DIY eGPU experiences

[ruhtraeel]

Alright, my Casegears 480W power supply doesn't do anything when I plug it in and turn it on. It's not hooked up to anything. Is it supposed to do something?

[User Retired 2]

Alright, my Casegears 480W power supply doesn't do anything when I plug it in and turn it on. It's not hooked up to anything. Is it supposed to do something?

You need the SWEX or paperclip trick to switch it on, even if you switch it to "on" on the rear switch.

When I tried Both Ati 5770 or Nvidia GTX 470 external cards, Ati did not work, Nvidia had Code 12, but worked when you restarted with Express card inside (but only external monitor worked) Now I took out 4GB of memory, Now both internal and External Nvidia Cards work. But perfomance depends which is primary display. Lets say External is primary, then 3dMark Vantage shows 7800 score, if you make Internal primary it gives like 3000. And comparement RE5 Test, 1280x1024 full details 16Q AA all settings maxed. On Internal 24FPS, External 51FPS.
..

Also how do I make it work with 6GB Memory not 2GB? thanks.

Whatever is set as your primary display is what the application will use as the gpu to run on. So set the primary as the HD5770/GTX470 to get the faster speed.

You'll need the DIY ViDock Setup v1.x to get the HD5770/GTX470 working with the internal gpu, using 6GB. It can be done because a similar specced Sony Z119 did it here. I understand the 1.0c grub was giving you some issues. Keep an eye for the 1.0d release any day now which uses the original grub4dos as default in case the beta was problematic for you.

[ruhtraeel]

Alright, so it will work once I hook it up to the PE4H, right?

Just making sure it isn't DOA.

[bunnybare]

Removed: wrong thread.

[Oddious]

Alright, so it will work once I hook it up to the PE4H, right?

Just making sure it isn't DOA.

If you use the SWEX and plug that into the 20/24 pin MB connector of the PS then flip the swith in the back then the PS then it should turn on. if yo uwant to test the PS now you can use the paper clip trick as Nando said.

http://deron.meranda.us/create/case-...wer-bypass.jpg

[lazyboy]

Your 1920x1080 benchmarks would be faster than 1680x1050 if done using x1E link mode.

I don't understand why, but you're right.
I'm getting almost the same performance (or better) with a much higher resolution. How odd.







The 1680x1050 pics can be seen here.

[clmerlin]

Finally something promising! I received the error 12 thing today and know I need to do that boot disk thing with the thumb drive. Now the problem is that I download the file and save it to C:\DIYVdock and run the program to make the disk and I get this fun little error that says " There are no valid DOS system files at the specified location." Am I misreading the instructions? Or is this something that has been posted about before that I just didn't see in my hours of pouring through this? Atleast there is a glimmer of hope now though, right?

[ithildin]

These are my preliminary benchmark results comparing Radeon HD5750 1GB vs Geforce GTX 465 1GB on a DIY Vidock x1E setup.

Notebook
Brand/Make: Asus F8SA/P
CPU: Core 2 Duo T9300 2.5 GHz
RAM: 4 GB DDR2 (667 MHz)
Platform/Chipset: Santa Rosa/Intel PM965, ICH8M
Laptop Graphics card: ATI Mobility Radeon HD 3650 1GB
OS: Windows 7 Professional x64
Drivers:Catalyst 10.6, Forceware 257.21

DIY Vidock
Graphics cards:

• Sapphire Vapor-X Radeon HD 5750 1GB
• Palit Geforce GTX 465 1GB

Adapter: PE4H v1.0 + EC2C
Connection: PCI-e 1.0 x1E (see notes here)
Power supply: 400W ATX PSU

Here's the Everest shot showing the Palit GTX 465 up and running at x1E:


3dmark 06
DirectX 9 - Fullscreen - 1280x1024 - Default settings - Feature tests

3DMark06	3DMark score	SM2.0	SM3.0
HD5750 x1E	10392	4986	4899
GTX465 x1E	5239	2144	2000

3dmark Vantage
Performance test - DirectX 10 - 1280x1024 - Graphics test only

3DMark Vantage	GPU score
HD5750 x1E	5442
GTX465 x1E	6472

Resident Evil 5 Benchmark
Fullscreen - 1280x800 - High settings

RE5 (variable)	Avg fps (DX9)	Avg fps (DX10)
HD5750 x1E	65.0	44.9
GTX465 x1E	69.1	65.0

Devil May Cry 4 Benchmark
Fullscreen - 1280x800 - High settings

DMC4 (scene 4)	Avg fps (DX9)	Avg fps (DX10)
HD5750 x1E		43.84
GTX465 x1E		70.19

Unigine Heaven 2.1
Fullscreen - 1280x800 - DirectX 11

Unigine Heaven	FPS	Scores	Min FPS	Max FPS	Tessellation
HD5750 x1E	26.1	657	16.6	54.8	disabled
HD5750 x1E	16.3	411	10.5	49.4	normal
GTX465 x1E	16.5	415	10.1	64.2	disabled
GTX465 x1E	15	377	9.1	63.4	normal

Battlefield: Bad Company 2
Fullscreen - 1280x1024 - Low settings

BF: Bad Company 2	Frames	Time (ms)	Min	Max	Avg
HD5750 x1E DX9	10033	140538	49	125	71.39
HD5750 x1E DX10	6911	153124	27	80	45.133
HD5750 x1E DX11	5727	167017	22	66	34.29
GTX465 x1E DX9	8641	154833	33	98	55.809
GTX465 x1E DX10	8570	157929	21	102	54.265
GTX465 x1E DX11	4876	136310	21	52	35.771

For this one I just drove a tank around Arica Harbour on an empty server, demolishing the MCOM A building, circling around the back of the compound and doing some damage to walls and watchtowers while switching viewpoints. I tried to retrace my steps for every iteration reasonably close.

I still need to fill in some blanks and rrun some more tests but here are my observations so far:

• DX9: HD5750 performs better than the GTX 465
• DX10: GTX 465 performs better than HD5750
• DX11: HD5750 is better optimized for Unigine Heaven but chokes with tessellation. In-game tests are inconclusive

Unfortunately the GTX 465 died on me before I could collect results for other game benchmarks and from its performance at x1. I will resume testing once the replacement card arrives.

[User Retired 2]

These are my preliminary benchmark results comparing Radeon HD5750 1GB vs Geforce GTX 465 1GB on a DIY Vidock x1E setup.
..
I'll update this post sometime tomorrow with fresh info.

Your great post above has been linked off the first post News section.

I hope the RE5/dmcv4 results are not due to NVidia specific optimizations. NVidia DX10 results showing GTX465 being ~50% faster than a HD5750. Need some more games to get a feel of what's the overall performance difference.

Can you also do x1 1.0 tests? I'm working with a PM55+GTX465 users who has switched to x1E and sees no performance difference. Wondering if x1E is specific to ICH7/8/9M chipsets, or could it even be a handshaking issue with ATI cards, with the ATI card switching to a lower performance mode that does not apply to NVidia cards? Your results will tell us then if x1E does increase performance in NVidia cards and so then Intel chipset have a legitimate issue in delivering proper x1 1.0 performance.

BTW - 3dmark vantage is showing you are running the HD5750 with 870/1430 clocks. If so, that's a pretty big overclock from the stock 700/1150 levels. Meaning you're running more of a HD5770 than a HD5750.

[Oddious]

Can you also do x1 1.0 tests?

Nando, I will have my new benchamarks ready from my 9800gt posted later today.

DMC4, RE5, furmark, and Heaven and , all showing and increase in x1E vs x1 on my toshiba
3dmark vantage will also be added but not showing and increase for some reason.

all test were run in Dx10 mode and default settings for the benchmark.