HTPC - NSK2400, P5E-VM HDMI, E2160 * BIG DUCT !

Show off your quiet rig.

Moderators: NeilBlanchard, Ralf Hutter, sthayashi, Lawrence Lee

Post Reply
Jay_S
*Lifetime Patron*
Posts: 715
Joined: Fri Feb 10, 2006 2:50 pm
Location: Milwaukee, WI

HTPC - NSK2400, P5E-VM HDMI, E2160 * BIG DUCT !

Post by Jay_S » Mon Jan 07, 2008 9:18 pm

Small build log and gallery for my new HTPC.

Ingredients:
CPU: Intel E2160
CPU Cooler: Scythe Nina Mini
Case fans: 2 Yate Loon D12SL-12 both hardwired @ 5V
MB: Asus P5E-VM HDMI
RAM: Geil PC6400 - I've had great experience with Geil, and they're cheap for Micron D9GMH
Case: Antec NSK2400
PSU: Stock Antec 380W
Hard Drive: Seagate 7200.10 320GB - horribly loud seeks
Optical: Samsung SATA DVD-RW
Keyboard/Mouse: SpecResearch 01027, aka: XGene Imperial 2.4G Wireless Keyboard+Trackball. I love this KB/Mouse combo. Note the optical trackball, L-click trigger, r-hand scroll wheel. Brilliant.

Now, if you've seen one NSK2400 you've seen them all. So no external pics. Hell, not many internal pics either. Instead, some important ones.

Originally, the CPU area it looked like this:
Image
(That's my girlfriend's finger pointing at the new MB she got me for Xmas)

Yes, average stuff. But I wanted to overclock a little and I wasn't terribly impressed with the 100% load temps at stock speeds (1.8GHz). Core Temp 0.96 reported 58C peak temps with prime 95 stressing both cores. Overclocked to FSB 266, Core Temp settled on 72C. This is not good, as I'm shooting for FSB 333 = 3.0 GHz.

It ocurred to me that the simple "air dam" that comes with the NSK2400 was insuficient to re-direct air through the CPU HS. So I devised...

PLAN 1 - AIR BAFFLE:
A simple, 2-sided cardboard air dam with some silicone gasket materiel to sqish between the 120mm fan and the top of the case - keeps it from sliding around inside the case.
Image

Which installs like this:
Image
Image
Image

As you can see, this prototype is hacked out of the Scythe Ninja Mini retail box, and its dimensions were constrained by the simensions of the box. Not an ideal solution, but I decided to test it anyway. I hoped this would force more air to get sucked through the HS. But my load temps didn't change at all! So..

PLAN 2 - BIG DUCT:
I constructed a new air duct out of a corrugated plastic lawn sign I found in the basement of our house. This duct encloses both 120mm fans and the CPU heat sink.

The basic pattern:
Image
Image

Installed:
Image
Image
Image

This forces almost all air through the HS. Now 266 FSB load temps peaked at 62C - a 10C reduction! Not bad!

But the air restriction is audible with everything else off in the theater. It's a quiet whisper, whereas before I could only hear HD seeks. The whisper is completely drowned out by the fan on my projector, though.

Other thoughts:
1) I bought a IR thermometer hoping to measure my CPU temps. This is basically useless with the MB installed in a case. I can't get the thermometer in there and aimed at the base of the HS.
2) However, I learned that my NB is REALLY hot - 56C - much hotter than the Ninja mini. I am going to remove the big NB heatsink to check out the TIM or whatever Asus stuck in there.
3) Core Temp 0.96 revised the Tjunction-max for L2 stepping Allendale core processors from 85C to 100C. I've concluded that this is in no way meaningful as a measurement of absolute temperature. More useful is the delta to Tj-max. Whether Tj-max is 85C or 100C, I'm aiming to keep my delta higher than 20C.
4) This MB overclocks like crazy. I adjusted nothing more than FSB to hit 333. I flirted with this speed for a little while before backing down to FSB 266 for general use. All other BIOS settings were left on "auto".
5) EVERY TIME I switch inputs on my projector away from and back to the HTPC, my desktop resolution resets to 800x600. Annoying.

The IGP X3500 seems unable to properly render mpeg-2 video in VMR9 mode. Overlay and VMR7 work fine, but VMR9 has jaggies/macroblocking-like effects. Very strange.

MY FAVORITE FEATURE THOUGH, is the ability to write custom resolutions and timings to the registry with the DTD Calculator tool described in this avsforum.com post. After adding a 1280x720, 48Hz mode, I can swith between 60 Hz (for ATSC HDTV) and 48 Hz (multiple of 24 fps for film-based DVDs) right inside the Intel graphics properties app. This is essentially a free, lightweight replacement for powerstrip, which doesn't support IGP.

Tomorrow night I'll do some more temp testing at a variety of FSB frequencies up to 333 MHz.

Thanks for looking!
Jay

yamahaSHO
Posts: 426
Joined: Sun Mar 07, 2004 4:38 am
Location: Bellvue, Nebraska
Contact:

Post by yamahaSHO » Mon Jan 07, 2008 10:57 pm

If you sealed off the gaps on on the sides of the heat sink, you'd probably see even more of an decrease in temperatures.

I like the idea of using both fans for the duct. I made a duct to use the rear-most fan. It ducts between the two fans and curves around to the front side of the Ninja.

Jay_S
*Lifetime Patron*
Posts: 715
Joined: Fri Feb 10, 2006 2:50 pm
Location: Milwaukee, WI

Post by Jay_S » Tue Jan 08, 2008 6:30 am

yamahaSHO wrote:If you sealed off the gaps on on the sides of the heat sink, you'd probably see even more of an decrease in temperatures.
Yes, I thought of that. But that would have required cutting very complex shapes around the ram and the rear panel connector backshells. A possible accidental benefit of the current duct is that air CAN leak in underneath the sides, providing some airflow to the VRMs and RAM.

I made the top just a little too wide. You can see how the sides do not fit at right-angles to the top - they get "pinched" in toward the bottom. I'll correct this when I make the final duct out of clear acrylic.

Thanks for the compliments,
Jay

Jay_S
*Lifetime Patron*
Posts: 715
Joined: Fri Feb 10, 2006 2:50 pm
Location: Milwaukee, WI

Post by Jay_S » Tue Jan 08, 2008 5:55 pm

Temp and power draw testing.

Test notes:
- Core Temp 0.96 Tjunction ref: 100C
- Temps were recorded for the hottest core only.
- Load temps were recorded after 10 minutes.
- Idle temps were recorded in the windows desktop.
- Power measurements taken at wall with Kill-a-watt meter.
- Ambient air temps taken with IR Thermometer aimed at dead airspace.

Code: Select all

 Off: 4W
Boot: 101W (Highest draw noted)

Test 1
Ambient temp: 18C
                     FSB: 200 MHz, 1.8GHz CPU
  Idle - windows desktop: 64W (multiplier drops to 6 - 1.2 GHz CPU)
     P95 both cores 100%: 96W
Core Temp 0.96 Idle temp: 32C
Core Temp 0.96 Load temp: 54C

Test 2
Ambient temp: 19C
                     FSB: 266 MHz, 2.4GHz CPU
  Idle - windows desktop: 64W (multiplier drops to 6 - 1.6 GHz CPU)
     P95 both cores 100%: 104W
Core Temp 0.96 Idle temp: 33C
Core Temp 0.96 Load temp: 59C average, some 61C peaks

Test 3
Ambient temp: 19C
                     FSB: 333MHz, 3.0GHz CPU
  Idle - windows desktop: 65W (multiplier drops to 6 - 2.0 GHz GHz CPU)
     P95 both cores 100%: 116W
Core Temp 0.96 Idle temp: 34C
Core Temp 0.96 Load temp: 65C average, some 68C peaks
Problems! After 17 min P95 had a rounding error! Then wouldn't reboot properly. After hard reboot, there was a post message, something like "Overclocking error, enter setup ..." So I'm back down to 266MHz FSB for the time being. Sheesh - I won't deny that there were a few minutes of pants-sh1tting fear there...

I don't think it was CPU related. 65-68C are not absurdly high Tjuntion temps. And my mem timings are loose (5-5-5-15). Mem voltage is on auto, though. maybe I shold bump it p a bit. But I think it's that damned NB cooler. I really need to remove it an have a look at it.

Jay

Zig
Posts: 9
Joined: Wed Oct 31, 2007 5:06 am
Location: France, Lyon

Re: HTPC - NSK2400, P5E-VM HDMI, E2160 * BIG DUCT !

Post by Zig » Wed Feb 06, 2008 2:46 pm

Jay_S wrote: 5) EVERY TIME I switch inputs on my projector away from and back to the HTPC, my desktop resolution resets to 800x600. Annoying.
I have the same issue with this Asus board connected to my Sony HDTV using HDMI.

Every time I switch inputs on my LCD TV away from and back to the HTPC, the desktop resolution resets to 1024*768 (I'm using XP here).

That makes me crazy :?

Have you been able to find a solution for this problem ?

Jay_S
*Lifetime Patron*
Posts: 715
Joined: Fri Feb 10, 2006 2:50 pm
Location: Milwaukee, WI

Re: HTPC - NSK2400, P5E-VM HDMI, E2160 * BIG DUCT !

Post by Jay_S » Fri Feb 08, 2008 7:36 pm

Zig wrote:Have you been able to find a solution for this problem ?
Unfortunately, there is no good solution yet. If you have the time, there are two threads worth reading in the AVSForum.com HTPC forum:

1) Custom Resolution Tool for Intel Graphics: Easier Overscan Correction

2) Intel G35 motherboards

The guys to pay attention to are Archibael (an Intel employee) and WoOzy (one of the authors of the custom resolution application).

I think consensus at this point is that Intel needs to release a new GMA x3500 driver.

You could also try this - turn your HTPC on last, and power it off before powering off your TV. I seem to have better luck with resolution persistence this way.

Do you experience any macro-blocking-ish artifacts using VMR9 rendering in playback? My picture is perfect with Overlay and VMR7, but with VMR9, the picture gets screwy.

I don't think GMA x3500 is really ready for mainstream adoption. I'm most likely going to get a low end PCIe card (ATI HD3450) once more reviews come online.

Good luck,
Jay

Hyperfuse
Posts: 1
Joined: Sun Feb 10, 2008 6:37 am
Location: North Carolina

Post by Hyperfuse » Sun Feb 10, 2008 7:13 am

Thanks Jay! I had been wondering both about this board's video decoding and overclocking with one of the 21xx series on it. If and when you decide to go with an ATI 2000 series for an alternative decoding solution, let us know how that works out, too.

Thanks again,

Dylan

Jay_S
*Lifetime Patron*
Posts: 715
Joined: Fri Feb 10, 2006 2:50 pm
Location: Milwaukee, WI

Post by Jay_S » Sun Feb 10, 2008 7:42 am

Dylan,

In all honesty, I'm trying not to go with an external video card. I love the low power consumption of the current setup. It's plenty powerful for everything I'm doing with it (SD DVD upscaling, HDTV watching and recording). Except for the resolution persistence and vmr9 rendering issues, this board is awesome.

In my area, six TV networks broadcast in HD, with a 50/50 split between 720p and 1080i. This board has NO problem with decoding and playback of either resolution. I use a Hauppauge WinTV-HVR-950 USB ATSC tuner, the Nvidia Purevideo Gold decoder and the free WatchHDTV application to watch and record HDTV. For some reason, I cannot view & adjust the PureVideo decoder's properties while watching live TV through WatchHDTV, but I can while playing back recorded shows with Windows Media Player 11.

The point of all this detail is this: when I check and un-check "Use Hardware Acceleration" there is a noticeable difference (10-15% or so) in CPU usage in Windows task manager. So, the PureVideo decoder can use features of Intel's GMA X3500. Overall CPU utilization is around 15-20% during playback of (approx 18Mbps) HDTV recordings while using PureVideo's HW Accel. I think this is great.

I'm also using the free AC3Filter for SPDIF pass-through to my receiver. This results in negligible CPU usage.

Right now, I'm trying to find the best way to archive my HDTV recordings. I have been recording hour-long shows, that get cut down to about 40 minutes once I remove the commercials. The resulting file sizes are between 4.5-5.2 GB. So some will fit on 4.7 GB DVD-R blanks, and some will not. I've tried a variety of transcoders, and H.264 is by far the most impressive. But it's still a step down from the original HDTV recordings. I wish there was an application that could transcode based on a desired output file size. Maybe one exists, I'm pretty new to transcoding.

Jay

Tzupy
*Lifetime Patron*
Posts: 1561
Joined: Wed Jan 12, 2005 10:47 am
Location: Bucharest, Romania

Post by Tzupy » Sun Feb 10, 2008 9:30 am

I have three suggestions that should help you get better temps:
1) If your Mini Ninja is mounted with push-pins, remount it with bolt-through (use SCURK or TR kit).
2) Replace the stock northbridge heatsink with a TR HR-05 IFX or similar.
3) Change fans to Scythe Slipstream 1,200 rpm.

Jay_S
*Lifetime Patron*
Posts: 715
Joined: Fri Feb 10, 2006 2:50 pm
Location: Milwaukee, WI

Post by Jay_S » Sun Feb 10, 2008 10:05 am

Tzupy wrote:I have three suggestions that should help you get better temps:
1) If your Mini Ninja is mounted with push-pins, remount it with bolt-through (use SCURK or TR kit).
2) Replace the stock northbridge heatsink with a TR HR-05 IFX or similar.
3) Change fans to Scythe Slipstream 1,200 rpm.
Tzupy,

I totally agree with suggestions 1 and 2, but how would switching to Scythe fans increase cooling?

I assume you're recommending a bolt-through mounting system to increase contact pressure. I'm not concerned too much with the contact pressure of Intel's push-pin scheme. I've read lots of material on contact pressure, and while I agree it's important, I actually think there's a hell of a lot of pressure generated by scythe's push-pin system. My MB bends quite enough as it is! I would consider switching to a bolt-through system out of concern over the push pins wearing out and un-clipping. If I ever wanted a bolt-through system, I'd probably just make one. And if I were going to go to all that effort, I'd probably also lap the IHS and the ninja.

As of right now, I'm pretty content with my temps and overclock.

Tzupy
*Lifetime Patron*
Posts: 1561
Joined: Wed Jan 12, 2005 10:47 am
Location: Bucharest, Romania

Post by Tzupy » Mon Feb 11, 2008 4:18 am

The bolt-through mounting, with a backplate, prevents the motherboard from bending.
The push-pins do apply a lot of pressure, but IMO the contact with the CPU is worse due to the mobo bending.
Several people on these forums have reported at least 5C improvement after switching to bolt-through.

The Scythe Slipstream fans do provide more airflow per rpm than any other fans.
I don't yet have one because they are not yet available in my country, but I'm very interested in them.
Compared with the Yate-loons you would probably get 15-20% more airflow for the same noise.
What I'm unsure about is how well would the Slipstreams handle the restriction created by your duct.

Post Reply