List of Intel Titan Ridge Thunderbolt 3 Devices  

Page 4 / 4 Prev
  RSS

karatekid430
(@karatekid430)
Estimable Member
Joined: 1 year  ago
Posts: 134
October 15, 2018 2:38 am  
Posted by: kevin335200

First thanks for your reply. Yes i'm new here and what i'm trying to do is just gathering some information to have a better understanding of this card, because i'm curious about how well can it be supported in Hackintosh. I may build a new Z390 desktop with this card for Hackintosh purpose.
If Titan Ridge behaves the same in PCH and CPU, why it seems like hotplug works with 1950X but not fully with 2400G? And why it doesn't need to jump pins with 1950X build?
Other questions are MacOS related, maybe i should buy it before asking...
Thanks

Hi,

If you are using Z390 then chances are it supports Titan Ridge natively, which might mean it places constraints on which slots are supported. With Alpine Ridge, it would simply be undetectable in an unsupported slot. But Titan Ridge supports OsNativePciEnumeration=True and will show anyway, and sort of work, but not as well as if it were in the official slot which will take care of the loose strings in BIOS.

It is easy to find out if Thunderbolt AICs are supported (check if BIOS reserved bus numbers for Windows or Linux with no kargs). But idk about Titan Ridge in particular. I do not own a Gigabyte H370/B360 supported motherboard to find out.

I have absolutely no interest or experience in Apple, though, so I cannot offer Hackintosh-specific advice.

Anyways, try disabling Thunderbolt support in BIOS (or Fully Disabled option if available) and see if both PCH and CPU slots work the same. Which Z390 motherboard do you have? I knew they are almost out but I did not realise people already had them.

Edit: I did not notice the question about pins and Threadripper. The answer is idk because I do not own AMD (yet - I have been wanting to since the first Ryzen but wanted Thunderbolt). Either they were lucky, or happened to have devices plugged in. Or maybe Threadripper does something different in BIOS or in general that means it works. Perhaps there is a small window where the Titan Ridge cards naturally sleep in startup and maybe this coincides with the detection on Intel systems. Maybe Threadripper takes more time to get through BIOS and gives Titan Ridge more time to wake up. Perhaps it is all a misconception and it happens the same on them all - there will not be that many people who own and have tried it with both personally. Perhaps people with Intel systems can enable a 10 second BIOS delay in the BIOS settings and report if that alleviates the need for the jumper and/or devices connected at boot? Cheers


ReplyQuote
wimpzilla
(@wimpzilla)
Reputable Member
Joined: 2 years  ago
Posts: 395
October 16, 2018 10:32 am  

@karatekid430

From what i know hardware side, the OS always come after basic cpu and system initialization.
When powered up, the MB will begin by firing up all the power lines, checking them one after the other.
If these are ok it begin to enable sequentially each devices on the board that need to be powered, these will return back a PWR GOOD signal if fine.
It's during this very early power lines check, that the PCH will also wake up, check and begin initialize the CPU to be able to receive it's 1st lines of code.
Then the MB with all the devices ready sending back a POWER GOOD signal, begin reading the bios to boot up.
At this point the PCH would have finished the CPU initialization, the CPU now being able to read the bios code.
In this process the MB also begin to allocate the memory for the hardware and all the information needed to be able to run the OS kernel.
At this point the OS kernel and it's stuff are loaded into the memory, the part of the memory that run the ring 0 privileges i suppose.
It's only at this point that the kernel will begin to boot, leveraging the devices, doing all it's stuff until you get the OS UI.

It is hard to change what the system see at boot, even before the OS kernel load, some of these setting are not meant to be changed at all.
Hope it also give you a peak of what one can and can't do OS side, since your computer is already running even before the bios begin to load.
Remember the PCH and CPU initialize very early presumably to boot up the CPU own kernel and all the ME stuff.

The best way imo to overcome these issues, is to mod the bios and/or the microcode used to run the devices.

2012 13-inch Dell Latitude E6320 + R9 [email protected] (EXP GDC 8.4) + Win10
E=Mc²


ReplyQuote
karatekid430
(@karatekid430)
Estimable Member
Joined: 1 year  ago
Posts: 134
October 19, 2018 1:27 am  
Posted by: wimpzilla

@karatekid430

From what i know hardware side, the OS always come after basic cpu and system initialization.
When powered up, the MB will begin by firing up all the power lines, checking them one after the other.
If these are ok it begin to enable sequentially each devices on the board that need to be powered, these will return back a PWR GOOD signal if fine.
It's during this very early power lines check, that the PCH will also wake up, check and begin initialize the CPU to be able to receive it's 1st lines of code.
Then the MB with all the devices ready sending back a POWER GOOD signal, begin reading the bios to boot up.
At this point the PCH would have finished the CPU initialization, the CPU now being able to read the bios code.
In this process the MB also begin to allocate the memory for the hardware and all the information needed to be able to run the OS kernel.
At this point the OS kernel and it's stuff are loaded into the memory, the part of the memory that run the ring 0 privileges i suppose.
It's only at this point that the kernel will begin to boot, leveraging the devices, doing all it's stuff until you get the OS UI.

It is hard to change what the system see at boot, even before the OS kernel load, some of these setting are not meant to be changed at all.
Hope it also give you a peak of what one can and can't do OS side, since your computer is already running even before the bios begin to load.
Remember the PCH and CPU initialize very early presumably to boot up the CPU own kernel and all the ME stuff.

The best way imo to overcome these issues, is to mod the bios and/or the microcode used to run the devices.

The initialisation does not *typically* change, but it can, and should (imo, the firmware should do next to nothing).

Because in the unsupported setup with Titan Ridge, the BIOS is not allocating / reserving bus numbers for Thunderbolt, the OS must do it. Linux can do this with kargs "pci=assign-busses,hpbussize=0x33". This works, but it is difficult for the OS to reserve extra bus numbers after it has already assigned the rest. When Linux first boots, it can be instructed to override the BIOS. It is at this point the Titan Ridge needs to be present - because if it is not, the OS does not see a PCI hotplug bridge, and does not reserve resources.

Everything you say is true in BIOS-supported setups (official Thunderbolt implementations), but I prefer my way without a single scrap of firmware support - and I can just hope that Windows eventually gains a workaround to detect Thunderbolt Titan Ridge cards when enumerating and reserve bus numbers. Because fixing it in BIOS / microcode is just a short term fix.

For some reason, Linux still has a problem assigning some BARs - but that does not seem to stop most devices working. The only thing that does not work is my eGPU which loads amdgpu.ko and then trashes the mm/pat.c +549 (page tables). When I reported a bug they blamed the BARs - but they do not stop other things from loading, and if they were a problem, a graceful driver would check if they are allocated and abort without messing up the running kernel. I played with it and other times it does stop gracefully with BAR errors. Which makes me believe the BARs are not the root cause.

If anybody knows how to kexec to Windows from Linux then there is a fair chance that I can solve the issue for Windows. I did kexec from Linux to Linux and that greatly tidied up the BARs, and kept the correct Thunderbolt bus number reservation in the second kernel, which did not have pci=assign-busses passed. Assuming Windows never overrides what is done for it (the evidences suggests they do not) then we might be able to boot to Linux which can reserve bus numbers and then kexec straight into Windows.


ReplyQuote
karatekid430
(@karatekid430)
Estimable Member
Joined: 1 year  ago
Posts: 134
October 19, 2018 1:56 am  

Yesterday I had an interesting thought: Thunderbolt add-in cards have everything needed to behave as high-performance video capture devices. Thunderbolt can supposedly encapsulate packets from anything. As it is, those packets are PCIe and DisplayPort. And things do not appear to be baked in hardware because they have been able to scale from Thunderbolt 1 to 3 and also add DP1.4 support with Titan Ridge.

So theoretically, you could flash a NVM firmware update to make it dump DisplayPort packets to the NHI and back to the host through PCIe. A kernel-mode driver could pick them up and pass them to userspace. If there is anybody reading who is talented with reverse engineering, this would be an interesting project for those people.

Theoretically if Intel mandated that laptops are required to feed the GPU output through the Thunderbolt controller (no connection to laptop display) and have the Thunderbolt controller connected to the display, then the external graphics video loopback could have no performance penalty. Titan Ridge can inject DisplayPort streams anywhere on the device chain, so there could be a DisplayPort input on eGPUs, to attach to the graphics card output with a short cable, which goes back through Thunderbolt as DisplayPort (instead of PCIe) and straight to the laptop display, bypassing the laptop's CPU and GPU entirely. The Thunderbolt controller would just feed the integrated graphics straight through to the monitor when such a loopback was not set up. The main issue here is what happens if there are multiple Thunderbolt controllers. It could be that there needs to be a multiplexer controlled by the Thunderbolt controllers instead, to switch the laptop display between the integrated graphics and multiple Thunderbolt controllers.

Note in the diagram that as it is, the internal graphics already has a connection to Thunderbolt for outputting DisplayPort, but the point is that it would no longer have its own direct link to the display (unless the multiplexer were to be used).

If Intel is reading this, please take this idea and find a way to support it. It could dump at least 4K 60Hz, but possibly 2x 4K 60Hz assuming that can fit into 32Gb/s PCIe 3.0 x4. It has the benefits:
- More goodwill to customers who see you are giving more features for their money
- Gain notoriety among gaming and streaming community
- For allowing a USD $99 add-in card to outdo dedicated capture cards worth thousands (I have seen that DisplayPort 4K 60Hz units are super expensive)
- If there are separate DisplayPort inputs and outputs on the Thunderbolt controllers (and they are not just one set of pins used in reverse) then even laptops could expose a standard USB-C port next to the Thunderbolt for feeding DisplayPort into the laptop to use it as a monitor (value adding) Edit: or just make it so the Thunderbolt 3 ports can also behave as DisplayPort alt-mode sinks, as well as sources, even better

The eGPU loopback idea does not make use of the video capture concept, but they would go hand in hand - allowing DisplayPort to be fed back into the laptop's monitor, or allowing the DisplayPort to be captured and recorded by the CPU through PCIe.


theitsage liked
ReplyQuote
wimpzilla
(@wimpzilla)
Reputable Member
Joined: 2 years  ago
Posts: 395
October 22, 2018 10:34 am  

@karatekid430

I understand, the only way i know for enabling devices after boot have been posted in this forum, is to edit the NVRAM of the bios to enable the TB board before loading the OS.
If one can't mod the bios itself, at least you can maybe get your hand on the setting that the machine set up preparing for the OS, stocked in the NVRAM.

To give you another example, like the BAR the MSR register are locked in the bios and also refer to PCH settings.
So as i said before these are the kind of setting you would get access by not only modding the bios but also modding the ME.
I did not spend a lot of time when modding the Intel ME, looking at security features related to TB, but pretty sure the core of the TB firmware is shared between the TB, the bios and the ME.

I fought a lot with MRS register at boot to unlock overclocking capabilities on locked platforms, but without a bios or ME mod, these setting cannot be changed once the OS loaded, at least not by someone with basic prog understanding like me. 
I suppose some security settings also can't changed that easily, especially if regarding the ME and TB platform.
Hope you will be more lucky and find a way to enable the TB at boot before the OS launch.

2012 13-inch Dell Latitude E6320 + R9 [email protected] (EXP GDC 8.4) + Win10
E=Mc²


ReplyQuote
karatekid430
(@karatekid430)
Estimable Member
Joined: 1 year  ago
Posts: 134
October 23, 2018 4:39 pm  

If anybody cares I successfully installed NVM36 update from Dell 9575 onto Dell 9370 (they use the same controller and I did it in the past by installing NVM30 from 9575 on my machine).

There are dozens of firmwares in the updater this time, I had to pick carefully with hexdump to find the one with 15D2 controller and NVM36.

19472 = offset of controller PCI ID
19528 = offset of NVM version

Which left 0x080D_secure.bin from Source/Intel/group/20180731_02_Laferrari_36/0x080D_secure.bin
Used Linux to flash it
Do at your own risk


ReplyQuote
karatekid430
(@karatekid430)
Estimable Member
Joined: 1 year  ago
Posts: 134
October 23, 2018 4:44 pm  
Posted by: wimpzilla

@karatekid430

I understand, the only way i know for enabling devices after boot have been posted in this forum, is to edit the NVRAM of the bios to enable the TB board before loading the OS.
If one can't mod the bios itself, at least you can maybe get your hand on the setting that the machine set up preparing for the OS, stocked in the NVRAM.

To give you another example, like the BAR the MSR register are locked in the bios and also refer to PCH settings.
So as i said before these are the kind of setting you would get access by not only modding the bios but also modding the ME.
I did not spend a lot of time when modding the Intel ME, looking at security features related to TB, but pretty sure the core of the TB firmware is shared between the TB, the bios and the ME.

I fought a lot with MRS register at boot to unlock overclocking capabilities on locked platforms, but without a bios or ME mod, these setting cannot be changed once the OS loaded, at least not by someone with basic prog understanding like me. 
I suppose some security settings also can't changed that easily, especially if regarding the ME and TB platform.
Hope you will be more lucky and find a way to enable the TB at boot before the OS launch.

As I said, I already found the way. Using jumper lead and that makes it enabled before OS launch. Linux handles it well with hot-plugging support, Windows not so much (works but no hot-plugging).
It might be tacky, but all it will take for it to be perfect is for the kernel developers of the respective operating systems to patch the kernels to reserve resources when it is detected. Everything is there now. Titan Ridge really turned out to be everything I was hoping for.

In bad news, of the estimated 40 or so Z390 motherboards, I can only find two with onboard Thunderbolt. The Asrock Z390 ITX motherboard is still using the crippled JHL6240 controller like the previous three iterations (seriously, what is their problem?) and the other one is the Gigabyte Z390 Xtreme with an unknown dual-port controller (no reviews yet). A fair few of them have AIC support but Asus is absolutely shocking with only two with AIC support and none with inbuilt. I believe Gigabyte has 9 with AIC support.


theitsage liked
ReplyQuote
karatekid430
(@karatekid430)
Estimable Member
Joined: 1 year  ago
Posts: 134
November 1, 2018 4:14 am  

The new Mac Mini is likely Titan Ridge (but nobody knows for sure as it uses Intel UHD 630 graphics which limit to DP1.2, so we cannot tell by the advertised DisplayPort version).

The curious thing is that the only part that matches Apple's description (i7, 6-core, 3.2GHz base, 4.6GHz boost, Intel UHD 630 graphics) is the i7-8700 which is socketed. If this turns out to be the case, it would be interesting to see if it could be upgraded to the 8-core version. It has socketed RAM. On forums there is speculation about whether the SSD is upgradable (whether it is replaceable, and if it is, whether the T2 chip would get in the way).  https://forums.macrumors.com/threads/2018-ssd-user-upgradeable.2150882/

If they are all upgradeable then it would actually be quite a nice machine despite the steep prices and the fact that it is from Apple (even if you erase OSX, it is still an Apple in firmware and will be unpleasant for other operating systems).

Edit: Silly me, when customising the system, in "Which processor is right for you?", it says "All Mac mini computers feature Intel’s eighth-generation (‘Coffee Lake’) desktop-class processors with support for fast 2666MHz DDR4 memory and Intel UHD Graphics 630.". So it must be the socketed i7-8700.


ReplyQuote
joevt
(@joevt3)
Estimable Member
Joined: 1 year  ago
Posts: 236
November 1, 2018 9:02 am  
Posted by: karatekid430

The new Mac Mini is likely Titan Ridge (but nobody knows for sure as it uses Intel UHD 630 graphics which limit to DP1.2, so we cannot tell by the advertised DisplayPort version).

The new Mac Mini is the least expensive product with a Thunderbolt 3 port. I have a couple Mac Minis connected to TVs in the house but they are OLD (Core Duo processors). For those reasons I ordered one. iFixit will probably have pictures of all the components eventually.

The new MacBook Air is probably also Titan Ridge. I suppose we can assume that all new Thunderbolt products will use Titan Ridge. The Mac Mini and MacBook Air both use Intel graphics, so they are both stuck at DisplayPort 1.2.

I think the new LG 34WK95U-W display uses Titan Ridge. I believe it requires DisplayPort 1.4 for full resolution (5120x2160). This will exclude Intel graphics or Alpine Ridge Thunderbolt controllers. I don't think it supports dual link DisplayPort 1.2 over Thunderbolt like the larger resolution (5120x2880) LG UltraFine 5K does. The 34WK95U-W also has DisplayPort 1.4 input so you can ignore the Thunderbolt input. It can get USB input from either USB-B (USB 2.0) or USB-C (USB 3.0). It is unclear if full resolution is possible with USB-C when the USB input is also set to USB-C because it could use USB 3.0 (most likely) which would allow only two lane DisplayPort 1.4 connection instead of four lane (which would still allow USB 2.0 - much like how the LG UltraFine 4K uses USB-C with USB 2.0 for four lane DisplayPort 1.2 connection). USB-C input uses the Thunderbolt connector - which means this display most likely uses the Titan Ridge Thunderbolt controller. Unlike Alpine Ridge, Titan Ridge has the ability to take Thunderbolt or USB-C input and also supports DisplayPort 1.4.


ReplyQuote
wimpzilla
(@wimpzilla)
Reputable Member
Joined: 2 years  ago
Posts: 395
November 3, 2018 9:13 am  

@karatekid430

Not sure i would bet on upgradable features when speaking about Apple.
Since Apple decide it's own cpu TDP, it could be that the cpu is simply soldered, without the need of a socket.
The same apply to both memory and storage space.

Would wait and check the motherboard before even thinking about, just to leverage to which point the product is really upgradable.

 

2012 13-inch Dell Latitude E6320 + R9 [email protected] (EXP GDC 8.4) + Win10
E=Mc²


ReplyQuote
karatekid430
(@karatekid430)
Estimable Member
Joined: 1 year  ago
Posts: 134
November 13, 2018 10:29 am  

Okay, Gigabyte are rude on Facebook and do not respond, days after they have seen the message. Even if you are pointing out a typo on their website for their benefit (there is no such thing as DSL7540 Thunderbolt controller).

Anyways, I found another way to find out the information I needed. I extracted the IFR from the Z390 Aorus Xtreme BIOS ROM and it is without a doubt Titan Ridge on board.

We have found our first and only motherboard with Titan Ridge onboard, out of about 40 Z390 motherboards. If the motherboard manufacturers were dogs, then Asus definitely gets the cone of shame. Out of all of theirs, two have AIC support and no onboard. Gigabyte has about nine with AIC support.

Anyways, the IFRs are remarkably different to what I am used of seeing - and include options for VT-d IOMMU protection.

0x40035 Form: Thunderbolt(TM) Configuration, Form ID: 0x27AF {01 86 AF 27 11 1B}
0x4003B         Setting: Discrete Thunderbolt(TM) Support, Variable: 0x4EC {05 91 19 1B 1A 1B A5 27 01 00 EC 04 14 10 00 01 00}
0x4004C                 Default: 8 Bit, Value: 0x1 {5B 06 00 00 00 01}
0x40052                 Option: Disabled, Value: 0x0 {09 07 04 00 00 00 00}
0x40059                 Option: Enabled, Value: 0x1 {09 07 03 00 00 00 01}
0x40060         End of Options {29 02}
0x40062         Suppress If: {0A 82}
0x40064                 Variable 0x27A5 equals 0x0 {12 06 A5 27 00 00}
0x4006A                 Setting: TBT Vt-d base security, Variable: 0x57E {05 91 1B 1B 1C 1B A6 27 01 00 7E 05 14 10 00 01 00}
0x4007B                         Default: 8 Bit, Value: 0x0 {5B 06 00 00 00 00}
0x40081                         Option: Disabled, Value: 0x0 {09 07 04 00 00 00 00}
0x40088                         Option: Enabled, Value: 0x1 {09 07 03 00 00 00 01}
0x4008F                 End of Options {29 02}
0x40091                 Setting: Thunderbolt Boot Support, Variable: 0x4F0 {05 91 1F 1B 20 1B A7 27 01 00 F0 04 14 10 00 02 00}
0x400A2                         Default: 8 Bit, Value: 0x0 {5B 06 00 00 00 00}
0x400A8                         Option: Disabled, Value: 0x0 (default MFG) {09 07 04 00 20 00 00}
0x400AF                         Option: Boot once, Value: 0x1 {09 07 65 1B 00 00 01}
0x400B6                         Option: Pre-Boot ACL, Value: 0x2 {09 07 64 1B 00 00 02}
0x400BD                 End of Options {29 02}
0x400BF                 Suppress If: {0A 82}
0x400C1                         True {46 02}
0x400C3                         Setting: Titan Ridge Workaround for OSUP, Variable: 0x515 {05 91 21 1B 22 1B C9 04 01 00 15 05 14 10 00 01 00}
0x400D4                                 Default: 8 Bit, Value: 0x0 {5B 06 00 00 00 00}
0x400DA                                 Option: Disabled, Value: 0x0 (default MFG) {09 07 04 00 20 00 00}
0x400E1                                 Option: Enabled, Value: 0x1 {09 07 03 00 00 00 01}
0x400E8                         End of Options {29 02}
0x400EA                         Setting: Tbt Dynamic AC/DC L1, Variable: 0x518 {05 91 25 1B 26 1B CA 04 01 00 18 05 14 10 00 01 00}
0x400FB                                 Option: Disabled, Value: 0x0 {09 07 04 00 30 00 00}
0x40102                                 Option: Enabled, Value: 0x1 {09 07 03 00 00 00 01}
0x40109                         End of Options {29 02}
0x4010B                 End If {29 02}
0x4010D                 Checkbox: Wake From Thunderbolt(TM) Devices, Variable: 0x4F1 {06 8E 23 1B 24 1B CB 04 01 00 F1 04 10 01}
0x4011B                 End {29 02}
0x4011D                 Setting: Security Level, Variable: 0x4F5 {05 91 27 1B 28 1B A8 27 01 00 F5 04 14 10 00 04 00}
0x4012E                         Option: No Security, Value: 0x0 {09 07 29 1B 00 00 00}
0x40135                         Option: User Authorization, Value: 0x1 {09 07 2A 1B 30 00 01}
0x4013C                         Option: Secure Connect, Value: 0x2 {09 07 2B 1B 00 00 02}
0x40143                         Option: Display Port only, Value: 0x3 {09 07 2C 1B 00 00 03}
0x4014A                         Option: USB Docking Only, Value: 0x4 {09 07 2D 1B 00 00 04}
0x40151                 End of Options {29 02}
0x40153                 Suppress If: {0A 82}

Full IFRs - https://1drv.ms/f/s!AuN0OlwU3-gMjNFADUJ1g58AtXdk4A


ReplyQuote
wimpzilla
(@wimpzilla)
Reputable Member
Joined: 2 years  ago
Posts: 395
November 13, 2018 11:26 am  

Pictures of the motherboard pcb with the soldered IC being a Titan Ridge class or never happened. 😀 
Code lie often too much, more often that what hardware does.

2012 13-inch Dell Latitude E6320 + R9 [email protected] (EXP GDC 8.4) + Win10
E=Mc²


ReplyQuote
karatekid430
(@karatekid430)
Estimable Member
Joined: 1 year  ago
Posts: 134
November 13, 2018 1:59 pm  
Posted by: wimpzilla

Pictures of the motherboard pcb with the soldered IC being a Titan Ridge class or never happened. 😀 
Code lie often too much, more often that what hardware does.

I would bet a lot of money on it. I have dumped IFRs from a lot of motherboards and every single one prior to this has been Alpine Ridge and they have all been exactly the same. This is markedly different. It is missing stuff that was always there. And added new stuff I have ever seen before.

Reviews are hard to come by for this motherboard in particular, and most motherboard reviews do not cover important details like this.


ReplyQuote
wimpzilla
(@wimpzilla)
Reputable Member
Joined: 2 years  ago
Posts: 395
November 13, 2018 3:02 pm  

I checked too but i also couldn't find a decent picture, unfortunately.
But i agree it's encouraging if you find out so much differences looking at the bios.

The motherboard website mention the audio IC name but not the TB IC one, sometimes i don't understand what's wrong about simply declaring basic specs.

2012 13-inch Dell Latitude E6320 + R9 [email protected] (EXP GDC 8.4) + Win10
E=Mc²


ReplyQuote
theitsage
(@itsage)
Famed Member Admin
Joined: 2 years  ago
Posts: 3081
November 13, 2018 3:21 pm  

@karatekid430 Nice find. One interesting thing I noticed looking through the manual is that there's no video input. This may be a sign the on-board Thunderbolt controller is Titan Ridge. Alpine Ridge requires video input for the on-board Thunderbolt port(s) to provide monitor output. I wonder that means it's possible to use a discrete GPU for monitor output through this motherboard too.

Best ultrabooks for eGPU use

eGPU enclosure buying guide

82 external GPU build guides


ReplyQuote
joevt
(@joevt3)
Estimable Member
Joined: 1 year  ago
Posts: 236
November 13, 2018 6:13 pm  
Posted by: theitsage

@karatekid430 Nice find. One interesting thing I noticed looking through the manual is that there's no video input. This may be a sign the on-board Thunderbolt controller is Titan Ridge. Alpine Ridge requires video input for the on-board Thunderbolt port(s) to provide monitor output. I wonder that means it's possible to use a discrete GPU for monitor output through this motherboard too.

Alpine Ridge and Titan Ridge have the same requirements for video input, just look at the GC-ALPINE RIDGE and the GC-TITAN RIDGE add-in cards. If there's no video input, then they are taking input from the CPU's DisplayPort outputs (Intel graphics) or they don't do video.

The motherboard specs says it supports DisplayPort 1.2. If it's using Titan Ridge, then this is an indicator that the inputs come from Intel graphics which doesn't support DisplayPort 1.4.


theitsage liked
ReplyQuote
wimpzilla
(@wimpzilla)
Reputable Member
Joined: 2 years  ago
Posts: 395
November 13, 2018 6:25 pm  

I tried to lurke at some naked pcb pictures and videos, you can find some here:


But unfortunately the camera never really focus enough on the chip to read the part number.
Tho one can leverage how it's mounted and wired on the pcb.

Looking at the package part number disposition and fonts it could be a Titan Ridge.
I mean it seems to me that the JHL7540 have a bigger dot on the IC corner than a DSL/JHL6540, so who knows.
Usually the DSL/JHL6540 have a smaller package corner dot, if i'm not mistaking here.

2012 13-inch Dell Latitude E6320 + R9 [email protected] (EXP GDC 8.4) + Win10
E=Mc²


ReplyQuote
karatekid430
(@karatekid430)
Estimable Member
Joined: 1 year  ago
Posts: 134
November 14, 2018 12:47 am  
Posted by: wimpzilla

I tried to lurke at some naked pcb pictures and videos, you can find some here:


But unfortunately the camera never really focus enough on the chip to read the part number.
Tho one can leverage how it's mounted and wired on the pcb.

Looking at the package part number disposition and fonts it could be a Titan Ridge.
I mean it seems to me that the JHL7540 have a bigger dot on the IC corner than a DSL/JHL6540, so who knows.
Usually the DSL/JHL6540 have a smaller package corner dot, if i'm not mistaking here.

OMG THEY PUT A JTAG HEADER FOR THE THUNDERBOLT CONTROLLER.

I have the MSI Thunderbolt M3 AIC and it has JTAG. Using continuity tester, I found it does not connect to the flash chip. Is that not what a JTAG is for? To re-flash the chip and un-brick? Either way, I want to know how to use it. Because if I can get a card that can easily be flashed, I can get serious about firmware modding. I always wanted to turn an add-in card into a Thunderbolt 3 peripheral but I tried on my Asrock AIC. It appeared for a second as the external device, but never worked after power cycling. It was a nightmare to unbrick. I have a feeling that despite the firmware I chose using the same Thunderbolt 3 controller, perhaps the USB-C port controllers were different - and hence it could not be woken up with a connection.

We can ask the person on that channel to send us some nude pics of the motherboard, leaving nothing to the imagination with the sexy Thunderbolt 3 controller. 😀

Posted by: joevt
Posted by: theitsage

@karatekid430 Nice find. One interesting thing I noticed looking through the manual is that there's no video input. This may be a sign the on-board Thunderbolt controller is Titan Ridge. Alpine Ridge requires video input for the on-board Thunderbolt port(s) to provide monitor output. I wonder that means it's possible to use a discrete GPU for monitor output through this motherboard too.

Alpine Ridge and Titan Ridge have the same requirements for video input, just look at the GC-ALPINE RIDGE and the GC-TITAN RIDGE add-in cards. If there's no video input, then they are taking input from the CPU's DisplayPort outputs (Intel graphics) or they don't do video.

The motherboard specs says it supports DisplayPort 1.2. If it's using Titan Ridge, then this is an indicator that the inputs come from Intel graphics which doesn't support DisplayPort 1.4.

Well my Gigabyte Z170X Designare has two Thunderbolt 3 ports on board (DSL6540 controller like the GC-ALPINE RIDGE) and has a single Mini DisplayPort input for one of the Thunderbolt 3 ports. The other gets its video from the motherboard. It is unclear what will happen if you attach a device with two monitors (although I own HP Z-docks so I could try). Also, I wonder if you attach two low-res monitors if it will just take them both from one input using MST.

Remember that Titan Ridge can inject DP into Thunderbolt at any position on the daisy chain - hence how the Blackmagic eGPU manages to drive Thunderbolt 3 displays. Perhaps Intel has relaxed the requirement in light of the new abilities. Honestly they needed to relax some requirements. Like how they relaxed the thing where eGPU could only have a single port

a) Technically vendors could probably have certified the device as an external PCIe enclosure and just happen to have the double-width slot and massive PSU - unless there is some other gotcha which says external PCI enclosures cannot supply more than some amount of power

b) Users should be able to make their own decision between performance and convenience - they might use the GPU for compute tasks with almost no bandwidth used, and have room to blow the bandwidth on other devices

I would like to use my eGPU with 10GbE chained after it. It would also allow me to charge my laptop with 100W and then chain devices with garbage power delivery (most of them) afterward. Despite the bandwidth not changing with chain length (I tested with two HP docks), the latency gets worse and that is what impacts performance. Although I wonder if Vulkan is affected. It cuts down on CPU involvement.

 

OKAY NOW FOR SOME BAD NEWS

I am still sure this Z390 Aorus Xtreme is Titan Ridge. And that is why I am angry. Because the BIOS still contains the tbtsmm and tbtdxe modules in UEFI. They are clearly not needed (the Titan Ridge runs on any system, as we have shown) but are still there. They could just be vestigial, or stubs / placeholders. But I was really hoping that there would no longer be traces of BIOS support. And no, this board cannot support Alpine Ridge - as it does not support add-in cards. Perhaps they just handle little things like RTD3 power saving. Either way, they need to go.

If anybody wants to read up on System Management Mode - go ahead. It is scary how much power that has over the computer, and you just do not know what is in there, because the BIOS is not open-source or audited. I hope that the SMM module turns out to be not important for Titan Ridge (as I thought before) because that will mean with BIOS modding we can just shut down SMM entirely.

Does anybody know of any modern motherboards that do not have AMI BIOS? They all seem to use that. Any Phoenix (acquired AWARD)? It would be interesting to see an alternative because I wonder how many of the problems and shortcomings of the motherboards we see today can be attributed to mistakes by AMI.


ReplyQuote
karatekid430
(@karatekid430)
Estimable Member
Joined: 1 year  ago
Posts: 134
November 14, 2018 1:23 am  

The IFR for the Gigabyte Z390 Aorus Xtreme also has USB-C BSSB switch (possibly for debugging, it has to do with DCI - direct connect interface). There are also direct references to USB-C kernel / platform debugging support.

The IFR also contains hints at support for two Thunderbolt controllers "DTBT Controller 0 Configuration" and " DTBT Controller 1 Configuration". This might not mean anything - as *all* Thunderbolt Alpine Ridge motherboards had settings for AIC support and location, regardless of if they actually supported AICs. So it might just be included and not used. But I hope it points to the possibility of more motherboards with support for multiple controllers.

The fact that it does not contain any references to AICs hopefully means that now the AICs can work on any port and will be auto-detected. Somebody will have to test it. But I wonder if this motherboard will unofficially work with GC-TITAN RIDGE with functional hot-plugging….

Edit: It legit contains options for prevention of DRAM Row Hammer attacks! Although, no SR-IOV, which is present in some X370 AMD mainstream motherboards that cost a fraction as much (for example, the BIOSTAR ITX, and a heap of others).

Edit: I just saw a way on Intel boards to possibly enable Titan Ridge on any motherboard. I never considered it before. There are options for "Extra Bus Reserved" and resources on various busses. Perhaps with Thunderbolt support fully disabled in BIOS, and these options tweaked, it could reserve the required resources for Windows. I do not see the equivalent options on AMD boards. 🙁

Not a long-term solution - the BIOS should butt out and leave things to the operating system. But it would be awesome in the mean time. If it worked, then somebody with much more knowledge could see if there is another way to hack those options into AMD systems.


ReplyQuote
wimpzilla
(@wimpzilla)
Reputable Member
Joined: 2 years  ago
Posts: 395
November 14, 2018 1:44 am  

I couldn't find any other pictures, giga sent the board and pictures to it's "influencer reviewers" first and the second video is from one of their engy.
There is still no decent review or hand on with the bare naked pcb, we will have to wait a bit a suppose.

Hope you are right about the added bios settings, as said above i do not trust too much the software in 1st place.
JTAG connector is usually connected to the chip to program it, you can find some common JTAG to usb adapter to read the IC.
As you said, not sure tho these are working, usually it got disabled in some ways because obvious security reason.

2012 13-inch Dell Latitude E6320 + R9 [email protected] (EXP GDC 8.4) + Win10
E=Mc²


ReplyQuote
karatekid430
(@karatekid430)
Estimable Member
Joined: 1 year  ago
Posts: 134
November 14, 2018 1:53 am  
Posted by: wimpzilla

Pictures of the motherboard pcb with the soldered IC being a Titan Ridge class or never happened. 😀 
Code lie often too much, more often that what hardware does.

My apologies, you could be right. The Asrock Z390 Phantom Gaming-ITX/ac uses the crippled Alpine Ridge JHL6240 (they must have stockpiled them and are trying to get rid of them). It has the same updated BIOS of the Gigabyte Z390 Aorus Xtreme.

The Gigabyte Aorus Xtreme could still be Titan Ridge, but we will have to wait for the "Actually Hardware Overclocking" YouTube channel to respond.

That would explain why the tbtsmm and tbtdxe modules are there - if it has backwards-compatibility with Alpine Ridge.

I am still hopeful, though!


ReplyQuote
wimpzilla
(@wimpzilla)
Reputable Member
Joined: 2 years  ago
Posts: 395
November 14, 2018 1:59 am  

My 1st thought would be to make compatible with other devices that already mount a Titan ridge, if the IC is an Alpine Ridge.
Hope he will answer you, not sure he would give you any information if giga don't wanted to.
Unfortunately one have to rely being Sherlock Holmes to get the basic specs of a feature on a 600e board!
I mean for the price i truly hope you are right and giga updated to a Titan Ridge, i mean i found crazy to not know this kind of stuff when one is ordering a 600e board!

2012 13-inch Dell Latitude E6320 + R9 [email protected]bps-mPCIe (EXP GDC 8.4) + Win10
E=Mc²


ReplyQuote
karatekid430
(@karatekid430)
Estimable Member
Joined: 1 year  ago
Posts: 134
November 14, 2018 2:02 am  
Posted by: wimpzilla

My 1st thought would be to make compatible with other devices that already mount a Titan ridge, if the IC is a Alpine Ridge.
Hope he will answer you, not sure he would give you any information if giga don't wanted to.
Unfortunately one have to rely being Sherlock Holmes to get the basic specs of a feature on a 400e board!
I mean for the price i truly hope you are right and giga updated to a Titan Ridge, i mean i found crazy to not know this kind of stuff when one is ordering a 400e board!

Here's another Z390 with inbuilt Thunderbolt 3 that was either released later than the rest, or completely missed by me when I was searching.

https://www.gigabyte.com/us/Motherboard/Z390-DESIGNARE-rev-10#kf

Again, we will need somebody who owns it who can check in Device Manager, or somebody who posts good PCB photos.

Edit: Is this enough proof for you that the Gigabyte Z390 Designare has Titan Ridge? BAM CONFIRMED


ReplyQuote
wimpzilla
(@wimpzilla)
Reputable Member
Joined: 2 years  ago
Posts: 395
November 14, 2018 2:10 am  

This is what i read on the product page of the Z390 Desig, someone explain what really this mean?

DisplayPort Input
Ultimate digital creative pen display experience
Z390 DESIGNARE motherboard offers the first ever DisplayPort input for external graphic card. Makes the possibility to take the latest GEFORCE® RTX DP1.4 8K resolution into the onboard ThunderboltTM. Provides the most natural experience of creation for professional artists and designers.

Something new marketing side?

Edit: Maybe but still here again i don't understand why not release the TB IC specs, on a board marketed to pro costing 300e .

2012 13-inch Dell Latitude E6320 + R9 [email protected] (EXP GDC 8.4) + Win10
E=Mc²


ReplyQuote
tristank
(@tristank)
Active Member
Joined: 1 year  ago
Posts: 12
November 14, 2018 10:32 pm  

Gigabyte just confirmed me that the Z390 Designare runs a Titan Ridge Controller and they also told me that none of their Z170/Z270 motherboards supports the GC-Titan-Ridge AIC. 

On a side note: Does someone know if the new asus workstation monitors like the PA27AC  are equipped with the JHL7440 Titan Ridge Controllers? Same question for the new Akitio Thunder3 Dock Pro?


ReplyQuote
joevt
(@joevt3)
Estimable Member
Joined: 1 year  ago
Posts: 236
November 15, 2018 12:39 am  
Posted by: karatekid430

Well my Gigabyte Z170X Designare has two Thunderbolt 3 ports on board (DSL6540 controller like the GC-ALPINE RIDGE) and has a single Mini DisplayPort input for one of the Thunderbolt 3 ports. The other gets its video from the motherboard. It is unclear what will happen if you attach a device with two monitors (although I own HP Z-docks so I could try). Also, I wonder if you attach two low-res monitors if it will just take them both from one input using MST.

The Thunderbolt controller will pipe the iGPU to one display and the dGPU to the other display even if they are connected to a single Thunderbolt port. MST will only happen if you use a MST hub or daisy chain a DisplayPort display to a DisplayPort display that supports MST and has a DisplayPort output. All the displays in the MST chain will be connected to one of the sources: either the iGPU or the dGPU.

Posted by: karatekid430

Remember that Titan Ridge can inject DP into Thunderbolt at any position on the daisy chain - hence how the Blackmagic eGPU manages to drive Thunderbolt 3 displays.

We don't know that the Blackmagic eGPU couldn't have implemented this feature using an Alpine Ridge chip.

Back to the discussion about the Gigabyte Z390 Aorus Xtreme, I don't know why we care if the Gigabyte Z390 Aorus Xtreme has Titan Ridge or Alpine Ridge since it doesn't have any DisplayPort inputs and therefore cannot support DisplayPort 1.4. Is the hope that there will be a future Intel CPU with integrated graphics supporting DisplayPort 1.4 that will work in the same CPU socket?

Posted by: wimpzilla

This is what i read on the product page of the Z390 Desig, someone explain what really this mean?

DisplayPort Input
Ultimate digital creative pen display experience
Z390 DESIGNARE motherboard offers the first ever DisplayPort input for external graphic card. Makes the possibility to take the latest GEFORCE® RTX DP1.4 8K resolution into the onboard ThunderboltTM. Provides the most natural experience of creation for professional artists and designers.

Something new marketing side?

Edit: Maybe but still here again i don't understand why not release the TB IC specs, on a board marketed to pro costing 300e .

So the Z390 DESIGNARE has a DisplayPort input to the Thunderbolt controller. The Thunderbolt controller must be Titan Ridge to support DP1.4.

It sucks that there's only one DisplayPort input. It means that it can't support dual link SST 5K displays with a single Thunderbolt port.

I don't know what "Ultimate digital creative pen display experience" means. Marketing noise to be filtered out by not-dumb people.

I don't know why they mention "external graphics card" with regards to the DisplayPort input. Any graphics card can connect to the DisplayPort input, whether it be internal, external, CPU, GPU, other computer, game console, iPad Pro, whatever.

Posted by: tristank

Gigabyte just confirmed me that the Z390 Designare runs a Titan Ridge Controller and they also told me that none of their Z170/Z270 motherboards supports the GC-Titan-Ridge AIC. 

The GC-TITAN RIDGE sort of works in my GA-Z170X-Gaming 7 like it sort of works in my Mac Pro 2008. I don't know if firmware/software changes can make it work as well as the built-in Alpine Ridge. They can't say it's supported until those firmware/software changes are made which will probably never happen.

Posted by: tristank

On a side note: Does someone know if the new asus workstation monitors like the PA27AC  are equipped with the JHL7440 Titan Ridge Controllers?

I don' think it has Titan Ridge. If it does, then it seems none of it's features are used (well, I'm not sure if having the USB ports controlled by either Thunderbolt or a USB type B cable is one of the features of Titan Ridge).

The display supports HDR400 which I think requires DisplayPort 1.4 but the display is only 2560x1440 and the manual seems to imply that HDR only works through one of the three HDMI inputs. The manual does not mention DisplayPort 1.4 anywhere. The manual does not say that the Thunderbolt port can accept a USB-C input. The manual contains this nonsense line about the Thunderbolt ports: "The other serves as a DisplayPort signal output only which can be enabled to support Thunderbolt daisy chaining." How can they call it Thunderbolt daisy chaining if the other port can only output DisplayPort? The manual says the DisplayPort and Thunderbolt inputs can't be used simultaneously (for PIP mode).

Posted by: tristank

Same question for the new Akitio Thunder3 Dock Pro?

If this used Titan Ridge, then the DisplayPort output would support DisplayPort 1.4 instead of just 1.2.


ReplyQuote
Page 4 / 4 Prev