Can todays Thunderbolt controller drive a Retina display?

[Semester]

A killer "feature" for a new mac pro would be if it could be bought along with a retina display. I've seen some people hoping for it this summer, but is that possible with this years technology?

One selling point of the next generation thunderbolt controller (coming 2014) is that it can drive a 4k television. If Apple were to double the resolution of their current thunderbolt display the display would have more pixels than a 4k television. So I guess todays thunderbolt controller can't drive a 27" retina display. Am I correct?

 

[Giuly]

Todays Thunderbolt controllers feature DisplayPort 1.1a, so they top off at 2K resolutions.

Though I'm still puzzled how Apple intends to offer Thunderbolt on the Mac Pro at all, and I guess so is Bob Mansfeld.

 

[mm201]

But is it a hardware limitation or can a firmware update provide DP 1.2 support?

 

[theluggage]

So I guess todays thunderbolt controller can't drive a 27" retina display.

Maybe not, but a 21.5" display with the same 2560x1600 resolution as the 13" rMBP could probably be sold as 'retina' on the grounds that the typical viewing distance is greater, and would be within current interface capabilities. The text would be rather large in "best for retina" mode, but with that resolution, the scaled modes are going to look pretty good.

I have to say, though, when I stare at my 27" cinema display, "gosh, this really needs more resolution" is not the first thing that goes to my head.

Todays Thunderbolt controllers feature DisplayPort 1.1a, so they top off at 2K resolutions.

Well, there are 2 DisplayPort 1.1 channels available, so perhaps they could be combined. Or some of the data could be sent over the PCIe link.

Though I'm still puzzled how Apple intends to offer Thunderbolt on the Mac Pro at all, and I guess so is Bob Mansfeld.

Maybe all, some or fewer of:

Option 1: Base model comes with integrated graphics on the motherboard that are good enough for anything short of pro-graphics applications. Assume anybody buying a specialist pro graphics card is also going to buy a specialist pro display - or put a secondary DisplayPort 1.2 input on any new TB display.

Option 2: Offer PCIe graphics cards with internal DisplayPort outputs and run flying leads to the thunderbolt controller on the motherboard.

Option 3: Offer graphics cards with thunderbolt controllers.

 

[thekev]

But is it a hardware limitation or can a firmware update provide DP 1.2 support?

This suggests there is a bandwidth increase coming in 2014 as I thought. Redwood ridge provides displayport 1.2. I initially thought it was falcon ridge that added that. Intel has a really silly naming convention.

Well, there are 2 DisplayPort 1.1 channels available, so perhaps they could be combined. Or some of the data could be sent over the PCIe link.

Given that no one has tested that, I wouldn't bet on it. I'm not sure that the current ones can actually allocate data channels to displayport. Displayport 1.2 compliance is supposed to be the first thunderbolt chip revision prior to a bandwidth increase for PCIe 3.0. I really don't think the first thunderbolt chips will see 4k.

 

[g4cube]

Current thunderbolt has DP 1.1a pass-thru.

Haswell has support for DP 1.2 in the 4600 graphics controller. If used with current Thunderbolt controllers up to and including Cactus Ridge, still limited to DP 1.1 performance.

Redwood Ridge Thunderbolt controllers will start appearing along with Haswell computers, and are capable of supporting DP 1.2. Most likely support of the max resolutions will steal bandwidth from other Thunderbolt peripherals. Maybe the new Apple Thunderbolt displays will be waiting until Haswell Macs are announced.

Not enough information available about Falcon Ridge to make any assumptions other than to know not to expect any solutions until 2014.

 

[mm201]

If each of the TB lanes can be used to drive a separate DP 1.1 signal, they could use strip-based input, like the IBM T221 did, only through a single cable. A smart driver can expose it as a single monitor and synchronize the vertical refresh.

Alternatively, there could be some "thunderbolt to DP 1.2 adapter" that constitutes two thunderbolt lanes into a single DP 1.2 signal.

 

[potatis]

According to this, http://www.emsai.net/projects/widescreen/bandwidth/, DP 1.2 (= Redwood Ridge = Haswell macs) will allow for max 4000x2400@60 Hz. So for current Thunderbolt macs, no retina display, for Haswell macs, max 24".

 

[Giuly]

According to this, http://www.emsai.net/projects/widescreen/bandwidth/, DP 1.2 (= Redwood Ridge = Haswell macs) will allow for max 4000x2400@60 Hz. So for current Thunderbolt macs, no retina display, for Haswell macs, max 24".

'Retina' of 27" displays is pretty much 3840x2160.

Maybe all, some or fewer of:

Option 1: Base model comes with integrated graphics on the motherboard that are good enough for anything short of pro-graphics applications. Assume anybody buying a specialist pro graphics card is also going to buy a specialist pro display - or put a secondary DisplayPort 1.2 input on any new TB display.

Option 2: Offer PCIe graphics cards with internal DisplayPort outputs and run flying leads to the thunderbolt controller on the motherboard.

Option 3: Offer graphics cards with thunderbolt controllers.

I had that one, but recent rumors have a bit more out-of-the-box thinking applied. Option 1 and 3 are not allowed by Intel (yet), and Option 3 is also bad business, defeating the purpose of buying a Mac Pro for the Thunderbolt port as every Windows PC would have them running with cracked Bootcamp drivers one week after the release.

 

[potatis]

'Retina' of 27" displays is pretty much 3840x2160.

Apple's definition of retina is to double all the UI elements in OSX so that would make thing kind of large, like on a 1080p 27" display which is 82 dpi.
But a retina 27" would also be possible with the Haswell macs, although with a bit less screen estate than the current 27" which is 109 dpi.
A retina 27" in 16:9 format and 189 dpi could be 4448x2498@60Hz which is 20 Gbit/s.
Things would be the same size as the old LED 24" which is 95 dpi.

 

[Giuly]

Apple's definition of retina is to double all the UI elements in OSX so that would make thing kind of large, like on a 1080p 27" display which is 82 dpi.
But a retina 27" would also be possible with the Haswell macs, although with a bit less screen estate than the current 27" which is 109 dpi.
A retina 27" in 16:9 format and 189 dpi could be 4448x2498@60Hz which is 20 Gbit/s.
Things would be the same size as the old LED 24" which is 95 dpi.

Apple's official definition of retina is pixels so small that it wouldn't make a difference if you'd put more of them on the display because your eyes can't tell the difference.

The @2x stuff applied to iPhones, Macs can drive displays of any resolution and set the DPI independently.

The 13" Retina MacBook Pro for example doesn't have 2x the resolution of the the 13" cMBP.

3840x2160 is actually 4x the pixels of 1080p while the current TBD at 2560x1440 has 4x the pixels of 720p, so the whole thing is equal to the move from 720p to 1080p.

 

[theluggage]

Option 1 and 3 are not allowed by Intel (yet),

For Option 1 I meant 'integrated on the motherboard' - i.e. what the 15" MBPs and iMacs already do. I don't think Intel have any prohibition on that.

and Option 3 is also bad business, defeating the purpose of buying a Mac Pro for the Thunderbolt port as every Windows PC would have them running with cracked Bootcamp drivers one week after the release.

Can't see why this would worry Apple - its not as if they have an exclusive on Thunderbolt anyway, and AMD/NVidia would want to produce their own versions of the cards.

 

[potatis]

The @2x stuff applied to iPhones, Macs can drive displays of any resolution and set the DPI independently. The 13" Retina MacBook Pro for example doesn't have 2x the resolution of the the 13" cMBP.

Yes it does, 2x in x and y direction respectively that is. cMBP: 1280 x 800. rMBP: 2560 x 1600.
A standalone retina display would for certain be designed for OSX HiDPI mode which uses 2x the size of UI elements in y and x directions. Or 4x if you think that way. They won't do a screen that is physically meant to be a "scaled" resolution, but a "best (retina)" resolution only.

 

[Giuly]

Can't see why this would worry Apple

It equally worries Intel, as Thunderbolt is a reason to buy a new computer. Hence, no Thunderbolt controllers on graphics cards to retrofit it.

Yes it does, 2x in x and y direction respectively that is. cMBP: 1280 x 800. rMBP: 2560 x 1600.
A standalone retina display would for certain be designed for OSX HiDPI mode which uses 2x the size of UI elements in y and x directions. Or 4x if you think that way. They won't do a screen that is physically meant to be a "scaled" resolution, but a "best (retina)" resolution only.

Right, it is twice the resolution on the rMBP. Yet, it's not twice the size of the old 15" HiRes display either.

Still, it can be any size in-between, it doesn't have to be 2x. A 1080p display looks as good as a 1440p one, and so will a 3840x2160 display. Everything except for graphics is Display PostScript and scales to whatever you want, and scaling down graphics doesn't result in a loss of quality, upscaling does.

 

[theluggage]

It equally worries Intel, as Thunderbolt is a reason to buy a new computer. Hence, no Thunderbolt controllers on graphics cards to retrofit it.

...but no high-end graphics cards with Thunderbolt = no chance of Thunderbolt taking off as a display connection standard. Intel have a window of opportunity to position Thunderbolt 2 as the go-to connector for 4k video.

and scaling down graphics doesn't result in a loss of quality

Yes it does - apart from the obvious "you're reducing the resolution - duh!" bitmaps optimised (anti-aliasing*, thickening thin lines etc.) to be displayed at 32x32 will potentially look better than bitmaps optimised to be displayed at 100x100 and subsequently downscaled to 32x32. If you do have to re-size, up or down, then using a factor of 2 will give the best results.

(* one way of doing that is to render it at 2x resolution then downsample - but if you're doing that you don't anti-alias the 2x version first)

Also, some software first renders vector images to an internal bitmap - not all software is smart enough to read the screen resolution and adjust the size of these internal bitmaps to match, so the OS has to upscale them. Again, using a 2x factor gives the best results.

...however, these could be pretty picky differences if the scaling is done using good algorithms and the target display is high enough resolution: certainly the "scaled modes" on the rMBPs look pretty good.

 

[Giuly]

...but no high-end graphics cards with Thunderbolt = no chance of Thunderbolt taking off as a display connection standard. Intel have a window of opportunity to position Thunderbolt 2 as the go-to connector for 4k video.

DisplayPort 1.2 is the connector for 4K video, and Thunderbolt 2 is compatible to it.
Intel cares to sell the CPU in your new computer, Apple and every other computer manufacturer cares to sell those computers. Whether Drobo or LaCie sell Thunderbolt or USB 3.0 drives is rather secondary from their point of view, as long as they sell Thunderbolt drives for those of us who want them at all.

 

[tintinV]

First, I'm not a genius in this topic but I see that the Macbook Pro with Retina display comes equipped with the Pericom PI3VDP12412 (according to techrepublic) which is a x4 lane displayport 1.2 switch.

Does this mean that it can theoretically support a 4k monitor in the future over 1 thunderbolt port?
Or is it still limited internally by the thunderbolt controller (which i suppose it's intel)?

 

[Fishrrman]

"But is it a hardware limitation or can a firmware update provide DP 1.2 support?"

No expert here, but I don't believe you're going to see any software/firmware update that will upgrade existing Mac Thunderbolt ports to the DP 1.2 standards.

Probably the ONLY way to get such newer standards, will be to buy newer hardware with the support built into it.