THE BOARD: Biostar TF7025-M2
-Form: mATX, 9.6" x 9.6", 24-pin ATX power
-CPU: Socket AM2, Athlon 64/X2/FX, 1000Mhz Hyper Transport (2000 MT/s)
-Chipset: NVIDIA GeForce 7025 Northbridge, NVIDIA nForce 630a Southbridge
-Memory: 4 x DDR2, dual channel, up to 8GB DDR2-800
-Expansion slots: 1 x PCI-E 16x, 1x PCI-E 1x, 2 x PCI
-Storage: 1 x IDE channel (2 devices), 4 x SATA2 (3Gb/s), SATA RAID 0/1/0+1/5 Matrix RAID
-Video: NVIDIA GeForce 7025
-Other onboard features: Realtek ALC888 8 channel audio, Realtek RTL8110SC 10/100/1000 LAN, 3 x USB 2.0 internal headers for 6 extra ports
-Rear panel connectors: 2 x PS/2, 4 x USB 2.0, 6 x audio, S-Video, VGA (D-Sub), DVI, RJ-45
Retail Box Contents:
-Rear I/O shield
-80-wire IDE cable
-SATA data cable
-SATA power adapter cable
AMD X2 3600+ Brisbane (AM2, 1.9Ghz, 65W, 1.300V)
Arctic Freezer 64 Pro PWM
Biostar TF7025-M2, Shipping BIOS: 430
1GB Corsair Value Select DDR2-667 (128MB allocated to video memory)
Hitachi 250GB 8MB SATA2 HD
LG 16x DVD-ROM IDE
Allied AL-D320EXP 320W PSU
No case, open testing environment, ambient temperature: 24C
Windows XP Pro SP2
The board came in rather unnecessarily large box - there was not a lot inside. The accessories included were lacking, especially when it came to cables. They were however nicely tucked into a complimentary nylon bag. The board itself has bad layout. Most troubling is the 24-pin ATX connector to the left of the CPU socket, obstructing airflow. Secondly, the SATA ports are aligned in a horizontal row for some reason, making for awkward cabling. There was plenty of room around the CPU socket though, so most third party heatsinks should be compatible. Strangely, the CPU fan header is located to the left of the socket so I had to turn the fan on the Arctic Freezer upside-down to get the connector to reach. There are 1 x 4-pin and 3 x 3-pin fan headers. The chipset heatsink is larger than the one on the TA690G, which is a good thing, but still somewhat inadequate. It got very hot during testing, even when idle, so an undervolted fan hanging over it should definitely be considered. That said, it did not cause any stability issues. There are power and reset buttons right on the board on the bottom right corner. It's a nice touch - I'd like to this on all motherboards one day.
Good overclocking menu. Almost limitless range of CPU voltages (but no underclocking available) and CPU frequencies to choose from. Full set of memory tweaks available and memory voltage starts at 1.95V (probably to reduce compatibility issues) and goes up to 2.50V. The CMOS Reloaded Menu allows you to save an extraordinary 100 presets. What impressed me most: if you enter settings that make the system unstable, upon reboot it will temporarily assign default values and stop at POST with an error message. I never had to reset the CMOS and presumably, one would never have to.
The BIOS has some elementary fan control: you can specify at what temperatures the CPU fan starts up and kicks into full gear. Unfortunately the CPU temperature it reports is way off. It had a reading of 40C at bootup and 60C during load even though the heatsink was barely warm.
BIOS flashing on the board is supported via floppy only, though it does not have to be bootable. If you are so inclined, the latest BIOS' are available at Biostar's Chinese site: http://www.biostar.cn/supports/BIOS.asp?vID=32&SID=32&MID=26&Value=561
Simply amazing for a mATX board. Highest HTT/FSB obtained (with all voltages set to stock) was 320Mhz (dropped multiplier to 4 and reduced HT multiplier to 3 as well to keep it under 1000Mhz). The CPU hit a wall at around 295Mhz x 9.5, but a little extra voltage got it stable enough to run SuperPI. Bottom line, most likely no matter what CPU you use, the board will not limit your overclock. I settled for 2.7Ghz (285Mhz x 9.5 with all stock volts) for an impressive 42% increase.
SPCR-SPECIFIC TOPICS OF INTEREST
Unfortunately you cannot undervolt in the BIOS. That leaves a lot of Linux users in the cold. For Windows users, CrystalCPUID works. The lowest multiplier I could use was 4.5 (from 9.5). 4 would cause the system to crash (though you can use 4 in the BIOS) I used multipliers of 9.5, 7, and 4.5 and systematically found the lowest stable voltages I use with HTT at 285Mhz. I also measured power consumption directly from the wall socket at idle and with Orthos running (small FFT setting).
1.28Ghz, 0.850V: 48W idle, 57W during Orthos (Small FFT)
2.00Ghz, 1.025V: 53W idle, 72W during Orthos (Small FFT)
2.71Ghz, 1.275V: 66W idle, 112W during Orthos (Small FFT)
Power consumption would've probably been better if I used a decent power supply. I have no information on the efficiency of the Allied power supply I used. The Enhance 80+ power supply I wanted (ST50EF-PLUS) is still on backorder.
With Fortron 400W
1.28Ghz, 0.850V: 43W idle, 53W during Orthos (Small FFT)
2.00Ghz, 1.025V: 47W idle, 67W during Orthos (Small FFT)
2.71Ghz, 1.275V: 57W idle, 102W during Orthos (Small FFT)
With Enhance 500W
1.28Ghz, 0.850V: 34W idle, 41W during Orthos (Small FFT)
2.00Ghz, 1.025V: 37W idle, 55W during Orthos (Small FFT)
2.71Ghz, 1.275V: 48W idle, 86W during Orthos (Small FFT)
System Off: 9W
Suspend S3(STR): 11W
Background applications: CrystalCPUID, Speedfan, NOD32, necessary drivers applications
Time to enter hibernate after a fresh boot: 11 seconds
Time to exit hibernation to desktop responsiveness: 15 seconds
Bootup time @ 2.71Ghz: 28 seconds (time from power button pressed to desktop responsiveness)
The BIOS' SmartFan seems to work adequately. A true SPCRer however, will want full customizable control, so I'm happy to report that two fan headers are fully controllable via Speedfan (vs. the TA690G's none). The 4-pin header controls 3-pin fans without problems.
GEFORCE 7025 INTEGRATED VIDEO PERFORMANCE
High Definition Playback:
WMP11: Windows Media Player 11 (ffdshow rev.1240, Quicktime Alternative 1.81 installed)
VLC: VLC Player 0.8.6b
VLC+CCPUID: VLC Player 0.8.6b with CrystalCPUID running in the background (Multiplier Management DISABLED)
Playback of H.264 encoded MP4s. CPU Usage figures follow:
X2 3600+ @ 2.7Ghz/WMP11: 9-19%
X2 3600+ @ 2.0Ghz/WMP11: 14-24%
X2 3600+ @ 1.3Ghz/WMP11: 18-36%
X2 3600+ @ 2.7Ghz/VLC: 0-2%
X2 3600+ @ 2.0Ghz/VLC: 0-13%
X2 3600+ @ 1.3Ghz/VLC: 16-24%
X2 3600+ @ 2.7Ghz/VLC+CCPUID: 9-18%
X2 3600+ @ 2.0Ghz/VLC+CCPUID: 13-21%
X2 3600+ @ 1.3Ghz/VLC+CCPUID: 21-32%
X2 3600+ @ 2.7Ghz/WMP11: 16-32%
X2 3600+ @ 2.0Ghz/WMP11: 22-50%
X2 3600+ @ 1.3Ghz/WMP11: 30-52%
X2 3600+ @ 2.7Ghz/VLC: 12-25%
X2 3600+ @ 2.0Ghz/VLC: 13-33%
X2 3600+ @ 1.3Ghz/VLC: 24-50% (some stuttering)
X2 3600+ @ 2.7Ghz/VLC+CCPUID: 18-35%
X2 3600+ @ 2.0Ghz/VLC+CCPUID: 18-44%
X2 3600+ @ 1.3Ghz/VLC+CCPUID: 31-58% (lots of stuttering)
Comparisons with other systems:
G7025: Biostar TF7025-M2 with 128MB allocated to video, 1GB DDR2
X1250: Biostar TA690G with 128MB allocated to video, 1GB DDR2
7950G: eVGA 7950GT 256MB, 2GB DDR1
Opt. 165 @ 2.5Ghz/7950G/VLC+CCPUID: 14-18%
X2 3600+ @ 2.4Ghz/G7025/VLC+CCPUID: 9-24%
X2 3600+ @ 2.4Ghz/X1250/VLC+CCPUID: 25-38%
Opt. 165 @ 2.5Ghz/7950G/VLC+CCPUID: 18-31%
X2 3600+ @ 2.4Ghz/G7025/VLC+CCPUID: 19-40%
X2 3600+ @ 2.4Ghz/X1250/VLC+CCPUID: 36-52%
Note: You may be wondering what CrystalCPUID has to do with HD playback. I encountered a strange quirk: when CrystalCPUID was left on running in the background (even with Multiplier Management disabled), VLC had a harder time playing back H.264 content, using up more CPU cycles. I was able to recreate the effect on my Opteron 165/Geforce 7950GT system, but to a lesser degree (few percent at most). It did not affect WMP11 or any other programs as far as I could tell.
5400 points in 3DMark2001, 1500 points in 3DMark03.
First off let me say it was a completely hassle-free setup/installation (as it should be). There were no strange unexplainable quirks or instabilities; it was rock solid throughout testing. I also had no driver issues whatsoever, unlike what I went through with the TA690G. It's kind of sad that AMD/ATI cannot produce a better chipset for its own platform than nVidia. Feature-wise, the TF7025-M2 is fully loaded, only lacking Firewire and HDMI (HDMI is available on the slightly more expensive TF7050-M2).
After days of testing I can only fault the board on three issues: the lack of undervolting options in the BIOS, the poor layout, and the somewhat inadequate chipset heatsink. In all other areas, it was simply top-notch. Performance was excellence once overclocked - many times during testing when I had to restart the system, I missed the window of opportunity to enter the BIOS because the POST screen was gone before the monitor came out of standby. It's amazing what $185CDN can get you these days ($80 for the motherboard, $70 for the CPU, $35 for the RAM). Kudos to Biostar.