Hi Marko,
Technically, even the x2 lane configuration does provide up to 40Gbps overall. It’s not intuitive. Essentially, PCIe signals are just one of several that are sent within the overall Thunderbolt 3 cable. Most notably, there’s also DisplayPort signals which use separate wiring and could utilize all the remaining Thunderbolt bandwidth.
Imagine two collapsible hoses within a larger, outer hose carrying up to 40 gallons of water per second. The inner PCIe hose needed for eGPUs can be connected to a larger x4 or smaller x2 faucet. The larger faucet could be opened to allow up to nearly 32 gallons per second while the smaller could open to almost 16 gallons per second. GPUs rarely need to open these faucets all the way and do their work just fine with just a trickle of water and an occasional gush coming through. Most of the hard work coming out of modern games is generated within the beastly GPU itself, just taking quick instructions from the CPU. But even in those rare instance that a faucet is fully opened, there remains space for DisplayPort signals too, with an overall maximum of 40 gallons per second possible regardless.
So, Dell and other manufacturers really aren’t misadvertising nor does fewer lanes prevent you from benefiting from a eGPU. There may be a few instances when having x4 lanes might expedite something within the graphics system, but overall the impact in actual games is small. You will still see FPS approaching those seen in a desktop system. Actually, you won’t notice the half bandwidth until other technologies come along like super fast external RAID storage. Even the GTX 1080 simply doesn’t use all that much still.
I’m new and I’m hoping to add to this thread. I just installed a new dell precision 7510 along with a razer core egpu and a nvidia gtx 1080 founders edition. While the the 7510 supports a 4x thunderbolt link and the cuda-z results look promising, the card otherwise doesn’t work. Meaning even with the onboard nvidia m2000m disabled the other benchmarks don’t seem to be using the gtx 1080 and must still be using the igpu (intel p580). Almost like the cuda cores are available to the system but the display isn’t. Connecting an external display (dell p2715q) causes the local display to lock up in different ways. For example the cursor can move between screens but the system is otherwise locked up. Here the cuda-z and hwinfo results.
For some reason I can’t upload the hwinfo results. Not sure why.
While the the 7510 supports a 4x thunderbolt link and the cuda-z results look promising, the card otherwise doesn’t work. Meaning even with the onboard nvidia m2000m disabled the other benchmarks don’t seem to be using the gtx 1080 and must still be using the igpu (intel p580). Almost like the cuda cores are available to the system but the display isn’t. Connecting an external display (dell p2715q) causes the local display to lock up in different ways. For example the cursor can move between screens but the system is otherwise locked up. Here the cuda-z and hwinfo results.
There is an existing Precision M7510 eGPU implementation at the following URL. Yes it has a x4 3.0 link to the TB controller.
https://egpu.io/forums/implementation-guides/dell-precision-7510-akitio-node-galax-gtx-1080-hof-its-working/
What’s interesting is your CUDA-Z performance has double the host-to-device bandwidth shown in that link.This is worthy of an implementation guide as a M3510 too also had the same halve host-to-device bandwidth. You are getting full Tb3 performance proving the Precision M7510 (and likely M3510) have proper TB3 hardware implementations.
As to your question about driving the internal LCD, the dGPU needs to be completely disabled to facilitate that. To do that go to Device Manager-> View Devices By Connection. Find your dGPU then directly above it will be a PCIe Port or Bridge. Right click to disable it along with which the dGPU will disappear. Reboot and test some apps to confirm eGPU acceleration.
While the the 7510 supports a 4x thunderbolt link and the cuda-z results look promising, the card otherwise doesn’t work. Meaning even with the onboard nvidia m2000m disabled the other benchmarks don’t seem to be using the gtx 1080 and must still be using the igpu (intel p580). Almost like the cuda cores are available to the system but the display isn’t. Connecting an external display (dell p2715q) causes the local display to lock up in different ways. For example the cursor can move between screens but the system is otherwise locked up. Here the cuda-z and hwinfo results.
There is an existing Precision M7510 eGPU implementation at the following URL. Yes it has a x4 3.0 link to the TB controller.
https://egpu.io/forums/implementation-guides/dell-precision-7510-akitio-node-galax-gtx-1080-hof-its-working/What’s interesting is your CUDA-Z performance has double the host-to-device bandwidth shown in that link.This is worthy of an implementation guide as a M3510 too also had the same halve host-to-device bandwidth. You are getting full Tb3 performance proving the Precision M7510 (and likely M3510) have proper TB3 hardware implementations.
As to your question about driving the internal LCD, the dGPU needs to be completely disabled to facilitate that. To do that go to Device Manager-> View Devices By Connection. Find your dGPU then directly above it will be a PCIe Port or Bridge. Right click to disable it along with which the dGPU will disappear. Reboot and test some apps to confirm eGPU acceleration.
Hey there! I recently got my new Dell Precision 7510 and I’m very happy with it. I’m also planning to get an egpu although I’m not yet sure which to get. I really would like to double it as a Docking Station so some USB Ports and an Ethernet Jack would be nice. Since I ordered mine I’ve read a lot of forum threads and did some research.
When i read about the XPS 15s TB Controller isn’t connected with 4x Pcie3 I was worried about the 7510. But, like some other posts I can confirm that the Controller is conected with 4x Pcie3 as I can tell from hwinfo64. Since it worries me a bit, I would like to find out why there are these strange numbers about the host-to-device bandwidth in the configurations here ( https://egpu.io/forums/implementation-guides/dell-precision-7510-akitio-node-galax-gtx-1080-hof-its-working/ )
Since I have no egpu Box at hand currently I have to rely on your information.
I also noticed that the “new” 7520 was release a few days ago and the Dell Support Page provides a Intel Thunderbolt 3 Firmware Update which is NVM 12 while the 7510s is NVM 16.
There are going on a few strange things with these Thunderbolt 3 Firmware Updates across the Industry. Are there so many different controllers around? I can only think of 4 different from TI???
While the the 7510 supports a 4x thunderbolt link and the cuda-z results look promising, the card otherwise doesn’t work. Meaning even with the onboard nvidia m2000m disabled the other benchmarks don’t seem to be using the gtx 1080 and must still be using the igpu (intel p580). Almost like the cuda cores are available to the system but the display isn’t. Connecting an external display (dell p2715q) causes the local display to lock up in different ways. For example the cursor can move between screens but the system is otherwise locked up. Here the cuda-z and hwinfo results.
There is an existing Precision M7510 eGPU implementation at the following URL. Yes it has a x4 3.0 link to the TB controller.
https://egpu.io/forums/implementation-guides/dell-precision-7510-akitio-node-galax-gtx-1080-hof-its-working/What’s interesting is your CUDA-Z performance has double the host-to-device bandwidth shown in that link.This is worthy of an implementation guide as a M3510 too also had the same halve host-to-device bandwidth. You are getting full Tb3 performance proving the Precision M7510 (and likely M3510) have proper TB3 hardware implementations.
As to your question about driving the internal LCD, the dGPU needs to be completely disabled to facilitate that. To do that go to Device Manager-> View Devices By Connection. Find your dGPU then directly above it will be a PCIe Port or Bridge. Right click to disable it along with which the dGPU will disappear. Reboot and test some apps to confirm eGPU acceleration.
Hey there! I recently got my new Dell Precision 7510 and I’m very happy with it. I’m also planning to get an egpu although I’m not yet sure which to get. I really would like to double it as a Docking Station so some USB Ports and an Ethernet Jack would be nice. Since I ordered mine I’ve read a lot of forum threads and did some research.
When i read about the XPS 15s TB Controller isn’t connected with 4x Pcie3 I was worried about the 7510. But, like some other posts I can confirm that the Controller is conected with 4x Pcie3 as I can tell from hwinfo64. Since it worries me a bit, I would like to find out why there are these strange numbers about the host-to-device bandwidth in the configurations here ( https://egpu.io/forums/implementation-guides/dell-precision-7510-akitio-node-galax-gtx-1080-hof-its-working/ )
Since I have no egpu Box at hand currently I have to rely on your information.
I also noticed that the “new” 7520 was release a few days ago and the Dell Support Page provides a Intel Thunderbolt 3 Firmware Update which is NVM 12 while the 7510s is NVM 16.
There are going on a few strange things with these Thunderbolt 3 Firmware Updates across the Industry. Are there so many different controllers around? I can only think of 4 different from TI???
Edit: I messed up the last point about the Thnunderbolt 3 controllers which come from Intel of course! Sorry for that! I also looked it up and there are 5 avalible currently.
The latest NVM Version I could find as an Update is 21 (Dell XPS 15). But the latest Intel Thunderbolt 3 compatible NUC is still on Version 18 which is quite confusing!
It might be worth noting that Dell has published an official-ish list of which laptops are using 2 lanes or 4 lanes, and are keeping it up to date:
http://www.dell.com/support/article/en/en/frdhs1/QNA44089/thunderbolt-3-40gbps-data-transfer-rate?lang=EN
I wouldn’t expect any changes in mobos for current models, just have to wait for another cycle.
I’ve read through this thread and some other threads and websites on the topic of the Dell TB3 PCIE 3.0 x2 limitation. I’m still a bit confused as to the practical ramifications of this.
For example, is the gimped Dell TB3 on the XPS 15 9550/9560 able to provide decent gaming frame rates on an external 1080P monitor?
I am considering the 9560 along with a Razer Core, Akitio Node or Mantiz Venus. I gather that there’s no point in buying a GTX1080 and playing on the internal monitor due to a possible bottleneck. But what about a 1050Ti/1060/1070 using an external FHD gaming monitor attached to the eGPU directly? Am I likely to get 60+ fps on modern titles such as Overwatch? I’m guessing the answer is yes, since TB2 implementations seem to be able to do this, but I wanted to verify.
If anyone has used their 9550/9560 with an external FHD monitor, can you please report some gaming FPS results?
I tested Battlefield 1 ( copy of a friend ) on my 9550 (6700HQ TS-3.1Ghz) on an external 1080p Monitor with a Zotac GTX 1060-6 on HIGH preset the other day.
I played an hour on some different conquest maps with over 40 concurrent players.
CPU Load was at 60% to 80% with some 100% spikes for less than a second(?).
FPS were at 55 average and 49 lowest.
I tested Battlefield 1 ( copy of a friend ) on my 9550 (6700HQ TS-3.1Ghz) on an external 1080p Monitor with a Zotac GTX 1060-6 on HIGH preset the other day.
I played an hour on some different conquest maps with over 40 concurrent players.
CPU Load was at 60% to 80% with some 100% spikes for less than a second(?).
FPS were at 55 average and 49 lowest.
Not bad at all, thanks! I think the gimped TB3 may well work out fine for what I want with an external monitor. Now to figure out if I should save some money and go with a 1060 because a 1070 might be bottlenecked by the x2 lanes on an external 1080p monitor. Was also thinking of buying a new GTX1080 with the recent price drops and use it for my desktop rig, and hand down my own 1070 for the eGPU…
Great video and well explained:
ϟ AKiTiO Thunder2 + EVGA GTX 1060 6GB SC Gaming (macOS Sierra 10.12.4 and Windows 10)
MacBook Pro (Retina, 15-inch, Later 2013) 3.2GHz Quad Core Intel i7-4750HQ / 8 GB 1600 MHz DDR3 / 256GB SSD + 1TB
✪ mini eGPU ● PCI Express vs. Thunderbolt ● Mac CAN game ● Gaming Laptops vs. MacBook Pro with eGPU
Just thought i’d add to this thread with my real world findings, i have an XPS 9550 and tested with a colleagues Akitio Node and ASUS Strix DC2 1060 6GB card – using external DELL monitor plugged in via HDMI directly into the gfx card.
Just ran a few quick tests with this setup:
-Mass Effect Andromeda, 1080p, on Ultra settings: Ran at a consistent 61 fps
-Heroes of the Storm, 1080p, on Ultra settings: Ran at between 140fps and 220 fps depending on the intensity of what was happening.
The FPS takes a hit if i disable the internal 960M, and run on my internal screen, but i think that’s to be expected.
Edit: forgot to mention I used a 2m Startech Thunderbolt 3 cable – rated at 20gbps.
Thank you for the update yiannisoc. I’ve been slammed with work so I haven’t had a chance to run tests myself but I’m assuming running at 4k will be around the same fps. Has anyone here given it a shot?
I’ll be trying it on a 4k tv, probably next week when i’m home, and i get my personal Node up and running (issue with the PSU atm). I’ll be sure to post back here when done.
Working