AOpen XC Cube EZ65: A Quiet, Powerful SFF

Complete|Mobile Systems
Viewing page 5 of 6 pages. Previous 1 2 3 4 5 6 Next

HIGH PERFORMANCE CONFIG

These comments refer to the following configuration:

  • Intel P4-2.8C CPU (800 MHz FSB)
  • OCZ PC-3200 256MB EL DDR Platinum Dual Ch SDRAM Memory (2 sticks)
  • nVidia GeForce4 Ti 4800SE 128MB VGA card
  • Hitachi / IBM Deskstar 7K250 250G Serial-ATA hard drive (HDS722525VLSA80). This is Hitachi's highest performance non-SCSI model, a 8 MB cache, 3-platter design

Obviously this is only one of many powerful component combinations that could be used in the XC Cube. It's merely to see how well it ramps up with higher performance components.

A NOTE on the TIGHT AGP SLOT -- The available room is quite tight, but the largish Ti4800 card installed OK. The clearance between the trace side of the card and the end of the hard drive is small -- probably not much more than half a centimeter. The biggest problems are:

1) The increased turbulence noise from the fan, which will be jammed up very close to the side of the case -- again, not more than maybe 1/2 a cm. The total clearance between the edge of the AGP slot and the side of the cover is about 5/8". It's tight but should be enough for some large passive HS on VGA cards meant for one slot.

2) Card removal. Given the lack of room to maneuver, there is only one way: Tilt it back so the ports on the VGA card are free of the opening in the case, then pull up on the back of the card.

A. Power, Temperature and Fan Speed

The AC power draw was measured using a Kill-a-Watt AC power meter. The other data in this table was obtained with the AOpen SilenTek thermal fan control utility. Motherboard Monitor could not be configured to work properly with this system. The ambient temperature was 20~21°C during testing. SilenTek was engaged in manual fan speed mode with the CPU fan set to 40%.

Activity
AC Power
CPU Temp
Sys Temp
CPU Fan RPM
HDD Temp
Idle
81W
33
35
2100
42
folding@home
125W
53
53
2380
42
CPUBurn
150W
59
58
2480
43

All temperatures in °C

Again, the accuracy of the temperatures reported are in question. One wonders whether System temp is really the CPU internal diode temp... and the CPU temp is the readout from the in-socket temp sensor?

It's clear that the combination of higher power CPU and VGA card have caused a big jump in AC power draw. Temps under load are up across the board, but still not cause for concern. The Hitachi HDD temps are much higher than the Samsung, which is a known cool operator. Still, low 40s are OK for a HDD.

B. Performance


Neither the CPU nor memory scores are unexpected, but it's difficult to accept the drop in HDD performance shown here. It is obviously not right; even if the Samsung and Hitachi performed the same (which they do not), the 8 MB cache in the Hitachi accounts for 30% real-world improved performance.


Almost 5X the score of the integrated graphics and double that of the ATI9100 VGA in the Shuttle Zen. It's about the score that an ATI 9600 Pro card typically reaches.


CPU performance is right up near the very top of the ladder.


Memory performance is very close to the reference levels for the 865PE board; perhaps a bit more tweaking would get it all the way there.


Ditto the multimedia benchmark.

C. Noise

Not surprisingly, in this high performance configuration, the AOpen XC Cube is not so quiet. This Ti4800 card by itself has a fan that measures 39 dBA/1meter. The PSU and CPU fans both have to ramp up higher to evacuate the 50% increased heat in the system. Plus there is the added 2-3 dBA of noise from the Hitachi HDD.

Mic 1m from
Idle
Max
HDD Peaks
Front
34
39
40
Left
35
40
40
Right
35
39
40
Top
34
39
40
Rear
36
42
42

* All measurements are in dBA @ 1 meter
* Max is measured after >20 minutes of CPUBurn stress testing, which heats the CPU more than any other program or utility we know of.
* HDD peaks is the peak noise during hard drive defragmentation.

I was able to effect about a 3~4 dBA reduction in noise by breaking into the 12V feed for the VGA card fan and inserting a Zalman Fanmate1 voltage controller to reduce the fan voltage down to 7~8V. This made the overall noise closer to being acceptable for a hardcore PC silencer. It would probably be acoustic nirvana for a hardcore gamer. Note that for the sake of time, the effect of HDD suspension was not recorded. Suffice it to say that the overall impact was not as significant as in the low noise config, simply because of the higher noise level we're dealing with.

Overclocking was tried, just a minor 5% boost of the FSB to 210MHz, with a tiny increase in Vcore voltage. The OC to 2.94 GHz was made painlessly on the first try. Here is a quick SANDRA CPU benchmark at that clock speed. OC-savvy users are surely going to play much more than this with this system's range of BIOS settings.

Finally, the system was settled back to normal speed, but undervolted this time, down to 1.45V from the default 1.525 Vcore, which helped to drop temperatures by a few degrees. It remains perfectly stable folding@home at that setting, running a bit cooler and quieter than at stock.



Previous 1 2 3 4 5 6 Next

Complete|Mobile Systems - Article Index
Help support this site, buy from one of our affiliate retailers!
Search: