Gigabyte Releases Thunderbolt 5 Add-in Card with PCIe 4.0 Interface

[TheLostSwede]

Gigabyte has quietly added the first Thunderbolt 5 add-in card to its website under the rather unimaginative name of Thunderbolts 5. The card itself is nothing remarkable when compared to Thunderbolt 4 add-in cards, although it sports a third mini DisplayPort input for unclear reasons, as well as two USB Type-C outputs. This time around Gigabyte has added support for more power though, as the card features a pair of what appears to be 6-pin PCIe power connectors and the card supports up to 100 W USB PD charging. Courtesy of the new JHL9580, or Barlow Ridge controller, the card offers support for DP 2.1 with resolutions of up to 8K at 60 Hz, as well as data speeds of up to 80 Gbps, or 120/40 Gbps in asymmetric mode.

Intel's JHL9580 controller has a PCIe 4.0 x4 host interface, the same as ASMedia's ASM4242 USB4 host controller and it's a big step up from Intel's previous Thunderbolt 4 controllers that were stuck with PCIe 3.0. Note that the JHL9580 is also offered in a Thunderbolt 4 version known as the JHL9540 which also uses a PCIe 4.0 x4 interface and offers DP 2.1 support, but slower speeds. Both of the new Barlow Ridge controllers also support USB 3.2 Gen 2x2 (20 Gbps), something previous Thunderbolt controllers didn't. Intel has a list price for the JHL9580 of US$19, which is just over US$8 more than the JHL8540 Maple Ridge controller which launched in 2020, as such, expect Thunderbolt 5 add-in cards to come at an even higher price point than previous Thunderbolt 4 add-in cards. Gigabyte might be first out of the gate, but we expect most motherboard makers to follow suit with products of their own before the end of the year.



View at TechPowerUp Main Site | Source

 

[Cheeseball]

Ah this is good. JHL9580 should be able to handle proper PCI-E 4.0 x4 bandwidth (~63 Gbps) for PCI-E tunneling.

Now I just need to wait for ASRock's version and hope it works with X870E (aside from the on-board ASM4242).

 

[TheLostSwede]

Ah this is good. JHL9580 should be able to handle proper PCI-E 4.0 x4 bandwidth (~63 Gbps) for PCI-E tunneling.

Now I just need to wait for ASRock's version and hope it works with X870E (aside from the on-board ASM4242).

So far, it's Intel only from what I can gather, but I guess we'll find out soon enough.
Also note the motherboard still need Thunderbolt headers for these add-in cards to work, which the X870/E boards appear to be lacking at least...

Edit: Some bad news for you, ASRock doesn't seem to have Thunderbolt headers on any of their boards with USB4 or Thunderbolt 4.

 

[Cheeseball]

So far, it's Intel only from what I can gather, but I guess we'll find out soon enough.
Also note the motherboard still need Thunderbolt headers for these add-in cards to work, which the X870/E boards appear to be lacking at least...

Edit: Some bad news for you, ASRock doesn't seem to have Thunderbolt headers on any of their boards with USB4 or Thunderbolt 4.

I know, it sucks balls. LOL

Most likely because they're relying on the built-in USB4 now. Which is fine for my current use case.

 

[TheLostSwede]

I know, it sucks balls. LOL

Most likely because they're relying on the built-in USB4 now. Which is fine for my current use case.

At least the ASM4242 gives you the full 40 Gbps, well, maybe not for Thunderbolt, but it's faster than the old Intel chips.
The new chips from Intel should do a lot better, but I'm curious how Intel is going to show their Thunderbolt 5 chip do 80 Gbps, when the PCIe interface is limited to ~64 Gbps...

 

[Tek-Check]

Note that the JHL9580 is also offered in a Thunderbolt 4 version known as the JHL9540 which also uses a PCIe 4.0 x4 interface and offers DP 2.1 support, but slower speeds

So, JHL9440 and JHL9540 have the same technical capabilities as JHL9480 and JHL9580 (Intel website), but are hard capped at 40 Gbps in implementation, right?

 

[ypsylon]

Wow, it would be double nice if this works on old TRX40 platform (Designare & Extreme have 5-pin TB headers).

 

[StimpsonJCat]

So Intel sell these chips for $19 in standard 1000 unit reels/trays, so the likes of Gigabyte and the rest are most likely buying them for $12 a pop, maybe less if they have a good relationship. The PCB cost to a large OEM is probably $3, and we add the rest of the components to the card, as well as the decorative stuff and the bracket, and retail packaging, and we are look at an approx BOM cost of about $35 at the most, and another $3 per card shipping/handling costs. Intel has already done the PCB layout, R&D, drivers etc. These cards would sell with a nice profit for about $100 - I wonder what the retail price actually will be?

 

[TheLostSwede]

So, JHL9440 and JHL9540 have the same technical capabilities as JHL9480 and JHL9580 (Intel website), but are hard capped at 40 Gbps in implementation, right?

Based on the information available, yes, that should be the case.

This is from earlier this year and suggests both use the same 13 x 13 mm packaging.

Details of Intel's Barlow Ridge Thunderbolt 5 Controller Leaks

Serial leaker @yuuki_ans on X/Twitter has released details on Intel's upcoming Barlow Rridge Thunderbolt 5 controller which will be known as the JHL9580 or JHL9540 depending on the SKU. The good news is that Intel has finally moved to PCIe 4.0 for the bus interface, which was expected due to the...

www.techpowerup.com

Wow, it would be double nice if this works on old TRX40 platform (Designare & Extreme have 5-pin TB headers).

Check the cables on Gigabyte's website, as they appear to have changed for older Thunderbolt cards.
Unfortunately, there's no compatibility list and no manual so far.

So Intel sell these chips for $19 in standard 1000 unit reels/trays, so the likes of Gigabyte and the rest are most likely buying them for $12 a pop, maybe less if they have a good relationship. The PCB cost to a large OEM is probably $3, and we add the rest of the components to the card, as well as the decorative stuff and the bracket, and retail packaging, and we are look at an approx BOM cost of about $35 at the most, and another $3 per card shipping/handling costs. Intel has already done the PCB layout, R&D, drivers etc. These cards would sell with a nice profit for about $100 - I wonder what the retail price actually will be?

1,500 units on a reel. I very much doubt they get that big of a discount with a newly launched product, as Intel doesn't really offer discounts on newly launched products.
None of the distributors I checked offer more than single reel pricing at the moment and the per unit pricing through distribution is US$22.64 per chip if you buy a reel or if you buy 200 units. I would guess more like US$50 total cost to the manufacturers at the moment.
As for MSRP, I'm guessing these will launch for US$150-200-ish, unfortunately. Many Thunderbolt 4 cards are listed for US$230+ still. MSI appears to be the cheapest one, as they have a sale on for $50 on their own online store for the MS-4463 v3.0.

 

[Tek-Check]

This is from earlier this year and suggests both use the same 13 x 13 mm packaging.

One bizarre thing is that the controller does not support DP54 mode (UHBR13.5), only DP40 and DP80.
This means that connecting AMD 7000 client GPU would drop the speed from DP54 to DP40.
Why did they do this?

 

[lexluthermiester]

Something stands out to me in those photos and article: 2x 6pin PCIe connectors. Really?!? Does this card really need 200w+ of power?

 

[Nanochip]

ASUS listed its thunderboltex 5 Barlow ridge card at least week ago. Also, many of the ASUS z890 motherboards (like say the Apex) that have onboard TB4 also have a TB header presumably for connecting the ASUS Barlow ridge card. Previous ASUS motherboards with onboard thunderbolt (like the z690/z790 hero or formula) did not have a thunderbolt header. So at least there seems to be some progress there on z890 with supporting both onboard and discrete thunderbolt. I guess the onboard solution is using the controller embedded in the arrow lake CPU.

Hopefully the next Lake embeds TB5. After us paying guinea pigs aka beta testers help to debug Barlow ridge.

ThunderboltEX 5｜Motherboards｜ASUS Global

Intel® certified Thunderbolt™ 5 card with dual USB Type-C® ports, 120Gbps bandwidth, supports 3x 8K displays, and up to 130W fast charging.

www.asus.com

 

[SRB151]

Wow, it would be double nice if this works on old TRX40 platform (Designare & Extreme have 5-pin TB headers).

I have a TB4 Maple Ridge card, could never get the TRX40 Designare to recognize it, wound up putting the TB3 card back in.

However, the TRX50 Aero has TB4 built onto the motherboard, but doesn't advertise it as such since they did not get Intel to certify it. My understanding is that it works with all the hardware, just won't support some of the new features they announced (two pc's with TB connection, etc.). Works fine with the Apple TB1 adapter and Sonnet pcie expansion enclosure.

Something stands out to me in those photos and article: 2x 6pin PCIe connectors. Really?!? Does this card really need 200w+ of power?

It's not for the card itself. It's for the Power Delivery pass through for any TB devices that may need it.

 

[kapone32]

Releasing a PCIe card when they themselves reduce PCIe lanes for the sake of M2. Why are we ok with losing the flexibility of PCIe to gain more M2 slots? I know it is a nothing burger if you are on a HEDT. I guess these are for people that want TB for whatever reason on Desktop.

 

[ypsylon]

Check the cables on Gigabyte's website, as they appear to have changed for older Thunderbolt cards.
Unfortunately, there's no compatibility list and no manual so far.

Before writing previous piece I visited the GBT website. Yeah I don't expect they'll make it compatible, but there is always hope. Both 3 & 5 pin cables were delivered with older TB cards from GBT (including one added to DesignareTRX40). I wouldn't put past any corporation changing pinout on the header out of spite, but if all 5 lines are the same then its simple task of swapping the lines in the plug. I never encountered old 3-pin header in the wild, but perhaps its found in some SFF or something like this.

 

[StimpsonJCat]

Why is Thunderbolt still such a hot mess?

 

[TheLostSwede]

Before writing previous piece I visited the GBT website. Yeah I don't expect they'll make it compatible, but there is always hope. Both 3 & 5 pin cables were delivered with older TB cards from GBT (including one added to DesignareTRX40). I wouldn't put past any corporation changing pinout on the header out of spite, but if all 5 lines are the same then its simple task of swapping the lines in the plug. I never encountered old 3-pin header in the wild, but perhaps its found in some SFF or something like this.

I could just be that they've changed the supplier/design of the cables, but at least one of the two looks different. However, it doesn't seem to be different on the motherboard side, but we won't know until they put up a compatibility list.

Why is Thunderbolt still such a hot mess?

Ask Intel and Apple, they're the ones in charge. At least people can't blame the USB-IF this time around.

One bizarre thing is that the controller does not support DP54 mode (UHBR13.5), only DP40 and DP80.
This means that connecting AMD 7000 client GPU would drop the speed from DP54 to DP40.
Why did they do this?

Where did you see that?

It's not for the card itself. It's for the Power Delivery pass through for any TB devices that may need it.

Or just for charging your phone/tablet/laptop/etc.
Loads of people here have complained about USB4 not supporting higher Wattages, well, this is the solution for some them at least.

 

[lexluthermiester]

It's not for the card itself. It's for the Power Delivery pass through for any TB devices that may need it.

You missed the point. What would need that level of power that someone would plug into the back of their PC? Seriously, who's going to do that and not use the charging adapter the device came with?

 

[Tek-Check]

Where did you see that?

Intel's spec for Barlow Ridge controllers list only UHBR10 and UHBR20 modes.
The same applies for iGPU on Meteor Lake, Lunar Lake and Arrow Lake in the sections for USB-C subsystem and graphics display.

UHBR13.5 is explicitly mentioned as NOT supported in official documentation for 200S (p.147) and 200V (p. 173). It's weird, considering the fact that upcoming Battlemage GPUs have 13.5 mode listed in Linux drivers.

200S	200V

 

[TheLostSwede]

Intel's spec for Barlow Ridge controllers list only UHBR10 and UHBR20 modes.
The same applies for iGPU on Meteor Lake, Lunar Lake and Arrow Lake in the sections for USB-C subsystem and graphics display.

UHBR13.5 is explicitly mentioned as NOT supported in official documentation for 200S (p.147) and 200V (p. 173). It's weird, considering the fact that upcoming Battlemage GPUs have 13.5 mode listed in Linux drivers.

200S	200V
View attachment 368289	View attachment 368290
View attachment 368291

So in other words, Thunderbolt 5 and Intel's new CPUs are not DP Alt mode compliant or even fully DP compliant. Good going Intel.

 

[Tek-Check]

So in other words, Thunderbolt 5 and Intel's new CPUs are not DP Alt mode compliant or even fully DP compliant. Good going Intel.

Not sure about Alt. With Hayden Bridge retimers, some systems should be able to run DP40 in Alt more over TB4 cables, but this is to be confirmed.
But yeah, UHBR13.5 does not seem supported on any product.

 

[billEST]

So Intel sell these chips for $19 in standard 1000 unit reels/trays, so the likes of Gigabyte and the rest are most likely buying them for $12 a pop, maybe less if they have a good relationship. The PCB cost to a large OEM is probably $3, and we add the rest of the components to the card, as well as the decorative stuff and the bracket, and retail packaging, and we are look at an approx BOM cost of about $35 at the most, and another $3 per card shipping/handling costs. Intel has already done the PCB layout, R&D, drivers etc. These cards would sell with a nice profit for about $100 - I wonder what the retail price actually will be?

same magin in car , tv .. oled panel its 80/150 e out the fabric