List of Intel Titan Ridge Thunderbolt 3 Devices
 
Notifications
Clear all

List of Intel Titan Ridge Thunderbolt 3 Devices  

 of  25
  RSS

itsage
(@itsage)
Illustrious Member Admin
Joined: 4 years ago
 

@joevt3 Thank you for the correction. It was my mistake. I had in mind multiple DisplayPort streams over TB3 and wrote MST. The correct type of connection is as you said "dual-link SST". It also makes more sense for the sole DisplayPort on this AIC to be output rather input. 

external graphics card builds
best laptops for external GPU
eGPU enclosure buyer's guide

 
2020 13" MacBook Pro [10th,4C,G] + RX 5700 @ 32Gbps-TB3 (Prototype eGPU Enclosure) + macOS 10.15.4 & Win10 1903 [build link]  


ReplyQuote
joevt
(@joevt)
Noble Member
Joined: 3 years ago
 

I've received a couple GC-TITAN RIDGE. So far I only took some photos and tested DisplayPort x 2 to Thunderbolt conversion (no PCIe) and two DisplayPort to USB-C conversion (no USB). Those two functions worked while the GC-TITAN RIDGE was in the IT-GO "ADP-099-31" without any internal connections (none of 6 pin power x2, USB 2.0 header, and AIC header were connected).

The DisplayPort port is an output. It does not seem to support DP++. A passive DisplayPort to HDMI cable that worked with my Mac's NVIDIA GeForce GTX 680 does not work with the GC-TITAN RIDGE, which means HDMI requires an active adapter.

I need to test other risers, try it in my GA-Z170X-Gaming 7 motherboard (no Thunderbolt add-in card header but may still kind of work in the PCIEX4 slot under certain conditions - like using Linux).

PCB GC TITAN RIDGE
PCB GC ALPINE RIDGE
GC x RIDGE Size Comparison
GC x RIDGE ports
GC x RIDGE ports in IT GO

Mac mini (2018), Mac Pro (Early 2008), MacBook Pro (Retina, 15-inch, Mid 2015), GA-Z170X-Gaming 7, Sapphire Pulse Radeon RX 580 8GB GDDR5, Sonnet Echo Express III-D, Trebleet Thunderbolt 3 to NVMe M.2 case


itsage liked
ReplyQuote
karatekid430
(@karatekid430)
Estimable Member
Joined: 3 years ago
 

I have an assignment to do, so I cannot write about this for long. But this could be big.
I received 2x GC-TITAN RIDGE AIC. I have so far tried with Gigabyte Z170X-DESIGNARE (which has AIC header and inbuilt DSL6540 with two ports).

In an hour or so, this thing has done what I have struggled to do with my various Alpine Ridge add-in cards for months.

A heap of firsts:
- First time I have ever seen OsNativePciEnumeration=True in WMI Explorer in Windows
- First time I have gotten working devices off three Thunderbolt 3 controllers on the same system simultaneously
- First time I have gotten three Thunderbolt 3 controllers to be properly detected, period
- First time I have gotten a working Thunderbolt 3 connection from CPU lanes
- First time I have managed to get Thunderbolt 3 devices approved when attached after boot in an unsupported PCIe slot (although it was slightly messy and messed up the lspci -vt a little)
- First time I have gotten a Thunderbolt 3 AIC to be detected with Thunderbolt 3 support disabled in the BIOS (it killed off the DSL6540 in the motherboard, but the 2x JHL7540 stayed detectable and working)

The last point is by far the most important. It means it almost certainly work in an AMD system, even without the header (although early indications are that it might need a Thunderbolt device attached at boot to have bus numbers reserved properly, and hot plugging will give varying results, perhaps even nasal demons).

I might finally be able to ditch Intel CPUs. Now, if somebody with a Threadripper system can get some GC-TITAN RIDGE and tell us if they work with the VROC-style PCIe bifurcation, that would be amazing! As in, get M.2->U.2 adapters then U.2 -> PCIe slot adapters and attach the GC-TITAN RIDGE to M.2 slots of VROC card, in the one CPU PCIe slot. Or in short terms, do they work with passive PCIe bifurcation?

The strange thing is that the officially supported motherboards are not even the ones that enthusiasts are interested in. H370/B360. In the officially supported PCIe slot of this motherboard (Z170), the GC-TITAN RIDGE did have a lot of trouble enumerating hot-added devices after boot. I have the Asrock Z270M-Extreme4 to test with, but only the one CPU and a lot of study to be doing. But at some point, I will move the CPU over and see what behaviour changes with another vendor.

Linux actually had a better time with Thunderbolt support disabled in the BIOS. The OSC error message (possibly does not like my motherboard firmware, or the firmware does not like it) from dmesg was replaced by a message indicating native PCI enumeration had been granted by the platform. It seemed to try to hot-add a fourth Thunderbolt device (first three were at boot time) but ran out of bus numbers. If you have the skills to modify a Linux kernel to reserve bus numbers on PCI slots given in /proc/cmdline at boot, then it might just work perfectly on any motherboard. It is too early to be sure, though. It is extremely promising - at the very least, it opens some interesting possibilities.

At some point I will want to do a full write-up and perhaps some YouTube videos on this, but for now, I will leave you with this to mull over.

Oh, by the way, the GC-TITAN RIDGE came with NVM23 which is pretty sad, given my Dell 9370 had NVM23 from the factory, and has since had updates for NVM28 and NVM33 (JHL6540). The external GPUs supported flag is set to "yes". USB support seems to work perfectly, without a hitch, with hot-plugging. Thunderbolt Networking appears to work perfectly, with hot-plugging. Power delivery works. I have not tested DisplayPort and alternate modes, but I suspect they will work. A quirk with Windows is that it shows SL1 security even when SL2 was set in the BIOS.

Please keep reporting new Titan Ridge devices and if you are getting on of these cards, please consider doing some experiments and publishing your results. Cheers!

42204022 716628722020908 3358050949523308544 n
42186182 300176234109469 4414425136959062016 n

To do: Create my signature with system and expected eGPU configuration information to give context to my posts. I have no builds.

.

ReplyQuote
itsage
(@itsage)
Illustrious Member Admin
Joined: 4 years ago
 

@joevt3 @karatekid430 Thank you for sharing the photos and reports. Your first impressions seem very positive that Titan Ridge AIC will enable TB3 capability on a lot more systems. This is a big deal. I have one on the way and will test with a Threadripper build.

external graphics card builds
best laptops for external GPU
eGPU enclosure buyer's guide

 
2020 13" MacBook Pro [10th,4C,G] + RX 5700 @ 32Gbps-TB3 (Prototype eGPU Enclosure) + macOS 10.15.4 & Win10 1903 [build link]  


ReplyQuote
karatekid430
(@karatekid430)
Estimable Member
Joined: 3 years ago
 

It appears hot-plugging works on one big condition: that at boot, on each port, you have at least enough devices (daisy chain length) that you will want. On that port, and that port only, you will be able to hot-plug devices up to and including that original daisy chain length. If you attach six Thunderbolt devices to each and every port at boot (with four ports, 24 devices), then you will have full functionality (at least afaict) because it should reserve the full number of busses. If you can modify the operating system to somehow use non-contiguous bus numbers (somebody more knowledgeable about PCI, is this actually possible with the way the hardware works?) or to reserve the full number at boot regardless, then it will work on any system (again, only at least as far as I can tell).

Here is a juicy picture from before I disabled Thunderbolt support in BIOS earlier on.... The only system I am aware of with six ports previously is the trashcan and those were only Thunderbolt 2. There is also the "about" section from the Thunderbolt software.

42194205 252512828798329 2595660822457876480 n
42098409 499342257220198 7730508144261988352 n

Last thing to note: it would appear that sometimes Linux can force enumeration (on the condition of available bus numbers) if it does not happen upon approval (such as if the card does not have the GPIO header attached). "echo 1 | sudo tee /sys/bus/pci/rescan". This never seemed to help before, with Alpine Ridge controllers. Oh, another one worth mentioning is that I swapped out X550-T2 with the GC-TITAN RIDGE in the PCH slot, and I cannot get them both detected in CPU slots. Perhaps in the PCH slot, I can get away without having header attached, but not in the CPU slots. The AIC in the top CPU slot has the header attached, which likely forces power during early boot. I have found in the past that connecting two AICs to same header (either Gigabyte y-cable or using a motherboard with both the new and old style headers, which are wired directly together, only meant to use one at once as per the manual) can work to get them awake for detection, but can cause confusion later on, ending up with one's PCI devices in the other's tree (or them both mushed together).

To do: Create my signature with system and expected eGPU configuration information to give context to my posts. I have no builds.

.

itsage liked
ReplyQuote
itsage
(@itsage)
Illustrious Member Admin
Joined: 4 years ago
 

That's remarkable. I have been scratching my head trying to get the GC-ALPINE-RIDGE AIC working together with the single on-board TB3 on my X99 Designare. Like you said, the AIC has a tendency of disabling the on-board controller.

external graphics card builds
best laptops for external GPU
eGPU enclosure buyer's guide

 
2020 13" MacBook Pro [10th,4C,G] + RX 5700 @ 32Gbps-TB3 (Prototype eGPU Enclosure) + macOS 10.15.4 & Win10 1903 [build link]  


ReplyQuote
karatekid430
(@karatekid430)
Estimable Member
Joined: 3 years ago
 
Posted by: itsage

That's remarkable. I have been scratching my head trying to get the GC-ALPINE-RIDGE AIC working together with the single on-board TB3 on my X99 Designare. Like you said, the AIC has a tendency of disabling the on-board controller.

Thanks. But no, the AIC would have been undetectable, full stop. The only way I could get Alpine Ridge detected on any unsupported slot, regardless of whether the system was Thunderbolt ready or not, was to get it running in supported system in a PCIe riser, disconnect the PCI from that computer, keeping the power connected, and hot-plugging into unsupported system, and doing PCI rescan. But it quickly caused the Thunderbolt driver to become confused, and only certain vendors of AIC were vulnerable (cannot remember now if it was Asrock or Gigabyte, but I think it was Asrock).

Edit: to elaborate, very few motherboards have support for onboard Thunderbolt and AIC. I think both are Gigabyte. Mine has two onboard, and the other has one onboard. I was not aware of mine for ages, and bought it second hand off Ebay when I found out. Possibly because it started off with USB 3.1, later upgraded to Thunderbolt 3 with a firmware update.

What motherboard did you have? Did it actually have the header? Anyways, some time I will start testing on other systems. I do not own an AMD yet - despite me wanting to so badly. The only reason I broke my vow never to have an Intel CPU again was because of Thunderbolt.

OMG, how did I forget to mention? Although OsNativePciEnumeration works, it does not use RTD3, at least on this Z170X-DESIGNARE (F23g BIOS). There are hidden IFR options that can be tweaked with EFI shell for that exactly, but a) I have little time during semester and b) given the remarkable ability of GC-TITAN RIDGE to (almost) function independently of BIOS, I doubt it will even matter. And although it reports it is not RTD3 capable, it does not appear from the PCI tree when no devices are detected, which is an indicator that it might actually be using it (or partial power savings).

@joevt3 any idea why it has USB 2.0 header? I am still stumped. My best guess is something to do with RTD3 and detecting power. Especially as it does not show up as a USB device in Linux or Windows....

To do: Create my signature with system and expected eGPU configuration information to give context to my posts. I have no builds.

.

itsage liked
ReplyQuote
itsage
(@itsage)
Illustrious Member Admin
Joined: 4 years ago
 

@karatekid430 I have the Gigabyte GA-X99 Designare EX and it has a header. When it first came out the port behaved as USB-C only. Once Gigabyte got Thunderbolt 3 certification, they released the firmware updated to make it Thunderbolt 3. What's nice about this board is that the on-board Alpine Ridge controller attaches directly to the CPU. When I attempted to add the GC-ALPINE-RIDGE AIC, the system could only see the AIC and no on-board TB3.

Not sure how it'll go with my 1950x build. I used a mATX board to fit the OMEN X case. Therefore the number of slots is limited to 3. You brought up an intriguing idea of using adapters for U.2 port. I have already used the 3x M.2 + NVMe adaper for a four way 960 EVO RAID0 setup. There are more PCIe lanes available yet which is unbelivable on this platform and price range. Here's a quick photo. I've since upgraded to 4x 8GB RAM modules to enable quad channel (write speed reaches 100 GB/s). The CPU cooler is my only last component to sort out. All existing coolers has their weakness. This may have to go custom loop eventually.

external graphics card builds
best laptops for external GPU
eGPU enclosure buyer's guide

 
2020 13" MacBook Pro [10th,4C,G] + RX 5700 @ 32Gbps-TB3 (Prototype eGPU Enclosure) + macOS 10.15.4 & Win10 1903 [build link]  


ReplyQuote
karatekid430
(@karatekid430)
Estimable Member
Joined: 3 years ago
 
cmd2
cmd

Back to one on the PCH, that works better. I still cannot get both running on the CPU slots. Might have something to do with how they share lanes if both populated. Plus the X550-T2 might have been keeping the lanes awake when the GC-TITAN RIDGE was not. Anyway, this is with Thunderbolt support disabled in BIOS, which is why I have hope that this will at least bring broken Thunderbolt functionality to AMD in the mean time. Since everything works perfectly except PCIe, it might just be worth having to reboot to attach devices.

@itsage if it does not work in your Threadripper then you might need to find a way to trick them into thinking they have TBT header. The force power is just a GPIO set high, afaik - not some fancy encoding. So attaching the correct pins to 3.3V (or maybe 5V, measure the voltage on the GPIOs in a real setup) might be enough.

Anybody into trashcans, there is a Reddit forum asking if anybody got this working with Hackintosh. Perhaps somebody could lend them a hand?  https://www.reddit.com/r/hackintosh/comments/9g64mw/gigabyte_gctitan_ridge/

To do: Create my signature with system and expected eGPU configuration information to give context to my posts. I have no builds.

.

ReplyQuote
itsage
(@itsage)
Illustrious Member Admin
Joined: 4 years ago
 

@karatekid430 The Mac Pro trashcan does not have any PCIe slots so I'm not sure how one would go about testing this AIC. Did you mean the Mac Pro tower? Those have PCIe 2.0 slots. I have a couple and will test when the GC-TITAN-RIDGE arrives.

external graphics card builds
best laptops for external GPU
eGPU enclosure buyer's guide

 
2020 13" MacBook Pro [10th,4C,G] + RX 5700 @ 32Gbps-TB3 (Prototype eGPU Enclosure) + macOS 10.15.4 & Win10 1903 [build link]  


ReplyQuote
 of  25