DIY eGPU experiences

Is your power brick 12V?

 

Is there any point in getting the new hdmi cable together with the purchase of a pe4h v2.4 atm?

 

I setup my diy vidock a few months ago, and so far it is doing great.
However, I have one problem. I don't know why but I need to fully shut down each time i unplug or plug the vidock into my laptop. Is there a way for the laptop to automatically resume from the external graphics card and the internal graphics card?

 

I'm sorry to get "dirty" this thread, but I'm happy to announce that we reached the first stage for ViDock for Light Peak:

Village Tronic made an Open Letter to see how many people want ViDock for new Z and want to know what kind of specification would love.
We have to reach 50 people Who would buy it (or at least the module I think), has to write its own preferences HERE -> https://www.facebook.com/topic.php?uid=89321949134&topic=17792

For references, take a look at the thread that triggered this opportunity http://forum.notebookreview.com/sony/600942-vidock-4g-light-peak-z21-possible.html

 

Tried two different ones, the one from my htpc, both directly and via the 4 pin floppy connector, and a universal one set to 12v.

Also tried the PSU from my server now, no luck either.

 

hwtools indicate the new cable just allows longer cable length, up from ~2m to 5m. I haven't required anymore than 1m so this new cable makes little difference to my setup other than it's flat so occupies a lower profile on my desk. The main feature desired is pci-e 2.0/expresscard 2.0 compliance for Series-6 chipsets which is still not there as yet.

Thank you for advising of the commercial developments for Light Peak. No dirt here at all. Whoever gets a Thunderbolt product out will bring much joy to the Apple community and same for the Sony Z2 community with a Lightpeak product.

A Lightpeak DIY eGPU?

On the DIY eGPU side, I have contact hwtools who have previously advised that they are working on a Thunderbolt product as detailed in eGPU PCIe 2.0 and Thunderbolt update. DIY eGPU set out to originally be a low cost, community driven solution for performance enthusiasts.

Briefly, Thunderbolt is copper based so once pci-e 2.0 cabling and Thunderbolt pinout has been sorted, it should be a relatively straightforward to update a PE4H to be a x4 Thunderbolt device.

Thunderbolt development is a stepping stone to Light peak since a x4 2.0 TB product would just need to be adapter to use an optical TX/RX pair. That could be done in a specific product OR via a Lightpeak to Thunderbolt adapter which does the optic to copper conversion. I personally think a Sony-specific LP to TB adapter would be a great solution as then any Thunderbolt device (storage, video) could be used on a Z2. Or the eGPU itself could be the adapter if it provided daisy chaining for other TB devices. Sony's proprietory LP implementation is probably their way of circumventing any Apple licensing issues in using Thunderbolt connectors.

I emailed hwtools yesterday to see if they'd be interested in making a LightPeak PE4H product and/or a Lightpeak to Thunderbolt adapter. The Z2 s deserving of such products to complement it's highest performance in smallest chassis design.

I'll update the NEWS section on the first page once I know more details of hwtool's response. If we have a firm go-ahead, interested Z2 owners may wish to liase in the design and track development of such product.

I'd suggest boot via DIY eGPU Setup 1.x. My system then can standy/resume with the internal GPU and eGPU working without a problem. I can also 'eject' my eGPU and hotplug or attach using sleep/resume method. Even Optimus internal LCD mode works after such a resume BUT that mode won't engaged if I attach my eGPU after boot or after eject.

 

How would the performance of a card be affected by the PE3A? I mean, the card could just get its power from the psu, the only thing which wouldnt be powered would be the pciadapter? Maybe the performance loss is not as much as the performance gain due to pci 2.0? or will the card not start at all?


edit: i know the PE4H 2.4 comes with a fuse bypass jumper , does the latest PE4L also come with one of those or do you still have to manually solder a connection on?

 

2011-5-8: hwtool's (now bplus) new flat cable product

If you click on the bplus NEWS link, on the right is a new product PCIEMM-030 A-500A. The "A" distinguishing it's part number from previous rounded cables. This is the new flat cable that ranges in lengths from 30cm to 500cm. Running wires in parallel with GND shields as they have now done being the best way of getting reliable high speed transmission.

So we see now that it's not vaporware.

I have received mine now. See pictures below comparing the 30cm new "flat" cable versus the previous 30cm round cable. bplus informed me here this is a better cable so can now have a cable up to 5 metres long but failed pci-e 2.0 signalling. I also received a special 15cm one (not on their website, PCIEMM-015A) that would have the greatest likelihood of working on a Series-6 chipset at full x1 2.0 speed.



UPDATE: My testing on a Series-6 HP Probook notebook finding this cable won't negotiate a pci-e 2.0 link.

Why I recommend the non-flat cable

The non-flat cable is easier to work with, bending to accomodate so is recommended over flat cable. The flat cable isn't pci-e 2.0 specced as yet so only real reason to get it would be if wanting over 2m of cable length or if were somehow threading it in your system where a thinner profile is preferred. Not many users would be in that situation so the non-flat cable would be better suited for most users.

 

Well I'd like to join the quad core and optimus band wagon so I am thinking of getting a dell latitude E6520 and picking up a 2920xm ES off of ebay, and 8GB 1600mhz DDR3 off of newegg for it. Keeping my C2D cool at high speeds is a huge pain, and I can realistically only do 3.4ghz without having a loud fan blowing underneath it, and even that is not prime95 stable (overheats and throttles). Ivy bridge got delayed until April if not later, and there's a decent chance that current SB systems would support it anyway (Intel said IB will be drop in compatible with SB systems).

Nando, you list the E6520 as having 4GB+ ram issues, which I highly doubt. You have the E6220, the 12 inch version of the E6520 as compatible with a TOLUD of 3.25GB. The E6520 even has a dGPU option available, and I plan on getting the version with an iGPU only, which should avoid any allocation problems.

I ruled out most other systems for either being too expensive, having too small of a screen, ugly imo (lenovo), and likely lacking the cooling to handle a 2920xm ES (someone ran prime95 and furmark with the dGPU option on an E6520 without the system throttling, which is more than could be said for a lot of laptops). Note that the ES version of the cpu is less than 1/3 the price of the retail version, but slightly slower (2.4ghz) and less power hungry (45W opposed to 55W).

I have 3 spare mHDMI cables and fully plan on experimenting with cutting cables shorter and trying to beef up the shielding to get the 5.0gt/s rate running with optimus. Please don't delete the reference to my little scaling analysis thing because at the bottom of that post I coppied down which 4 wires are used for PCI-E, and I threw out the index card I wrote those down on.

I thought of waiting for Thunderbolt, but it seems that with the way thunderbolt works that it would most likely require a new PE4H, and with Ivy Bridge delayed until several months into next year, that's a long time to wait.

Also I emailed Abo, the setfsb author asking about supporting the SB clock generator. If he responds to me I will post the response on the forums.

So I was just wondering what your thoughts are. The main problem I have with the E6520 is that there is no form of bios crisis recovery, meaning that there will be no experimenting with flashing an x2 link or trying to enable multiplier based overclocking in windows. If you guys know of cheap optimus compatible 15 inch systems with good cooling, and an expresscard slot, please let me know.

 

Have you already tested the 15cm one?

 

The E6520 you are proposing would give the most versatility in extracting peak performance from a x1 2.0 eGPU, offers business grade durability and often can be had for great prices from the Dell Outlet. It's really quite a bargain choice.

I have no confirmation that the E6520 will have TOLUD issues other than a E6220 has a sufficiently low TOLUD to allow an eGPU to work. Does the E6520 uses the same configuration as the E6220? Not sure - I'd advise asking in the Dell forums for a pci-e bus Device Manager screenshot for 100% confirmation. Previous E6510 and E6500 had a TOLUD of 3.5GB so Dell may have continued or scrapped that.

You could also consider geting the NVidia NVS4200M E6520. I found while perusing the Dell area that the E6520 has a rare Optimus and muxed switchable graphics setup. The bios setting to discrete actually disengages the iGPU completely and wires the dGPU to the LCD, so you can get full functionality in Linux. Its worth noting that I was able to get a disable the NVidia dGPU on a Series-6 system using DIY eGPU Setup 1.x here ( confirmed by wicked20's XPS15 L502 [email protected] implementation) which provided two benefits:

(1) can hijaak it's PCI-e space for use by the eGPU. Means if the system has a dGPU then there is no TOLUD issues

(2) then only the single NVidia eGPU is present so Optimus functionality is engaged using it. If the dGPU is active then it's given Optimus functionality and the eGPU was runs as an extra adapter with no Optimus pci-e compression no internal LCD mode - pretty much as if the desktop driver was loaded.

One new experimental feature since that testing was done is a Video cards->dGPU on-demoted that moves the secondary bus of the dGPU beyond that of the eGPU. I have no means to test it but the hope is that the eGPU will be given priority by the NVidia driver to function in Optimus capacity. If it works, then the dGPU could still remain active in the background (assuming there is pci-e space to accomodate the dGPU + eGPU) to provide Physx or engage as the LCD driving adapter in a muxed switchable setup if switched to dGPU mode. This would however mean booting via eGPU Setup 1.x which you'd be a pro at now. If you did want to maximize CPU performance and keep the system cool, then maybe the iGPU version is best?

As for hacking the bios for x2 2.0, we still haven't got a x1 2.0 Optimus confirmation because of supposed cable problems. If that works fine, then just like previous examples, x1.Opt and x2 deliver similar performance levels so no real need to try to hack the bios to do x2 2.0 (non Optimus). If x1 2.0 causes Optimus to not engage, something NVidia could have done to the driver, then you'd want to do x2 2.0 non-Optimus to get the most performance possible on the platform.

RE: pci-e 2.0 cabling, you missed my post by 4mins Advise getting the newest 15cm cable to test with the E6520.

If you want to start hacking the bios then best thing to do is to unsolder the 8-pin bios chip, put a socket in it's place that can allow easy remove and reprogramming of the chip and use a USB eeprom burner (~$30 off ebay) to program it. Please PM if want more discussion on that.

I suggested users wanting to confirm if cabling is the problem to consider modding a PE3A with input voltage for the slot (+3.3V) since it uses no cable.

PE4L doesn't need the fuse bypass since it doesn't use that fuse arrangement.

It arrived today and all the systems in the house are only pci-e 1.0 capable. Hmm.. where to find a Series-6 system with an expresscard to test the 15cm cable?

 

^ no friend with x220 for example?

 

After a month, I received my Gold Plated 1080P HDMI Mini V1.3 M-M Connection Cable - Black (1.5M-Length) - Free Shipping - DealExtreme cables yesterday. If I know HWTools will produced the new flat cable (PCIEMM-030A-500A), I would have wait. Anyway, now I can moved my eGPU to the back of my notebook, freeing some space. The new mini-HDMI cable does not running in parallel with power cables but they crossed. The distance between them is 1 inch when they crossed. Is that OK & free from interference from power cable?

I can hotplug the eGPU without using the sleep/hibernate trick (eGPU automatically take over dGPU as primary device, shutting down the internal monitor). I can eject eGPU using "Safely Remove Hardware and Eject Media" in the taskbar (dGPU take over as primary device & internal screen lit). However, in the event of power failure, my notebook freezed (I think so since both keyboards - internal & external - doesn't responding). dGPU failed to take over.

 

Can anyone clarify the uses of WWAN please other than having a mobile service provider give you internet

I may need to get an adapter so that I can put the PM3N in it's place

 

If you move your wifi card into the WWAN slot and the system doesn't pick it up, as I expect will happen, then the WWAN slot only has USB/mSATA lines. In which case you'd need to remove the wifi card and put a PM3N in it's place since it's the only place you have pci-e lines.

To run concurrent wifi you'd then need to go external USB OR you might have luck getting wifi-in-WWAN (USB) going if Lenovo haven't whitelisted the slot.

 

If the problem is the cable, isn it possible to make hwtools.com to create a board connected to express card with 2 separated 1x 1.0 channels connected to another board with 1 channel which go to the Video Card?

 

So I tried 3 different power sources, 2 different laptops and I'm getting nothing. No new device found in the Windows Device Manager, no bootup errors, no errors anywhere.

Made a DIY ViDock Setup 1.x bootdisk, it finds nothing. It tells me to set a 7/15s delay and do a cold boot, how do I even set this delay? Can't find much in the FAQ and the PE4l manual doesn't mention it either. Is it those tiny flip switches on the pe4l? If so in what position should they be for a 7-15s delay.

Devcon result:

Code:

ACPI\PNP0F13\4&D3590E2&0
    Name: PS/2 Compatible Mouse
    Device is currently using the following resources:
        IRQ : 12
ACPI\PNP0000\4&D3590E2&0
    Name: Programmable interrupt controller
    Device has the following resources reserved:
        IO  : 0020-0021
        IO  : 0024-0025
        IO  : 0028-0029
        IO  : 002c-002d
        IO  : 0030-0031
        IO  : 0034-0035
        IO  : 0038-0039
        IO  : 003c-003d
        IO  : 00a0-00a1
        IO  : 00a4-00a5
        IO  : 00a8-00a9
        IO  : 00ac-00ad
        IO  : 00b0-00b1
        IO  : 00b4-00b5
        IO  : 00b8-00b9
        IO  : 00bc-00bd
        IO  : 04d0-04d1
ACPI\PNP0100\4&D3590E2&0
    Name: System timer
    Device has the following resources reserved:
        IO  : 0040-0043
        IO  : 0050-0053
        IRQ : 0
ACPI\PNP0103\0
    Name: High precision event timer
    Device has the following resources reserved:
        MEM : fed00000-fed003ff
ACPI\PNP0200\4&D3590E2&0
    Name: Direct memory access controller
    Device has the following resources reserved:
        IO  : 0000-001f
        IO  : 0081-0091
        IO  : 0093-009f
        IO  : 00c0-00df
        DMA : 4
ACPI\PNP0303\4&D3590E2&0
    Name: Standard PS/2 Keyboard
    Device is currently using the following resources:
        IO  : 0060-0060
        IO  : 0064-0064
        IRQ : 1
ACPI\PNP0A03\FF
    Name: PCI bus
    Device is currently using the following resources:
ACPI\PNP0A08\0
    Name: PCI bus
    Device is currently using the following resources:
        MEM : 000a0000-000bffff
        MEM : 000d0000-000d3fff
        MEM : 000d4000-000d7fff
        MEM : 000d8000-000dbfff
        MEM : 000dc000-000dffff
        MEM : 000e0000-000e3fff
        MEM : 000e4000-000e7fff
        MEM : c0000000-feafffff
        IO  : 0000-0cf7
        IO  : 0d00-ffff
ACPI\PNP0B00\4&D3590E2&0
    Name: System CMOS/real time clock
    Device is currently using the following resources:
        IO  : 0070-0077
        IRQ : 8
ACPI\PNP0C02\1
    Name: Motherboard resources
    Device has the following resources reserved:
        MEM : fed1c000-fed1ffff
        MEM : fed10000-fed13fff
        MEM : fed18000-fed18fff
        MEM : fed19000-fed19fff
        MEM : e0000000-efffffff
        MEM : fed20000-fed3ffff
        MEM : fed45000-fed8ffff
        MEM : ff000000-ffffffff
        MEM : fee00000-feefffff
ACPI\PNP0C02\2
    Name: Motherboard resources
    Device has the following resources reserved:
        IO  : 0061-0061
        IO  : 0063-0063
        IO  : 0065-0065
        IO  : 0067-0067
        IO  : 0070-0070
        IO  : 0080-0080
        IO  : 0092-0092
        IO  : 00b2-00b3
        IO  : ff00-ff0f
        IO  : ffff-ffff
        IO  : ffff-ffff
        IO  : 0400-047f
        IO  : 0500-057f
ACPI\PNP0C04\4&D3590E2&0
    Name: Numeric data processor
    Device has the following resources reserved:
        IO  : 00f0-00f0
        IRQ : 13
ACPI\PNP0C09\4&D3590E2&0
    Name: Microsoft ACPI-Compliant Embedded Controller
    Device is currently using the following resources:
        IO  : 0062-0062
        IO  : 0066-0066
ACPI\PNP0C0A\1
    Name: Microsoft ACPI-Compliant Control Method Battery
    Device is not using any resources.
ACPI\PNP0C0C\2&DABA3FF&2
    Name: ACPI Power Button
    Device is not using any resources.
ACPI\PNP0C0D\5&2F008F20&0
    Name: ACPI Lid
    Device is not using any resources.
ACPI\PNP0C14\0
    Name: Microsoft Windows Management Interface for ACPI
    Device is not using any resources.
17 matching device(s) found.

Ordered a expresscard->usb3.0 card which I should be getting tomorrow so I can test if there's something wrong with the expresscard slot, but I doubt it since it didn't do anything on two systems.

 

Hello. Do you have any other suggestions or I go back to the GTX580 instead of GTX590?
Bye and thank you.

 

Which laptops did you use? Brand & model? Desktop graphic card?

 

I don't know. Maybe you can try "Save diags" function in the Setup 1.x to get the system snapshot, post it here & ask nando4 to check it for you.

 

The laptops I tried were a MSI GX640 and a Lenovo 3000 N200, the MSI is were it should run on as that's my own laptop . Both are running Win7 pro.
The desktop graphics card is a ASUS ENGT520 (NVIDIA GT520). The card itself works great in two other systems (my htpc and server) so it isn't defect.

 

This guy have the same problem with 12V adapter but no problem with ATX PSU. However his jumper settings are different though (look in the video description).
‪Problem - PE4L 1.4c with 12V adapter‬‏ - YouTube

Does both LEDs light up when you tried with 12V adpater or ATX PSU?

 

Hi NewNet,
Did you try compaction w/ 2 GB of RAM? I notice that I can get 1024MB free w/ 2 GB compare w/ 256 MB free w/ 6-8GB RAM.

Thanks,

 

Thanks again Nando.

Could somebody ask them or someone to make or where to find adapters for the Apple TB's active cable, such as mini Displayport to mini-hdmi male? Or make similar active cable for mini hdmi connectors?

Thanks.

 

I've suggested bplus consider integrating the use of the Apple's $50 Thunderbolt cable with mini displayport ends for the PE4L/PE4H to get pci-e 2.0 speed. Making their own cable to mimic Apple's would require getting Gennum chips who advised me:

Gennum’s pinout is only available under NDA, and only after you have received the Thunderbolt Interface Specification from Intel.

NewNet and I have done lengthy troubleshooting with the GTX590 already via PM but could not get around the error 43. GTX580 works fine.

The only thing we haven't done is save a 16kb dump of the GTX590 PCI space on a working desktop system, then load that in the notebook. It may well be that the internally SLIed GTX590 requires some additional initialization that simply isn't being done on a notebook.

Try a ATX PSU first. It's the least troublesome power source to test a PE4L with since it provides a 12V and 5V input. The 5V input is down regulated to 3.3V.

An external 12V AC adapter can only supply 12V so that is why a jumper is set so 3.3V (@1A) comes from the expresscard. The official spec for pci-e says up to 3A can be drawn so such a setup is underspec and may or may not cause issues depending on how much 3.3V power the desktop video card actually draws.

Also set the the PCI reset delay switch to 0s (far left) position.

 

I see. NewNet latest post doesn't show he/she already got help.

BTW, Axeia already tried 380watt Seasonic PSU & still not working.

 

So I got a reply from someone with an E6520. It's TOLUD is 3GB, so I went ahead and ordered a base model with the top end screen for $950 including tax. (Since when were computers taxed!)

I also ordered an 2920xm ES from ebay for $306, and 8GB of 1600mhz ddr3 from newegg for $65. Dell only offered 1333mhz ddr3 even though Intel's site says 1600mhz is supported and many people have purchased it. It hardly cost anything to go from 1333 to 1600 so I lose nothing but performance if it only runs at 1333mhz.

Unfortunately Dell has the estimated shipping date of August 22nd. Much experimenting will be done with mHDMI cables. pci-e 2.0 WILL happen one way or another.

 

I've got a ThinkPad x220 (8GB) + GTX 460. I ran into the Code 12 issue, so I'm using Setup 1.x. I think I've done everything correctly, but I'm stuck on how to finish up the setup with chainloading to Win7. When I select that from the menu, I am brought to a grub prompt and I'm not sure what command to enter. Can anyone advise here?

Also, both bridge compacting and 32-bit compacting resulted in no solution found. But 36-bit compacting appeared to work. Can anyone with an x220 verify that that sounds right?

Thanks in advance!

 

Two questions:

(1) Did you flash a bios older than 1.12 that uses a lower TOLUD. X220 bios newer than 1.12 sets TOLUD=3.5GB when you have 4GB or more installed so has no room to accomodate a eGPU. If you are using that bios then previous X220 implementations linked from the mai page indicate no need to even run Setup 1.x

See Request to fix TOLUD in newer BIOS files - Lenovo Community

(2) If you still require Setup 1.x then are you using UEFI/EFI boot? This is sometimes referred to GPT partitions. If so, the DIY eGPU Setup 1.x isn't setup to chainload those. I'm evaluating the best way of adding that feature. Could just re-do your partitions using a traditional scheme for the time being.

Congratulations on the order. Can't wait to see what performance you get. Your OC/d GTX460 setup should be the first to break 3dmark06 of 20k.

To do full spectrum pci-e 2.0 testing consider obtaining (1) a PE3A (+mod for input power) and (2) a new 15cm flat cable.

 

Could someone with experience ordering from harmonic inversion tell me how long it would take the PE4H to be shipped to the continental US? Is it faster than from HWtools?

 

I recall HIT does 3-day ground shipping. They charge $10 for it. bplus takes the same time or perhaps 1 day longer and they charge $22 for it. A PE4L is significantly better value from HIT being $70-shipped with SWEX. bplus charging $87-shipped for the same.

 

Why do you keep recommending the PE4L while all the setups that I see use a PE4H?

 

Because it's cheaper & can deliver the same functionality. However, I recommend PE4H. Mainly because you can do more than x1 link with compatible notebooks.

 

Hello everyone,

Have anybody had luck getting 1.2 opt set up working on a laptop with discrete Radeon gpu? I think I remember reading about the difficulty of disabling Radeon gpu a few pages back; I was wondering if there has been any progress.

My laptop HP Elitebook 8560P came with Sandy Bridge i7 dual processor and AMD Radeon HD 6470M. It seems that Intel HD 3000 is disabled/inactive in this machine - it doesn't show up in device manager nor I can install Intel graphics driver. There seems no options to disable discrete card in Bios (latest ver.) Expresscard slot can configured to gen 2 in Bios.

quick spec: http://h18004.www1.hp.com/products/quickspecs/13947_na/13947_na.pdf

edit: Apparently full expresscard 2.0 speed of 5Gbps cannot be achieved via express card slot. see image below
Given this, what would be the best solution to extract the maximum performance from an external graphics card? Thanks in advance

 

I've the sliders set to the far left. The adapter can do 7.0A.
The Zotac GT520 draws 31 watts tops, mine should draw slightly less as it doesn't have a fan. Almost all my test have been using this adapter via the 4 pins FDD as I don't get any of the leds light up otherwise.

If that sounds confusing, see the pic. It probably does a better job explaining it than me.

With the adapter connected to the 12v DC plug none of the leds light up at all. Using the ATX plug with the jumper in either position or no jumper I get the D2 Ext led (minus / one furthest from the FDD plug) to light up.
With both jumpers connected both leds light up.
With the adapter dc plug the usb port doesn't work either, with the ATX plugged in I can use the USB port on the pe4l but so far not the pci-e slot.
What's the desired situation, should both light up?

 

^what kizwan said

The problem you've identified is there is no active Intel HD3000 on your system, so Optimus will not work. It is not possible to enable it to work either.

The HD6470M dGPU is pretty low on performance so even a x1 NVidia or AMD eGPU would perform better plus it will improve yet again once we have x1 2.0 hardware going. However, the NVidia Optimus driver will not engage so you can't get the pci-e compression that elevates Optimus implementation's benchmarks even beyond x2 (non-Optimus) levels.

The tech in your 8560P is a basic dGPU-only setup compared to the switchable, battery life extending tech of some other systems like a Thinkpad T520/W520 or a Latitude E6520. The Latitude's muxed setup being the most versatile from the lot.

If x1.2Opt is your aim then the best bet getting there would be do a systemboard swapout. Maybe HP would be willing to do it for you if there was some misconception the system came with Intel HD3000 + HD6470M switchable graphics?

The 12V DC setup should see D1 light up if JP4 is set. But anyway.. it's easier to work with the ATX PSU until it's all working OK. The LEDs confirm it's powered up OK. So the USB lines works but the pci-e lines do not.

Try to hotplug your expresscard and/or the mini HDMI cable to see if that gets it going. Worth also checking what port the expresscard is and making sure it isn't disabled in the bios. We did have that occur at one stage way back. Given the video card is confirmed OK in your previous post here, that then leaves the mini HDMI cable/EC2C/PE4L, or the expresscard slot as being suspect. It's unlikely that two expresscard slots would be faulty so it's the PE4L-EC2C+mini HDMI cable that are the likely problem. I'd suggest trying another mini HDMI cable first and take it from there.

 

Let me get this straight, doing more than x1 link meaning crippling your laptop and basically converting it to a freaking desktop since its not gonna be moving soon while connected to another mPCIe slot via the miniHDMI on the laptop besides the expresscard to get double bandwidth ?

 

Thanks for the quick reply nando4. Amazed by what you and other members have done with this very informative thread on egpu!

While I chose the laptop with no gaming in mind (but soon I realized that I can't pass up Diablo 3!), it's disappointing see HP resorting to a less sophisticated graphics system when more flexible switchable gpu set up is implemented in the segment competitors. But beside this issue I'm very satisfied with the laptop.

Forgive me if this has been asked before.. what is your speculation on egpu solution via USB 3.0 in near future?

 

When one opted to use eGPU, he/she already converting his/her laptop to a partial desktop computer. The most convenient setup is to use ExpressCard for x1 or x1.Opt configuration. Of course, x1.Opt will deliver better performance than x1 link however not all people have laptop with Intel HD Graphics or Intel 4500MHD. So, what they did is improvised. It's not convenient but one can put his/her laptop to hibernate, disconnect the mini-HDMI cable from mini-PCIe card (or disconnected the PM3N cards altogether), put back wireless card, close the bay & go mobile. Yes, it's a lot of work but for enthusiast it won't be a problem. I don't mind use the two available mini-PCIe ports to get more bandwidth. Hence, why I chose PE4H. If your laptop have Intel HD Graphics or Intel 4500MHD & ExpressCard slot, it will be much better because you'll only need to use ExpressCard slot to do x1.Opt link with Nvidia GPU. However, on some notebooks still need to put them to sleep/hibernate before unplugging the eGPU. My Dell is one of the laptops who can support hotplugging the eGPU. (Actually I can enabled hotplug capability on any of the PCIe ports.)

Anyway, setting x2 link, meaning using the mini-PCIe ports won't crippled my laptop. It still a mobile computer. I still can go mobile with it. Even though I didn't do x2 link setup yet but I'm confident it just take me less than a minute to unplug the eGPU. If you think all of this inconvenience, there nothing I can say to make you think otherwise. I respect everyone opinion & choices.

 

and I saw you waiting for pci-e 2.0 compliant cabling.
What's this? It's important for me or not need in my case?

(Laptop is lenovo g560 I5 2.67GHZ 3GB RAM)

Thanks

 

Hi Gil3,

You don't need to wait for PCIe 2.0 kit. It only useful for Sandy Bridge notebook.

 

OK
Thank you!

 

It could be great setting x2 link if you don't use your lappy for mobile purposes.
I'm actually thinking of selling my ASUS G51J-A1 (i7 720qm, gtx260m) and getting a sandy-bridge with a intel HD to get it work with x1.Opt, but only after I check it's performance with a SSD hd.

btw this thread is awesome.. keep the good work.

 

Hotplugging didn't change anything.
Got my Expresscard -> USB 3.0 today and the expresscard slot works just fine, so it isn't blocked in the BIOS and is functional. Sent a mail to the seller requesting a replacement, got it off ebay as it's cheaper and the shipping costs are much much lower, also comes with the ATX flip switch as a nice little bonus.. kind of regretting that choice now though .

Appreciate the help, guess I'll head to the shops to see if they sell a mini-hdmi cable here but I doubt it.. they're years behind with everything.
[edit] As expected, no mini-hdmi cable to be found.. time to find a webshop[/edit]
[edit2] I'll wait for the ebay seller to reply, mini-hdmi cable is €20,- (that's 28.53 usd)[/edit2]

 

Is the seller at the link below:
PE4H + EC2C ExpressCard to PCI-E Adapter V2 stock 10pcs | eBay
Reliable?

It's the PE4H ver 2.0a which is what I need Based on what was told to me

 

Same guy (company?) where I got my pe4l from, same price too! Mine seems to be defective, going by his ebay feedback score it does seem to be reliable. I'll post again once he responds to my replacement request.

 

Hi guys. I did a lot of reading before jumping into the band wagon of eGPU. I need some advice before proceeding before I proceed with this project.

I'm currently having a Dell Inspiron 1440 with these specs

CPU: Intel Core 2 Duo T6600 2.2 Ghz
GPU: Ati Radeon Hd 4330
Memory: 3gb

I have a few question.
1. Aprrently using Everest software to check my PC details, I have only 1 PCI-e express slot which is currently in use by the dGPU. So should I just get PE4L or better to go for PE4H? Or am I looking at the wrong on checking how many PCI-e slot I have?

2. How do I check my TOLUD? I have been spending an hour trying to look for it but I don't know how.

3. Well I'm getting a
a. ATX PSU = Coolermaster GX-650W (well 550 and 650 it's just 50 bucks different here in terms of my currency so not much a big of deal.) GX-650W - Cooler Master
b. Sapphire ATI Radeon HD 6790
c. Either PE4H OR PE4L?

Question is if I do get PE4H, if the cable in the future does support 2.0 and I change my laptop to get express card 2.0, with the PE4h will give me 2.0 speed? Or do I need to reinvest? The currency exchange rate to USD is high, so I have to weigh my purchase properly.

Any other thing I should look out for?

Thanks guys for helping and I really appreciate for those invested time building this article and thread.

 

Some USB 3.0 (5GT/s) to pci-e chips will be released to the market later this year. bplus have expressed interest in creating USB 3.0 eGPU hardware. USB 3.0 is on many systems so it would be a great solution if a desktop video card can be made to works without latency.

A PE4L is smaller and can accept 12V AC adapter input, but needs props to stand on a desk. A PE4H is larger, is self supporting uses 15-19V AC adapter input, can allow x2/x4 links with some notebooks and costs more. Both work great with an ATX PSU. If money is tight the PE4L will do the job but a PE4H sits easily on my desk and so I prefer it.

The Core2Duo Inspirons don't have any TOLUD issues. You could even load them up with 8GB of RAM with no problems. They have a somewhat rare 36-bit root bridge so the latest Win7 ATi/NVidia driver can automatically allocate the eGPU to 36-bit space. No error 12 with those. The 1440 offers quote a decent package at a low buy price.

The Inspiron 1440 implementation I did with the 4500MHD (iGPU) model had the the expresscard slot as port5, so can do a x1E link to with an ATi/AMD card giving some decent performance. x2 is not possible on those systems.

bplus are testing different cables to achieve x1 2.0 compliance . I would like to say that a new cable in the future is all that is required but there is still the posibility that the EC2C or PE4H/PE4L itself may also be the speed limiting culprint. Consider too that there are plenty of Series-5 or older systems that can use the pci-e 1.0 gear so it would get a resale value if you wanted to update to a pci-e 2.0 kit when it's released.

 

Quite interesting.... as you know, Sony Light Peak is a modified USB port, but we know that it can work even like a normal USB 3.0 port. What do you think about this?? When that port works like a USB 3.0 "talks" with TB/LP chip or there are additional connection soldered on PCB towards the northbridge?

 

the problem is that USB3 has a theoretical limit of 5gbps, not that it actually uses it, its as fast an esata port which is 3gbps.