will be a better match the to bulk caching that the 24GB of the P6000 is doing. They probably aren't going to top all of the Nvidia options, but there are options that are better than the 580X ( taking $/performance off the table ; long term support may be factor is still rocking a 780. ).

Both of those probably should outperform the RX580. ( the 'consumer' 5700 and the WX9100 work in eGPU set ups https://support.apple.com/en-gb/HT208544 , so pretty high chance they will also work inside the new Mac Pro also. )
 
deconstruct60 said:
The tested Radeon Pro cards are missing the WX9100 ( Vega 64 with 16GB VRAM) and the new W5700 ( Navi and 8GB VRAM). The W5700 has about twice the memory bandwidth of the WX7100 ( which pragmatically replaces) and the 16GB VRAM will be a better match the to bulk caching that the 24GB of the P6000 is doing. They probably aren't going to top all of the Nvidia options, but there are options that are better than the 580X ( taking $/performance off the table ; long term support may be factor is still rocking a 780. ).

Both of those probably should outperform the RX580. ( the 'consumer' 5700 and the WX9100 work in eGPU set ups https://support.apple.com/en-gb/HT208544 , so pretty high chance they will also work inside the new Mac Pro also. )
Click to expand...

I agree that WX9100 should be included. That page last update on August 26, 2019.

However, the result still looks strange to me. WX8200 shouldn't be that weak. And WX7100 should be just a bit slower than the RX580. I wonder if the Radeon Pro driver (Windows 10) is working properly.
 
I’m in the same search for a graphics card. If you can upgrade to a 7.1, go for it. It would be a huge upgrade for you.

If you want to stick with your 5.1 for a little longer, I would invest in a card that I could transfer to a 7.1 in a couple of years ahead.

I would get a cheap boot graphic card to keep around and a Radeon VII or Vega frontier edition or Vega 64 as my main card.
 
I am using RX580 in a MP 5,1 dual X5680 3,33GHz. It's usable. But the system does (did?) have a weak point with 2D draw performance. I don't think you could improve that performance only with a better GPU. Just as explained above, it just must be CPU single core limiting. ArchiCAD is quite good at using all the cores, not only with rendering (which uses them all). It uses the available processor cores at background for updating the multiple view/layout/detail tabs open at any time. But it does not use multiple cores like that for generating 2D viewport. And that's too bad. We have no fix for it with cMP's I am afraid.

In 3D window/3D work you get very good performance with almost any decent/recent GPU (Like the RX580). You don't need the most powerfull card to work in 3D . It's most of the time really smooth and there is no waiting there. It's the 2D, I am feeling the same pain every day. All the elements and their detail redraw must depend on CPU single core speed, that's what I think.

Which version of ArchiCAD you are using by the way? Release 23 had a promise to reach even more cores for performance. I have used AC21 lately, and now 22 recently. I'll move to 23 really soon, but at the time I can't say if there is any 2D performance increase to be seen with AC23 with 2D draw. I have AC23 installed, but it has to mature before using it for production - an old habit of mine, and others too I believe.

I've got Cinema 4D too. ProRender hasn't been the real savior with me, at least not yet. I do hope it delivers some benefits later on, I mean regarding what there is to come with GPU rendering developments.

OT/
I myself can't justify the price of a Mac Pro. I now have a side project, a Z440, and I'm upgrading it towards it's higher end specs at the time (E5-1680v3, 128GB RAM, NVME, nVidia GPU). I think I have to try windows properly, not just tinkering with it a little bit like I've been doing for a couple of years now.

Almost all my necessary software is now cross-licenced, so no big monetary hit there:
ArchiCAD (cross licence with dongle or softkey or floating with licence server)
Cinema 4D (kind of cross licence, can run 1 with mac and 1 with PC)
Rhinoceros 6 (from now on it's cross licenced with Cloud Zoo)
Solibri (floating cross licence)
LibreOffice (free for every platform)

Only Affinity products has to be purchased for windows separately (Photo, Designer, Publisher). That's gonna be something like 100€-150€.
But today and tomorrow I am still using my Mac Pro 5.1 with Mojave at my home office. And it still does it's job, i'd say.
/OT, sorry for that..

I've been thinking a Vega56/64 upgrade too, but as I said, I don't believe it brings speed to be noticed with AC 2D performance. At least not that much.
 
I have thought about that too. I allready took notice of tonymacx86.com for example.

It's just that I've got this architectural business running fine right now, and would like to be on the 100% legal side of the fence. Just want to keep everything running and by the book as far as possible.

Wouldn't mind doing something like that as a hobby though, that would be perfectly ok. Who would even care.

Anyway, Thank You very much. I might want to ask something later, for my next hobby I guess ;)
 
  • Like
Reactions: Flint Ironstag
You must log in or register to reply here.
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.
Top Bottom
Back