Project BSG Scorpion Fleet Shipyards (Lian Li PC-A77B)

[cstkl1]

bro did u see on ek website
they now said asus and msi .. the backplate doesnt support it
i am wondering did u tell eddy about ure asus the 2gc's

also hmm can u pin point to me exactly where was that pin thing that u bent it

[luke997]

bro did u see on ek website
they now said asus and msi .. the backplate doesnt support it
i am wondering did u tell eddy about ure asus the 2gc's

also hmm can u pin point to me exactly where was that pin thing that u bent it

Thanks for the info, that's interesting!

Well that tells me 2 things.

1) Only ASUS and MSI's GTX 480 are beefed up camparing to all the rest - with those 6 chips on the back near the VRM area.
I'm not sure what they are for but I can guess they wouldn't add it just to make Eddy's life harder :-)
So... thumbs up!

2) This incompatibility can be easily overcome by using washers underneath each standoff - this raises backplate just enough over the chips so there's no contact as well as over those wires near the capacitor in the middle of the card.

So not sure why Eddy would not attach the washers and comment in the instructions to use it on Asus & MSI?
I'll send him an e-mail about it.

3) Wires - there's no need to bend those wires if you use washers from the start, I bend it only as this was my first step to prevent shorting, before actually raising backplate sligthly.
Just to be safe however, once you have mounted the backplate check for the spacing across the areas. Just in case there would be some very oddly cut wire from those caps.

I think the difference between those cards is worth to spread on the forum, perhaps someone will be able to get all manufacturers, compare and test.

[cstkl1]

hmm becarefull of that 3 screws in the middle that connects the backplate to the block
loosening it was what killed my card as ure contact with ure mem is less.

i would think just cutting a hole at the top of the plate 6 squares for the chips would do like what evga did for the gtx470 backplate

[jasonelmore]

wow nice build. so how much did all of that stuff tally up to? $$

[Camino-Palmero]

Great system - I like it!

[luke997]

Finally got some time to test GPUs so I've been benching in Vantage all last night and this morning and here are results:

1) Overclocking
a) Best performance vs voltage.
It was surprising to find out that the sweet spot for my GPUs was 860 core, 1950 mem and just 1.025v (that's stock voltage on 1 card and just 1 notch up for 2 others)

Amazing performance with such voltage, average ~44200 Vantage P (no PhysX) from 3 runs.
Temp delta ~ 10C

b) Best overclock
Just a stunt, no point running it otherwise. 900 core, 1950 mem, 1.138v

Interestingly, average Vantage P from 3 runs = 44300 (no PhysX)
Temp delta ~ 11C

c) High overclock, low voltage - best Vantage results
Again, just for fun setting. 880 core, 1950 mem, 1.075v

Average Vantage P from 3 runs = 44530 (no PhysX)
Temp delta ~ 12C

d) 24/7 settings
I'll be cooking BIOS with NiBitor 5.6 shortly - @850/1950 @1.025

2) Power consumption.
This was interesting. I've used monitor at the power socket, measured for whole system - including monitor, wheel, sound, 2xD5 @ 5 and 19xApache @ full bore.
Below results in Vantage, first max recorded (always last feature test), then max in GPU tests (always Calico) than average spotted throughout in watts.
They are far from being accurate, I have no way to get the results into log and calculate.

(a) settings
1200/1130/850
(b) settings
1280/1220/930
(c) settings
1330(!)/1260/980

[luke997]

Quick update - got almost all the parts in - including Bitspower SLI links as EK link will go, most probably for good if the flow will be significantly improved (yes Waterlogged, I hear you - but still waiting for the RMA T-Virus.

Slim chances but I hope to get new version with screwed in caps

In the meantime I've joined XS HWBot team and finally got over 46K Vantage P (PhysX OFF)

46159 3D Mark Vantage P PhysX OFF

CPU @ 4.7, GPU @ 890/1975

[luke997]

This morning I was curious to see how far I can get this CPU, quite pleased with the result


@ 4.926Ghz

[josk1]

that's banana dancinging good!!!!

[CptDreadFlint]

Great looking build, luke! (You had me at "Shipyards" when you first started)
Huge fan of the amount of black on black with black components. You made it work so well together! (gunning for the same look myself)

Congratulations and enjoy your beautiful system!!

[-TYPHOON-]

looks like you're rig of the month at Guru3D
http://www.guru3d.com/article/guru3d...nth-june-2010/

[luke997]

looks like you're rig of the month at Guru3D
http://www.guru3d.com/article/guru3d...nth-june-2010/

Yeah, it was a very pleasant surprise!

Great looking build, luke! (You had me at "Shipyards" when you first started)
Huge fan of the amount of black on black with black components. You made it work so well together! (gunning for the same look myself)

Congratulations and enjoy your beautiful system!!

Thanks. You have a great project yourself!

[luke997]

Quick 3D Vision Surround feedback to keep all the info in the worklog.
Note this is very skim test as I don't have much time - so without any scientific method, just a quick gameplay to see how it runs by the feel (3 scale - slideshow, semi-smooth, smooth), no benchmarks.

Overall, I'm very impressed considering this is beta.
Some dissappointment with NFS Shift, may be 3-way SLI issue though, will experiment with it more.

1) Setup
Very easy, by configure button for multi display in the panel.
You can assign displays as you want as well as configure bezel space.
Worked out 114 pixels for AW 2310 (by eye).

2) Metro 2033
Can run only up to 5040x1050. I quickly figured out why . Tricky setup
2D - no problems, great
3D - AWESOME! Had to scale down to DX10/medium/AAA/4AF to be smooth

3) Batman AA
5760x1080 fine, easy setup
2D - no problems, great
3D - AWESOME! Had to drop though from 16AA to 2AA to be smooth / semi smooth at times

4) NFS Shift
5760x1080 fine, easy setup
2D - no problems, great
3D - slideshow, unplayable. Will try out 2GPU + PhysX setting later. A shame as I like this game a lot

5) Dirt 2
5760x1080 fine, auto setup
2D - no problems, great
3D - very good

6) Mass Effect 2
Doesn't work in surround, scales screen very badly. Again a shame as I like it a lot.

7) Unigine Heaven
5760x1080 fine, easy setup
2D - fine, had to drop bells&whistles though
3D - had to drop bells&whistles a lot, spectacular in 3D but crashed on me at some point.

8) Crysis Warhead
5760x1080 fine, easy setup
2D - Epic win Enthusiast settings, though no AA
3D - not tried yet,

Will have a look at Just Cause 2, Prototype, L4D, L4D2, & Stalke Pripyat once I have more time.

Quick desktop screen so you know it is real


http://rapidshare.com/files/40388265...und_screen.jpg

[luke997]

I'm kicking off Rev 1.1 this weekend.

Modifications are:

1) Replacing SLI EK Bridge for BP SLI crystal links
2) New MCR220 rad in the back
3) Adding 1.5mm spacers to stop fittings in the GPU blocks.

RMA'd T-Virus is nowhere to be seen so it'll have to wait for Rev 1.2

New rad should help temps a bit, as well as the spacers and BP links by improving flow - afterall it replaces 14x90 bends (12 in the EK bridge and 2 between CPU and Mosftet_2) into 1 or 2 (inlet to the first GPU block - no way around it, not enough space between card and PSU for anything else).

No pics - it didn't happen - so there you go

[Mudgey]

Hi Luke, awesome set up you have there.

I had bought the 3D vision kit last year and loved the extra depth and immersion it added to the games, however every game I played had some problem or another. I am extremely impressed when it does work and really want this kit to work! If you have the game could you let me know what it is like playing with Bad Company 2? This is the game I mostly play currently.

Many thanks,
Mudgey

[luke997]

If you have the game could you let me know what it is like playing with Bad Company 2? This is the game I mostly play currently.

Many thanks,
Mudgey

Here are the ones that work well and I have a ton of fun with:

Just Cause 2 (overall, the best one I would say)
NFS Shift
Batman AA
Metro 2033
L4FD
L4D2

All those games I play at max depth and tweaked convergence.

I still have Prototype & Stalker CoP to try, heard they are good.


Quick feedback on the software I've used:
Arcsoft TMT:
- photos 2D to 3D - works quite well - actually better then I've expected, especially on photos where there are objects closer & further to the screen.
- HD clips 2D to 3D - works very well on still camera or with little shaking. Not so well when the image is shaking

Overall me & wife watched our holiday pics and clips and it was definitely overall good experience, on some photos/pics you felt like being there.

CyberLink PowerDVD
- DVD 2D to 3D - as HD clips for TMT. Watched District 13 nice experience.
Anything like genuine 3D movies but if you don't expect miracles it's quite cool.

[luke997]

Rev 1.1 done - with mixed results.

The only upside was that I didn't break or flood anything

As for the flow with new SLI links - it actually dropped to 160l/h (from 200l/h)!
Simple conclusion - in this loop 14x90 bends doesn't matter significantly while adding MCR220 does.

But...I'm going to learn the lesson and plan for the next move

What's next? Some of you must be laughing now :applaud: yes I'm going to replace 2xD5 with 2xDDC and see how it's going to work in this mile long loop.

Tomorrow I'll post temps - it should be better still with the new rad even though the flow dropped a bit.

Now few quick shots to show the changes.

[Church]

luke997: BTW, why don't you connect gpu-s in parallel? IIRC recent Gabe's tests shown better results that way, as flow resistance should drop greatily, especially in your 3-way-sli case.

[luke997]

luke997: BTW, why don't you connect gpu-s in parallel? IIRC recent Gabe's tests shown better results that way, as flow resistance should drop greatily, especially in your 3-way-sli case.

Good question! I'm always opened for suggestions so will certainly consider.

For my setup, mimicking EK parallel is a right way? So basically adding 2 more BP bridges where now stop fittings are?

I also understand that by splitting water it halves flow in both "streams" - I'm worried that half of the poor flow I now have would actually be worse for the temps (my logic behind it is that going from 2gpm to 1gpm is fine while from 0.7 to 0.35 might be not good).
Is this correct or not?

I'm the "measure twice cut once" type of guy so have to think it through well before implementing

[Church]

luke: imho your D5 don't have as much head pressure as DDCs to handle as good much resistive blocks in single loop, so while it's true that in parallel connection each from your 3 gpus will get 1/3 of overall flow it will get compensated partially by more overall flow because of dropped restriction in gpu full cover loop part, and temps of gpu blocks are not that much dependent on flow. Well, you can always experiment, and if you get better temps for cpu & gpu in serial, so be it. But why not try parallel connection aswell, especially if it has worked better for some?

[ScottALot]

Wow the tubing and fittings around the CPU are very confusing, but it must work well!

[Waterlogged]

luke997: BTW, why don't you connect gpu-s in parallel? IIRC recent Gabe's tests shown better results that way, as flow resistance should drop greatily, especially in your 3-way-sli case.

+1, this is why your flow dropped. Switch it before going to DDC's.

[luke997]

But why not try parallel connection aswell, especially if it has worked better for some?

+1, this is why your flow dropped. Switch it before going to DDC's.

I guess this will be Rev 1.2

Is the way I've described it earlier good way to implement this or is there any better?

[Alexandr0s]

Unless you want to use EK's parallel connector block, your way is great .

[luke997]

Knock knock to my door this morning - and surprise, surprise, RMA'd T-Virus just came from US!

So I'm getting onto Rev 1.2 now - new res, perhaps parallel for GPUs and I'll have chance to replace dead battery

Will keep you posted!