Titan ITX status. (placeholder)

[Lazzer408]

Have you tried something similar to my design? Yes
Any problems with cooling? No
And if the GPU is facing down, how is the cooling? Great
Is there enough room under the video card and the bottom of the case? Yes
I'm guessing all your video cards were using pci express 16x, right? Yes
Different cables helped or it didn't matter? Didn't matter
A problem of length maybe? Not sure
Or not enough shielding on the ribbon cable, a problem of quality? Yes
After all, if the normal riser cards work fine (as i assume from your design), then it's something to do with the ribbon cable used in flexible riser cards, but what? Lack of ground plane


My PCIe extension was much shorter that what you intend to use. The first sample I recieved didn't have any shielding, much like an IDE cable. Without any shielding, it wouldn't work at all and if the board managed to post, it revert back to the onboard video as if the gfx card wasn't even installed. I tried placing aluminum tape to shield both sides of the ribbon and although this helped significantly (bios would find it every time) there were still problems with stability and some cards still didn't work at all. Symptoms would be the computer hanging, bsod, video goes black, reboots, or powers on with no video. It has to do with signal losses, echoing, and cross-talk. Some cards can apperently handle a little noise, some don't. I've seen someone use an extension that was 12" long and it worked fine for their setup but this was back in the GeForce 5k/6k series days. The riser cards I've got my hands on all have a ground plane layer in the pcb and some even have additional capacitors on the power rails to provide a little extra filtering. Lets say a card's ground reference floated above zero, say 0.5v, and the logic signaling is supposed to be between 0 and 1v, what you wind up with is signal switching between .5 and 1v. This is one possibility that would reek havoc on a digital circuit but I'm not a digital signal expert. Interestingly, Thermaltake uses a ribbon in their Mozart SX (VC7001SNS) chassis.

The toughest thing about designing a product is the trial and error and the learning curve. If you have engineering experience it's alot easier but there's still hurdles to overcome. If this chassis was a standard ATX chassis it would be a cakewalk but it's not. I've had to nudge things .005" here and .020" there and really know beyond a shadow of a doubt that everything is going to function as designed before I drop a huge investment on parts. The devil IS in the details. I've seen him!

 

[Lazzer408]

Top and bottom covers for one of the prototypes are finished. I still have the faceplate and internal brackets to do.

 

[t_ski]

subbededededededdeddddededdededededed until the alcohol waerassasds offfffff

 

[Lazzer408]

subbededededededdeddddededdededededed until the alcohol waerassasds offfffff

I thought you never finish anything? I bet you finished a few tonight!

 

[t_ski]

two really strong onez. lotsa kettle one + peach ssssschnnnnnnnnaaaaaaaaaaaaaaaaaaaaaaaaaaaappppppppppppppppppppppssssssssssssssssssssssssssssssssssssssssss + un pequito jugo de naranjajajajajajaja

 

[DanD3n]

I've seen someone use an extension that was 12" long and it worked fine for their setup but this was back in the GeForce 5k/6k series days. The riser cards I've got my hands on all have a ground plane layer in the pcb and some even have additional capacitors on the power rails to provide a little extra filtering.

I think my pci express flexible card has a ground plane, i can see that the rings around the mounting holes on the pcb are made of copper, so there should be a copper ground plane. Maybe. No capacitors, though. But it works fine and, as i've said, i have a medium GPU from the latest generation nvidia cards (560TI)...Maybe i got lucky with a quality cable? Or maybe when my next upgrade will come, it will not work; and i can't be sure right now, without trying this particular cable on multiple cards. :sigh:

Anyway, about your case, where do you place the SFX PSU? I'm guessing it's on the mobo side, with the exhaust slots on the side (or top?). And under the dvd drive there should be the 3.5" hdd, right? Without an internal picture available, i can only imagine the internal layout.

About the GPU card, have you tried in your designs to rise it to the top of case, letting more space at the bottom, making possible for mounting bigger, custom, heatsinks? And you could try isolating it from the rest of the case, so the hot air wouldn't go in the cpu/hdd/psu area. Something similar was done the new Alienware X51 slim case, but it's not very efficient because i don't think they made it that way for the cooling, but for the ease of installation. But the idea remains.

ps: are the sides made out of solid blocks of extruded aluminium? And the back and front plate are screw mounted directly on those two plates, right? Isn't this more expensive to make? It's certainly better than other mounting designs, easier to assemble and much more solid all around, but i think it will add to the costs (and weight). Btw, when do you think (or hope) it will be available for sale?

 

[Lazzer408]

The PCB is likely to have proper shielding. It's the ribbon itself that causes the problems.

Custom heatsinks are a small market. It's impossable to design a product to suit everyone's needs. My design is intended to support a wide range of retail graphics cards and I'm sure there's many cards out there that would have cooling issues in this chassis. Mainly those with large orb-style coolers that just blast heat 360deg hoping the case fans can get it out of the chassis. It's best to use a graphics card that vents most of it's heat out the rear slots.

The extrusion is fairly cheap to manufacturer but it's a large investment to begin production. The die cost alone can be $1500-$2000 and you have to buy 1000-2000lbs of aluminum billet to be extruded. If this makes 2000pcs, for example, then the die cost is diluted to $1 a piece. That's not so bad. Add the extrusion time, length cut, deburing and finishing, and anodizing and it starts to add up quickly but is still relatively cost effective especially in low profile enclosures. As you have pointed out, the fit and finish of the final product using extrusion can not be beat by folded sheet metal.

The design of the "Titan ITX Mini" is complete. I've given the CAD files to various fabrication shops to see who can do what and at what cost. Some of the components are made in China (power switch/LED, fan, fan Y-cable, fan controller, power supply, screws) and it takes alot longer to hear from them. Everyone is on vacation for Chinese Newyear. The extrusion, sheet metal, and final assembly is all made in America by American companies. The country may be in need of face lift but you simply can not compair the quality of some of the samples I've recieved. They all do very good work. I would like to see this hit the shelfs, or at least my website store, by summer so long as politics don't bring me down.

 

[Lazzer408]

Which screw do you like? The hex cap screw recessed flush into the face, or the flat head screw flush with the face?

 

[Norton]

How about these?

 

[Lazzer408]

How about these?

View attachment 45480

Sure, torx is an option. That's just a matter of swapping out the screws. I have to machine the faceplate and choose which countersink gets chucked in the CNC. I can countersink a hole deep enough that the large head of a cap screw can sit in, or countersink a bevel for a flathead screw. You can always swap out a philips for a torx once you recieve it but the screw head will have to be one style of countersink or the other.

 

[Norton]

Sure, torx is an option. That's just a matter of swapping out the screws. I have to machine the faceplate and choose which countersink gets chucked in the CNC. I can countersink a hole deep enough that the large head of a cap screw can sit in, or countersink a bevel for a flathead screw. You can always swap out a philips for a torx once you recieve it but the screw head will have to be one style of countersink or the other.

IMHO the black screw complemented your aluminum panel and Torx is a much hardier fastener than phillips head.... Phillips head fasteners usually get chewed up pretty good after a half dozen or so tightenings.

 

[Lazzer408]

IMHO the black screw complemented your aluminum panel and Torx is a much hardier fastener than phillips head.... Phillips head fasteners usually get chewed up pretty good after a half dozen or so tightenings.

I'll agree with that. The retail product will be anodized black. Some chrome screws would accent the black well. Philips is always handy for those without tool boxes. Also, there's no reason to ever remove the faceplate. The entire assembly slides out the back for service.

 

[Norton]

I'll agree with that. The retail product will be anodized black. Some chrome screws would accent the black well. Philips is always handy for those without tool boxes. Also, there's no reason to ever remove the faceplate. The entire assembly slides out the back for service.

Chrome on black definitely nice

If cost is similar- Torx will likely make manufacturing cleaner (same idea regarding tightening- less quality control work checking for chewed up phillips head screws)

 

[Lazzer408]

Price isn't a deal breaker on a screw. I'm not concerned about a few pennys on only 4 screws per unit. Which one of these do you like?

 

[Norton]

Price isn't a deal breaker on a screw. I'm not concerned about a few pennys on only 4 screws per unit. Which one of these do you like?

http://www.modelsport.co.uk/_images/products/normal/15457_01.jpg

http://www2.gpmd.com/image/h/hpic1539.jpg

I like them both but the bottom ones in a countersunk hole gets my vote

 

[Lazzer408]

The button head you posted isn't bad either. The chassis has rounded edges all around with a rounded power button and a radius around the power button and around the optical slot. It might blend well with everything. I'll order a few of each and see how it looks. Thanks for the input.

 

[stupido]

My vote would go to the flat ones if they are the same color as the aluminium panel. But... (there is always one but(t) ) the round one might look better is used as decoration (they should be different color than the aluminium plate)...

maybe direct comparison will tell the best...

 

[DanD3n]

The PCB is likely to have proper shielding. It's the ribbon itself that causes the problems.

When i replaced mine, i simply took one from an old IDE cable and cut it to size. Applied 3 layers of alu shielding and it worked. If other video cards will not work, then it seems that not all video cards are built the same, and some are more sensible than others to the signal level that travels through a ribbon cable. In your flexible riser cards try-outs, haven't you encounter one cable that worked on all the video cards tested?
I have second thoughts of my design, because of the uncertainty of future video card upgrades, caused by this flexible extension.

I like them both but the bottom ones in a countersunk hole gets my vote

+1
I also like the bottom ones, they add to the industrial, built like a tank look.

 

[Lazzer408]

When i replaced mine, i simply took one from an old IDE cable and cut it to size. Applied 3 layers of alu shielding and it worked. If other video cards will not work, then it seems that not all video cards are built the same, and some are more sensible than others to the signal level that travels through a ribbon cable. In your flexible riser cards try-outs, haven't you encounter one cable that worked on all the video cards tested?
I have second thoughts of my design, because of the uncertainty of future video card upgrades, caused by this flexible extension.

That's why I'm not using it. If there was so much as ONE issue what so ever with it, then it's out. There were many issues. 1: Not working 100% with 100% of tested cards. 2: Have to tape every one of them. (try that with a 1000 pieces.) 3: Ribbon cables can fail espicially in a friction type connection and the number of wires multiplies the failure rate. 4: Cost. (I couldn't find them cheap enough.)

In your design, you have one thing over mine when it comes to this PCIe riser and that is it's away from many of the interference sources. It's also up against the bottom side of the motherboard tray. Can you get your hands on a bunch of cards to test?

If you thought "good enough", your not finished. When you think "Perfect", your finished.

 

[DanD3n]

3: Ribbon cables can fail espicially in a friction type connection and the number of wires multiplies the failure rate. 4: Cost. (I couldn't find them cheap enough.)

I think that once it's mounted in the such a case, it's rarely moved again. It's the same as with older IDE hdd cables, but those got much, much more friction, due to hdds being swapped more frequently. They're quite resistant, iirc. About cost, in my search i found that it varies quite much, from vendor to vendor. I payed 20 usd on mine (it was bought from amazon, but i saw it in other places, too)...

In your design, you have one thing over mine when it comes to this PCIe riser and that is it's away from many of the interference sources. It's also up against the bottom side of the motherboard tray.

I was even thinking sandwiching the cable between two metal plates (currently only one, between the mobo and video card), so it's shielded from both sides. Add this to the triple alu tape i already applied, i think it will be overkill.

Can you get your hands on a bunch of cards to test?

Unfortunately, not, but i'm thinking about it.

 

[Lazzer408]

I was even thinking sandwiching the cable between two metal plates (currently only one, between the mobo and video card), so it's shielded from both sides. Add this to the triple alu tape i already applied, i think it will be overkill.


Unfortunately, not, but i'm thinking about it.

One layer should be fine and if you can extend a tab off the end of it to grab one of the mounting holes (to ground the tape) that would be even better. If it was 1 card out of 100 that caused problems I may overlook it, but it was more like 1 out of 10. It was 2 out of 12 actually. I think one was the 8800 (or 9800) and the other was an older all-in-wonder I had laying around. One has PCIe power connections and the other does not but both failed. The newer ATI/AMD 4k 5k 6k cards didn't have any issues. I didn't want to accept any issues with any cards and in my case it was easier to go with another option like the riser card. I've had zero issues with those.

I should mention that the problem showed it's head right away. It didn't take 3 hours of Furmark to crash it. If the BIOS detected the card it and managed to boot, it would fail as soon as the driver loaded.

 

[Lazzer408]

The first faceplate came home tonight. I still have work to do but here's a progress shot.

 

[Lazzer408]

Slowly but surely. This is a cheap display but it shows the option. Anyone know any good displays? One with an IR reciever would be best.

 

[DanD3n]

Matrix Orbital LCDs are pretty neat, esp those VFD.

 

[Lazzer408]

Matrix Orbital LCDs are pretty neat, esp those VFD.

Thank's. I'll look into it.

EDIT - Price? o.o I'll keep looking.