DIY eGPU experiences

Alright hey guys so I ran 3dmark06 again with a non-dead PSU, and a tiny bit of overclock on the GPU...these were my results:
NVIDIA GeForce GTX 460 video card benchmark result - Intel Core i5-2410M Processor,LENOVO 4286CTO score: 15341 3DMarks

CPU unfortunately appeared to be throttling, so I lost a few points on the CPU score there, but I feel like my score is still too low...hmm. Will run again with fans on max, see if that does anything.

 

With either driver, you gotta "safely remove" the graphics card before unplugging. You don't need to sleep the laptop after you have safely removed it, just unplug the EC and power off the eGPU.

 

Wait really? I found that it wasn't necessary with the mobile drivers, as long as the computer went into standby...

 

Try 3dmark Vantage, cause 3dmark 06 is a lot CPU dependant.

 

Do you guys think there is a chance to use only a XBOX 360 psu (175W) for a GTX 560 ti (consuption 170W)?

 

XBOX360 PSU got 5W @ 5V, so you actually have 170 vs 170W then. Better get the 203W PSU, should work unless you do some serious overclocking. Note that the reference design of the 560Ti already draws 193W peak in Furmark:

http://ht4u.net/reviews/2010/gesamtvergleich_leistungsaufnahme_grafikkarten/index5.php

I also use a 203W and they are super efficient. Running a GTX460 (160W top-end) in furmark for an hour, the PSU got handwarm. Did some mods with the cooling system tho (better heatsinks, only passive cooling).

Love if you report back how it runs, perhaps if the setup becomes unstable in furmark (nevertheless, games should work fine) you can do some undervolting and push down the amps a bit.

Disclaimer: All personal experience, YMMV!

 

Thanks for your interest. I am also scouting but nothing so far...
If anything occurs pls inform.

 

Hello everyone!

I did a good review of my eGPU experience with my Lenovo X220 + GTX 560 Ti, but some tests was not done like 3dmark 11 and Heaven bench.

So, I did them and I edit my post here :

http://forum.notebookreview.com/gam...851-diy-egpu-experiences-439.html#post7694162

I also added a battlefield 3 beta test.

 

Atticus, did you experience any sort of noticeable mouse lag in the BF3 beta? I am not completely sure but I might have a small amount and I am thinking it may have something to do with the eGPU setup I have. Of course your CPU is better than mine, and it could be some sort of CPU issue. I am not sure if it is actual lag, or just me not being used to the weapon handling mechanics of the game.

 

MMmm... mouse lag..?! No sorry..

I tried the game maxed and I saw mouse lag there..but it was all the game that lagged!lol demanding game!

But it is a Beta, so we will see. I am sorry, I can't help you.

 

Is the BF3 beta playable @ 1920x1080 and low details?
20-30FPS avg @ high/ultra is a bit choppy, but for myself I don't need quality
Would be awesome, if you could test that. Might get a GTX560Ti again for my x220t again

 

Hello nando4, kizwan today I received ASUS HD 6750 1Gb DDR5 Graphic Card. I'm starting all over again from scratch. First thing I want to ask is which cable I should use to supply power to eGPU? I'm confused since I have three options:

1. External 5V/12V power supply cable x1 - Comes with PE4H Kit.

2. Floppy Adapter - Comes with Corsair CX430

3. 1 x Power cable - Comes with Asus HD 6750 1Gb DDR5 Graphic Card

I'm little conscious with this point since my XFX ATI Radeon HD 6570 DDR5 2Gb Graphic Card is damaged and still in the service center.

Thanks.

 

It really doesn't matter how you power the PE4H as long as you use the floppy connector. I personally use the cable that came with the PSU, but using any adapter will be fine. You also really can't go wrong when connecting the 6 pin connector since you have a single 12V rail PSU. Make sure use the 4 pin floppy connector and the 6 pin connector.

Basically just plug things in any way that the connectors fit and power will not be an issue.

 

Thank You Khenglish for reply. But I don't get what is meant by '6 pin connector'?

Thanks.

 

@pappu6600,

Like Khenglish said, you'll need a floppy power connector for the PE4H & 6-pin PCIe power connector for Asus HD 6750. The cable came with PE4H & Asus HD 6750 are adapter in case your PSU don't have that kind of connector. Since Corsair CX430 PSU already have floppy & 6-pin PCIe power connector, you don't have to use the adapter that came with PE4H & Asus HD 6750. BTW, with Corsair PSU, the 6-pin PCIe power connector is labeled with "PCIE" on the cable/connector.

 

Honestly it is hard to tell. In operation metro the outdoor area hovers at around 30 fps on mediumish settings at 1080P. Inside the tunnels is better with the framerate hitting 50 pretty often, but the flames and other particle effects lower the FPS some. However, I am not entirely sure if this is due to the x1 interface itself. In BFBC2, I would average 35 fps for periods of time on all high settings at 1080P (with 50+ fps common as well). However, the GPU usage would only hover around 60%, which I attributed to the x1 bandwidth bottlenecking the card. In my desktop, my GTX 560 Ti very rarely goes below 95% usage in BFBC2. However, in BF3 the gpu seems to pretty much stay at 99% load the entire time.
This load is being measured by EVGA Precision, and I am not exactly sure how it measures it; if a saturated x1 link somehow shows up as 99% load then I would not be able to tell how much performance I would be missing, but if the load percentage was independent from the pci-e lane I would expect to have around 60% usage, which I am not seeing. Of course I only have an i5 CPU, so that is probably holding be back some as well.

Most importantly, the demo seems to be only DX11 if you have a DX11 card; I can't find any way to force DX10 which I think would provide a significant performance increase, as DX10 played a lot better in BFBC2 with my eGPU setup than DX11 did. Also, if the apparent mouse lag wasn't there it would be perfectly fine and playable; I don't notice any sot of mouselag in BFBC2 despite the low framerate. It could be just because the weapons have a bit more realistic handling in BF3 but I am not really sure. Because this is a beta, there could also be some issues with my particular mouse or game issues that are causing this.

 

Source

- adding Thunderbolt to ur existing system will be VERY pricey (even missing out the coasts for the pcie expansion box ...)

Edit:
Interesting review of a Thunderbolt product

 

Hello Kizwan, I tried to run eGPU last night with original drivers. eGPU gets detected in dos, device manager, catalyst manager. In display settings, I selected to 'extend my desktop to this monitor' and I saw laptop wallpaper in crystal clear quality on Samsung Sync Master B2030 LCD.

That is it! There after don't see anything like desktop icon, navigating folders, game running on laptop. LCD screen remain static to Laptop wallpaper.

I tried to uninstall original driver and installing DNA driver for XP. But same sequence repeats. I don't understand what to do now, please help.

Thanks.

 

@pappu6600, I don't have XP here but probably you'll need to set one of the display to be primary in Control Panel or Catalyst Control Center.

 

By default 'Plug and Play Monitor on Mobile Intel(R) 965 Express Chipset Family' is primary. I don't get 'one of the display to be primary'. Should I make Asus HD 6750 as Primary?

Thanks.

 

I guess you should make primary the monitor connected to your egpu, so the 6750 one. Correct!

By the way, I notice that the PE4H has two jumpers next to the floppy power connector. What are they used for?

 

That's the fuse bypass jumper introduced as of PE4H 2.4, effectively ding the same thing as Panzer's fuse bypass. The fuse bypass only necessary on high power eGPUs like GTX465+, GTX570+ and probably a heavily overclocked GTX560Ti. Without it, peak 3D activity would draw excessive current causing the fuse to shut down resulting in a black screen/BSOD.

 

@pappu:
yes, in "display properties" you need to make a checkmark at "use this device as primary screen" (or similar).
Then games will run on the egpu screen.

The desktop icons will move to your egpu screen, if you deactivate the internal screen. When you reactivate it, the icons stay on the egpu screen.

 

Thank you noric, gilipollas for kind reply. I activated Asus HD 6750 eGPU by check marking extend my desktop and make it primary option and I also deactivated default intel chipset iGPU by clearing check mark ( primary monitor and extend desktop options). Finally I saw Laptop screen going blank and Desktop screen showing all items.

Is it normal laptop screen goes blank and desktop screen remain active. I'm asking this because on forum I saw screen shots showing both screen active.

Also please guide me on how to use eGPU on internal lcd.

Thanks.

 

@pappu6600: I don't understand very well your status, but I think that if you now have your external monitor primary and displaying the main desktop (i.e. taskbar, icons, etc.), you should extend the desktop to secondary monitor (i.e. internal lcd).

Be aware that in some notebooks, such as my Asus, there is a keyboard shortcut to switch among several monitors settings: in my case, by pushing Fn+F8 you go from internal only to external only to both screens active. When both screen are active, by default they display the same image (i.e. desktop with taskbar, icons: no primary-secondary) but you can open the display properties menu and set the primary monitor and extend the desktop to the secondary.

Hope I have been able to explain what I mean!

I see, thanks! So, if I want to use a lower end gpu do I need to take off the 2 jumpers or leave them in place?

 

By doing so display get reverted back to internal lcd but it automatically assigns internal lcd as primary and external lcd as secondary. Again situation is same I see only wallpaper on external lcd and internal lcd as usual.

I use dell vostro 1510 and keyboard shortcut is Fn+F8. But when I push these buttons I get the message "Presentation Mode is: OFF No External Display is Attached"

Thanks.

 

deleted since it hadnt been useful ...

 

Nando,

just posting to update my score. I was wondering if my score is low, but looking at the list, it looks like it is the norm for my card. I don't game, but what I can see is that if I play movies, I get horizontal tearing in fast action scenes. I am running 2560x1600. Eventually I want a card that does 1x dual link DVI and 2x single(or dual) link DVI.

 

You could try using a lower resolution when playing videos.

 

Playing 2560x1600x32-bit color movies @30FPS requires 470MB/s bandwidth. A x1 1.0 link provides 250MB/s. You do not have sufficient bandwidth for playback. So then the only way to playback the movie without choppiness is:

- add more bandwidth: use a x2 link if your system is capable (likely not)

- scale down the resolution (1080P@30FPS=238MB/s) or run 16-bit color (2560x1600x16-bit=234MB/s).

- get a x1.Opt capable notebook (4500MHD, Intel HD or HD3000 iGPU). x1.Opt uses pci-e compression so can do FHD playback without choppiness.

 

For some reason certain streaming videos (some on youtube and some on other sites) only seem to play at about 11 fps instead of the 24 - 30 that they should be when run in full screen mode. This doesn't happen for all youtube videos, and happens only when they are playing in full screen, but regardless of resolution. If it happens with a video it happens regardless of if I watch it in 360P or 1080P, so I doubt it is a bandwidth issue.

Also, if I download a 1080P video from youtube that is causing me trouble, it plays perfectly fine without lag when I play it in VLC.

Any ideas?

 

Today I played BF3 beta with my GTX460 at x1.2 Opt and I am happy to say it was very smooth However the game itself was a bit disappointing, felt like Crysis 2 demo; very console like. Hopefully they fix that.

 

What sort of FPS were you getting and at what settings? Also what CPU do you have?

 

I'm not sure what FPS, but it felt completely smooth so >30fps? 1920x1080 res, don't know what the settings were since I never found them. I have a Core i5-2410m, 4GB memory. I assume this is what low settings looks like though...

 

Do you get any mouse lag at all? For some reason my setup doesn't do to well regardless of what settings I use and there always seems to be a hint of mouselag. Your CPU is a little better than mine and if you have a working 1.2 optimus setup then you will have more bandwidth than me; I wonder if that could be enough to make a significant performance increase.

 

Hi!

Sorry, but have you already sent your performance score with GTX 460 on 1.2 Optimus? If yes, I guess I missed it. I remember you showed how you achieved 1.2 opt, but that was with a weaker card (perhaps a 430?), wasn't it?

D

 

Full Youtube / Flash Hardware acceleration for smooth fullscreen video playback

I have noticed that stuttering in Youtube fullscreen playback occurs if one of the acceleration methods are inactive as shown in the following pic:


Debug dialog can be activated by right click on the youtube video - show video information.

If you open the stream in fullscreen and click on the timeline navigation bar, both accelerations get activated and the video plays smoothly without stuttering:


We can further investigate this. System info: i7-620 w/ GTX460 @ x1.Opt, Firefox 4.0.1 and 7.0.1, Flash version:

 

I noticed the same thing. Many videos(not necessarily voutube) or streams show about 10fps at full screen at 2560x1440. When lowering the resolution to 1600x900 or similar the videos play smoothly at full screen.

VLC is working fine at fullscreen at 2560x1440 with any kind of video.

 

Hey, yea I did. I made a few posts with links to my 3kmark06 scores, although my CPU was throttling so I could've gotten a bit more out of it...I don't really see that big of a difference between my setup and similar setups, but then again my CPU is only an i5-2410m while a lot of people are running i5-2520 or i7-2620.

I'm not sure of mouse lag, I don't feel like there's anything wrong in particular.

 

I don't know when I can test Battlefield 3. The download keep restarting. At one time it is 70%, later it 40% again.

 

Ok, I found it! So it is around 15,5k. That is the same as with x1.1 opt. Shouldn't it be more?

 

The CPU is throttling & affected the 3dMark06 result. That is why the score is a bit low. If serialk11r can run 3dMark Vantage or 3dMark11, we might be able to see the correct score.

 

Isn't 15K kind of high got a gtx 460 with x1 optimus? I have a GTX 560 Ti and I could only manage a 13.5K and my CPU definitely wasn't throttling, so I would think it wouldn't perform much worse than a 2410m.

 

I added some pictures of my setup and from Battlefield 3... and more!

Take a look here : http://forum.notebookreview.com/gam...851-diy-egpu-experiences-439.html#post7694162

 

In the first post there are two different system with i5-2410M, GTX460 & GTX560 respectively (x1.1Opt). Both score 15K with 3dMark06. One system with i5-520M, GTX460 @x1.Opt score 13.8K (3dMark06). Assuming CPU is not throttling, i5-2410M have an advantage over i5-520M. While i5-520M will Turbo Boost up to 2.933GHz with single core, i5-2410M most of the time will Turbo Boost up to 2.6GHz with two cores.

When comparing serialk11r's score with other system with i5-2410M + [email protected], this show serialk11r's score is not what we expected from x1.2Opt setup. CPU throttling might explain why the score is low.

 

I monitor my CPU speeds and when I am gaming or doing benchmarks BOTH cores are ate at least 2.7 Ghz. The performance increase from a 2410 would be from the sandy bridge architecture, not the raw clock speed.

 

That exactly I'm trying to explain. i5-520M will Turbo Boost on single core most of the time, while i5-2410M will Turbo Boost with two cores most of the time. (I didn't include the clock speed this time.)

 

Excellent. I'm still downloading. Already downloaded 4.6GB. How big is the setup file?

 

[message deleted]

 

will this work with a optima laptop that has a nvidia 7300 go