Thunderbolt 3 add-in cards are PCIe 3.0 x4.I've tried searching the thread, but is the Titan Ridge card a PCIe x4 card, and limited to ~20Gbit over TB3 because of the Gen2 slot in the cMP?
Thunderbolt 3 add-in cards are PCIe 3.0 x4.
There are ways to add PCIe 3.0 slots to a Mac Pro to get the full performance of the Thunderbolt 3 controller. Anything with a PCIe 3.0 switch with a PCIe 2.0 x8 or greater upstream link could be used to do that. This usually means a two or four M.2 or U.2 carrier PCIe cards with adapters or a PCIe expansion box. Expensive.
There are ways to add PCIe 3.0 slots to a Mac Pro to get the full performance of the Thunderbolt 3 controller. Anything with a PCIe 3.0 switch with a PCIe 2.0 x8 or greater upstream link could be used to do that. This usually means a two or four M.2 or U.2 carrier PCIe cards with adapters or a PCIe expansion box. Expensive.
That is exactly where I was going with the question. I was curious if anyone had used a PCIe switch to get a Gen3 x4 slot out of the deal.
I looked at amfeltec, my goto for oddball risers, and they don't do a M.2 to PCIe x4. There are items like these that are close: https://minerparts.com/product/pcie-4x-to-ngff-m-2-adapter-with-4-pin-power-cable/
Now to just wait for someone with a switch card and a Titan Ridge card to give it a go.
Edit: Finally going through and reading all the posts. I see this was an early thought.
Not just a thought - there are pictures! I'm thinking that an M.2 to U.2 adapter with a U.2 cable and a U.2 to PCIe x4 adapter would be better than what I tried with the amfeltec (similar to wjesse's Highpoint) because the U.2 cables are more flexible and can be detached from both ends.I see this was an early thought.
You're looking at the wrong devices. Even a real Mac shows 2.5 GT/s x4 for those devices (but real Mac's don't list those devices in the PCI section). Those devices are internal to the Thunderbolt controller, so Intel doesn't bother with setting the link speed register to something reasonable. You have to look at the speed of the upstream Titan Ridge bridge. The lspci command will give the current link speed values - or look for my lspcitree.sh script. In my Mac mini 2018, The upstream Titan Ridge bridge has an Alpine Ridge 4C 2015 device id [8086:1578]. The NHI is 15eb and the USB controller is 15ec just like in your screen shot. Downstream bridges are 15ea. Or you could try a benchmark. Anything over 1000 MB/s will prove that it is at least 5.0 GT/s. Anything over 2000 MB/s will prove that it is 8.0 GT/s.I tested my GC-Titan Rideg with HightPoint SSD7102, like this
Unfortunately, the link speed shows 2.5GT/s and link width is X4, use a PCIe switch to get a Gen3 x4 seems not useful?
I didn't warm boot from Windows 10.
Please share your experience if the card came out. I look forward.Still waiting for the delivery. It is shipped, but it will take another few weeks to arrive!
Or you could try a benchmark. Anything over 1000 MB/s will prove that it is at least 5.0 GT/s. Anything over 2000 MB/s will prove that it is 8.0 GT/s.
I tried four of those ADT-Link risers in the amfeltec. I think my problem was that folding the cables too much reduced their quality. Or maybe spacing four of them at 0.8" was adding too much stress to the cables. I cut a rectangular hole in the back of the amfeltec for the cables to exit the case (since the amfeltec is low profile, there was plenty of room). I think a M.2 to U.2 adapter would be better, because then you can buy any length U.2 cable, and the cable is more flexible, and can be disconnected at either end - but the U.2 connector on the amfeltec side might take more than one slot of space? The HighPoint SSD7120 probably solves that problem. On the other side, the U.2 to PCIe x4 adapters I've seen seem to be wider than necessary.What the hell, I'm going to hop in and join the party. Going to try the Amfeltec squid with the Titan Ridge card.
I'm liking the ADT-Link riser that wjesse tried. They have that riser with a right angle x4 connector, so I can try mounting the Titan Ridge card upside down in the 4th slot, and draw power for the riser from the SATA connector right above it. I can use a blank PCIe card for support.
Probably steal USB 2 for one of the TB3 connectors from the Bluetooth jack since I don't care about bluetooth.
I've got an Apple Thunderbolt display and TB3 <-> TB2 adapter laying around at the office.
What would be the best way to test link speed of TB3? I was thinking about a TB3 to NVME adapter, but I don't know which might be reliable.
Probably steal USB 2 for one of the TB3 connectors from the Bluetooth jack since I don't care about bluetooth.
Nice. I'm currently working on a hub solution for taking USB 2 from the bluetooth header. I don't believe that this has actually been done yet. A possible complication is that the header does not provide 5V but rather 3.3V. Hopefully, the Titan Ridge only needs the data lines... If you try this, please let us know!
I'll have to put a meter on that connector, and see if it is 3.3v.
I tried four of those ADT-Link risers in the amfeltec. I think my problem was that folding the cables too much reduced their quality. Or maybe spacing four of them at 0.8" was adding too much stress to the cables. I cut a rectangular hole in the back of the amfeltec for the cables to exit the case (since the amfeltec is low profile, there was plenty of room). I think a M.2 to U.2 adapter would be better, because then you can buy any length U.2 cable, and the cable is more flexible, and can be disconnected at either end - but the U.2 connector on the amfeltec side might take more than one slot of space? The HighPoint SSD7120 probably solves that problem. On the other side, the U.2 to PCIe x4 adapters I've seen seem to be wider than necessary.
About the Apple Thunderbolt display: I'm not sure it works properly with the GC-TITAN RIDGE. I believe Thunderbolt 1 devices do work (Apple Thunderbolt to FireWire or Gigabit Ethernet adapters) but maybe there's an issue with PCIe + DisplayPort traffic over Thunderbolt 1 from a GC-TITAN RIDGE? I don't know for sure that there is a problem.
Amfeltec support is wrong. The problem with the gen 3 was that it ran at gen 1 speed in the Mac Pro but there is a workaround for that to make it run at gen 2 speed, and the workaround is not required anymore when you use the Mojave firmware update for the MacPro4,1 and MacPro5,1. Read the amfeltec thread or the Highpoint thread. The gen 3 cards are faster because while the upstream in a Mac Pro must be gen 2, the downstream (NVMe drives) can be gen 3, giving up to twice the performance of NVMe drives in a gen 2 only card.It looks like you are using the gen 3 Squid (I see the fan). I grabbed the Gen 2 squid (gen 3 nvme connections) because amfeltec support said it was faster than the gen 3 model in the classic Mac Pros. Might be interesting to see if it behaves any different.
Yeh, I had a small argument with the sales guy at Amfeltec. I had made a stupid assumption the gen 2 card was still gen 3 downstream.
I believe there are apps for Windows and MacOS that can change the brightness of the LG UltraFine 5K using DDC/CI (over DisplayPort) instead of HID (over USB).No brightness/sound/video settings in Mojave or Windows
I believe there are apps for Windows and MacOS that can change the brightness of the LG UltraFine 5K using DDC/CI (over DisplayPort) instead of HID (over USB).
USB requires PCIe communication - which should work in Windows, but maybe Windows doesn't have a LG UltraFine 5K driver for brightness control? Check the Windows Device Manager to see if the PCIe devices of the LG UltraFine 5K are working (including the USB controller). Or maybe having to power cycle the display means the PCIe devices can't work without proper hot plug support. Maybe the PCIe devices will appear if you don't power cycle the display - use the second display to check the Device Manager. Maybe the display can be power cycled after the computer starts up, but before Windows loads?
The LG 34WK95U-W 34" 21:9 UltraWide 5K is a DisplayPort 1.4 display. It also supports dual cable DisplayPort 1.2 through Thunderbolt like the UltraFine 5K. The display is only 2160 tall instead of 2880, but this allows it to use 10 bpc where as DisplayPort 1.4 would only 2880p to do 8 bpc (DisplayPort 1.4 only has 75% the bandwidth of dual cable DisplayPort 1.2). DisplayPort 1.4 with Display Stream Compression (DSC) is required for higher bit depths / resolutions / refresh rates. I don't know of any displays using DSC and I don't know if MacOS or Windows drivers supports it.You're probably right! I've decided to return the Titan Ridge and the Ultrafine already though. So I can't check. Not worth loosing boot screen and the unaccelerated Mojave support. Will hold off until there are 5K using a single DP 1.4 connection instead.
Might reconsider if Apple decides to launch an aluminium 6K display on WWDC. That one will probably be thunderbolt only..
It's worse than that. It's 6 bit with FRC. It can't do 10 bpc with DisplayPort 1.4 at 5K without Display Stream Compression.This monitor is only 8 Bit: https://www.displayspecifications.com/en/model/c0a11069