DIY eGPU experiences

It looks that your expresscard is port 1. But the best to know is to put something in then recheck with AIDA64.

But it looks it could be possible to do a X2 link, but the problem is you may need to remove the WLAN Card that is on port 2... meaning that you won't have wifi anymore :/ (But if your expresscard is port 4 and not port 1 that would be easier, however I think it's port 1)

But you're not doomed ^^ I get great performance with X1.opt much more than GTX 485M

 

Hey zyrn,
I wish you get success. And if you are successful just let me know.

Because it looks like you and i have the same chipset.

Can you please tell me that do you have have HDMI port in your laptop?


How to connect the the EC2C with the laptop??
thanks.

 

Put it in the ExpressCard slot if you have one?

 

I have been bitten!!!... by the DIY ViDock bug

It makes me do all sorts of crazy things like build a sweet [albeit not quite finished] steel enclosure that was designed perfectly if I don't say so myself (well, I am saying so )
I 'blame' Nando4 for my catching this bug thanks to you Nando!

Enjoy:




netting in at 4 Lbs 12 Oz (as it was in the last picture there); The enclosure itself is made of some sheet steel (I think stainless steel) that was folded into a box at 5.25" x 4.5" x 10.5" (at the long end) and 7" (at the short end); it has the pci section of a desktop case. It has a Corsair 430W PSU that I stripped to [mostly] the basics, and then spliced the remaining cabling to be only just long enough; and a full EVGA GTX460 1GB SuperClocked video card. The whole thing is currently quietly whirring next to me

Internals of the dock ATM:


To see more of my DIY ViDock setup, please check my Flickr account (Agent--9) and my thread over on TPCR http://forum.tabletpcreview.com/hardware/40268-diy-vidock-tablet-pcs-experiences-benchmarks-setup-ect.html

 

what are people's plug and play experiences with this? I've read differing experiences from making express card hot swappable and very simple to having to reboot and go through all these different steps to get it working

 

But how do i get to know whether i have an express card slot or not??

 

I was under the impression that it was not possible to route the output back to your internal screen. Is this no longer the case?

Sheer curiosity, as I have a laptop with internal GPU that works well for my mobile gaming needs, and a desktop for when I'm home.

Good thing it's not a PlayStation, otherwise Sony might sue you!

 

Finally got it to work but it has its glitches. I have two screens attached using the DVI and displayport. The displayport one flicks about every mintue. The DVI is connect to SP230W running at 2048 by 1152. The displayport is attached to a U2211h.
I ran 3dmark06 with the sp2309w on and u2211h off and got a score of 11623.

update:
I changed the cables and now the SP2309W is using a HDMI cable and the U2211H is using the DVI cable and the flickering has stopped. I didn't have an extra displayport cable so I am not sure if it was the cable or something else that was causing the flickering.

 

No, unfortunately Vostro 3300 does not have an HDMI port. The EC2C goes to the expresscard slot afaik. I will not probably do it in the next few weeks tho so it will take some time for you to get my results.

Expresscard slots look like this: 34mm/ 54mm.

 

It works on the internal screen of your laptop if you have one with Intel Integrated graphics (the X3150, 4500MHD, or Intel HD) as this says DIY ViDock Experiences - Optimus


my DIY setup can do display on the internal screen of my lowly SL9400 based computer and play most games acceptably smoothly (no it is not at 200FPS but it is smooth enough for my needs, wants, and expectations from such a underpowered system like mine) at the very highest settings and still stay under 70 C on the laptop [70 for the Intel gfx chip, but the processor stays under 60 most of the time], and under 60 C on the dock; but I can also output to either DVI or HDMI displays off the video card's output.

 

What steps are required to set up a x2 or x4 connection? I have all the parts and cables, but I can only get x1 to work.

I get the activity led to turn on only on the PM3N attached to port 1.

Thanks.

 

Nice enclosure. My 460 is still basking in the fresh air of my room, sucking some of the laptop's heat exhaust due to the short cables I needed to do high pci-e overclocks. Luckily the laptop's internal cooling sucks (combination of poor/cheap design and a nick in the heatsink where it makes contact with the CPU die), so it isn't much heat (most gets blown out by a fan on a P1 heatsink that's stuck directly onto the laptop's internal heatsink).

How do you get the card out of there if you wanna replace it BTW?

 

That sucks that your laptop overheats like that, have you considered getting a replacement heat sink from ebay or some such online retailer? though with it running at 3.51Ghz I guess it is prone to running hot no matter what heatsink it has.



I sadly, cannot OC my PCI lanes because my PLL chip is a piece of junk that is totally locked down but then again I'm not stuck using shorter cables, so maybe its not so bad

Thanks!
I designed the case so that I unscrew the 4 screws on the 'side', and the one on the 'bottom', that will allow me to pull one side out slightly [the side that has the 4 screws on it] which releases pressure from the PCI stack [the metal stack which is cut out of a old desktop tower -it is held in there mostly by pressure, well, a LOT of pressure thanks to the way the sides are leveraging it] thus allowing me to slide the entire PCI stack out [with GFX card screwed down, and the PSU 'attached' to it; all as one piece]. Though I think I might have to unsolder the AC power lines to remove it all the way, but that just adds 10 seconds to the process, and I am not planning on removing the card anytime soon.

The PCI stack acts like a sled, holding the gfx card and psu in place [like this pic, but with the PSU shifted over into place -it sorta wedges onto the pci stack, and like this it holds on without anything wiggling]


you can see how the pci stack fits inside the sheet metal housing, and how the housing would squeeze it as to hold it in place once that bottom end is secured


It isn't nearly 'user friendly' when servicing it, but that also means it isn't very 'thief friendly' either, which is a huge plus for me.

 

first of all thanks very much to all The ones involved in this great project.

I have a Toshiba A 665-3D laptop with i7-740qm 1.7 GHz quad proc and native nvidia gts-350m. i was hoping to get a significant boost with a gtx 460 x2 setup as the card is about 5 times more powerful than my original according to specs. i play all my games in 3d on my projector at 1024x768 120 hz and, i guess it may be due to the 3D effect, but even @ x2 the pci bandwidth still seems to be a serious limitation to the card potential as i get roughly the same perf as with my discrete gpu. (i checked under gpu-z to confirm the link was really x2)

my setup is expressed card+mpcie on port 3+4 (i had to unplug wireless adapter on port 2). the setup can be an issue because i cannot hotplug my mpci-e pm3n and when plugging mini hdmis after x2 setup the card is only recognized once in tenth...i though pm3n v1.2 might solve this but what is strange is that x1 setup on the mpci-e port always works! i can plug / unplug the m-hdmi as many times as i like and presh f5 under setup and it is there...is it normal ?

looking fwd to optimus supporting x2 and non intel discretes!

 

GTX 460 x2 is a definite overkill as it comes to connecting it via expresscard slot. Even in a normal computer at x8 it gets bottlenecked pretty heavily so no wonder really.

 

My experience with x2 is that only the 3dmarks show significant limitation from the x2 link. Real games don't have as many sudden camera angle changes, and I think Nvidia's driver optimizations are what cause the bulk of the poor performance at x2. Also your CPU's turbo clock when running in dual core mode is a lot more important than having 4 cores. From my understanding, games have a hard time using more than 2 cores. If your dual core mode isn't much higher than your base of 1.73ghz then your cpu is a big bottleneck. I actually see more performance gain in a lot of situations from increasing my CPU clocks than increasing PCI-E clocks, and my cpu is 3.06ghz stock (6MB Penryn C0, which is roughly equivalent to a 2.7ghz i7 in games).

 

hallo,
i have a problem with this system
i already follow the instruction but after i resuming the windows, my laptop didnt detect my VGA at all (i use Asus F3SC plus GTX 460 and W7 32bit Games edition) even in device manager, and fan speed didnt slower when i resume the windows
i tried to attached usb to check and it detected in my laptop

any solution with this?

 

1. Make sure everything is connected.
2. If you are using mPCI-E, make sure that the slot you are using is fully wired. Some slots looks like mPCI-E, but in reality only have the USB connections wired and not the PCI-E. Put your wireless card or something in the slot and see if it works.

 

Are there any progress on x1.opt links and intel series 6 chipset? I know ppl are testing, but does it look promising or not?

 

I think the HDMI cables can't handle the increased data rate or something...I'm waiting until Kepler to do a ViDock though, hopefully it'll be sorted out by then.

 

i'm using express card slot..
my friend said, the slot works because the lamp on the PE4L is on, adan when i test with usb attached on my PE4L it also detected
i dont have an express card devices for now

maybe the windows? should i change to XP ?

 

still not working. Is there something with ati+nvidia setup that I'm not understanding? All I get BSODs caused by atikmdag.sys (usually with "Driver unloaded without cancelling pending operations" error, sometimes "irql not less or equal" error message).

My system detects both cards and lets me install drivers but after that it's just bsod. Nothing but bsod.

Dell M1530 (T7250,4GB RAM,Nvidia geforce 8600M GT) + 400w psu + PE4H + EC2C + Asus radeon HD 4870 1GB + HP h2207w display (with dvi->hdmi adapter). There are many M1530 users here, some with working ati + radeon setups. Anyone having same kind of problems?

I've tried:
- removing ram to 2GB
- setup 1.x compactions
- different wddm 1.1 drivers for both cards, xp drivers for external gfx card
- disabling, enabling vid cards
- clean win7 install
- removing wifi (PCI port #2)
- reseating vidock and reconnecting cables

I'm going to try next:
- test dock and psu with different laptop
- bang my head against a wall
- sell those parts, give laptop to my little sister and get a desktop pc.

If my default (any way to make it not default? no?) gfx card is disabled or without proper drivers - vga, no aero, no bsod, no external gfx (cannot change display device) and ati ccc complains about incompatible graphics card.

If internal graphics card has drivers installed: Fan starts when connected and stops after 5 seconds -> atikmdag.sys causes bsod.

Should external monitor activate just after connecting vidock to laptop? Because that's when my system crashes.

 

@pirkko

Exactly what PSU are you using? saying '400W PSU' means jack squat (what matters is the Amperage on the +12 rail, not the overall rated Wattage), can you post a picture showing the side of the PSU that has a little chart on it?

Also, do both the little green LED's light up on the dock connector? also, do you have all the power plugs that are needed plugged in? (you need that small 4 pin connector plugged into the dock connector in order for it to work)

 

Thanks for response! I've been a little suspicious with that (no-name) psu. Got it for 20€ from previous owner.

PSU seems to be this model:
MODECOM - producent sprzêtu komputerowego

Both LED's are lighting up (I'll doublecheck this back home) and I've got all required power plugs connected (two 6-pin for vid and a 4-pin for PE4H).

 

If the PSU is indeed the model you have mentioned above, then it seems to be slightly underpowered. The 4870 needs about 30A on the +12V rail and your PSU has 2x12V rails, totaling 30A which means it will be drawing maximum power from the PSU to run.

Now if your PSU is not (power-)efficient (which is mostly the case with generic PSUs) then your card might be underpowered. If you are overclocking then you would need more amperage on the 12V rails. Could you try this setup with another PSU?

 

Win7 is known to be the best OS for this, so changing to Xp shouldn't improve anything.

If that isn't the problem then I think you have defective hardware. Unfortunately there are a bunch of pieces that can be bad, ex the cable, the vid card, the PE4L, and the expresscard connector. I have heard of bad PE4Ls before so i would try to get a replacement for that first.

 

Finally: why 36-bit PCI allocation fails on many systems

The reason + potential solutions as to why 36-bit PCI allocation doesn't work on Dell Latitudes or MSI CR720 is here. These systems then have no free 32-bit PCI space to accomodate a 4GB DIY ViDock so can only do a 3GB installation at the moment. Worthwhile for affected users to investigate the proposed solutions in that link.

Last Setup 1.0f help doco request

NBR DIY ViDock users are encouraged to improve the help documentation supplied with Setup 1.x to cater for all skill levels. here is the latest 10f-pre6 pre-release version with a revised menu structure and caching goodies to play with.

Seems we have a shy bunch on here as I've not received any response so far. Please PM any help documentation you'd like to be integrated into the next Setup 1.x public release.

 

how can i check that parts?
i mean the defect one, like VGA or express card or PE4L

 

Does anyone else also have a PE4H vs PE4L experience.. ?

 

Uhhh 30A would be 360W. The 4870 has 2 6-pin power connectors, so the most it could draw would be 225W. Please don't spread misinformation...

That said, a good quality PSU is a good idea, I wouldn't trust a sketchy looking old PSU. For one thing it would probably have very low efficiency, and heat and noise could become an issue.

A PSU I've looked into for this is the Dell RM112, rated at about 255W. It's made by Delta I believe, and is 80 plus Gold certified, and should be adequate for most DIY ViDock setups as cards running between 120 and 200W would keep the PSU in its most efficient range.

 

HI everyone!
I'm sorry cause my bad English (I'm from Croatia).
I saw this topic few days before and only thing I can say about it- this is AWESOME!
Vidock looks like perfect option for me. I'm pretty sure I'm going to buy PE4H, but I'm confused with expresscard and mPCI slots, so I have 2 questions:
1st Is expresscard x1 connected with fermi graphic card (optimus link) bottleneck for graphic card?
2nd When can we expect new version of diy vidock with expresscard2.0/3.0 support (approximately; is it about 20 days or about 3 months)?

I hope you understand my English...

 

Aaaaaw yeaaaah! I bought decent psu, hooked it up and after some tinkering it works! Crysis 2 suddenly became playable- it runs smooth even with "hardcore" video settings. Not too shabby for 3 years old laptop, is it? thank you everyone! I'll post some benchmarks with x1E mode enabled later.

Maybe I should order parts for 2x setup next...

 

Nice to hear! Moral of the story...don't skimp on the PSU!

I'm going to try a x2 M1530 eventually - the PM3N v1.2 is out, but not yet bundled with a PE4H (Which you can't buy separately) - and HWTools are very slow responding to emails...

 

I don't have an Optimus setup, but the performance hit due to reduced bandwidth compared to x16 is very small. There is no difference between using an expresscard slot or mPCI-E slot. There is also a small performance loss due to the CPU needing to spend some time compressing the PCI-E data, but overall Optimus offers near x16 performance.

The problem with expresscard 2.0 currently is that the cables used cannot transmit the higher frequency signal well enough to run a graphics card. Otherwise people would be using it already. The problem would be even greater for version 3.0, which will not be released for at least a year, probably longer.

 

The only way that I am aware of is to replace a part at a time and see if it works. Like I said, the most likely culprit is the PE4L, but another graphics card or cables would probably be the easiest to test. The only other thing I can think of is that the PSU you are using is so inadequate that it can't run the card at all, but usually when power is an issue the card will be detected, but hang the computer at some point.

 

I'm still having this problem.

I bought a new PSU as I thought that may be it, but it's not.

Any help?

 

Hi.
Great to see an easy and relatively inexpensive solution to a hardcore gaming PC. I'm quite new and am interested in building something like this but I just have a few questions:

1. How big of a hit (percentage wise) does vidock via expresscard(x1 optimus) have in comparison to a true desktop pci-e x16 channel?
Big gains between x1, x2, x4 and x16?
2. My current laptop has an ati 3650 chip but if I build an a vidock platform with an Nvidia card, would I even be able to run OPTIMUS?
3. Does the system have to be rebooted or shut down entering and exiting vidock?

I already have a 28" LCD but obviously cannot run games at a high resolution due to the graphic card on my current T500.

 

Wouldn't this overheat the GPU, as aside with the psu?

At least put a carbon fiber for heat reflection i guess.
Even if there is a fan/,.

 

So anyone up to date on news about thunderbolt ports being able to be used for external graphics solutions?

 

Nah it looks fine, the intake for the video card is closer to the outside fan. The heat generating components of the PSU don't share the same air and are blown out the back separately, so both video card and PSU get plenty of fresh air.

 

No, the GPU runs at the same temp [about 60 C max, usually in the 50's; Idle at 22-mid 30's depending on where it is] as it was when it was in open air, the casing on the card even stays cool; at most only a few small components on the PSU ever heat up (and even then, not by much, remember all that is drawing power is the GPU); then there is the large 120mm fan that is always pulling fresh air into the case and the only way for it to get out is at the other end (which means it has to go over the PSU and through/ under the gfx card -below the card is a .5" gap or so, and there is a small gap at that end on the exhaust side- so everything has fresh air constantly being pulled across it)... but of course the exhaust on the gpu is warm, so that can't be covered up safely [duh]

The only place I can feel any heat after using it for a gaming session, is at the exhaust side of it, and even then only the GPU vent is warm.


I made the whole thing so that there is no downside to any of it: no adverse heat, no shaking about, easy to use and setup, relatively small [I sacrificed some size to get cooling that I am more comfortable with]; though it is not 100% finished, it needs to be cleaned up/ polished, have a thinner 120mm fan secured in place, and have a place to easily plug a external mini HDMI cable (I will be getting a female to female mini HDMI adapter so I can put the super short cable inside it, with the adapter's end available to plug in a mini HDMI cable of my choosing between the dock and EC adapter -also makes it easier and safer to take with/ transport)... and of course I will put a fan grille on the 120mm fan

 

I gues the ATX Power Switch won't work with a modular PSU?

(like this one: Newegg.com - ePOWER EP-450CD 450W ATX12V / EPS12V Modular Juice Box- Dedicated Graphic Card and CPU power Supply - Power Supplies)

 

I have a Sony Vaio VPCZ1390S with an Intel HM57 Chipset on the South Bridge. It has three PCI-E 2.0 x1 ports. Port 1 is occupied by a wireless card, port 2 is occupied by a memory card reader (built-in to the laptop) and port 3 is free (expresscard slot). The North Bridge has a PCI-E 2.0 x16 port with a GeForce GT 330M graphics card attached.

I read in the first post that in order to create an x2 link, you would need to connect adjacent ports (ie., p1+p2, p3+p4). It looks to me like I can't create an x2 link, even if I disabled the memory card reader. Can anyone confirm this?

Alternatively, is there anything I can do with that x16 port, via disabling the 330M card on it?

If not, I guess I'll wait for USB 3.0 display adapters.

Thanks!

 

So I got my previous problems sorted out, now I've got two more.
1) VGA output doesn't work.
2) The PSU, card and laptop all shut off under medium load. The laptop actually restarts, but the card and PSU both need to be turned off for a few minutes. Neither the PSU or card even get hot.

The card is a GT430 and the PSU is a Codegen 450w at 16A. It's similar to the one Nando is using.

 

Takaji, it is not practicable and most likely not even electrically possible to wire the GT330M's x16 port to the outside. However, you could replace the wireless card with a mPCIe adapter. The ricoh combo chip on port 2 sits on a daughterboard (connected to the motherboard via a ribbon cable) so could be unsoldered/permanently disconnected and another x1 port could theoretically be rooted out if we could get hold of schematics or reverse engineer signal lines. By doing so, you can create a x2-system with the Z (port1+port2)

As you can see in the performance charts, the best option however on a Z is x1.Opt with a Fermi card. AFAIK, you won't get a very noticable performance gain by changing from x1.Opt to x2. I think you can also do a x1E setup with an ATI card through expresscard with setup 1.x/port4 enable. If you actually want to use such external graphic solution with a Z, usb3.0 hooked up via EC won't give you same performance since its PCIe-USB3.0 controller already does conversion job and will bottleneck your setup afterwards.

 

How long is your mHDMI cable? What Jumper/Switch configuration do you have on your PE4L? Are you sure PSU is powered on without interrupts (Paperclip loses contact or so, if you are using SWEX, ignore that question)? Have you checked your windows event log to figure out if there is some shutdown signal? Did you try to install an older NVidia driver?

 

The mHDMI cable is the standard one that came with the PE4L, so around a foot long. I'm using a paperclip for now as the seller forgot to include the SWEX however it is on it's way. The paperclip is in there firmly, so I'm sure it's not that. I haven't checked the logs yet actually, I'll have a look in just a second. I'm using the latest Verde driver (270.61) with Nando's nvam.inf.

I've got 22 of these.

 

Hmm, I'd suggest you check your RAM with memtest86 or something first. I guess we need to narrow the cause down first. Do you have the opportunity to try a fresh win7 setup? Do you have a second GPU laying around for testing? Did you wire up the PE4L correctly? What is the Jumper configuration? Both LEDs on? Try to pull JP4 out before assembling to the GPU, if both LEDs come on with that, give it a run.

 

Careful with your terminology, modular means you can disconnect the power cables. That's a dedicated PSU for graphics, and one of the comments says it turns on by molex connected to the primary PSU. If you use that thing you can have your notebook feed power to it via USB possibly, although if it works off the 12V signal then it gets a bit complicated.

All that said, 25 bucks is way too cheap for a PSU, that looks very sketchy and will run hot and loud, not to mention probably delivering pretty poor power and having low efficiency.

 

I don't really have the opportunity to try a fresh installation and I don't have a second GPU (I come from a family of laptop lovers ), The PE4L is wired correctly, both LEDs are on, but when I remove JP4, the light for mains power comes on, then when I plug it into my computer, it turns off and the other one comes on.