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.
Why is there no eGPU Thunderbolt out yet?
- Thread starter IceHot
- Start date
- Sort by reaction score
A few demos showed up in September with no estimates for releases nor even when a launch is planned. It seems at least 3 companies work on it but no one got it to work flawlessly just yet.
Usually between the first demo and being able to buy one has to expect at least about 6 months. I would expect them next summer. Spring if you are lucky.
If they show up even later than next summer I suspect there is some technical issue that cannot be fixed without some support from Intel or the GPU companies. Which might mean it will take even longer or until there is TB 1.1 out.
Afaik a big problem is the entire plug-play nature. Keeping the system from crashing when you pull the cable or damaging important data. Also getting it to run without having it ready at boot.
I suspect drivers are a bitch with this one and previous such attempts came from AMD/ATI which had the know how. Now it is normal companies like MSI that are supposed to work out the kinks with what seems limited support from the GPU companies.
BTW even once they are released I wouldn't be surprised if they only work on Windows. It seems not too easy or else they would be done already. The OSX drivers probably only Apple could really fix and on OSX nothing like AMD Dynamic Switching or Nvidia Optimus exists. I would suppose some such driver model would be necessary to work.
In bootcamp it should work as far as I go that would be all I need anyway.
Maybe this is one of the problems that most Thunderbolts are on Macs but actually getting those things ready to work in OSX is the hardpart that they try to figure out. Maybe they want it to work in OSX too at launch and since nobody is out the door yet they prefer to finish up before release. None of the Demos afaik would work on anything but Windows.
Usually between the first demo and being able to buy one has to expect at least about 6 months. I would expect them next summer. Spring if you are lucky.
If they show up even later than next summer I suspect there is some technical issue that cannot be fixed without some support from Intel or the GPU companies. Which might mean it will take even longer or until there is TB 1.1 out.
Afaik a big problem is the entire plug-play nature. Keeping the system from crashing when you pull the cable or damaging important data. Also getting it to run without having it ready at boot.
I suspect drivers are a bitch with this one and previous such attempts came from AMD/ATI which had the know how. Now it is normal companies like MSI that are supposed to work out the kinks with what seems limited support from the GPU companies.
BTW even once they are released I wouldn't be surprised if they only work on Windows. It seems not too easy or else they would be done already. The OSX drivers probably only Apple could really fix and on OSX nothing like AMD Dynamic Switching or Nvidia Optimus exists. I would suppose some such driver model would be necessary to work.
In bootcamp it should work as far as I go that would be all I need anyway.
Maybe this is one of the problems that most Thunderbolts are on Macs but actually getting those things ready to work in OSX is the hardpart that they try to figure out. Maybe they want it to work in OSX too at launch and since nobody is out the door yet they prefer to finish up before release. None of the Demos afaik would work on anything but Windows.
From what I see there isnt so much problems in that regard usually with the diy proposition
and that is more firmware than anything
Not a problem with the interface.
More of a problem of how willing intel is to provide small companies developer rights for thunderbolt.
More of a problem of how willing intel is to provide small companies developer rights for thunderbolt.
What do you mean by that? TB is basically PCI-Express, so a fairly simple controller/switch should do the trick.
Seems optimus can be convinced to work with a DIY eGPU.
http://forum.notebookreview.com/e-g...851-diy-egpu-experiences-123.html#post6542661
To feed back the video signal to the internal monitor.
Obviously limited to the GPUs that Nvidia supports for Optimus on both ends. Usually some Intel + some Nvidia external. Yet AMD has effectively the same thing in drivers so it should work there too.
http://forum.notebookreview.com/e-g...851-diy-egpu-experiences-123.html#post6542661
To feed back the video signal to the internal monitor.
Obviously limited to the GPUs that Nvidia supports for Optimus on both ends. Usually some Intel + some Nvidia external. Yet AMD has effectively the same thing in drivers so it should work there too.
The testing that have been done until now have all been on a thunderbolt to pci card adapter right? Wouldn't that limit the speed to the pci card? the max eGPU 3dmark 11 scores i have seen hovers around the 6000 mark, and that is with pci card. i Wonder how much a thunderbolt gpu could push it? 
probably because it will cost 1000+ still waiting on a cheap thunderbolt harddrive enclosure as long as they are still 160 bucks I wont pop for one.
The testing that have been done until now have all been on a thunderbolt to pci card adapter right? Wouldn't that limit the speed to the pci card? the max eGPU 3dmark 11 scores i have seen hovers around the 6000 mark, and that is with pci card. i Wonder how much a thunderbolt gpu could push it?
As far as the video card is concerned, it is plugged into PCIe.
However it is not 16x, much slower.
BUT. This is why video cards have local memory on them. To avoid pushing traffic over the PCIe bus. If you're streaming live data over the bus, performance goes to hell anyway, whether its 8x or 16x. If you get a card with plenty of memory the performance should still be pretty good unless the game has too many textures or whatever to fit into video memory all at once.
i dont understand **** about this thunderbolt external gpu thingy. Everybody has their own opinion on how fast it will be, but how can anybody know when there hasnt been any benchmarks whatsoever ???? why cant apple make an eGPU dock, for gods sake it would sell like grass.. goodnight pepps
Simple, because, just like the 17" MBP's, it would be used by a very small minority of clients. Why invest millions in R&D, setting up production, advertising and such for a few thousand sales a year? The return on investment for Apple just isn't there.
With the right advertising i forsee this to be very succesful. Apple could call it something along the lines of Gamebox or whatever, im sure it will sell. Problem is, normal non geeky people dont know that a "box" that could increase graphical performance by 10000% is possible...
With the right advertising i forsee this to be very succesful. Apple could call it something along the lines of Gamebox or whatever, im sure it will sell. Problem is, normal non geeky people dont know that a "box" that could increase graphical performance by 10000% is possible...
My point still holds in my opinion. 99% of geeks will tell you if you want to be gaming, a laptop isn't the right platform, a mac laptop is worst'
People dont know what they need, its just the like ipad man.
If you're going to be tethered to a desk with you eGPU that'll need a power supply of it's own, you're better off building a desktop PC for the price of the eGPU...
With the right advertising i forsee this to be very succesful. Apple could call it something along the lines of Gamebox or whatever, im sure it will sell. Problem is, normal non geeky people dont know that a "box" that could increase graphical performance by 10000% is possible...
That's because 10000% isn't possible. Given that games are gpu bound right now, it could improve it. You'd give up some performance due to the bandwidth available through thunderbolt. You can't expect native framerates. Right now a big problem would be thunderbolt certification and testing on the Windows side. You need to figure out a case that works and can be worked into the price. Standard gpus are not hot pluggable, so if plug and play is considered a necessary feature, they will pick pick up costs tweaking firmware. It could be $200 over normal graphics card prices with lower performance while lacking official OSX support. When has Apple ever pushed graphics upgrades? They would prefer you buy a new notebook or the upgraded imac. This would go against their current business model. Right now Apple has their RAID + docking station with mostly existing ports. Thunderbolt went in wherever mini displayport existed. I'm not convinced they care about it past that point.
Sony had an external graphics card for the Vaio Z over a year ago.
If Apple wanted it, it would have happened by now.
http://www.pcgamer.com/2011/09/20/hands-on-with-sonys-external-graphics-card-its-superb/
In fact even earlier than that
http://www.pcgamer.com/2011/06/30/fancy-an-external-graphics-card/
If Apple wanted it, it would have happened by now.
http://www.pcgamer.com/2011/09/20/hands-on-with-sonys-external-graphics-card-its-superb/
In fact even earlier than that
http://www.pcgamer.com/2011/06/30/fancy-an-external-graphics-card/
Last edited:
It seems to me that the entirety of TB is a niche market. Currently (and it has been out for quite a while now) anything TB is very expensive to the average consumer, so the only people buying and using TB tech are professionals, for the most part. That would lead me to believe that sales of anything TB related are low which leads me to conclude that ROI for Apple regarding TB is a fail at this point. I thing TB will stay a niche market as long as USB 3 is around and possibly go the way of the dodo once USB 4 rolls around.
From what I understand, a lot of the problem seems to be coming from exorbitant licensing fees being demanded by Intel in order for companies to create thunderbolt compliant peripherals. AFAIK it's just not possible for manufacturers to create thunderbolt products that are price-competitive when compared to USB 3.0 and so this restricts any potential sales and means they're reluctant to release thunderbolt products. In other words, Intel is killing its own technology.
No market for an external gpu?
That isn't market research. You found something on the internet. If you searched for donkey pr0n, I bet you'd find that too. I've read many times that Apple pushed for an exclusive release for the first year. I'm really not sure that's true. If it is, they weren't working in your best interest. You absolutely will not see the push for such a thing from the Mac side of peripheral devices. It's a tiny market. You're mainly looking at those with a 2011 to 2012 macbook pro or air. Minis sell because they're cheap. The number of people buying them outside these boards that would spend $600 or so to upgrade graphics is likely slim. 27" imac owners are also unlikely targets. You're throttled somewhat going through thunderbolt, making the gains a bit slim relative to the price.
This is the kind of thing that would take off on Windows first where a larger gaming market exists. They didn't even see thunderbolt until this year, and it takes a lot of work to get get these things to operate seamlessly on either platform, as graphics cards are not normally plug and play devices. You'd want to be able to recognize the device without rebooting. Ideally if the market is there, you'd see these things tested and produced by the same companies that distribute graphics cards today. Just don't expect the push to come from OSX.