Sprocket Redux

[lynx]

My cover is blown.

Damn.

One fan too many?

[clocker]

Ouch.

[brotherdoobie]

Very nice, Clocker. I am getting ready to build my next computer. The last one I built is a little old in the tooth. (Athlon XP "2000" cpu. 2003) You have given me some ideas.


Peace bd

[SingaBoiy]

Damn. Nice job

@Duffman - Surely you mean "Good Riddance"

[clocker]

@Duffman - Surely you mean "Good Riddance"

Sprocket says she would prefer "Ride of the Valkeries" for the soundtrack.

Personally I think that's a scosh pretentious so don't shoot the messenger.

[clocker]

While I've been preparing to build the new base for the Stacker ("preparing"= waiting for Wed., my day off) I've been looking into tweaking my setup's performance and ran into an interesting bit of info about RAID.
Several people were complaining about sub-standard throughput using the nVidia RAID controller and looking for relief.
After a great deal of experimentation, one guy seems to have found a workaround.
Turning off "read caching" under the controller properties makes a significant difference...here's an example:
This is my original setup (read caching enabled)...

And the identical array with the caching disabled...

Note that the average read went up @20% while CPU utilization was cut by half.

Not sure yet what to make of the extreme spikiness of the second chart, whether that is significant or not I don't know, but the improved speed is very welcome.

It also turns out that I have two generations of Raptor.
The latest (presumably) support "command queuing" while the earlier versions do not.
When I rearrange my drives I'll be sure to pair the drives together and see if that makes much difference.

The base is coming along nicely, it's been glued and is in clamps now.
I may have made a strategic procedural error, time will tell.
I decided to mark/drill/cut the top plate of the base for fitting to the Stacker floor after it was assembled and that might be a problem.
1/2" plexi is a whole different animal than the thinner stock I'm used to, especially when jigsawing- it likes to overheat and remelt behind the blade.
I'm going to try a slower feed and (possibly) some coolant/lubricant (water with a bit of dishsoap) to avoid this condition.

Wish me luck.

[clocker]

K then, the new arrangement is in place.
Fabrication was not as bad as I'd feared, but I did screw up my dimensions a bit which required one extra mounting plate for the Torin blower.

Here is the base mounted with the blower...

The blower is significantly quieter than the twin San Ace fans I was using before.
It moves less air, but at higher pressure and no dead spot so the trade off is about even I think.
This is shot from the case interior showing the secondary mounting plate for the blower...

Sitting on the case floor is the third plate to which the radiator is affixed.
This is my original Black Ice Extreme rad...smaller but less dense than the two heatercores I have so airflow through the fins is less restricted.
Time will tell if this setup is as good as the bigger rig...

And the whole loop in place...

After assembly I saw some minor changes that will simplify the tube runs, but I'm going to test her as is and see before I drain/reconfigure the loop.

Early results show this to be almost comparable to the heatercore rig with a lot more free space in the case.
Next I might try running two rads in series passively.

[ApacNTS]

*drool* that's just inspiring.

[Duffman]

"It's something unpredictable, but in the end its right"

[clocker]

... but in the end its right"

Well, Sprocket (Stacker version) is about halfway through the evolution I originally envisioned.
Even I'm not sure what "the end" is going to look like.

The next step is a custom PSU and a custom reservoir.
The PSU is basically just two units (Seasonic 600 and Antec Neopower) in a single enclosure with modular wiring.
The rez will be a 12" x 3" acrylic tube.

On with the show.