- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
I have been told that the reason a peripheral Thunderbolt 4 drive enclosure (PCIe NVMe M.2 2280 M Key) provides a link speed test result of only 770MB/s to a Thunderbolt 4 Host, is because the Thunderbolt 4 controllers only supports one (1) PCIe 3.0 lane (PCIe 3.0 x1), as compared to the two (2) or four (4) PCIe 3.0 lanes supported by Thunderbolt 3 controllers.
Is this correct?
The vendor for the Thunderbolt 4 drive enclosure (PCIe NVMe M.2) also stated that the link speed test results, to a Thunderbolt 3 Host, for their other equivalent Thunderbolt 3 drive enclosures, have not been an issue because Thunderbolt 3 supports PCIe 3.0 x2, x4 lanes for Thunderbolt 3 M.2 2280 NVMe peripheral devices.
The vendor will not disclose whose Thunderbolt 4 Controller chip-set they are using.
But referencing the Intel JHL8440 Thunderbolt 4 (peripheral) Controller Product Specifications, at:
I find the following:
Supplemental Information
- Description: + Thunderbolt™/USB4 peripheral support at 40G
+ Native USB Type-C interface capabilities: USB2, USB3 (10G), DP1.4 Alt-mode
+ Tunneling capabilities (32G PCIe, USB3(10G), 2 displays (up to DP1.4)
+ Other native interfaces: x1 PCIe 8GT/s, 1x USB3 (10G)
I/O Specifications
- Port Configuration: Quad
This description for the Intel JHL8440 states:
1) Tunneling capabilities (32G PCIe ...
and
2) Other native interfaces: x1 PCIe 8GT/s...
Does "x1 PCIe 8GT/s" mean that this Thunderbolt 4 controller only supports one (1) PCIe 3.0 lane in the "link" to the Thunderbolt 4 Host controller?
Would this be why the Thunderbolt 4 drive enclosure (PCIe NVMe M.2) is only providing a link speed test result of 770MB/s to the Thunderbolt 4 Host?
If not, will someone please explain what is meant by the reference "x1 PCIe 8GT/s" reference?
It is my understanding that 8GT/s is the raw data transfer speed, in this case of 8Gb/s. Correct?
And the "32G PCIe" refers to the minimum data transfer requirement of Thunderbolt 4. Correct?
Also, would someone offer an opinion as to why a Thunderbolt 4 peripheral would have a 770 MB/s link speed test result?
In addition, I believe that the Thunderbolt 4 (Host) controller would be the Intel JHL8540 or the Intel JHL8340 controller.
If this is correct, should the Intel JHL8540 and the Intel JHL8340 controllers have the same specifications, as published for the Intel JHL8440?
I ask this because the information provided by the Intel JHL8540 and Intel JHL8340 Thunderbolt 4 Controller Product Specifications, at the following links are not the same as the Intel JHL8440.
and
Here is what is provided by the Intel JHL8540 and Intel JHL8340 Thunderbolt 4 Controller Product Specifications.
I/O Specifications
- PCI Express Configurations ‡ x4 Gen3
capabilities: ACS, FPB, PTM, P2P, 128b payload - Display Port 2 DP Sink, 1 DP SRC, DP1.4a tunnel/re-drive
x1/x2/x4 1.62/2.7/5.4/8.1
Thank you, in advance, for your time and assistance with addressing my questions.
Link Copied
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Hello @fp5849,
Thank you for contacting Intel Embedded Community.
"Thunderbolt 4 drive enclosure (PCIe NVMe M.2 2280 M Key) provides a link speed test result of only 770MB/s to a Thunderbolt 4 Host, is because the Thunderbolt 4 controllers only supports one (1) PCIe 3.0 lane (PCIe 3.0 x1), as compared to the two (2) or four (4) PCIe 3.0 lanes supported by Thunderbolt 3 controllers."
Seems correct, we have that with 1 lane, it can be transferred up to 985 MB/s at 8 GT/s, I find that the speed is accurate, because:
Version 3.x: 8 GT/s
x1: 985 MB/s
x16: 15.75 GB/s
https://en.wikipedia.org/wiki/PCI_Express
https://paolozaino.wordpress.com/2013/05/21/converting-gts-to-gbps/
So, in this case, let's do some conversions:
8 GT/s x (128b/130b) = 7.877 Gbps -> Converting bit to byte -> 7.877/8 GB/s = 984.62 MB/s
As for the test result being 770 MB/s I'm not sure, it seems there is something else to take into account, but the result seems realistic with 1 lane.
Now, the controller has 4 lanes more that can be tunneled through the Thunderbolt/USB4, and use a 4 lane PCIe, that is the 32 GT/s that is mentioned too, but as a PCIe Phy, it only has one lane, the other lines must be used through the USB4.
I think that the devices that can be compared together are these, as they are an Accessory Controller, that act as a Hub or a point of exit in the Thunderbolt/USB4 domain:
Goshen Ridge - JHL8440 -> x1 PCIe Gen 3
Titan Ridge - JHL7440 -> 4x PCIe Gen 3
The Maple Ridge controllers, the Intel JHL8540 and the Intel JHL8340, are Host Controllers, that act as a point of entry in the Thunderbolt/USB4 domain.
Also, you may use the Thunderbolt support if you need further help:
https://www.thunderbolttechnology.net/contact
Best regards,
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Diego,
Thank you for the explanation.
But why would Intel reduce the number of PCIe 3.0 lanes, in the peripheral (device) controller from a x4 PCIe End-point (JHL7440) with TB3 to a x1 PCIe End-point (JHL8440) with TB4?
Thanks again for your time.
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Hello @fp5849,
The JHL8440 has x4 PCIe TB4 but with the USB4 tunneled feature, the direct PCIe PHY is x1 lane TBT3. This decision could be a market decision being available the JHL7440.
Best regards,
- Subscribe to RSS Feed
- Mark Topic as New
- Mark Topic as Read
- Float this Topic for Current User
- Bookmark
- Subscribe
- Printer Friendly Page