eGPU Results - Clevo P870DM-G + GTX1080 + TB3 HDK / PE4C v4.1 (M.2 NGFF)

System :

17" Clevo P870DM-G
i7-6700K CPU
64GB RAM GSkill 2800Mhz?
512GB Samsung 950 Pro
GTX980 dGPU
Win10



eGPU Setup(s) :

PE4C V4.1

Inxtron TB3 HDK

Asus Strix GTX1080 OC

Spoiler: TB3 Setup Images - Prepare for ghetto overload

TB3 Results :






http://www.3dmark.com/3dm/15193649 -Standard, GPU = 18536

http://www.3dmark.com/3dm/15191071 - Ultra, GPU = 10007

http://www.3dmark.com/3dm/15193921 - Extreme, GPU = 5206

http://www.3dmark.com/3dm/15193762 - TimeSpy, GPU = 6855















PE4C V4.1 - M.2 NGFF -

Spoiler: PE4C Setup Images - Ghetto over 9000 ? Nah.

With the cables routed a bit more effectively :
http://imgur.com/a/Z0mx6




http://imgur.com/a/BtVwJ

CUDA-Z :

Spoiler: CUDA-Z

CUDA-Z Report
=============
Version: 0.10.251 64 bit http://cuda-z.sf.net/
OS Version: Windows x86 6.2.9200
Driver Version: 372.70
Driver Dll Version: 8.0 (6.14.13.7270)
Runtime Dll Version: 6.50

Core Information
----------------
Name: GeForce GTX 1080
Compute Capability: 6.1
Clock Rate: 1809.5 MHz
PCI Location: 0:5:0
Multiprocessors: 20
Threads Per Multiproc.: 2048
Warp Size: 32
Regs Per Block: 65536
Threads Per Block: 1024
Threads Dimensions: 1024 x 1024 x 64
Grid Dimensions: 2147483647 x 65535 x 65535
Watchdog Enabled: Yes
Integrated GPU: No
Concurrent Kernels: Yes
Compute Mode: Default
Stream Priorities: Yes

Memory Information
------------------
Total Global: 8192 MiB
Bus Width: 256 bits
Clock Rate: 5005 MHz
Error Correction: No
L2 Cache Size: 48 KiB
Shared Per Block: 48 KiB
Pitch: 2048 MiB
Total Constant: 64 KiB
Texture Alignment: 512 B
Texture 1D Size: 131072
Texture 2D Size: 131072 x 65536
Texture 3D Size: 16384 x 16384 x 16384
GPU Overlap: Yes
Map Host Memory: Yes
Unified Addressing: Yes
Async Engine: Yes, Bidirectional

Performance Information
-----------------------
Memory Copy
Host Pinned to Device: 2842.01 MiB/s
Host Pageable to Device: 2414.87 MiB/s
Device to Host Pinned: 2981.8 MiB/s
Device to Host Pageable: 2559.01 MiB/s
Device to Device: 111.991 GiB/s
GPU Core Performance
Single-precision Float: 9246.7 Gflop/s
Double-precision Float: 295.548 Gflop/s
64-bit Integer: 465.641 Giop/s
32-bit Integer: 2742.09 Giop/s
24-bit Integer: 2084.85 Giop/s

Generated: Tue Oct 04 23:08:26 2016

http://www.3dmark.com/fs/10425255 - Standard, GPU = 23143

http://www.3dmark.com/3dm/15359522 - Extreme, GPU = 10938

http://www.3dmark.com/3dm/15250559 - Ultra, GPU = 5451
http://www.3dmark.com/fs/10398914 - Ultra, GPU = 5635

http://www.3dmark.com/3dm/15360121 - TimeSpy, GPU = 7672











One of the advantages of the PE4C implementation is that i can still use the 980 as the dedicated card for the 4k Internal panel or as a CUDA card or just disable it.
Also unlike the TB3 setup, where i needed to turn the PSU off manually to have the TB3 device shutdown, the PE4C shuts down the GPU as soon as the system shuts down.







Post your questions peoples. Let me know if you need me to run any specific tests. But please be reasonable, im at work all day so have very limited amount of time once im back.


@Mr. Fox @Prema @Phoenix (<-Dont complain, thought it would be cool to show it to you ) @Papusan @Johnksss@iBUYPOWER @Meaker@Sager @Mobius 1 @TBoneSan @D2 Ultima @Q937 @ssj92

 

Reserved

 

ha, we were actually thinking about that exact setup @M.2 in order to make our Batmans eGPU capable very interesting to see that this actually gives better results than the TB3 enclosure. could be that the M.2 lanes are maybe more independent from other components than the TB3 port?

on the other hand, not a very ergonomic setup, to say the least @Keyboard half removed

since youre still able to use the internal 980, heres an interesting idea: asynchronous SLI via DX12 with external 1080

in any case, +rep!

 

nando seems to agree with. And the m.2 ports are connected directly, so no TB3 to PCIe conversion overheads. ( https://www.techinferno.com/index.p...ffm2-pe4c-v41-win10-bloodhawk/#comment-150544)

That is definitely one of the drawbacks. For now . But on our systems, either its a $1300 MXM card or living with the keyboard open, with a < $800 setup. Definitely not for most people though. Might be really helpful for the P775 / P770 folks.
These speeds made me wonder, if the Proprietary AGA connector is just a fancy looking M.2 port extension. Id actually be willing to buy one and tear it down if someone can get me the service manual with the pin outs. Maybe i can just use their connectors and make the AGA usable on other systems with the M.2 slots. Or even use those connectors to make an external port on our systems

I didnt know about asynchronous SLI! That is definitely interesting. Will look into it, whats the performance bump like ?

 

performance bump varies with DX12. heres an example with ashes of singularity:

https://www.computerbase.de/2016-02...a/4/#diagramm-mutli-gpu-3840-2160-high-preset

ive also read an article where the funny conclusion was that the fastest config was a mix of AMD + Nvidia GPUs cant find the link though...

 

I doubt that we'll see a schematic of this thing, or if we do, it wont be soon. The best way is to get one and reverse-engineer it. After all, the PCIe pins on the GPU side are not a secret Multimeter is your best friend.

As for the TB3 vs M.2, I'm pretty certain that it is indeed the overhead. TB3 is a multi-purpose port, where the M.2's PCIe is just that - PCIe communicating with PCIe devices.

 

Hmm. I'll probably pick one up next week. Let's see how it goes.

 

Just a reminder, the ZM is PCIe 2.0x4 (if memory serves), meaning we may take a hit whereas the DM will be able to recreate this with relative ease with the bottom open and no keyboard issue. Throw an AIO cooler on the card, maybe better fans on the cooler, and do some good works...

 

true, but then again that would be our best bet for an egpu setup, seeing as were lacking TB3. either that or go directly through the mxm slot.

Sent from my Huawei Mate 8 NXT-AL10

 

I definitely agree. My point is more that which card we pick might have to be based on the bandwidth limitations of 4 lane PCIe2.0, and if you have an NVMe drive, forget about it (said with NY accent). Meanwhile, the MXM would allow support for any card, just about...

 

yeah with those choices id say my priorities would lie with an nvme drive instead of an egpu still waiting for xpoint, although the 960 pro looks nice already. looking forward to the first reviews. according to specs, its about 40% faster on 4K QD1 than my 850 pro. still, xpoint should beat that by a multifold...

Sent from my Huawei Mate 8 NXT-AL10

 

Weirdly enough I haven't seen any performance dips even with a transfer running from my nvme drive to 2 external USB HDDs.
How many PCIe lanes does Skylake have for peripherals ?
The MXM PCIe bandwidth/lanes don't seem to affect the M.2 ones.

 

X point comes in December with a refresh a couple months later and is 5100Mb/s instead of the 3300Mb/s of the 960 evo, although that is possibly over the bandwidth of the PCIe3.0x4, meaning although faster, it may not be worth it (although sustained write on the 1TB 960 evo is 1200MB/s, which means xpoint Optane sustained write may be able to run higher and not be limited by bandwidth, making it much faster).

The ZM only is able to have one PCIe2.0x4, not two 3.0 like I think your P870DM has (ours splits to pcie2.0x2 I believe if we have two drives, don't know if yours does). So that could be part. I think the 4790k and 6700k both have 16 pcie lanes, but the motherboard could have a controller on the 870 for more lanes...



Sent from my SM-G900P using Tapatalk

 

Don't think it splits it. But I might be wrong. Haven't the time lately to check. Will look it up later tonight.

 

which is the best solution if i don't want the laptop to be all open and in a mess?

 

TB3.
get an Akitio TB3 PCIe Box and mod it to fit larger cards.
Or get the TB3 HDK like.mine and slap it in a SFF case.
And don't pair it with anything more than a 1070. That's a waste of money.

 

Very interesting results! Im amazed. Too bad youd have to have the keyboard open. Even though i treat my laptop as a desktop and have it stationed, i wouldn't want to have the keyboard opened up like that. Now if only there were some kind of sleek mod to make it work without keyboard off hehe...

Well i guess Akito it is. Is it hard to mod it to fit the 1080 GPU? Will there be any openings ? If so then i guess other one would be good enough.

Would 1080 really be a waste though? It squeze out a bit more performance than a 1070 right ?

I was playing Gow 4 last night. 1080p with everything maxed out except for 'Screen Space Reflections'. V-Sync enabled so was getting a solid 58 FPS or so. Enabled Screen space reflections from Ultra to Insane caused a drop to about 40 or so FPS.

I wonder if i had 1080 egpu if i could do higher res and maxed out settings.

Any chance you did a benchmark on Gow4 bloodhawk? They have the F5 option in video settings to run one in case you havent seen it.

 

Well having the keyboard open is one of the drawback I guess. For people like.me, who leave the system on the left with the wires running to the back and then under the table, it works just fine. Specially since I don't quite care about aesthetics.

If you get the Akitio Case , then es that will need a bit of modding for it to fit a 1080/1070.

For one, you will have to cut / fold the front of the case to be able to fit the length of the cards.
Second , you will need to find a way to supply enough power to the card. I don't know about the 1070's , buy a 1080 won't be able to draw enough power using an external DC power supply. You will need to make a 6pin 12V to Barrell connector cable and supply power to the GPU using the PSU as well. There are multiple ways of going about this.

As for 1070 vs 1080, you can easily squeeze out similar amount of performance using the 1070 with a slight overlook, maybe even at stock. Difference between a 1080 and Titan XP is about 10% over TB3. So you decide. Personally I find it a waste of money, if you really intend to use them only as an eGPU.

I'm still waiting on my Gears Of War 4 key, the one that comes free with the card, once that comes in ill run the tests for you.

 

Hmm i see. Thinking of just opting for the inxtron tb3 hdk, and slapping it in a sff case as i dont like the idea of cutting a case heh... Now if i use the inxtron, will i still need to make 6pin 12V to Barrell connector cable? How hard is that exactly? Is there a guide i can check out before i dive into this to make sure im some what capable? Lol.

Well i am doing the egpu route until i decide to build a desktop.... Sure would like to get more power via the m.2 slot but both my slots are filled with 950s....

Anyways thanks a lot for your help!

 

Nope with the HDK you can just use a molex to 6 pin cable to power it. Just make sure the polarities are correct on the molex end.

 

Would it be possible to to create an external PCI-express port by creating a connection from the M2 slot to a PCI-E cable or add an MXM PCI-E card?

 

Even though possible, it isnt that easy. There are people who are working on that. But if you use an extension to extend the MXM port to PCIe you will loose quite a bit of bandwidth.

 

@bloodhawk I see. Has anyone done a similar test as you with the 775DM3-G or at least tested the TB3 connection? It's hard to find things through the search. I'm about to order one now from HIDEvolution, so that's why ask.

 

Nope, AFAIK only Linus from Linus Tech Tips has tried a razer core with the P870DM-G. Other than that, none that i have seen online.
But with a system like the P775DM3, dont think you even need an eGPU.

 

Here is the benchmark run @ 1440p :

 

@bloodhawk, it looks like your TB3 benchmark results for the GTX 1080 is similar to the general GTX 980 TI benchmarks, true?

I have a P870DM with GTX 980 desktop and it looks like going the TB3 route with a GTX 1080 won't give me much gain according to your benchmarks.

 

I would put it at 1070/980ti OC.
If you want to use a 1070 , then it's worth it, a 1080 is a waste of money over TB3.
But even then you are gaining about 30%.

 

Woohoo! BGA for tha win!
http://www.3dmark.com/compare/fs/10388271/fs/10343976

 

LOL

btw . i managed to completely shut down the 980 thanks to @Prema

http://www.3dmark.com/compare/fs/10495738/fs/10388271

Fianl setup (for now)

 

We are getting one of TB3 based docking https://www.akitio.com/expansion/thunder3-pcie-box to test. We will test few VGA cards too to see if they work.

 

Nice.

I have been talking to Dan from Akitio for a while now, over at TechInferno. Their Node eGPU box will releasing soon (hopefull), that one will fit GPU's much better than the PCIe Box.
The PCIe Box might only be able to fit GTX 1060's (or smaller) i think.

 

We will get their Node eGPU too - we have good Test Lab at Eurocom to test all sort of new "stuff".

 

Wouldn't expect anything less.
But you guy` benchmarking guy needs a few lessons/tips from @Mr. Fox

The GTX 1070 benchmarks in in the P870Dm, arent quite where they should be

 

Oh, so now you want to be a hater! LOL
http://www.3dmark.com/compare/fs/10388271/fs/10495738/fs/10294987
J/K

Looking real good @bloodhawk I think i'm going to grab one of those with the 250 watt psu and add one of my Desktop psu's to it. Then test my 1070 on it.

 

Ouch. Imma go cry in a corner now.

 

I was hoping Akitio would get a bigger box as well. I like how the cases look. And preferably one with two pcie slots....

 

Only thing is that, from the information they have provided so far, it seems like the Node eGPU Case will be a yet again be a re implementation of the TB HDK.
Even the one that was demo'd earlier this year, had the exact same port configuration.

 

Hi @bloodhawk - for your setup in the picture, is that via TB3 or M2 for your eGPU?

 

That one is m.2.

 

Hi @bloodhawk , how much of an +% increase in FPS do you get on average playing games (1080 eGPU via m2 or TB3 versus Desktop 980). I am seriously considering a eGPU solution for 1440p gaming.

 

If you can maintain a full PCIe3.0 x4 Link, then the results will be same as desktop, obviously there will be variations based on how much and which processor are using.

For me going from GTX 980 to GTX 1080, gains were about 40%-50% increase in fps. This how ever will depend on the game as well. I mainly only play CSGO (capped at 300) , Overwatch and DOTA2.

 

@bloodhawk Does your Thunderbolt 3 Secure Connect utility say that external GPUs are supported, similar to what is shown on this compatibility page for the Node? My Zbook Studio sadly says that it is not, even though I am at the latest firmware / software version for the Thunderbolt 3 controller. I was hoping to try and pick up a Node to use with a 1070 in a couple of months...

 

But im not running the stock firmware. Also this shouldn't really make a difference tbh. Linus managed to get the core working on a stock P870DM-G. And the stock firmware says Eternal GPU''s arent supported. Same for the dell XPS 15/13 i think.

 

Oh, good.

Do you think there is much difference between a eGPU host machine that has dedicated graphics vs. integrated graphics? I am in the process of getting my machine swapped with another that has a FHD screen, and I can choose one that only has the Intel graphics instead of Intel + Quadro. I am thinking that a 1070 might work a little smoother if I do not have to have the Quadro drivers loaded.

 

Well that entirely depends on the machine. But an advantage would be that because of the presence of the internal graphics, you will be able to loop back from the GPU to the Internal display, there will be a bit of performance loss of course.
But again from what i have heard, the desktop processors hold PCIe bandwidth more robustly than the BGA counterparts. If you use M.2 i have a feeling the 6820HQ should be able to handle it depending on what other devices are connected to the system.

 

I hadn't considered M.2 because I wanted something that would work without having to reboot the machine.

Supposedly these Zbook's have two Thunderbolt 3 controllers, one for each port, so each can have its own PCIe 3.0 X4 bandwidth. In theory I could use both the HP Thunderbolt 3 Dock (which charges the machine and has port replication, and then an eGPU on the other port connected to my external displays at my desk, with a minimal performance hit.

Does the TB3 setup work ok if you connect it while the P870DM is sleeping, then resume the notebook?

 

@bloodhawk , I have the P870DM - where can I get such firmware
I plan to go the TB3 route rather than M2 route for now as I don't want to open the laptop. Do you remember how much of an FPS increase it is (on average) for the games you play? I'm guessing maybe 20%?

 

Don't think they have 2 controllers in a single device. They use the newer one which supports 2 ports over the same controller.
Just like the one in the new DM3/DM2 and the same one is on the TB3 HDK board.
Even If they were able to implement 2 controllers they would sap the heck out of available pcie lanes and clog peripherals.

As for sleep/wake, I haven't really tried. I never put my system to sleep, since I find it a waste of time and resources. It was useful when SSDs were not common and sleep/wake/hibernation saved a bunch of boot time. My system boots up in under 10 seconds.
So its either on and connected or turned off.
Sorry since this wasn't quite the answer you were expecting. But I can give it a shot for you next week. Going to be out for the weekend.

Well, it isn't available to the public. And don't think it will ever be. It was just part of small experiment / test. But you really don't need it, unless you have the Core , even then it's not quite necessary.
Well over TB3, running Overwatch @ 1080p on EPIC the frame rate jump was about 15-20fps, coming from the 980. But then again, Overwatch scales rather well.

 

I think you are right. I found a couple of images on ebay of the Zbook Studio system board, and it looks like a single TB3 chip on the top of the mainboard near both ports. Its not a big deal, because I can just as easily use a usb 3.1 dock.

I wish I could shut my machine down at night. I usually have a couple of VM's, Photoshop, Chrome, and 3-4 copies of IntelliJ running at the same time so reboots can be kind of slow even with the Toshiba XG3 SSD.

 

That's understandable. But even with the Razer Core hooked up to a Razer Blade, there are Sleep wake issues. I can't comment on my system just yet, since I haven't used sleep/wake with TB3 eGPU hooked up yet.