Is there any thunderbolt eGPU solution available

So i may buy the MBP if there is a thunderbolt eGPU solution that i coul use. How faster will thunderbolt be compared to the express card slot eGPU solution on the x220.

 

the 6750m on the latest MBP refresh is not going to run Metro 2033 maxed, obviously, but it is already fast enough to deliver decent framerates for most games at native resolution.

At least it's ranked higher than the 5830m, which works just fine for me for about any game I've tried at 1920x1080.

 

Sonnect Echo Express

 

Yeah i found that too but that has not been launched yet. Is that the only solution right now, any eta for it.

 

Don't know about MBP but isn't the new Z also supposed to have a external GPU connected via thunderbolt.
A question, do you think we will be able to use this sonnet stuff and have the display on the MBP screen itself or will we have to get a external monitor?

 

you would still need it more than probably

the Z is also a 2000 laptop, not to mention sony is also going proprietary again! and the gpu is not what was commented and that I advertised, its not going to be a 6770m, probably a 6630m\

I do hope that the concept spreads forward

 

Aren't MBPs also 2k$ laptops(if you consider the 15inch ones). The reason i mentioned the Z was cause it will have eGPU support from sony itself and will be able to run it on its own display, so chances are if we use these Sonnet expansion chassis with it there will be some way to get the eGPU to work with the laptop display, and i dont think it would be possible with the MBP.

 

The Thunderbolt port also carries a Displayport signal. Sony just needs to wire that as an input into the LCD.

The other way of doing it would be for the pci-e signal to carry the display image signalling and use the iGPU as the frame buffer, just like NVidia Optimus does. Problem with this approach is it uses a portion of the total bandwidth available for output display.

I'm sure Sony will use the Displayport approach to maximize bandwidth.

 

So that means that if someone replaces Sony's external GPU with their own thunderbolt eGPU solution they should be able to use it on the laptops display right. And as i sad such a thind would not be possible with the MBP.

 

If it provides the same input signalling into the Thunderbolt port then there is no reason why it shouldn't work. If using a desktop video card then the output DP port would need to be routed to be an input into the Thunderbolt port. I'm sure Sonnet will be aware of just what people want to put into their box and provide an adapter to do just that.

bandwidth: using an expresscard-based eGPU setup

Maximum bandwidth is x1 1.0 (2.5GT/s). This will double to x1 2.0 when hwtools finally provide a pci-e 2.0 compliant hardware for use on a Series-6 chipset. Or alternatively just use a Villagetronic solution which already claims pci-e 2.0 compliance.

Now with NVidia cards we've been able to get x1 pci-e compression going which approximately doubles bandwidth. So the best that we'll get will be approximately equal to x2 2.0 (10GT/s) with an NVidia Fermi card. If using the Optimus internal LCD mode then part of this bandwidth carries the display image data back to the iGPU for display.

bandwidth: using a Thunderbolt port

Sonnet suggest it's a x4 link. Intel say it's 10Gbps, which means x4 1.0. This is odd given that Series-6 chipset can do pci-e 2.0. So is it x4 2.0 (20GT/s) or x4 1.0 (10GT/s)? Now if it's x4 1.0, which I'm 90% sure it won't be, then it's going to be pretty much the same as the top-end x1.2Opt NVidia expresscard eGPU implementation.

A Thunderbolt port also has a Displayport input signal wire.

If it's x4 2.0, then I've read a report somewhere that x4 2.0 gives 95% of a x16 2.0 link's performance. So we are looking at very close to full desktop video performance. In addition, the separate DP input signal means none of this bandwidth needs to be used up to carry the final display image to the LCD.

Conclusion

A x4 2.0 Thunderbolt port eGPU implementation is superior in every single possible way: more bandwidth, designed for internal LCD mode, small port size.

Now just need notebook vendors to incorporate Thunderbolt across the board. The bean counters that have been dependant on consumers to upgrade their notebooks every two years to get better dGPUs won't like this since now they could upgrade their eGPU by just buying a new desktop video card.

 

^^^hear hear!

 

So if like the mac book air comes out with a thunderbolt port then it would be just as good as the 11 inch alienware?

 

MBA will weigh significantly less and potential to leave the m11xR3 in the dust when use a TB attached eGPU setup. The M11xR3 is a 2kgs (4.4lbs) monster.

 

thank you nando4, good to know that this new port actually has a use

 

I was super interested in Thunder bolt in the early days of the announcement, when they originally planned to have fibre optic cabling with rockstar speeds, but then they cheaped out and went with copper. As far as I know its slightly better then USB 3.0 I believe and no on really uses that yet. Unfortunately the push for anything after USB 2.0 has kinda been for enthusiasts. Its tough to switch something new when the average consumer has no clue about 3.0 and thunderbolt or really see the benifits in it, yet! Im sure there like DX9 there will be a point were people stop developing for 2.0 and switch to thunderbolt.

This interface would also make that eGPU a great alternative. It makes sense, the reson why most people buy seperate functional devices, if I only really play games on my big screen, whats the point of lugging the gpu around when I could have a lighter laptop for work and school, then snap it into a dock to get extreme gaming. Most people would argue at that point to just get a console though.

 

it just means that the signal converter, which is expensive for telecoms, could not achieve what they wanted

its a 10gb/s dual channel connection, you dont have to share the traffic with the info that is coming in, with the info that is coming out, its quite faster than the USB 3, which is theoretical 5gb/s with a single lane. SATA III is 6gb/s. Basically its the fastest I/O that we have that is simple and portable to use, pcie is still faster

Indeed I agree, and people havent stopped to develop for dx9

that is the idea behind the ultrabook concept from intel, Im anxiously waiting for it to arrive

 

NICCCEEE Drools. Can't wait if this is going to be a reality! Will spell the death of the desktop!

I had similar concerns but I think when fiber optics become cheaper you will see the return of Intel's 'Lightpeak'

 

fiber is cheap, the signal converter is the tricky part

 

When that happens.... Hello to GTX590 SLI on a i7 Macbook Pro.

 

Yeah....you can have your own mini-volcano right next to you for the handy hot-dog roasting multitasker!

Mr. Mysterious

 

Then there is possibility of this coming down the pike, albeit not terribly soon...

PCI Express cables could take us to 32Gbps speeds by 2013 -- Engadget

 

::drool::

Nando, why do you create such glorious visions of a bright thunderbolt future!
Now this is all I've been thinking about for the few days and can't get this potential out of my mind! I really, REALLY hope sonnet creates a perfect solution for an egpu for the thunderbolt.

Like everyone else, I wonder how this will work in terms of drivers on the macbooks and also the possibility of using the internal display.

 

Just a bump to see if there have been any updates on this situation

 

no there havent, only that village tronics is developing another one, I hope hey release it soon

 

False, there is one. Doesn't look like it's released yet actually, but they promise it's coming. Sonnet - New Product Announcement at NAB 2011

 

I guess what Mr. MM was saying was about updates, as i had asked about updates on the Sonnet product you mentioned.
So no more news of the sonnet, Nice to here that another company is developing a something similar, hopefully we will have a working solution by the holidays.

 

if you don't require gaming on the move then yes. otherwise 11 inch alienware is still the better choice as eGPU solution is not very portable (a desktop graphics card plus desktop PSU plus some cables)

 

There are users creating "portable" 12V/120W eGPU solutions using lower powered cards like a GTX550Ti and small AC adapters to drive them. The combo is not as portable as a m11xR.. but a Thunderbolt based GTX550Ti would deliver GTX560M performance levels so is considerably more powerful.

 

indeed and that is what I want, a more portable solution, thus the sonnet is one hell of a product for me, since it would drive probably a 5770 (apple drivers issue), and I would game when Im on vacation elsewhere.

thats what I meant, there are some NAS on development, also cams, and several other high bandwidth products.

There is the Pegasus NAS already available, more products are due to this summer. And next year.

 

BPlus Thunderbolt product, PE5H 1.0 + enclosure, has an ETA of end-November. See the Thunderbolt, USB 3.0, PCIe 2.0 eGPU update for details.

 

It's not that linear. we even have pci-e 3.0 x16 around because pci-e 2.0 x8 is already bottleneck for high-end sli cf setups. in the future it will be worse. just imagine making a egpu and being happy about it because you have let's say, 70% of the performance using a pci-e 2.0 x1 and then upgrading to a last generation high-end card like the new ATI just around the corner and jut having 50% of the performance... not good.
95% would be in perfect conditions where you would use it in a motherboard without any messy cables and with good drivers that really work without anykind of issue.
The best analisys I've found was this one: AMD Radeon HD 5870 PCI-Express Scaling Review - Page 25/26 | techPowerUp
you could add it to your tomshardware analysis since it's outdated and very biased.

Anyway these numbers are great. It shows a 2.0 x1 solution is hugely bottleneck. There's no 2.0 x2 data because there are no x2 slots. But according to some calculations most cards have a bandwidth of about 12-15Gbit/s from there the performance will drop a lot.
pci-e 2.0 x1 4 Gbit/s
pci-e 2.0 x2 8 Gbit/s
pci-e 2.0 x4 16 Gbit/s

imo this is the way to go. Problem is how to implement it since new sandy notebooks only have 1 or 2 mini pci-e slots and no expresscard.
Do you known any recent notebook with 3 or 4 mini pci-e slots?

we already have those cables. The spec has been around since 2007. No one implements it. Just like mxm cards that have like a decade and just 0.1% of the notebooks have it. and you can't buy an mxm card anywhere like you buy a desktop one.
The spec is so developed that you can even right now buy the cables. There are already companies who produce them.
It's always the same problem. no notebook manufacturer will ever adopt this solutions. It's always better for them for you to pay a full price for a new notebook then for just buying a new graphic card, external or not.

And this thunderbolt thing could help a bit but currently right now we still don't have anykind of solution. thunderbolt delivers great bandwidth up to 20Gbit/s bidirectional. That would do the trick. But not even new desktops have it. Maybe with ivybridge that will change that

https://www.computerworld.com/s/article/9220434/Thunderbolt_vs._SuperSpeed_USB_3.0

 

Sony's eGPU is weaker then the built in MBP 15 GPU.

 

Say what you will but the new 2011 iMacs include thunderbolt ports (yes they aren't full blown desktops, but hey its Apple )

 

I wouldn't expect much from eGPU's
AMD/Nvidia were already looking for a solution,
but it seems quite hard to learn the OS that the eGPU is the main GPU.

Also thunderbolt has limitations, don't expect a GTX580 or 7970 as an eGPU,
it will be more like a 6670 or 6770 (desktop)
Remember that not only the card have to be included but also an external PSU and cooling

 

I beg to differ. x4 2.0 pci-e link will give the dream of ~95% performance of a full x16 2.0 link in real world games. Only limitations being:

1. The Thunderbolt enclosure should use the full 20Gbps channel rather than 2x10Gbps and should be affordable. The MSI GUS II (max 150W) has had a rumored pricetag of $150.

2. there is sufficient power to run the video card of your choice (2). If want a GTX580 then will need a hefty 12V/304W rail on your PSU to drive it at full load

3. whether the notebook manufacturer wires the displayport back into the LCD via a mux, ala dedicated graphics. If not then can revert to using Optimus' frame buffer method to copy the display traffic back to the iGPU via the pci-e link but there is a bandwidth penalty for that.

4. Driver support. Win7 appears to have no issues but MacOS does as discussed.

 

Hi, nothing new about Thunderbolt eGPU (except MSI Gus as i want a gtx 680 which is, I guess, too big for that enclosure) ?

 

I remember reading through this forum last year and wondering if Thunderbolt eGPU solution would mature but it seems noone has found a way to capitalize on the martket yet.

Maybe Razor can create the ultimate LAN laptop by engineering there own eGPU and put a destop GTX 680 thunderbolted to a 17' laptop.

 

i will jump on with few completely noob questions:
1. Is it possible that in the year or two my 2011 mbp 15 will be enjoying some nvidia over9000GTX via thunderbolt and playing metro 2066 maxed out?
2. When is it likely to happen?
3. Is this going to be like — you buy some special adapter and insert desktop gpu and connect via TB or more likely like "whenever nvidia pleases to create egpu specifically for tb will get that without much options?

ty

 

We are probably waiting on some no name company from china to make this, isnt that what started the current eGPU craze?

 

Sony should just try to stream the graphics bandwidth of a gtx 590 into there Vaio Z.

 

the evolution of thunderbolt is marked by some key aspects:

1) bandwidth

2) cabling/Connector

3) price

Bandwidth

this is where the problem lies for us. while a pcie 2.0 x4 aint the end of the world in terms of bandwidth, and it isnt, just check benchmarks done with high end cards like the 480, and there is a negligible performace loss, the problem here is that thunderbolt should be used a docking solution as well, just connect things daisy chained and you are good to go, however the more things you connect the more that limited bandwidth is shared the slower things will be.

Clabling/Connector

Thats another bottleneck, while the copper cable is a good thing, since it packs power and data, its costly and limited. While we move to other techs like fiber, the cable will be more expensive unless we counter that with some interesting production costs.

The connector presents another chalenge, to be physically compatible with mdp, it needs to be well, physically compatible to it, and thus limits to what we can do in terms of engineering. We still dont know if the limitation of the 2 lanes is from the controller chip or the cable or the connector.

Price

While this is a high end piece, and should be, and will be, there are some marketing concerns due to it, adoption depends on the availability of options that there are for you to chose, and from the marginal utility.

Availability, its low as it was expected, thunderbolt was launched as a halo product in apple, simply to the market to gain knwoledge of the tech, and sincerely, there are a few firms that can compete in the idea of transmiting what a product does like apple. they know how to bring to people what their products do, and do it in a great way.

Marginal Utility is going to be one tough nut to crack, simply because there are other high speed interfaces available for people to chose, the idea of thunderbolt is to unify those. No it aint competing with USB3, it never was, it never will.

For example to have more than 4 USB ports in a notebook is a wasted space.

Why?

Simply because most people only use 2 USB ports, and that is not the usual as well. The usual is one, for the mouse. The other is there for you external discs, for your keyboard, and for whatever dumb product people come up with USB, like hand heaters, coffee heaters, microwaves, and so forth. Thus to have 1 USB is essential, 2 yes great, 3, good, 4-10 useless. there are diminishing returns for that, simply put.

Thus the main problem is to win the barrier of the change in tech for high professionals, that should have professional cameras with the thunderbolt tech to make the transfer much faster, to make available for that cad professional a gpu when he is rendering something at a table to make the job quicker.

And now we get to the other face of marginal utility, when is quick, quick enough?

Im thinking of going for a ULV processor in ivy, it should maintain my level of performance that I already have with this i5 2415m, which is plenty enough to run those pesky VMs that I run from time to time, the major demographics dont demand more than the ULV offers us, actually my father is rocking the same x61 that I used in the company 3 years ago (yes we worked in the same company he was in the joint venture and I was at the main). And while its slow, quite slow when compared to what I have and what he has for personal use, it still fulfills the requirements of what he has to do, which is to boot auto CAD and visualize something in there, to run emails, some excel sheets and some docs. Thats it. Exclude the CAD thing and add internet and you have what people do in their pcs. Sincerely core 2 duos are still quite good for what those people do.

Now my mbp with a SSD and 8gb of memory is what most people need? nope.

Do some like to game on their pcs, thus the egpu powered by thunderbolt would be great? Yes

Do I want one? Somewhat, maybe down the line.

Does thunderbolt equipped cameras, network clusters and so forth would be a great thing? Yes

Are there other hardware limitations that we should consider? Yes

What are those? HDD, peripherals and so forth.

The idea is to bring together high end tech not available for notebooks or that was available but with a bandwidth or compatibility problem. The egpu thread is filled with tweaks necessary for a egpu to run, this should be a plug and play thing, and thats where thunderbolt is supposed to shine. I dont want more hassle, I want something simple, I have problems that needs solutions Im not seeking to solve problems of that solution.

 

I feel like we need to reaffirm the adequacy of PCI-E 2.0 x4 bandwidth and end the misinformation:



Summary: You won't notice much difference in GPU-starved games. Bring out Thunderbolt eGPU!!!

 

Take anything said above with a grain of salt, because....

http://forum.notebookreview.com/5332448-post10.html

...of hilarious statements like that.

 

I would like to say just one thing. You don't have to compare the eGPU setup with a "desktop config". We all know that is a match lost before starting (except for portability). But you have to compare the eGPU with the notebook market, where the updates are limited and their price is not always so affordable. The eGPU is a suitable way to have a ultra thin (and lightweight) laptop for work + the possibility to have a good gaming experience only when needed.

Today there is NOT a single card (speaking about single chip) that uses the PCI 2.0 16x bus at its full bandwidth, so the problem is "non-existant" for the majority of the customers, because SLI/CF configs are about 5% of the global market.

And trust me, while: (points in brackets)

Performance
Notebook + eGPU (1) < Gaming Notebook (2) < Desktop (3) <--- NOTE: This statement isn't always true but assume that a gaming notebook always perform better than an eGPU setup

Price
Desktop (3) < Notebook + eGPU (2) < Gaming Notebook (1)

Upgradability
Gaming Notebook (1) < Notebook + eGPU (2) < Destkop (3)

Portability
Desktop (1) < Notebook + eGPU (2) < Gaming Notebook (3)

Now summing the points you have

Desktop: 10
Gaming Notebook: 7
Notebook + eGPU: 7

But if we give different weight to the details (that are dependant on the person) the situation changes;
P.E. For me having the "best performing notebook" while paying it 2000€ sounds silly, I prefer to have a "good" performer for 1000€ because after 2 year cycle the value loss will be less than a top tier machine and it will be easier to sell it (because not everyone would like to spend 1400-1500€ on a used machine).
If I can have a "top performer" at 1000€ like the possibility granted by the eGPU), I can sacrifice the portability a little but it's fully personal. So the analysis should remain between the notebook market and their solution, where the eGPU can easily compete (and with thunderbolt even more) with no hassles.

I suggest everyone to compare my gaming videos with the one mades by gaming notebook, looking at the price of the solution (mine was around 1500$ comprehensive of the GTX560); with thunderbolt they can ONLY improve.

 

First WINDOWS gaming laptop with Thunderbolt. First Asus G55VW Gaming Laptop Config Available for Pre-Order | Laptoping I guess more WINDOWS laptop with Thunderbolt will make this a reality.

 

hilarious is your post. you are being very silly. have you checked the DATE of the post? 09-21- 2009 <-

In that time the a 8600M GT would almost outperform a 8800GTX desktop in a pci-e 1x configuration. And it would cost like 600$.
we are in 2012 not in november 2009!
Actually if you had checked the date and my other posts and topics you would see many mentions to that tomshardware pci-e bandwidth tests as well as many others unbiased and independent tests.

 

Don't be offended. But it is really hilarious to see this post on the first page of the eGPU Experience thread.
Ain't it???

PS Off course what you say is right 2009 is 2.5 years back and things change from one day to another, although I think that the first configuration was with a 4670 @ x1 and I don't remember being outperformed by 8600M GT. Anyway, don't take anything personally.

 

800 freaking dollars? WTB MSI GUS

 

what for? you've got 7970

 

.. or can get a sh/Outlet Lenovo X220/T420/T520/E420/E520, Dell E6220/E6320/3450/3550 or HP 2560P/4530s/8460P/8560P + PE4L 2.1b + GTX460-336 + PSU for ~$800 and enjoy about half the bandwidth (x2 2.0) with the x1.2Opt pci-e compression engaged.

A Macbook Air TB Sonnect configuration would cost ~x2.5 - x3 more. That just doesn't make much financial sense to pursue.