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

devon807

macrumors 6502
Original poster
Hey everyone!! I hope you are having a wonderful day! Just wanted to share my Mac Pro project. This AI-Focued build is going to replace my old DL380 G10 with two Tesla P100s:
oldin.JPEG



The old G10 ran great for a while, but I needed better GPUs for fine tuning. The setup ran great for inference. For example, was able to run Qwen32b in RAM and had around 30 T/Ps, but was slow as molasses for fine tuning using HuggingFace transformers on PyTorch but I did want a big increase in inference as well.

The first step in this was to find out what platform would be the best for the build, I had settled on Threadripper, more specifically the 3970x with a good amount of Ram, that solution would require a custom build in a rack mount case, and the case I wanted was the McPrue Apollo R but then I realized, most TRX40 motherboards only have a few PCIe slots for full length GPUs. An alternative would be the real deal 7,1 Mac Pro, as I already own one and the performance is great, has plenty of PCIe lanes/slots, and is aesthetically pleasing. However, even though the Mac Pro has 8 PCIe slots, I may be limited to around 3 based on the 1400W PSU, and the spacing of 2.5 slot GPUs. So either way, If I want to run more than 2/3 i'll have to run them in a PCIe enclosure or Thunderbolt eGPU enclosures. With this in mind, I did look into custom cases/SuperMicro systems, but in my professional experience with SuperMicro I haven't had the best experience in terms of reliability, and the fans on certain GPU servers are insanely loud, some even have proprietary fan connectors, making that option a no-go.

So, first item on the list to acquire is the Mac Pro itself, as my OG Mac Pro is my daily driver, full of Radeon 6800s, 40GBe cards, and PCIe NVMe SSDs. I was able to find a good deal a few hours away, for a 16 Core,128GB RAM,8TB SSD, RP580 7,1 for $1500, which is a good deal for the system.
box.JPEG



Now that I have the system, now I need GPUs. This was a difficult decision to make, there are so many great option, including MPX modules, but for ease of use and widespread success, I chose the 'cop-out' option, the RTX 3090, and two of them. Was able to source one 3090 FE and one MSI 3090

The FE came with a dual 8 pin to 12 pin, and the MSI uses standard two 8 pins.
Also popped in a Intel XL710 to connect it to the 40GBe network.

gpus2.JPEG




After installing everything, I wiped the internal SSD, disabled SIP and startup protection, and installed RHEL 10, as RHEL/CentOS/Fedora are my preferred flavor of Linux, and installed the Nvidia drivers by adding the repo to yum.repos.d. Accidentally installed the Graphical install instead of basic server. But both are detected no issues:
Screenshot 2026-05-30 172508.png

So far, performance is amazing, training time from 4 days to 12 hours on the same dataset, 100+ T/Ps on 72b LLMs. I'm very impressed, I thought I would have to have three to get where I am now.


afterrack.JPEG



So what now?:

I would like to add a 3rd 3090,A 3090 Turbo which is exactly two slots as opposed to 2.5, however I would have to set the power limit on each 3090 to about 225W. Also might look into external PSUs/External PCIe expander enclosures.

Also an going to re-rack everything so both Mac Pros are at the top, and on the top of one another, the gap is ugly lol. I also need to clean the fan, please disregard the cat hair and dust :/
 

Attachments

  • fe.JPEG
    fe.JPEG
    1.2 MB · Views: 73
Very nice to see your setup. A good showcase of the 7,1s extended life, despite claims of it’s demise.

The build intrigues me. Isn’t the RTX 3090 FE card 336mm long? Looks like the fan rack is slightly bent, but nowhere as much as I would have expected.
I guess the rack version allows more wiggle room to accommodate slightly longer cards.
 
Very nice to see your setup. A good showcase of the 7,1s extended life, despite claims of it’s demise.

The build intrigues me. Isn’t the RTX 3090 FE card 336mm long? Looks like the fan rack is slightly bent, but nowhere as much as I would have expected.
I guess the rack version allows more wiggle room to accommodate slightly longer cards.
Thanks! and the 3090 FE is reportedly 313 mm long, but I still had to to slightly bend the fan rack. I think the Mac Pro can safely accommodate up to a 304 mm card.
 
I was wondering about your setup ever since you posted the pic in another thread. Are these R640, 740 servers or another model of PowerEdge?
 
  • Love
Reactions: devon807
Super nice. I'm still on the fence, but I think 4 x 3090 turbo limited to 225W each is going in my rack mount. Plus the 7800XT for dedicated gaming. External PSU. I'm super curious what real world network bandwidth you're getting. Does MacOS natively support the Intel 40Gb NICs? 👍🏿
 
  • Love
Reactions: devon807
Really impressive build! Huge respect for your hard work getting the 2019 Mac Pro running RHEL 10 paired with dual RTX 3090s for LLM fine-tuning and inference—cutting training runtime from 4 days down to just 12 hours is an outstanding achievement, your hands-on experimentation is incredibly inspiring for independent AI researchers like me.
I own another Mac Pro 7,1 configured with Xeon W-3275, 192GB RAM and AMD Radeon Pro 580 GPU, dual RTX A5000 GPUs and I plan to install Ubuntu for my own deep learning & large model inference work. Unfortunately I’ve run into a tricky installation issue: whenever I boot the Ubuntu installer from USB and select the install option after GRUB, the screen goes completely black instantly with no further output, making installation impossible.Have you or anyone here encountered this same black screen problem installing Ubuntu on Mac Pro 2019 with dual RTX GPUs? Any tips or boot parameters you could recommend would be much appreciated!
 
I’ve run into a tricky installation issue: whenever I boot the Ubuntu installer from USB and select the install option after GRUB, the screen goes completely black instantly with no further output, making installation impossible.

What port on the Mac is the display plugged into?
 
  • Love
Reactions: devon807
Really impressive build! Huge respect for your hard work getting the 2019 Mac Pro running RHEL 10 paired with dual RTX 3090s for LLM fine-tuning and inference—cutting training runtime from 4 days down to just 12 hours is an outstanding achievement, your hands-on experimentation is incredibly inspiring for independent AI researchers like me.
I own another Mac Pro 7,1 configured with Xeon W-3275, 192GB RAM and AMD Radeon Pro 580 GPU, dual RTX A5000 GPUs and I plan to install Ubuntu for my own deep learning & large model inference work. Unfortunately I’ve run into a tricky installation issue: whenever I boot the Ubuntu installer from USB and select the install option after GRUB, the screen goes completely black instantly with no further output, making installation impossible.Have you or anyone here encountered this same black screen problem installing Ubuntu on Mac Pro 2019 with dual RTX GPUs? Any tips or boot parameters you could recommend would be much appreciated!
Try it with just a single monitor plugged directly into the 580.
 
  • Love
Reactions: devon807
Super nice. I'm still on the fence, but I think 4 x 3090 turbo limited to 225W each is going in my rack mount. Plus the 7800XT for dedicated gaming. External PSU. I'm super curious what real world network bandwidth you're getting. Does MacOS natively support the Intel 40Gb NICs? 👍🏿
Thanks! The XL710 is natively supported, although MTUs over 2120 aren’t, but real-world speeds are the ball park of 1300/MBs-3130MBs, of course iPerf3 indicates full 40GBe with multi stream enabled but I haven’t see it transferring large files, max is around 3k MBs
 
  • Like
Reactions: Flint Ironstag
Really impressive build! Huge respect for your hard work getting the 2019 Mac Pro running RHEL 10 paired with dual RTX 3090s for LLM fine-tuning and inference—cutting training runtime from 4 days down to just 12 hours is an outstanding achievement, your hands-on experimentation is incredibly inspiring for independent AI researchers like me.
I own another Mac Pro 7,1 configured with Xeon W-3275, 192GB RAM and AMD Radeon Pro 580 GPU, dual RTX A5000 GPUs and I plan to install Ubuntu for my own deep learning & large model inference work. Unfortunately I’ve run into a tricky installation issue: whenever I boot the Ubuntu installer from USB and select the install option after GRUB, the screen goes completely black instantly with no further output, making installation impossible.Have you or anyone here encountered this same black screen problem installing Ubuntu on Mac Pro 2019 with dual RTX GPUs? Any tips or boot parameters you could recommend would be much appreciated!
I had this happen, after GRUB loaded I got a black screen, had to wait about 5 minutes and the installer popped up on the 580 display, just make sure your display is plugged into the 580 and not the A5000s
 
Really impressive build! Huge respect for your hard work getting the 2019 Mac Pro running RHEL 10 paired with dual RTX 3090s for LLM fine-tuning and inference—cutting training runtime from 4 days down to just 12 hours is an outstanding achievement, your hands-on experimentation is incredibly inspiring for independent AI researchers like me.
I own another Mac Pro 7,1 configured with Xeon W-3275, 192GB RAM and AMD Radeon Pro 580 GPU, dual RTX A5000 GPUs and I plan to install Ubuntu for my own deep learning & large model inference work. Unfortunately I’ve run into a tricky installation issue: whenever I boot the Ubuntu installer from USB and select the install option after GRUB, the screen goes completely black instantly with no further output, making installation impossible.Have you or anyone here encountered this same black screen problem installing Ubuntu on Mac Pro 2019 with dual RTX GPUs? Any tips or boot parameters you could recommend would be much appreciated!
The easiest way to do it, would to be to build a kickstarter file, inject it to the ISO, and modify the grub.cfg of the ISO, assuming you know all your SSD and network interface identifiers. Are you planning on using Ubuntu graphically? Or over the network?
 
What port on the Mac is the display plugged into?
I run a direct Thunderbolt 3 cable into the stock Thunderbolt dock built into the case. The catch is my monitor exclusively uses Thunderbolt inputs, and the AMD Pro 580X doesn’t feature a Thunderbolt port at all.
 
Try it with just a single monitor plugged directly into the 580.
OK,I'll try

I had this happen, after GRUB loaded I got a black screen, had to wait about 5 minutes and the installer popped up on the 580 display, just make sure your display is plugged into the 580 and not the A5000s
thanks, I'll try!

The easiest way to do it, would to be to build a kickstarter file, inject it to the ISO, and modify the grub.cfg of the ISO, assuming you know all your SSD and network interface identifiers. Are you planning on using Ubuntu graphically? Or over the network?
I plan to set up Ubuntu’s graphical desktop build, so hooking the display straight to the 580X should fix the black screen? The kickstart ISO tweak route is far too advanced for me—I’m completely new to this.
 
I run a direct Thunderbolt 3 cable into the stock Thunderbolt dock built into the case. The catch is my monitor exclusively uses Thunderbolt inputs, and the AMD Pro 580X doesn’t feature a Thunderbolt port at all.

yup, i would guess that's your problem - the necessary software that reads the return video stream from the GPU's MPX connector, and routes it to the TB port on the IO card isn't provided by your Linux installer.

To understand (approximately) the 7,1 from an architectural standpoint, video goes out to the GPU via PCI, and then to the GPU's onboard TB bus if the GPU has TB ports, or gets returned to one of the system's TB busses via the MPX additional slot, to be routed to a TB port.
 
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.