Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
External Thunderbolt PCI Expansion Chassis and Hub in Development
- Thread starter MacRumors
- Start date
- Sort by reaction score
That is a pretty good call, I was thinking adding an AMD 6770 would be a pretty darn good performance boost.
Ok, so you haven't seen the facts either then I take it..facts are, that even at 4x speed that Thunderbolt runs at its more than enough bandwidth for a 6970 or so to run at 95% full speed, so really...it IS fast enough..., at least, that's what I believe to recall correctly, indeed.
Pessimists
I would think that the MBP and 21.5" iMacs have the Lightning Ridge TB chips, but simply have all 4x channels wired to their single TB ports, whereas the 27" iMac has the Lightning Ridge TB chip as well, however they have it wired so each port gets 2x channels.
2 TB channels = 1 DisplayPort, so if the MBA can support only one extra display via DisplayPort/Thunderbolt with it's Eagle Ridge TB chip, how come both iMacs and the MBP can support 2 DisplayPort displays if what your saying, only the 27" iMac has 4x TB channels ie. 2 DisplayPort capability? Obviously Lightning Ridge is inside the MBP and both size iMacs, it's just the 27" has split the channels across 2 ports instead of one. Since the cabling is going to be copper, you could assume we can expect relatively cheap 'Y' adapters coming out which will split a 4 channel TB port into 2x 2 channel TB ports, or will it simply daisy-chain more devices? lol i don't know... all I know is that so far, I've only heard the Eagle Ridge chip (2x channels instead of Lightning's 4x) applies only to the new MBAs
I would think that the MBP and 21.5" iMacs have the Lightning Ridge TB chips, but simply have all 4x channels wired to their single TB ports, whereas the 27" iMac has the Lightning Ridge TB chip as well, however they have it wired so each port gets 2x channels.
2 TB channels = 1 DisplayPort, so if the MBA can support only one extra display via DisplayPort/Thunderbolt with it's Eagle Ridge TB chip, how come both iMacs and the MBP can support 2 DisplayPort displays if what your saying, only the 27" iMac has 4x TB channels ie. 2 DisplayPort capability? Obviously Lightning Ridge is inside the MBP and both size iMacs, it's just the 27" has split the channels across 2 ports instead of one. Since the cabling is going to be copper, you could assume we can expect relatively cheap 'Y' adapters coming out which will split a 4 channel TB port into 2x 2 channel TB ports, or will it simply daisy-chain more devices? lol i don't know... all I know is that so far, I've only heard the Eagle Ridge chip (2x channels instead of Lightning's 4x) applies only to the new MBAs
Thunderbolt interface doesn't support more than two channels so my guess would be that Apple is only using half the channels in MBPs and 21.5" iMacs. It's similar to SATA in Macs, the PCH provides more channels (i.e. ports) but Apple has decided not to use them. You can't really slam the all channels into one port
13" MBP uses the bigger TB chip (Light Ridge) but doesn't support more than one external display. This is because the Intel HD 3000 doesn't support more than two monitors so this isn't a limitation of TB, but the GPU.
Seems like a great solution for someone who doesn't want to have both a laptop and desktop work station.
The Macbook Air appeals to me because of the portability but it doesn't have the graphics power that my job requires. This looks like it could solve the portable workstation dilemma.
And the price point is excellent and affordable!
The Macbook Air appeals to me because of the portability but it doesn't have the graphics power that my job requires. This looks like it could solve the portable workstation dilemma.
And the price point is excellent and affordable!
What, Thunderbust or expensive Thunderbust peripherals?
Will there be any Pros left that these are marketed to?
A few Aperture users? Some Logic users composing iBieber masterpieces?
Maybe the remaining users of of Final Cut Double Plus Super iPro Maximus.
Both of those experiments are fundamentally flawed in different ways.
In the first, they are testing SLI configs. TWO cards with some of the data traveling over the separate SLI interface. So you split the data traffic ( to reduce congestion/bandwidth) to the cards and then use another bus to transfer data. I'm not sure how that is suppose to demonstrate there are not significant effects on reducing PCI-e bandwith to a single card since they are deliberately avoiding the bottlenecks!
In the second, they are using synthetic benchmarks to boost the results.
Don't look at the summary, look at the individual game results. More than a few show greater than 10% drop-offs when chop down to a x4 limits. One game, 'Riddick DArk Athena' is actually backward. Chopped down bandwidth gets faster throughput ( perhaps triggers a better caching mode or reduced gameplay that testers didn't note). Then synthetic stuff is more geared toward polygon rates and invoking commands inside the GPU.which won't matter as much when it comes to data throughput. Both of these latter two issue skew the results.
If you put a card in a TB expansion port you will get a "lessor" card than what you paid for. That's better than nothing, but "almost too small to measure" conclusions aren't warranted by these experiments.
Last edited:
What real-time game reads textures off the hard drive in a just-in-time fashion when it needs them? Main memory of a computer is typically measured in GBs these days. The memory of more video cards is less than that. Just as the textures are cached in VRAM they are typically cached in RAM first ( usually during the disco show the games put on when they first start up). As textures are bumped to the VRAM that frees up space to cache what hasn't been used yet into RAM.
Sure large VRAM (and main RAM) take the pressure off but still have a bottleneck if push too much texture data across the wires (e.g., sudden scene changes with a completely different background. )
These are good set-ups for "better than what is soldered to the motherboard" solutions. But enablers of the ultimate gaming rig ..... not really.
Throw in USB, Ethernet, and Audio ports (or at least have a way to run audio through the graphics card in the external case), and this would make the perfect companion to a MacBook Air. It would be the perfect mobile form factor and a decent desktop gaming machine all in one. The only problem is now I'm really wishing Apple put the full Thunderbolt chip in the new Airs, so they could use 4 lanes rather than only 2.
Funny, I was commenting on needing something like this for MBPros a few days ago in the MBPro thread.
Upgrading the graphics card is the biggest issue with the MBPro and iMac.
If this comes out, it's no longer an issue and you can upgrade your VC a few years later.
Upgrading the graphics card is the biggest issue with the MBPro and iMac.
If this comes out, it's no longer an issue and you can upgrade your VC a few years later.
Those are the prices of the current Expresscard 34 version.....
When, and if, they do a TB version I predict a price surely higher than that, around 400$.
If they provide a real boost in graphic performance i'll surely buy one ASAP...
I have always been very doubtful and the point of my previous link wasn't that there is no difference (I never said that). All comparisons are flawed because PCIe is used. 2.0 x4 slot is good for 16Gb/s while TB is good for 10Gb/s. There is already a big difference. And that is just the raw bandwidth, TB will add its own latency (although only 8ns) which may cause even further handicapping.
Like I posted earlier, the enclosures provide only 225W so high-end GPUs will be severely limited by that already (TDP can be surpassed by 20-30% when under maximum load).
Three issues.
space on the sides: there is only TB socket. (and have already tried the flawed 1 USB socket experiment )
space in the inside: not much room for the larger chip package.
costs: the $999 needs to match margins with rest of Macs.
The "smaller" TB chip is for less expensive solutions. You get less, you pay less.
It would not be surprising to see a later MBP to mutate to have two TB sockets (e.g., retire FW ) plus it makes for uniform parts across the MBP line.
Yes but in reality people are not going to shell out money for extra hardware where they have to resort to hacks to get none supported graphics cards to work. No self-respecting company would market a device that requires people to hack their system.
Thunderbolt interface doesn't support more than two channels so my guess would be that Apple is only using half the channels in MBPs and 21.5" iMacs. It's similar to SATA in Macs, the PCH provides more channels (i.e. ports) but Apple has decided not to use them. You can't really slam the all channels into one port
and where have you read that?
Right on Intel's site.
http://www.intel.com/content/www/us...thunderbolt/thunderbolt-technology-brief.html
Apple TB hub
Agreed. It's interesting that Apple, as an early adopter/promoter of FW and now TB, has only played around with the hub concept in the Apple Displays but never in a full-fledged docking device.
If TB truly can support multiple channels (in higher powered machines, anyway) you'd think this would be the time for them to jump into the game. They have to able to come up with something better (and sexier) than the Bookendz dock approach.
C'mon, Apple. Step it up!
Agreed. It's interesting that Apple, as an early adopter/promoter of FW and now TB, has only played around with the hub concept in the Apple Displays but never in a full-fledged docking device.
If TB truly can support multiple channels (in higher powered machines, anyway) you'd think this would be the time for them to jump into the game. They have to able to come up with something better (and sexier) than the Bookendz dock approach.
C'mon, Apple. Step it up!