DIY eGPU experiences

[andeelym]

Based on your previous reply here it looks like your USB controller is getting clobbered during PCI space compaction so is failing chainloading. You want to do the chainloading test before doing any compaction. Try the following:

1. Install latest Catalyst 10.6. It supports 36/64-bit PCI allocation so you may not even need DIY ViDock Setup 1.x and PCI compaction at all. If still get an error12 against your video card, then:

2. Ensure you can chainload into Win7 successfully. The MBR method should work *without* any modifications to the config files.

Joseph_Lin's Sony Z119 shown here just needed the underlying bridge configured. Can do that at the commandline:

1. boot with DIY ViDock switched off

2. Poweron the DIY VIDock then run command below, ensure it finds a solution.

compact preserveall import \core\devcon.txt makebatch %DRV%:\core\pci.bat
call %DRV%:\core\pci.bat

3. Then chainload into Win7 as done successfully above.

grub4dos mbr       OR
grub4dos w7

Oh ok, I didn't know that. Will unplugging all USB devices help in the compacting? Cos I have a powered 7-port hub connected at all times.

Ok, I'll try to install the latest drivers. Actually I already installed tge enzyme 1.4 beta that includes the catalyst 10.6 (i think). Also I installed the readydriverplus. Not sure will that help at all. Anyway, I am using my friend's old X1950Pro, will that cause a prob? Maybe I should just get a 5770 myself to try it out...

For the commandline (joseph_lin's solution), where do I run the commandline? After windows 7 boots up? Or using the DIYViDock Bootdisk and choosing commandline instead of menu-based?

Thank you for being the consultant at this thread so far. Even that I do not have it working now, I applaud your knowledge and your efforts in helping others here to realise their dreams (as well as mine).

Maybe this have a potential to be a commercial product, after finding or modding a case to house everything together. Just my bizman side thinking... Hee hee.

[ithildin]

Great. I did runs dmcv4 using DX9 and DX10 in x1E and found no performance improvements. Let's see if you have any better results.

I've posted my initial tests with Enzyme 1.4 beta 3 drivers over at Cell-systems forums. Basically I see the same as you - no significant performance improvements across the board. Disappointing...

[jamesbond007]

Yup same results here. Just tried the drivers did not notice any improvement.

[Z4gRoS]

Regards again, where I can ask for 2.0? I did not find it:-(other technical doubts, up to all the Watts it lasts? I have read of the solutions similiares and they come to 85W maximum, or it depends on the source being able? Performance of the card can I extract with this solution? 100 %, 70 %? Thank you!!! This one quite very good

[andeelym]

You got to state in the comments to merchant or additional comments in the shopping cart when you are ordering it.

You can always hook up a PC PSU whose wattages can go up to 1200W. But I read that there is a limitation on the gfx card. It cannot be those that uses too high of wattage. Its those nvidia top end ones.

As for performance, I am not veru sure. Maybe the veterans can advise on that.

[jamesbond007]

I believe the performance lost is 50-60% at x1 and about between 15-25% at x2. This is based on the pcie scaling done here http://www.techpowerup.comreviews/A...press_Scaling/

The performance lost at x1 2.0 which is equal to x2 1.0 is about 25% depending on the resolution. There is a nice summary at the end which sums everything up. They did a similar test on the gtx 480 here techPowerUp :: GeForce GTX 480 PCI-Express Scaling Review :: Page 1 / 25 but they were unable to do the x1 test but the rest of it has similar losses to the 5870 scaling so it could be safe to assume that the x1 might have the same losses.

As for the losses when done using the pe4h, you could refer to the performance table at the first page of this tread as there are many successfully implemented systems there.

However note that the cpu and also play quite a role in getting high fps especially in modern games. Don't expect to get the same performance compared to the desktop benchmark of a similar card based on the scaling test as the benchmarks done usually use much better cpu compared to to those available on laptops. So unless you have a blazing fast cpu you might get further drop in fps. Also as can bee seen in the performance table depending on the cpu and amount of ram u have you might get different results when compared to other set ups with the same card.

Example

A 5770 might get 50fps in far cry 2 when bench marked with a desktop while u might only get 30fps with the pe4h even with x2 although the actual loss due to bandwidth might only be 25% the lack of cpu power might further reduce the score. Also there are some set ups where although they have a less powerful card like a 5750 they still out perform a 5770 due to the lack of cpu power and ram on the 5770 set up.
Some other examples of game where cpu plays a role in fps.
http://www.pcgameshardware.com/aid,6...hots/Practice/
http://www.tomshardware.com/reviews/...s-i7,2360.html

As a summary the total performance improvement you will see will not only depend on the bandwidth but also the cpu and ram you have. The better the cpu and the more ram you have the the lower the losses you will see in the fps ignoring losses due to bandwidth limitation. However if you are just looking at the gpu alone then the losses should be like mention above, 15-25% for x2 and 50-60% on x1.

Losses aside you will still get massive improvement in fps that are even comparable with high end laptops with gpu such as 5870 at a fraction of the cost. plus there are evidence that the new gtx 400 series card are very bandwidth efficient so there might be even less loss due to bandwidth. So I would definitely recommend u get the pe4h.

[WoodroweBones]

Hey guys, I'm confused. Do you need a special port on your laptop?

I've got a Toshiba Satellite Pro L650: Toshiba Canada - mobile computing - notebooks - satellite pro notebooks - Satellite Pro L650

Can I use this dock?

[WoodroweBones]

Ok you need an ExpressCard slot... I dont think I have one

[jamesbond007]

Based on the spec on the link i don't think it can be done. Do you have an express card slot any where or do you have a empty mini pcie slot. If you have either one then you can use the pe4L or pe4h. If you do not have either one than you can't use this solution.

[Z4gRoS]

I believe the performance lost is 50-60% at x1 and about between 15-25% at x2. This is based on the pcie scaling done here http://www.techpowerup.comreviews/A...press_Scaling/

The performance lost at x1 2.0 which is equal to x2 1.0 is about 25% depending on the resolution. There is a nice summary at the end which sums everything up. They did a similar test on the gtx 480 here techPowerUp :: GeForce GTX 480 PCI-Express Scaling Review :: Page 1 / 25 but they were unable to do the x1 test but the rest of it has similar losses to the 5870 scaling so it could be safe to assume that the x1 might have the same losses.

As for the losses when done using the pe4h, you could refer to the performance table at the first page of this tread as there are many successfully implemented systems there.

However note that the cpu and also play quite a role in getting high fps especially in modern games. Don't expect to get the same performance compared to the desktop benchmark of a similar card based on the scaling test as the benchmarks done usually use much better cpu compared to to those available on laptops. So unless you have a blazing fast cpu you might get further drop in fps. Also as can bee seen in the performance table depending on the cpu and amount of ram u have you might get different results when compared to other set ups with the same card.

Example

A 5770 might get 50fps in far cry 2 when bench marked with a desktop while u might only get 30fps with the pe4h even with x2 although the actual loss due to bandwidth might only be 25% the lack of cpu power might further reduce the score. Also there are some set ups where although they have a less powerful card like a 5750 they still out perform a 5770 due to the lack of cpu power and ram on the 5770 set up.
Some other examples of game where cpu plays a role in fps.
Prototype: CPU Benchmarks, System Requirements and Screenshots - Prototype, Test, benchmarks - PC Games Hardware
Phenom II 955 Versus Core i7 920: Gaming Value Compared : Introduction: A Little Background

As a summary the total performance improvement you will see will not only depend on the bandwidth but also the cpu and ram you have. The better the cpu and the more ram you have the the lower the losses you will see in the fps ignoring losses due to bandwidth limitation. However if you are just looking at the gpu alone then the losses should be like mention above, 15-25% for x2 and 50-60% on x1.

Losses aside you will still get massive improvement in fps that are even comparable with high end laptops with gpu such as 5870 at a fraction of the cost. plus there are evidence that the new gtx 400 series card are very bandwidth efficient so there might be even less loss due to bandwidth. So I would definitely recommend u get the pe4h.

Thank you!!! Thank you!!! I from Chile, With my friend we want to buy a GeForce of high range but we will wait a bit in order that they lower the prices, there I write to them the results:-D... THANK YOU AGAIN!!!