Advice: i9 9900k Radeon VII or VEGA64 FCPX build

[MjayNL]

Hello guys,

I have some questions that every Hackintosh builder had in the beginning but I couldn’t find some clear answers.
@CaseySJ has a comparable build but some thinks are different.

Let me explain, I’ve been working with Macs since 1992 and spent a lot of money in Power Macs, iMacs, MacBooks and Intel Mac Pro’s.
My current Mac is an iMac 27” Late 2012, 3,4GHz i7 with 32GB Memory and an NVIDIA GeForce GTX 680MX 2GB. This system is still working fine for Photoshop and video editing 1080p and some 4K. But it’s time to buy or build a new system because my media is getting bigger and heavier.

I’m thinking about these components:

• Gigabyte Z390 DESIGNARE
• Intel i9 9900k
• HyperX Predator HX436C17PB3K2/32 32GB DDR4 3600MHz or 64GB
• Samsung 970 EVO 1TB (system) NVMe M.2
• Samsung 860 EVO 2TB (data) SATA SSD
• EVGA SuperNOVA 1000W G3
• NOCTUA NH-D15 cooler
• Sapphire Radeon VII or Sapphire NITRO+ Radeon RX VEGA 64
• abwb 802,11 AC Wi-Fi + Bluetooth 4.0 PCI-Express (PCI-E) bcm94360cd Combo
• Phanteks Evolv X
• iiyama ProLite XB2779QQS-S1 5K 27 inch monitor or BENQ 4K monitor

I read a lot about the graphic cards and about controlling the fan speed etc. My only wish is to work stable with FCPX.
I have 3 questions:

1. Is it true that with OSX 10.14.5 you don’t need to control the fan speed of the Radeon cards anymore?
2. Buy a the Radeon VII or the VEGA 64?
3. Should I go for an iMac i9 instead if I want to be sure of a stable system?

 

[thecrusher]

I'm looking at nearly the exact same build.

The advantage in going with Apple for stability (in my opinion) has more to do with the fact that you get guarantees from Apple than with the actual likelihood of issues, which can happen in either case. Both Apple and the hackintosh community can help you with issues, but Apple will replace hardware, and the hackintosh community will not.

The Radeon VII is a more powerful card without consuming as much actual power. The support is arguably more spotty because it is newer, but should improve (no guarantees).

Don't know about the fan speed.

According to Noctua, you're going to want the NH-D15S for that motherboard.

 

[thecrusher]

I have a question related to this build. It's my understanding that the Designare will use CPU PCI lanes for the M.2 slots, is this true? So if I use both M.2 storage and an expensive GPU, am I only getting 8 lanes for the GPU? Would this have any impact?

Or am I completely misunderstanding this? (I've never built a PC before and, while I am a software engineer, this hardware stuff has never been something I've needed to understand.

My desire would be to have a really fast GPU for when I game, and potentially for some rendering (although I haven't investigated software that supports GPU-based rendering). But I definitely do some projects now and then that would benefit from hi-speed I/O.

And for completeness, if I wanted to run two AMD GPUs with Crossfire, what does that mean for the M.2 storage? That I can't use it with the Designare motherboard? Or that the motherboard would use DMI for those lanes in that case? Or it would just catch on fire and explode?

 

[MjayNL]

Hey,

I bought and build my system, I'm still experimenting with some stuffs far so good.

This is what I got:

• Gigabyte Z390 DESIGNARE
• Intel i9 9900k
• HyperX Predator HX4 64GB DDR4 3600MHz
• Samsung 970 EVO 1TB (system) NVMe M.2
• 2x Samsung 860 EVO 2TB (data) SATA SSD
• Corsair RM1000x
• NOCTUA NH-D15S cooler
• Sapphire Radeon VII
• Phanteks Evolv X
• BENQ SW271 4K monitor

 

[thecrusher]

Well I researched a bit on some of my above questions. Some tests in 2016 showed only about a 1% or less difference in GPU performance in x8 vs x16 lanes. But in extreme cases there was a slight difference and it seems that a tiny percentage of the time you could max out the PCI bus.

In 2018, tests with an RTX 2080 Ti showed that the card is perhaps maxing the bus out a bit more often. Typical tests showed about a 5% difference with some as much as 10%. So it seems we've reached the point where using only 8 lanes on the GPU just starts to matter.

On the NVMe side, the difference is perhaps a bit more, with NVMe cards on the chipset PCI lines perhaps as much as 10% slower than CPU lines. Still though, that's still 6 times faster than SATA SSD cards and thirty times faster than spinning platters.

Does the Designare let you pick which PCI lines the NVMe uses, or does it always choose the CPU lines?