Gigabyte X299X - Catalina Support

[TheBloke]

I thought I should feel happy after fixed my mobo boot up and run normally and all setting seems settled! But It doesn't react like a 14 core machine when working on CLO3D program! Every task I apply need to wait for finish even worst than my last 4 core hack! Wondering anyone here like mine? Does it cause my main system on M2 sit in M2M shared bandwidth with GPU? My Cinebench CPU (Multi-Core) result 17484 / Geekbench 11006 isn't good or bad?
Hope someone can guide me out please!

I'm not really sure what's happening. Your CPU benchmark results sound reasonable. I'm currently getting 23202 from Cinebench R23, which is 1.33x times your result. I have 1.28 more cores plus an overclock, though not a huge overclock at the moment - I tuned down my overclock temporarily and also have the max temperature set to 90C which is thermal throttling me on my AIO.

So your Cinebench score sounds reasonable relative to mine. My current Geekbench 5.4 CPU score is 17456 which is 1.6x times your score. I imagine here my overclock is making a bigger difference, as Geekbench doesn't run hard/long enough to hit any thermal throttling. But your score still sounds reasonable for a 14-core CPU without an overclock. (My highest Geekbench score in past testing was 19063, but that included a higher CPU overclock plus a RAM OC with tightened timings.)

You have a 48-lane CPU, meaning an NVMe in slot M2M will reduce PCIEX8_1 (slot 2; second down from RAM slots) to x4 speed. So if you have a GPU in PCIEX8_1 then yes it might perform badly - it'd be like running it as an eGPU .

But surely you have your GPU in PCIEX16_1 (slot 1) or PCIEX16_2 (slot 3)? There's no resource sharing with the x16 slots, so you can run 2 x full x16 GPUs at once on this board, regardless of NVMe usage.

However it seems like this CLO3D software doesn't have any GPU acceleration on macOS, it's CPU only? (I Googled and their website says it only supports NVidia CUDA acceleration.) In which case I have no idea why it would perform badly given your CPU benchmark results seem fine.

In case there are other apps you're using which do use GPU, what are your Geekbench Compute GPU scores in Metal and OpenCL modes? My Vega 64 gets around 64000 in the Metal benchmark. I'd assume your 5700XT should be a bit higher than that.

Based on what you said I can't see any obvious problems. As long as your GPU is in one of the x16 slots you should get full performance out of it, and your CPU benchmark results sound OK.

Is it just CLO3D that performs badly? What about other software? Maybe there's some weird issue specific to that software.

Do you also have a Windows installation? If so, I'd do some testing in that also, to get a comparison. Install some of the software you use in Windows, and install some Windows monitoring software like HWInfo and check your core temperatures and other CPU/motherboard stats while running some high loads; maybe you're thermal throttling due to insufficient cooling?

You can also check temps in macOS using HWMonitorSMC2 - and also install Intel Power Gadget so that HWMonitorSMC2 can read all values from the CPU.

 

[TheBloke]

What CPU cooler do you recommend with the i9-10980XE?

I currently have the Coolermaster Masterliquid ML360R and that's the only modern cooler I have experience with at the moment. It's a decent AIO which has enabled me to overclock up to 4.6Ghz on all cores, albeit with thermal throttling in various situations. I got it because it was #1 in Tom's Hardware's 360mm AIO tests. It's an RGB cooler so it comes with all the RGB junk, but you can just not connect it and it's no issue.

(I was meant to be upgrading to a full custom water loop - the best option for a power-beast like this CPU - but 3 months later I'm still waiting for delivery. Never buy anything from watercool.de!)

The 10980XE's TDP is 165W, however with power consumption limits removed - and still at stock speeds - I believe it will draw more than that. Therefore I'd say that a 360mm AIO is definitely the preferred option if it's possible.

So if that rack case that byteminer mentioned can take a 360mm radiator I'd say go for a 360mm AIO. You should then be able to remove the power limits and run at stock speeds with no throttling, and also have a bit of headroom for an overclock if you should wish.

If it proves hard to fit a 360mm then you could consider a 240mm instead. That might still be fine at stock speeds, I'm not sure, but I think it'd definitely not be suitable for any overclocking.

I can't offer any advice on specific coolers as I currently only have experience of the ML360R, so I'd just say check the reviews and comparisons and find something that's highly regarded and will fit in your chosen chassis.

 

[BerndVP]

ok, AIO it will be then.
For my game rig I have 9900k overclock to 4.9 all core, rtx 2080 super all watercooled with an EKWB loop 1x 360 rad and 1x 120 rad

 

[TheBloke]

I have the F3e version, but I don't know if there are improvements compared to before, I have the same bugs that are currently also present on F3c.

Hey @LeleTuratti I'm still really curious about this BIOS F3E you say you have for the Gigabyte Designare 10G? Could you upload it here and give more details about where you found it? If there is another BIOS version after F3C I'd love to re-test some of the overclocking bugs I've experienced.

 

[BerndVP]

Ordered the board, it's gonna be a step by step build, so now the board, then cpu then ram

 

[LeleTuratti]

Hi, sorry for the late reply, but these days I am really very busy.
Fortunately, currently I have solved almost all the problems with this card.

The cause of the random KPs was a Quirks that was left over from the sample plist and I forgot to disable (it probably was: DisableRtcChecksum).

Several of us had this problem some months ago. It turned out that the SBUS SSDT that dolgarrean provided in the first post of this thread was blocking sleep. This can be diagnosed by checking the system logs for messages related to sleep, and looking for a line like "Sleep prevented by..".

Yeah thanks!! This solved the sleep problem!

This sounds like the good old time offset issue.

I solved it by removing all SSDs and HDDs, running a clear CMOS and with a clean install of Windows, then I set the UTC time (again).
For some reason, in fact, the time of the bios was freezed before, now the seconds run normally. I do not know the cause, perhaps after some KP.

Anyways, digressions aside, for your issue, since you're using Clover, the fix for this issue is fairly straightforward — switch to my patched fork of OpenCore.

I don't have much experience with Opencore, so I preferred to first find stability with a bootloader that I know well.
I can confirm that now everything works with Clover except the USB boot bug that happens every time you boot the system with a USB connected.

Clover 5131 X299X Designare 10G
I do not recommend to use it (USB boot bug), but for those interested I attach my EFI with Clover 5131.
It is ready to use, just modify TSCAdjustReset.kext according to your CPU and configure the SMBIOS.

OC 0.6.7
Yesterday I started the migration to OC starting from the @JTR version updated by @TheBloke (Thanks, it was a very helpful starting point).
I am still making the final adjustments in the few free minutes I have, but I think I have almost everything set up, except for a few cosmetic details.
When the work is done maybe I can upload the new EFI OC configured for those who, like me, do not use the integrated Intel wifi card and have an AMD RX 5000.

Where did you get this F3E BIOS for X299X Designare 10G? Do you have a link, or can you upload it here?

I can't find anything about F3E via Google. Latest on the official website is still F3C, and I don't see anything on the Gigabyte forum about any later BIOS'.

I got this BIOS from the Gigabyte team after pointing out to tech support that the whole web is talking about X299X card bugs.
Unfortunately this F3e beta BIOS has the exact same problems as the F3c version (USB boot bug, RAM bug, etc) and I haven't noticed any overclocking improvements.
For those who want to test this F3e beta, it is attached at the bottom of the post.


I am left with only one doubt, now that everything seems stable and working, in the coming weeks I will be re-assembling this build to include a custom loop, so with this opportunity I can patch the Thunderbolt firmware.
However:
- Is the firmware patch overwritten and lost with a BIOS update?
- Currently Thunderbolt already works, even with the hotplug. The only "flaw" is that devices connected at boot (like my TB Dock for example) must be disconnected and reconnected to be usable. This also happens in a Windows environment. It's normal? Could the cause be Apple TB3 to TB2 adapter?

 

[TheBloke]

For those who want to test this F3e beta, it is attached at the bottom of the post.

OK great, thanks. The bug I most care about right now is the VCCSA reset issue, where VCCSA goes back to Auto after every shutdown and on wake. Probably the issue is still there, but I will try F3E just to be sure.

I am left with only one doubt, now that everything seems stable and working, in the coming weeks I will be re-assembling this build to include a custom loop, so with this opportunity I can patch the Thunderbolt firmware.
However:
- Is the firmware patch overwritten and lost with a BIOS update?
- Currently Thunderbolt already works, even with the hotplug. The only "flaw" is that devices connected at boot (like my TB Dock for example) must be disconnected and reconnected to be usable. This also happens in a Windows environment. It's normal? Could the cause be Apple TB3 to TB2 adapter?

A few days ago I got my first TB3 device - the CalDigit TS3 Plus Thunderbolt 3 dock - so I now have a little Thunderbolt experience.

I've not yet flashed the TB firmware because I don't need it for the dock, and I'm not sure I'll ever need the extra features it brings. However I am certain that flashing the firmware is entirely separate to the BIOS. To flash the firmware you need to dismantle the board and flash a Winbond chip directly on the motherboard. It's not going to be affected by a BIOS flash.

As for the second issue, I had the same problem and it's easily solved: In the BIOS -> IO Ports -> Thunderbolt, enable "Thunderbolt boot support". You will no longer need to plug/unplug TB devices on boot, they will be auto-detected in both macOS and Windows. This also does not require the TB flashing.

The Thunderbolt is working pretty well, even without chip flashing, thanks to the Thunderbolt SSDT we are using, provided by dolgarrenan and based on CaseySJ's extensive Thunderbolt work.

 

[BerndVP]

If you don't need the Thunderbolt Bridge link and don't need the 'Thunderbolt BUS' for specific audio cards, then you're good to go with no flashing.

 

[LeleTuratti]

The bug I most care about right now is the VCCSA reset issue, where VCCSA goes back to Auto after every shutdown and on wake.

Are you referring to when the overclock settings are ignored and the motherboard boots with RAM @ 2133 MHz and CPU @ 4600 MHz? Yes, the bug still appears to be present.

As for the second issue, I had the same problem and it's easily solved: In the BIOS -> IO Ports -> Thunderbolt, enable "Thunderbolt boot support".

With Clover this setting enabled causes a KP when connecting a TB device.
Now with OC I activated it and the devices connect to boot normally.

However, there are pros and cons:
Thunderbolt boot support ENABLE: Devices connect and are hotpluggable only if at least one is connected and powered on at boot (A TB Dock is perfect for this). If no device is connected at boot the Thunderbolt ports will not work.
Thunderbolt boot support Disable: Thunderbolt ports always work, but the devices do not connect automatically on boot, you need to disconnect and reconnect them.
I currently keep the option enabled because I have a Dock which solves the problem.

If you don't need the Thunderbolt Bridge link and don't need the 'Thunderbolt BUS' for specific audio cards, then you're good to go with no flashing.

Yes, I use Thunderbolt for my UAD audio devices.
On system information the bandwidth is 2.5GT/s. Can I get faster speeds or latency improvements with the firmware patch?



One last problem I have been having for a long time is that sometimes the desktop background stays black after boot. The wallpaper settings are locked and do not reset the wallpaper.
The only solution is to logout and then logon, or use the terminal command "Killall Dock".
Seems to be a Catalina glitch, has anyone ever found a solution?

 

[TheBloke]

Are you referring to when the overclock settings are ignored and the motherboard boots with RAM @ 2133 MHz and CPU @ 4600 MHz? Yes, the bug still appears to be present.

No, the issue you're describing sounds like the standard boot failure issue: that when using Clover, or OpenCore without JTR's patch, boots that take longer than a certain threshold - which is easily triggered if USB drives are plugged in, or if the UEFI driver of a device causes delays (as happens with my X520 PCIe NIC) - cause an automatic shutdown and restart. Upon restart, all overclock settings are ignored and the system will be running at 3.0Ghz and 2133Mhz RAM.

The VCCSA issue is different: when the VCCSA voltage setting is changed from Auto, to for example 1.250V, this setting gets ignored whenever a shutdown or sleep/wake is done. It happens after both Windows and macOS shutdown.

After a full shutdown (but not restart), the BIOS setting still says 1.250V (or whatever), but VCCSA is actually running back at 0.9V. This is visible both in the BIOS (VCCSA is one of the values shown in the top right of each BIOS screen), and via any hardware monitoring software like HWInfo in Windows or HWMonitorSMC2 in macOS (there it will show up as 'VIN5').

When the issue occurs, VCCSA will show up as 0.9V instead of the value you set it to. I'm not 100% sure if this means it's actually 0.9V, or whether it's back to Auto.

In order to get it to re-apply, the user must change VCCSA to a different value (eg 1.245V) and then F10 in the BIOS. But it will only stay applied until the next shutdown or sleep/wake.

I used to have a tedious but effective solution for this: every time I did a cold boot, I'd go into the BIOS and flip the value between 1.250V and 1.245V (or vice versa), then F10 and boot normally. But more recently I've started making use of regular sleep and wake, which breaks that workaround because VCCSA also goes back to Auto (or 0.9V) when the system wakes from sleep.

However, there are pros and cons:
Thunderbolt boot support ENABLE: Devices connect and are hotpluggable only if at least one is connected and powered on at boot (A TB Dock is perfect for this). If no device is connected at boot the Thunderbolt ports will not work.
Thunderbolt boot support Disable: Thunderbolt ports always work, but the devices do not connect automatically on boot, you need to disconnect and reconnect them.
I currently keep the option enabled because I have a Dock which solves the problem

Ah yes, I see the same. I didn't notice before because like you I have a TB3 dock.

I believe it should all work fine if the firmware flash is done.

On system information the bandwidth is 2.5GT/s. Can I get faster speeds or latency improvements with the firmware patch?

I'm pretty sure that 2.5GT/s figure is incorrect. 2.5GT/s per lane is PCIe 1.0 speed. A x4 link at 2.5GT/s would be limited to about 10Gbit/s of bandwidth, and I've already done approx 13Gbit/s through my Thunderbolt 3 dock.

There may be speed or latency improvements from the FW flash, I'm not certain - dolgarrenan talked about something like that in his first post. I don't currently have enough TB3 hardware to fully test TB3 performance, beyond that 13Gb/s test (which I tested with 4 x USB3 to SATA adapters connected to SATA3 SSDs, each one connected to a different USB bus on my CalDigit TB3 dock, each copying a 15GB file to a RAM disk.)

There's a ton of info on this in CaseySJ's Z490 thread, so that would be the best place to look/ask to find out the finer details of TB firmware flashing.