HD Native Card-Motherboard Compatibility

[Digit Al]

My System

Hackintosh. If this makes you nervous. Don't read.

Asus Rampage 4 Gene. i7 3820 3.6 ghz 4 core.
HD Native card - Madi - RME ADI648 - Aurora - Big Ben Clock.
All works great. PT 11.3.2 8 instances of VePro 400 patches and running 30 instances of Lexicon PCM. All peachy.

Bought a Gigabyte Z390 M Gaming and an i9 9900k Cpu at 3.6 ghz.

Both run 10.14.3 Catalina or 10.13.6

Using the 12.6 HD driver.

With the Asus, Pro Tools CPU meter shows 4% when playing back.
With the Gigabyte, Pro Tools CPU meter shows 16% when playing back.

In Geekbench the Asus clocks at 13k, the Gigabyte at 35k, so the gigabyte is clearly faster. But Pro Tools performance is clearly slower.

Could this be a HD driver issue?

Using Pro Tools with a MOTU PCIe 424 card both computers have the same performance.

Again, exact same SSD with the exact same software, drivers etc.

I can't find it, but it seems to me that at a certain point Avid upgraded the HD native card, and I'm wondering if it's a compatibility issue with the new mother board, an issue with how Gigabyte deals with PCIe vs Gigabyte, or something like that.

The new system is way faster, but if the CPU load is already higher just playing back, I don't think it will do well with the patches playing through VePro on top of it.

Any thoughts?

[JFreak]

Might not be so simple.

Your i7 has 10MB L3 cache for 4 core operation and your i9 has 19.25MB for 10 core operation. As you may know, L3 cache is the "bridge between CPU cores and motherboard" and even if the cores would be many times faster, your motherboard is talking to L3 cache and cannot get any more throughput your cache is offering.

[Digit Al]

Also, I am looking at the specs for each CPU

From Intel's ark website:

"Max # of PCI Express Lanes
A PCI Express (PCIe) lane consists of two differential signaling pairs, one for receiving data, one for transmitting data, and is the basic unit of the PCIe bus. # of PCI Express Lanes is the total number supported by the processor."

Out of all the CPUs I have:

3820 has 10bm cache and the Max # of PCIe lanes is 40. PCIe v2.0

4930k has 12 mb cache and the Max # of PCIe lanes is 40. PCIe v3.0 Scaleability 1S Only

E5-2697 v2 has 30mb cache and the Max # of PCIe lanes is 40. PCIe v3.0 Scaleability 2S Only

BUT the 9900K only has 16mb cache and the Max # of PCIe lanes is 16. PCIe v3.0 Scaleability 1S Only

This would lead me to believe you're correct about the cache, but I think the number of lanes would matter too.

I also just found this link below. The x299 cpus are monstrously more expensive than the 1151 CPUs, but it seems like one can't get the same performance from the 1151 CPUs.

This really is a problem. Time to see what the return policies are for all this stuff.

While I researched it all before buying, I really kind of lucked into the i7 3820. It really has been a powerhouse for me for at least six years, and I've been running all the new legato instrument libraries without a hitch in VePro, which feeds into Pro Tools via AAX plugin, and can record 48 tracks simultaneously.

If you go to the passmark benchmark site, they give the overall scores, but also the single core score. I had always thought that the reason the 4930k at 3.4 ghz and E5-2697 at 2.7 ghz didn't perform as well as the 3820 for my Pro Tools computer was the slower clock speed. And I think that is a factor for the 2697, but the 4930k has less of a cache than the 3820, which I think affects it as well.

This again is why I'm asking if the HD Native card is a PCIe v1, v2, or v3? As you'll see according to the link, the throughput on each successive version of PCIe improved, so the number of lanes might not need to be increased.

I am going to be getting an Asus Prime MB used next week, to see if perhaps Asus handles it differently. But so far, it's not looking good for the 1151 socket set, which is too bad. I liked the mATX form factor and the built in graphics of the Z390 M board.

Now, here's the question, and maybe this ain't the place, but why would the other pcie424 card not show any difference in in the performance meter in Pro Tools?

But Thanks for the info, this answered a question I had wondered about for a while now. :)

https://silentpc.com/articles/perfor...ress-bus-lanes

[JFreak]

Specs like this is hard to find but on top of my head I remember HDN is physically PCIE 1.0 x4 card but only uses x1 lane.

[Digit Al]

Thanks again for the info :)

[JFreak]

No problem. Though the thunderbolt box was a bigger hit than PCIE card as anyone having PCIE most likely bought into HDX

[martthie_08]

interesting thread, just stumbled across this, thanks for posting!

Did you end up getting the ASUS mb working? any findings?

just to clarify - you're not on Catalina, are you? 10.14 is Mojave...

I myself have 10.14.6 running on old hardware (i7 4790K on a Gigabyte GA-Z87-UD7 TH) with HDX.
There's a HD Native PCI card on the way to my place and I'd like to set up a really fast second workstation for mixing, so this sparked my interest!

[Digit Al]

To recap

My System

Hackintosh. If this makes you nervous. Don't read.

Asus Rampage 4 Gene. i7 3820 3.6 ghz 4 core.
HD Native card - Madi - RME ADI648 - Aurora - Big Ben Clock.
All works great. PT 11.3.2 8 instances of VePro 400 patches and running 30 instances of Lexicon PCM. All peachy.

Bought a Gigabyte Z390 M Gaming and an i9 9900k Cpu at 3.6 ghz.

Both run 10.14.3 Catalina or 10.13.6

Using the 12.6 HD driver.

With the Asus, Pro Tools CPU meter shows 4% when playing back.
With the Gigabyte, Pro Tools CPU meter shows 16% when playing back.

NOW

EVGA X299 Micro with an i7-9800X 8 core 16 MB L3 Cache with 44, yes 44 PCI Lanes

So, unlike the 9900k with only 16 which were being used up by the VGA Card, this MB/CPU has plenty.

Still the same problem. With the Gigabyte, Pro Tools CPU meter shows 16% when playing back or even at rest.

It should not be this way. I'm at my wits end. From a CPU speed view, the 9900K is faster as it clocks out at about 35k on Geekbench, where as the 9800X is 30k.

As far as the L3 cache being the issue, I tested the old setup with a 4930k 6 core, which 12 mb L3 cache, and it still ran at 4% with the Asus Rampage IV Gene. So the L3 isn't the issue it seems.

Can the new motherboards really be this bad?

Any thoughts?

[martthie_08]

ok, 10.14.3 is still Mojave though

I've got a new second system set up with Gigabyte Z390 Gaming X and 9700K on 10.14.6 and HD Native PCIe on 2020.3.

Send me a test session and I'll let you know my System Usage.
I don't put too much trust in that meter though, it would be better to set up a heavy session for benchmark purposes. Is there something like this out there for ProTools.

btw, I purchased the 9700K specifically because I could never get ProTools hackintoshs running smooth with hyperthreading, so maybe try turning it off in the BIOS.

[Digit Al]

Hi sorry, It's running on 10.13.6, High Sierra.

Keep in mind also that all three systems are running the same exact system, and the 9900K and 9800X systems both used the exact same drive.

In other words. I installed the system and from that clean install, cloned to the Asus system, and the 9900K system. The 9800X system require no further mod to the install.