ViDock 4G for Light Peak (Z21) - POSSIBLE.

I don't understand the question?
I don't have a desktop, I have some crusty old servers, a Dual PII XEON and Dual PIII

 

^ Sorry.. What I mean is that: since the fact that ViDock in this case is connected via Express Card means that ViDock maximum bandwidth is about 2.5 Gb/s against PCI-E 1x 2.0 which is 4Gb/s and PCI-E 16x 2.0 which is 64 Gb/s.

This means that in this case, ViDock is running like a crippled 1x PCI-E 2.0 . We know that 1X 1.0 (express card case right?) it's 70-80% slower then the same eGPU going at 16X 2.0. We know also that 1X 2.0 is 50% slower; 4X 2.0 is slower 5-25% and 8X 2.0 is slower 0-10% (talking about actual high-end video cards).. I'd like to know how are your performance with that, since it's going like a crappled 1X. It should go slower 80% then at full capacity.

 

great idea, I'm in!

What could be the prize range of such a thing? (I don't know much about ViDocks)

 

from villagetronics the usual is 150-300, seeing that they are making this for apple too, its going to cost more than that I bet, and this is without the gpu

 

^ I don't think so. They're developing an ADDITIONAL module to make it ok for ThunderBolt. Price will remain the same for ViDock. Maybe a little higher for Light Peak due to special wire. Apple is not a problem here.

 

Just to be clear since I didn't see my name on the list, I am in on this for sure. As for the extras (2x vs 4x), I am sure you all know what I need better than I do.

 

^ corrected. Uhm, for example, I don't know XD what do you do with your laptop?

 

First post here.

The threads related to the Z2 and to possible extensions such as a Light Peak ViDock - are the most informative / thought-provoking I've seen in a long time: outstanding. I've now read through thousands of posts on the main thread...

Want to buy a Z2 pretty much loaded up, but with 256 rather than 512 GB. Would love gold, but can live with black.

Not a gamer but, like our Montpelier Maestro, I really want WQXGA capability - to the point that it's almost a deal-breaker. Waiting for a Z22 refresh (where it still would very probably not happen), or for the Z3 (might happen) - is not appetizing.

Bottom line: ViDock solution would trip the lever - I'd get it immediately.

To the questions:

1) What do you want to use the LP ViDock for?

One or (more likely) two WQXGA monitors. Audio card. Maybe SATA RAID.

2) Do you want 1 slot x4 or two slots x2?

Two slots x2 works for me.

3) Any other features you want?

Lots of ports, including eSATA, audio, etc.

Thanks to all who participate here - fantastic insights available nowhere else!

 

The main thing I want this for is for gaming, so a sound card and/or more usb3 ports would just be icing on the cake. For me, this would make my purchase of the Z2 a much longer term investment. Even if this would be a bulky setup, it would go with me to school on days where I have a lot of time to kill between classes to play the likes of Street Fighter, Diablo 3, and Battlefield.

Edit: I guess I would need it drive the internal display now that I think about it.

 

Thanks for support!

In any case, take a look at the first page.. on the bottom, there are two link about scaling performance and a very small resume.

 

1) To insert it with medium performance GPU, USB 3.0 card and Soundcard
2) 2 x2
3) To be able to drive the laptop display, to be able to charge the laptop from the dock

In full disclosure I basically copy and pasted what other people wrote but this is what I would use it for. Egg Salad!

 

!!!!!!!! FIRST OBJECTIVE ACCOMPLISHED

Village Tronic decided (in less then 48 hours) to make an Open Letter to Z community, to find out if there at least 50 PEOPLE who want ViDock for Light Peak. You can read the lettere here -> https://www.facebook.com/topic.php?uid=89321949134&topic=17792

They made more questions, more specific. So READ that CAREFULLY.

There's a copy of the letter in this thread first page.

Remember to WRITE your preferences at Village Tronic Facebook Page -> https://www.facebook.com/topic.php?uid=89321949134&topic=17792

 

I'd like to answer, but I don't have enough experience with the ViDock to make an educated response. Anyone else care to post their response here and I'll more or less mirror it myself?

 

^ first of all, tell me what are your purposes. I mean, what application you'd use if you would be able to use an external gpu.

Also, question 4) is about the entire chance to "personalize" ViDock.. there are different modules if you want to use Express Card or Thunderbolt (and I HOPE Light Peak) and the chance to plug or not usb 2 usb 3 and other ports. It's "like" actual modular power supplier

question 5) is about if you prefer to buy only the module (light peak in this case, if you already own a ViDock compatible) or if you want to buy also the chassis

 

@ Ung-Kung: write here the answers to the 5 questions you can find at the first page. I'll post for you in FB page

 

Posted my answer.

 

I have responded to your alert in of commercial development of the LP ViDock you post at DIY eGPU experiences, by advising what LP developments on the DIY side are available for Z2 owner's, quoted below:

 

already posted my answer on the fb

 

Didn't see it. Thanks for the hint! I think we can push for both the aim!! Obviously, I would prefer you solution cuz it would translate into a LP to TB adapter (which is better)!!

My question is.. even if hwtools would do it.. wouldn't there be license problem?

 

Good question. Simple answer is I don't know but can speculate. hwtools (TW) and amfeltec (CA) are selling "engineering test equipment" which I assume means they have some exemptions from compliance and/or licensing. It works in our favor because such gear is also substantially lower in cost than commercial eGPU solutions. As long as it produces a reliable working solution then I'm personally fine with it. For example, my DIY eGPU setup hasn't skipped a beat and cost less than half of what the commercial setup would cost.

I had thought that the Sonnet Echo Express would be the first to deliver but that doesn't appear to be happening and if it did, the $500 rumored RRP for an enclosure would scare folks away. Wonder if they too hit Thunderbolt pinout and/or licensing issues?

challenge #1: getting x1 2.0 working on the expresscard products

Now there's some technical issues as you can imagine. The first one is that neither the DIY nor the commercial solution have demonstrated a x1 2.0 working link of their respective expresscard 2.0 (copper) solutions on a Series-6 Sandy Bridge platform, eg: X220 + GTX570 + VD4+ here. That needs to be sorted first before have any chance of getting a x4 2.0 working.

hwtools are diligently working on using different cables. I've suggest to them that they consider integrating a (active) $50 Apple TB cable that can transmit at that speed. If you read that link you'll see they've had to add smarts to get reliable transmission over copper.

challenge #2: no Thunderbolt pinouts

The second one is there is no Thunderbolt pinout readily available. Unless one is made the only way of figuring out what signals are running where is to try to reverse engineer TB. It is a mini Displayport connector so it's pinout can give a lot of clues.

challenge #3: Sony's LP specification

Given the difficulty of obtaining Thunderbolt specifications from Intel/Apple, I wonder how hard it will be for Intel/Sony to give the LP specs?

Why I have a DIY community-driven bias

I'm all for any eGPU Thunderbolt/LP solution, commercial or DIY. I can't wait to see the equivalent of ~80% desktop x16 2.0 on a x4 2.0 eGPU performance on a tiny MBA or Sony Z2. The Z2 with the i7-26xx CPU making it the ultimate in mobile performance.

I am also biased in favor of budget DIY solutions. Why should $$ get in the way of some good FPS? Call that being a fiscal conservative, bang-per-bucker or a tight-waddle Hence I prefer DIY community driven solutions with hardware vendors only making the hardware to spec based on the communities requirements. This means then that nobody can claim ownership of the solution and therefore not charge exhorbitant prices for it.

In this way opening up to hardware and software to the public allows a much greater pool of innovation to draw from. This is what allowed the DIY eGPU [expresscard/mPCIe] version to fix many of the problems that plagued all previous eGPUs solutions, provide an avenue for individuals to share their experiences and build a table of performance demonstrated.

I'll get a tentative feel of where hwtools/amfeltec (DIY hardware) are at and suggest they read the threads to see there is a demand for these products. You can see that both commercial and DIY hardware vendors have quite a few techical challenges to make this happen. I'm hopeful there will soon be a DIY eGPU [Thunderbolt/Lightpeak] version that continues in that DIY eGPU [expresscard/mPCIe] spirit.

 

^ I completely agree. I've also wrote something to Sonnet saying that an adaptor would be great for us, and that we could be able to use their peripherics also. In addition I told 'em why I think it could be possible since the problem is to find out correspondence in pin/channel of thunderbolt and that light peak port and that the real difference is optical fiber against copper (infrared signals against electrical signal) so that it would be sufficient a fast infrared/electrical adapter to solve the problem. Who knows what they'll tell me...

As for challenge 1: express card has a different "channel" to pci-e than thunderbolt/light peak, right?? So maybe it won't be a problem.

In addition, Thunderbolt is 10Gb/s (20Gb/s bidirectional) so to be a 4X 2.0 it needs support of 4 PCI-e lanes... is it THunderbolt and Light peak connected to those 4 lanes?

If I'm not wrong, Village Tronic already has documentation from Apple about Thunderbolt...

 

I'll take a DIY solution if it is the only solution. When I wanted to pull the trigger on getting WQXGA going on my Z1 I just wanted to get it done. The ViDock was more or less plug and play. Now that I have a working solution I am more likely to experiment when it comes to the Z2/LP. I don't think the Z2 is going to immediately replace my Z1, it will be more like the UX and the Ps for me, basically it will be added to the toys that I don't need.

 

Learning how to convert from optical to copper from the PMD

The easiest way to do a Sony LP to copper adapter is to pull apart a PMD and trace the interposing chips that lead off the HD6650M to the dock connector and duplicate. Those chips would be converting the LP signal to x4 2.0. Unlike a TB eGPU passive adapter solution, this actual requires some real engineering work to figure out.

Now as to Thunderbolt pinouts and documentation, I believe neither is readily available. Anyone claiming to have it is requested to show proof as I believe it doesn't exist. What other possible reason exists then for neither Sonnet and hwtools having products already available for purchase? hwtool's having been waiting for pinouts to appear since June 28th.

Keeping in mind that TB/LP is a game changer favoring further notebook purchases and no longer ties users to needing to upgrade their notebooks to keep up with higher GPU demands every two years. x4 2.0 is 80% the speed of x16 2.0 so if more than acceptable. Commercial interests so far has stopped the widescale adoption of an external pci-e interface on notebooks probably for those profiteering reasons.

Example Thunderbolt technical implementation: extending DIY eGPU's PE4H

If we knew the pinouts and full specification of Thunderbolt then the current x4-capable DIY eGPU's PE4H require the following alterations to support Thunderbolt [based on it what we know so far - it's x4 2.0 pci-e and displayport):

+ use a cable that can transmit reliably at pci-e 2.0 speed. I'll once again highlight that neither commercial nor DIY expresscard solution has demonstrated a solution that can do this to date with Apple's $50 Thunderbolt cable the only verified pci-e 2.0 cable.

+ routing the PE4H's x4 lanes to the right mini HDMI connector (TB) pins and add other pci-e signalling lines.

+ add a Intel Thunderbolt controller to act as a mux/demux pair for pci-e and DP signals when attaching to a Thunderbolt port. This may not strictly be necessary depending on whether the receiving Thunderbolt port can process just pci-e signals as Apple tell us that Displayport devices can be directly attached, so may be possible. REF: http://arstechnica.com/apple/news/2011/02/thunderbolt-smokes-usb-firewire-with-10gbps-throughput.ars .

+ if the TB controller is added then also add a mini displayport pass thru port where the desktop video cards DP output is routed back via the TB port so could drive the internal LCD if the manufacturer provided a muxed solution to drive the internal LCD.

+ add a mini DP (TB) port that extends the pci-e/TB bus to allow daisy chaining of peripherals.

This TB passive adapter would be a breeze for a motherboard manufacturer to create if was given the pinout to do it and didn't have to wade through the licensing issues. So if anybody has such contacts with manufacturers, then I'd suggest giving them this info as a basis and let them know Apple and Sony folks would be very happy if they can create a product for their TB/LP systems.

 

Sounds awesome +rep

 

VERY INTERESTING.. apart the chance to contact manufacturers, which is difficult.

I've a question:

Are we sure that TB and Sony's LP are 4X 2.0 ?? I mean, I saw some benchmarks about Sony's Power Media Dock and even if it has a 6650M it's slower then the 6630M soldered in Sony Vaio S. This is probably due higher response time using external cable with Light Peak.. But we're talking about 12% slower Power Media Dock?????VAIO Z???????????????????
So.. is it possible that, at least in Light Peak, flash descriptor is set to grant only two lanes and not four? And if it's like this.. it's possible to enter flash descriptor and tell him to set 4X for Light Peak channel?

 

By the way guys, only 10 people wrote in FB page of ViDock.. get there..

 

What link width does Sony Z2's LP port use?

The Northbridge has a x16 port available that can be configured to host a GPU. I've seen that port configured as:

1) x16 - 1 dGPU
2) x8 + x8 - 2xdGPUs in SLI config
3) x8 + x4 - 1xdGPU (HD6750M) + MBP Thunderbolt eGPU port as shown here.

Previous gens of notebooks ran x8 1.0 SLI and had no pci-e bottlenecks so Sony have likely opted to run x4 2.0 from the Northbridge via the PMD to match that and Thunderbolt's bandwidth. That does then equate to 20Gbps (bidirectional), which is what LP is rated at.

You can confirm this is the case by asking a Sony Z2 owner to run GPU-Z and check the HD6650M's link width (eg:x4 2.0), AID64's Northbridge info for the 1:0.* ports or my favorite: 'lspci -vv' using Linux checking the 1:0.* ports like shown here.

 

This is odd.. as stated here, it's 8X (4X + 4X or other config) http://forum.notebookreview.com/gam...851-diy-egpu-experiences-247.html#post7027248 (HM67)

P.S. I don't think they connected all 16 lanes (which is impossible with HM67 right?) cuz there's the WiFi and the WWAN (present or not) option.. so at least one lane is meant for this, so at maximum it would get a 8X

 

Well, as I'm no expert to ViDocks I can't answer all the questions they want me to.
Only thing I can say is the main purpose I'll be using a ViDock for:
INSTEAD of an PMD to attach more monitors/monitors with higher res, having more ports (USB, eSATA, audio jacks), being able to put in an audio card. I'm not looking for a high end GPU, as I'm not going to be rending videos or playing games. Mostly use it business, academic and scientific related.

 

Which pci-e bus are the TB/LP and expresscard/mPCIe connected to?



The documentation you are referencing is from the DIY eGPU thread, pertaining to a x1 link off a PCH or Southbridge. It is important to note that a expresscard/mPCIe link and a TB/LP attach on different pci-e buses as shown in the above image (it's Series-5, but pretty is pretty much the same for the Series-6 Sony Z2).

A TB/LP hangs off the CPU where it can have up to a x16 link. A expresscard/mPCIe slot hangs off the PCH which is configured as x1, but in some specific cases can be ganged to x2 and x4.

 

......ah.... thanks!!! That's where I was wrong

I asked somebody who already has Z21 to run GPU Z under load. If ti comes that it's not 16X but 8X or badly 4X... is there something we can do??

 

If one looks at the CPU specs:
Intel® Core? i7-2620M Processor (4M Cache, 2.70 GHz)

It lists:

Anandtech specifies x4 for TB:
http://www.anandtech.com/show/4194/intels-codename-lightpeak-launches-as-thunderbolt

 

One reason for lower than expected performance is if it uses a muxless setup. There part of the the supposed x4 2.0 link is carrying the image display traffic that is rendered by the iGPU. Performance would get worse as resolutions increase. If it's a muxed setup then then it switches through the LVDS signals from the 6650M back to the internal LCD using wires so never encroaches on the x4 2.0 link.

I'd be guestimating that Sony did a muxless setup. You'd know for sure if a Z2 owner can disable the iGPU in Device manager and the internal LCD might switch but it still displays an image on the LCD in proper resolution (not VGA compatible mode). If it goes into VGA compatible mode then it's a muxless setup.

 

I don't think Villagetronic can be considered an innovative company, nor do they have ties to say Intel or other companies that work on Lightpeak. If they are / have, they would have jumped immediately on the Lightpeak + Thunderbolt opportunity. Instead, they prefer to wait for the wave of users to arrive. They're not going to be able to ride it when it arrives, they'll be too late. Because of this, Villagetronic will never get very big, nor influential. They'll always cater to the hobbyists consumers.

 

Since there's nothing on the market yet, every effort towards thunderbolt or light peak egpu is welcomed anyway.

As they say, they're already developing thunderbolt module.. time will say.

 

I'm waiting for answers by some users. Anyway, muxed or muxless, we would always have a loss of performance since the data/image is sent back with Light Peak, right?? I wonder what happens in a laptop as SA: hd 6630M is connected by pci-e bus.. but the data/image that would have to go to the screen.. are sent back to iGPU through PCI-E or are sent directely to screen?

-----------------------------

EDIT: I'm still waiting for answers for Z21 but I found out that HD6630M on SA is set as PCI-E 16X 2.0 and the dGpu goes in that way, full PCI-E 16X 2.0. While 6650M in Z21 goes with Light Peak so in any case its own speed will be 4X (or a little more... 4X it's 16 Gb/s but TB or LP can achieve 20 Gb/s bidir. so if PCI-E is set to grant 8X or 16X to LP.. it will go @20 Gb/s right?).
6630M in SA achieves 28.32 fps (avarage) when 6650M in Z21 achieves 24.86 fps (avarage) at high resolution 1600x900. With those resolution 6650M with PCI-E 16X it's 10% faster so it would get 32.152 fps. Taking into account that PCI-E 4X cut performance by 20% (but it varies from 5% to 30% depending on application or game) we obtain 25.96 fps. It's 4% over the benchmark but, even if scaling varies with applications, we can say that a lot of the loss performance is due to the difference in PCI-E bandwidth REALLY granted. It may be that, if the PCI-E set for LP is 8x or 16x that LP works "like" a 5X (20 GB/s) or still there's direct connection with screen.

 

An article discussing the forthcoming TB ViDock. Includes a link to an NBR post and a closing thought that people may eventually be able to build their own TB ViDocks.

External Thunderbolt graphics card for Macs will not blow you away | ExtremeTech

 

I'm about to make a post on facebook for the LightPeak Vidock for Z2.

Does he mean that Parts is like a single PCB like a PE4H ?
I'd prefer a "Product" version if my 3 slots GTX 480 could fit in it... but I don't think so

By the way, couldn't we benefits from Optimus tweak for bandwidth compression with such a product ? Just like I'm doing with my current eGPU x1.opt ?
Could it be possible to get a X4.opt with a LightPeak Vidock ?

 

They mean a module to add to your existing ViDock.. that doesn't exist so you need the entire product

I don't know.. I think it's possible since TB and LP connect directely to PCI-E bus. In addition, for LP we know already that PMD works on the laptop screen ( but we don't know if it's the iGpu who renders the final image or it goes directely to the screen through additional wires ) but maybe there would be drivers problem.

 

Thanks for explainations

 

NVidia's driver only engage the pci-e compression if it detects a x1 link as found on the DV4-2000 x1 and x2 eGPU testing.

We asked NVidia to allow pci-e compression to be an option on link widths other than x1 and for systems without an 4500MHD or newer iGPU present but have had no action from NVidia.

In the DIY eGPU thread we are trying to get a x1 2.0 link going and confirm if the same compression is enabled. If it does, then a x1 2.0 Optimus link approximates x2 2.0, or half of TB/LP. That performance level, at a substantially lower pricetag via say a Lenovo X220, might be more than enough for some ppl.

 

a Thunderbolt link mixes Displayport and pci-e traffic

Further to my previous, comments. We see at

Thunderbolt smokes USB, FireWire with 10Gbps throughput that a Thunderbolt controller is shown accepting displayport and pci-e traffic which it would de mux and send to the LCD and dGPU. It would mean the sender end would need to package that traffic to send that way using a Thunderbolt controller in order to be considered a complete TB implementation AnandTech - Eagle Ridge: The Cheaper, (optionally) Smaller Thunderbolt Controller .

The use of that Intel TB controller also means that a complete external video devices enclosure will need that same controller to take desktop video card DP and it's x4 2.0 pci-e signals, mux it and send it to the receiver or they will need to use NVidia optimus to carry that traffic over the x4 2.0 link to drive the internal LCD.

The Intel TB controller also begs the question whether a matched TB controller will be needed in any LP to TB adapter to be able to communicate to TB devices (eg: TB eGPU). I would assume not. If it is required then it's another chip that would be required in a LP to TB adapter and more $$ to Intel's stock price.

 

^ Why do you assume not?? How can an eGpu connect to a TB/LP port which has a TB controller, @ 4X 2.0, without its own controller and using TB cable?? TB cable reaches 20GB/s since data are packaged in some way (and there's the prescend of active chips).. or you say that maybe it's possible to achieve that data transfer without packaging with TB controller??

Anyway, if there's no need for an eGpu to have its own TB/LP controller... what we'll need to get an eGPU work with Sony's LP (THEORETICALLY)? An adapter supporting PCI-E 4x + a matched cable + infrared/electrical converter?

 

I'd gladly buy two ViDock units for my new Z2 if they get to support the LP.

However I do not have the facebook account so I cannot put my name on the list

 

^ get to the first page of this thread, read and answer questions. I'll post them with a permalink for you in fb page.

 

you are assuming that the 2 controllers can comunicate with themselves without major problems.

 

I reviewed a Thunderbolt-equipped Mac's Linux diags here but could not see the Thunderbolt controller in any of the output files. Meaning then need to do physical inspection to find out. It would help then if a a Z2 owners could establish:

(1) Is the Z2 doing a muxless or muxed HD6650M setup? Muxed would mean if the iGPU is disabled in control panel, the HD6650M display on the internal LCD would be taken out with it.

(2) Is there a Intel Thunderbolt controller in the PMD? Would require dismanting the PMD and look for an Intel chip in there.

Z2 owner can consider a mPCIe DIY eGPU solution

If you have a Z2 and can access the wifi mPCIe port, then you have the option of doing a mPCIe DIY eGPU Optimus solution for the time being until a some solution using the LP port is released. A Optimus eGPU via mPCIe would be running a x1 1.0 link for now and will outbench your PMD's HD6650M. x1 2.0 is in the works with bplus testing different cables to get it to work. Though Sony owners may be horrified at the lack of elegance of such a solution.

 

yeah but I think it's no possible with 2 different controller... TB/LP controller packs data .. so I think there's always the need to have a LP/TB controller on the enclosure/eGpu to work out.

Probably we need this: (1)eGPU <-> (2)PCI-E 16X @ 4X <-> (3)LP controller <-> (4)cable able to sustain 4X + (5) infrared/electrical converter per each channel able to sustain 10 Gb/s conversion per each channel <-> (6) RX/TX cable inside USB envelopement to plug into USB 3.0 PORT/LP.

I think that point 2 and 3 are not an enormous deal, in next months there will be out in some way specifications and pinout.. the worst,when exclusive will be off in 2012. point 4, the cable, it may be bypassed using TB directely. Point 5 it's the problem.. create such an adapter which converts the signal in the right way. It would be more easy to put a LP controller in the enclosure and use a USB cable with 2 rx/tx cable inside which is simpler. And there wouldn't be any problem of pin-outs since there are only 2 rx/tx cables..

no pmd for me..

No access without opening the entire laptop. And still.. I would have to plug uff WiFi adapter.. no way in such a laptop. I'll be waiting for USB 3.0 which is a start.
I would have no problem with lack of elegance.. the problem would be lack of wifi and open/close/open/close the laptop. XD

FROM a review: http://www.notegear.com/Content/con...opage=&TNum=1246&kind=2&SearchString=&Search= no sign of TB controller on that face... I think it's on the OTHER face. (still.. since this would be a controller that use rx/tx.. controller may be different from TB one.. right?)

http://weekly.ascii.jp/elem/000/000/047/47911/
http://weekly.ascii.jp/elem/000/000/048/48360/
these are images of LP inside the laptop.
*********************************************************************************************

SO: is there someone who wants to open up PMD and see if there's sign of TB controller? Which is probable since there are rx/tx wires.
in addition, is there somebody who can try to plug his own PMD, then disable integrated GPU in device manager and see if screen turns to VGA compatibility mode or remains/turns into a proper resolution?

 

By looking at the pictures there is no thunderbolt controller in there, there are 2 versions eagle ridge (mba) and light ridge (entire mac line up except the air), we have to remember that when we develop something there are at least 3 prototypes, this could have been a different prototype that led to the sony implementation, or a best case scenario its an eagle ridge (because its small, and there are some parts of the pictures that we cant see).

By no means Im trying to post negative things here, Im just looking and taking guesses, from what I see there would be a need for an entire unit to be developed for sony implementation

 

I'm in too. I'm not Z21 owner yet but soon(if there is nothing change).
1. 1 4x
2. for gaming and 3d engineering
3. usb3.0 dvd drive and others