Plug & Play External Graphics Card Solution for Precision laptops?

I'm toying with the idea for myself (for my M6700). I noticed that on eBay some sellers will include a power supply. It's basically a repurposed laptop power supply, so you don't have to have a full ATX power supply hanging around.
Example: http://www.ebay.com/itm/381874142524

 

also just remembered that my m6500 has express card expansion, so I wouldn't need to use WWAN connection for it

 

Let us know how it goes if you decide to try it. Since it has become pretty clear that a MXM GeForce 1060 won't work in the M6700, this is the next best option for me to upgrade the GPU... but if I do it, I won't be messing with it until summertime.

 

I'm going to order the "Beast" along with an inexpensive GPU. Want to see how it improves my nephews 2010 Dell Studio 1558 has i5 and a soldered ATI GPU next to the CPU on the mobo. His has an expresscard port. I'm interested to see if it improves some of his civV and civVI gaming and other games (if it works at all). I'm not sure how this will play out. His doesn't even support switchable graphics even though he has both a dgpu and igpu, uses ATI gpu exclusively. Not sure if we'll be able to utilize the egpu on internal or even on the external display. Have to find out. I'll try it on the m6500 too

I'm a bit confused on if I need a desktop ATX desktop power supply in addition to connection for a laptop AC adapter power supply. Or if desktop PSU is only required for the more power hungry GPU cards. I'm reading up on it before pulling trigger. I just read that others with studio 1557 have had confirmed success using the "beast" but I'm not sure to what extent.., so Studio 1558 presumably should work too.

@Aaron44126 I'm curious what restricting factors prevented the 1060 working in the 6700.

 

cleaning up my post and edits from earlier.

Here are the laptops I'll be testing. I may change my plans to use mPCIe connections from the eGPU adapter to the laptops instead of using expresscard as both laptops have chipsets with PCIe rev 2.0 but if the expresscard port of laptop is older rev 1.0 then the eGPU would only see PCIe rev1.0 x1 speeds from expresscard. We'll see.

2010 Dell Studio 1558 with i3-330M 6GB RAM and ATI Mobility Radeon hd 5470 1GB as the dGPU. (Intel HM55 chipset supporting PCIe rev2.0)

2010 Dell Precision M6500 with i7-920xm 24GB RAM with NVidia Quadro FX3800M 1GB dGPU. (Intel PM55 chipset supporting PCIe rev2.0)

​

Here is the eGPU I've selected:

RX 470 4GB

I ordered AMD because they have a nice price performance ratio and the RX 470 is much lower power consumption than prior gen AMD.

Nvidia has a fairly major issue relating to eGPU DIY. The latest nvidia drivers prevent use of their cards in mPCIe ports unless they are setup as "HOT PLUG" enabled. I believe expresscard ports do not have that restriction but WLAN mPCIe ports do. Meaning that you either use TB3, or expresscard, or download an older nVIdia driver without that restriction (and commit to never updating it) or you get a modded system BIOS allowing you to set your mPCIe ports as "HOT PLUG". It is believed by many in the eGPU community that nVidia did this to prevent eGPU DIY use. I don't have an opinion on that, but research "error 43" as it relates to latest nVidia drivers on eGPU.io to learn more if going with an nVidia eGPU such as 1050 or 1060... https://egpu.io/forums/expresscard-...-off-mpciengff-m2-egpus-error-43-with-372-90/
​

eGPU adapter and remainder of parts:

• EXP GDC 8.0 "Beast" with ExpressCard (I suggest BPlus PE4C v2.1 though as it allows for multiple connections/lanes to some laptops for PCIe v1.0x2 in laptops that don't have pcie2.o. Or even better a BPlus PE4H v2.4 which support 4 cable connections to 4 ports for up to x4, assuming the laptop has that many ports. SUggest study various BPlus eGPU adapters as they have many for many variations and performance)
• Dell DA-D2 12V 220W power supply (connects power directly to "beast" via 8pin)
• 6p to 6p+8p cables (provides power to eGPU from "beast" via 6 pin to 6or8 pins on the eGPU). Not needed if you have ATX PSU which would connect direct to eGPU 6or8pins or if you select a VERY low power GPU that doesn't have a 6 pin connector.
• i7-640M (confirmed to be compatible upgrade from the i3-330m in the Dell 1558) ordered from ebay for $50. because, why not?

That's it for now until all the parts arrive. There's a myriad of variables to consider and I won't know more until I actually begin testing. My goals are modest, hobby/fun/experimenting

​

 

well, I got my AMD RX470 yesterday in the mail. Turns out my two older laptops (M6500 and Studio 1558) only support PCIe v1.1x1 speeds via expresscard and mPCIe as they were first gen i-core CPU chipsets. They are marketed and show up in HWiNFO as being PCIe v2.0 but apparantly that is for power only and not for v2 speeds. However responses to my thread linked below...
https://egpu.io/forums/expresscard-...d-m6500-beast-expresscard-amd-egpu/#post-4434
...shows that there is hope in that I could use a BPlus PE4C v2.1 instead of the "EXP GDC beast 8.0" to utilize two ports of the laptop for PCIe v1.1x2 speeds. But that requires of course more money: $60 for the pe4c 2.1 with an HDMI to mPCIe cable. That is in addition to the HDMI-EC cable that I currently have.

or there is the BPlus PE4H 2.4 that allows up to four cable connections to the laptop.... WWAN, WLAN, FCM, plus ExpressCard slot. Those first three are all mpci-e/non usb (even the wlan) and coupled with the express card gives potential for v1.0x4 speeds in the M6500.

Those of you on later gen precisions such as m6600 and m6700 will have true PCIe v2.0 access with ability to use two connectors to the laptop for PCIe v2.0x2 speeds or using PE4H V2.4 possibly 2.0x4. Obviously I cant test that scenario because I don't have the appropriate laptop to do so. But I will test similar but at v1.1x1 and v1.1x2 (that last one will be after my other adapters arrive).

 

DELL M6500 i7-920xm with 24GBRAM and internal Quadro FX3800m 1GB
EXP GDC BEAST v8.4d with AMD RX470 4GB
HDMI connection from EXP GDC to the ExpressCard port on M6500


connected everything except between the laptop and the GDC (expresscard unplugged) prior to booting up. Then i made sure my dell BIOS was up to date.

After booting from the BIOS update I opened Device Manager. Expanded the device adapter tree. It showed Nvidia Quadro FX3800 as only adapter. Correct.

I connected expresscard into the M6500. The fans on the external RX470 began to spin. The external display did not show anything. Still only the internal laptop display. Then a new adapter appeared in the device manager under display adapters. MS Basic Video adapter but it had a triangle. I told it to update itself to latest driver.

At that point it also began to download latest AMD driver for the AMDRX470 automatically at same time it was installing the MS Basic video driver. It asked to reboot.

After that reboot I had both displays. The internal and the external. The internal is now driven by the MS Basic adapter (presumably via the intel gpu, not sure yet). The nVidia display adapter no longer appears in device manager, just the AMD driver.

I'm not sure if the Intel GPU is driving the internal laptop display or if the quadro is. HWiNfo shows the quadro is but that doesn't seem possible without the nvidia driver active.

The external is definately driven by the RX470 as it (the TV) is connected by HDMI directly to back of the 4X470 card on the GDC dock.

I'll just post my screen captures. Maybe you all can make better sense of it than I can:

 

and showing it connected

 

I'll post benchmarks in here after I run them. I need to first get a baseline using the built in quadro gpu...

but until then here is a simplified "benchmark" to look at using my rx470 for first time as an egpu
http://www.userbenchmark.com/UserRun/3171844

Edit: Okay, I'm back. Below is HeavenBenchmark using the rx470 eGPU on the external 1080p Sharp tv. I should be doing the baseline first with the quadro, but since I'm all setup with the eGPU I'm doing those.

Heaven Direct3d11 eGPU. The frames dropped most on transition just before last scene. Within a scene Avg was above 60 and often in mid 70s with a few jumps into 100s. But in the final scene transition it dropped drastically for a split moment (I'm wondering if CPU bottleneck there as it did this with dGPU later). I'm excited to try in a real game. I tested with Directx9 too but score was almost identical.


I'll test next with quadro dGPu. First to the external and then again to the internal display to get a baseline score.

 

baseline benchmarks using internal GPU that came built in the laptop. An nvidia quadro FX3800M.

I do not have an HDMI connector on the M6500 to test HDMI connection to the external 1080p Sharp TV as I did with the eGPU. So instead I took several baseline benches with the quadro dGPU as follows:

- dGPU FX3800M and Internal 1920x1200 display - no scaling in benchmark. ie. benchmark performed at 1920x1200



- dGPU FX3800M and Internal 1920x1200 display - scaled to 1920x1080 in benchmark




- dGPU FX3800M and External 2560x1440 dispaly via displayport - scaled to 1920x1080 in benchmark. This surprised me in that it was better than using internal display


Also worth mentioning now that I've gone back and forth a few times betweeen the dGPU and eGPU . To switch from eGPU to dGPU only requires unplugging the eGPU via ExpressCard. and restarting windows. Switching back to eGPU simply plug in the ExpressCard and reboot once more. In other words, no fiddling with device manager, re-enabling drivers manually or uninstalling reinstalling drivers.

My next test will be with the same eGPU setup (RX470 via ExpressCard) and an old Dell laptop with a dGPU ATI/AMD instead of the nVidia quadro that the M6500 has

Eventually I want to try this butdget eGPU setup using the newer precision 7710 but via an M.2 pcie port instead of ExpressCard. I know many are excited about TB3 eGPU, but this is more interesting to me . Requires purchase yet another eGPU adapter and cable connections for M.2 but is worth it as this is incredibly satisfying experiment.