|
Page 3 of 3
Overclocking and Performance
Test Components and Software: DFI Lanparty nForce2 Ultra B - Athlon Thoroughbred XP1700+ (JIUHB DLT3C 0310 XPMW), (2x256) 512MB of Corsair XMS 3500 (3502v1.1), ATI Radeon 9800Pro w/ ATI Catalyst 3.5 drivers(v6360), (1x) 20GB Maxtor DiamondMax Plus8 ATA133 HDD, Windows 2000Pro NTFS SP2 DX8.1.
Overclocking Setup: Fast CPU Command Decode (aka CPU Interface); 128bit Dual Channel memory bus; Ram timing 6-2-2-2.0 (for Abit NF7-s) and Ram timing 6-2-2-2-2.0 (for DFI LanParty Ultra B) are used for all overclocking speeds. All the benchmarks were run twice and obtained average scores.
3Dmark 2001 SE - Default settings
|
3Dmark 2001 SE |
|
12 x 200 FSB = 2,400MHz
|
15,139 |
|
11 x 225 FSB = 2,475MHz
|
19,195 |
|
10 x 250 FSB = 2,500MHz
|
19,788 |
Quake 3 Arena - Original Version; Low detail, Standard Setting
|
640x480 |
800x600 |
1280x1024 |
|
12 x 200 FSB = 2,400MHz
|
359.3 FPS |
349.6 FPS |
293.9 FPS |
|
11 x 225 FSB = 2,475MHz
|
429.8 FPS |
427.8 FPS |
376.9 FPS |
|
10 x 250 FSB = 2,500MHz
|
451.6 FPS |
449.8 FPS |
387.4 FPS |
Unreal Tournament 2003 - [H]ard|OCP version 2.1 (average FPS)
12 x 200 FSB = 2,400 MHz
|
Resolution
|
dm-antalus
|
dom-suntemple
|
dm-phobos2
|
dm-inferno
|
ctf-face3
|
ctf-citadel
|
dm-asbestos
|
|
640 x 480
|
371.6
|
277.0
|
278.7
|
333.2
|
242.5
|
317.6
|
326.4
|
|
800 x 600
|
370.9
|
276.7
|
279.2
|
329.4
|
242.3
|
313.0
|
326.0
|
|
1280 x 1024
|
250.2
|
237.8
|
265.1
|
176.9
|
228.8
|
207.1
|
321.2
|
11 x 225 FSB = 2,476 MHz
|
Resolution
|
dm-antalus
|
dom-suntemple
|
dm-phobos2
|
dm-inferno
|
ctf-face3
|
ctf-citadel
|
dm-asbestos
|
|
640 x 480
|
395.8
|
295.5
|
295.5
|
356.5
|
258.6
|
339.4
|
347.9
|
|
800 x 600
|
394.6
|
295.5
|
296.8
|
348.1
|
259.0
|
327.3
|
350.1
|
|
1280 x 1024
|
251.0
|
243.6
|
278.7
|
176.9
|
240.2
|
207.9
|
343.3
|
10 x 250 FSB = 2,500MHz
|
Resolution
|
dm-antalus
|
dom-suntemple
|
dm-phobos2
|
dm-inferno
|
ctf-face3
|
ctf-citadel
|
dm-asbestos
|
|
640 x 480
|
412.0
|
308.11
|
309.0
|
372.8
|
270.6
|
354.6
|
367.3
|
|
800 x 600
|
409.6
|
307.5
|
308.3
|
358.9
|
268.9
|
339.9
|
366.4
|
|
1280 x 1024
|
251.4
|
246.0
|
288.2
|
176.9
|
246.7
|
208.0
|
356.8
|
Calcum10 - CPU Calculation Mark
|
Clock & FSB
|
Calculation Mark
|
|
12 x 200 FSB = 2,400 MHz
|
20,637
|
|
11 x 225 FSB = 2,475 MHz
|
24,418
|
|
10 x 250 FSB = 2,500 MHz
|
24,762
|
Primordia ScienceMark 2.0
|
Clock & FSB
|
Calculation Time
|
|
12 x 200 FSB = 2,400 MHz
|
407.731s
|
|
11 x 225 FSB = 2,475 MHz
|
390.763s
|
|
10 x 250 FSB = 2,500 MHz
|
375.775s
|
PiFastV4.1 (Hexus PiFast) - Pi Calculation
|
Clock & FSB
|
10m digits calculation time
|
|
12 x 200 FSB = 2,400 MHz
|
60.85s
|
|
11 x 225 FSB = 2,475 MHz
|
57.49s
|
|
10 x 250 FSB = 2,500 MHz
|
55.18s
|
Sisoft Sandra - Memory Benchmark
|
Clock & FSB
|
Integer/Float (MB/s)
|
|
12 x 200 FSB = 2,400 MHz
|
3145 / 3009
|
|
11 x 225 FSB = 2,475 MHz
|
3529 / 3342
|
|
10 x 250 FSB = 2,500 MHz
|
3889 / 3663
|
Comparison Benchmarks Setup
Compared System Setup: Abit NF7-s V2.0 - Athlon Thoroughbred XP1700+ (JIUHB DLT3C 0310 XPMW), (2x256) 512MB of Corsair XMS 3500 (3502v1.1), ATI Radeon 9800Pro w/ ATI Catalyst 3.5 drivers(v6360), (1x) 20GB Maxtor DiamondMax Plus8 ATA 133 HDD, Windows 2000Pro NTFS SP2 DX8.1.
Same cooling applications and conditions were taken when testing each overclocked system setup.
Cooling setup: Custom watercooling and liquid(methanol & water)@ up to 150 litres, chilled by deep freezer; Two independent cooling loops for CPU and GPU; no radiator; Liquid temperature of below 0C.
* Note: Some of the benchmarks are not available from the Abit NF7s due to its absence. However, some of the benchmarks are included for both Abit NF7s and DFI LanParty Ultra B board.
Sisoft Sandra - Memory benchmark
|
Abit NF7-s V2.0
|
DFI LanParty Ultra B
|
|
Clock & FSB
|
Integer / Float (MB/s)
|
|
11 x 225 FSB = 2,480 MHz
|
3496 / 3301
|
3529 / 3342
|
|
10.5 x 230 FSB = 2,475 MHz
|
3572 / 3347
|
3612 / 3404
|
|
10.5 x 254 FSB = 2,672 MHz
|
N/A
|
4000 / 3771
|
Futuremark - PCMark 2002
|
Abit NF7-s V2.0
|
DFI LanParty Ultra B
|
|
Clock & FSB
|
CPU / Memory / HDD
|
|
12.5 x 220 FSB = 2,756 MHz
|
8393 / 6226 / 753
|
8465 / 6840 / 663
|
PiFastV4.1 (Hexus PiFast) - Pi Calculation
|
Abit NF7-s V2.0
|
DFI LanParty Ultra B
|
|
Clock & FSB
|
10m digit calculation time
|
|
12 x 225 FSB = 2,706 MHz
|
54.32s
|
53.90s
|
|
11 x 252 FSB = 2,778 MHz
|
N/A
|
51.25
|
Calcum10 - CPU Calculation Mark
|
Abit NF7-s V2.0
|
DFI LanParty Ultra B
|
|
Clock & FSB
|
Calculation Mark
|
|
12 x 225 FSB = 2,706 MHz
|
24,161
|
26,756
|
Highest Stable Clock speed
|
Abit NF7-s V2.0
|
DFI LanParty Ultra B
|
|
VCORE
|
2.1v
|
1.96v
|
|
Clockspeed
|
2,800MHz
|
2,740MHz
|
Based on the benchmarks, the DFI board is slightly faster clock for clock, FSB for FSB.
As you can see, the Abit NF7-s outperformed the DFI board in terms of higher total clockspeeds. The Vcore in the Abit board is very user-friendly. On the other hand, this DFI board isnt able to obtain very high clocks, even with the Vcore modifications on the board. There are few DFI boards out there, by luck, will accept considerably higher Vcores such as 2.15V or higher, but based on many user reports most of these boards do not seem to like high Vcores. There also had been rumors that the Over Voltage Protection (OVP) may kick in at 2.1V with the DFI board. The board however, was not even stable at 2.04v.
Highest Stable Clock speed
|
Abit NF7-s V2.0
|
DFI LanParty Ultra B
|
|
FSB in 2D
|
220 MHz
|
254 MHz
|
|
FSB in 3D games
|
201 MHz
|
248 MHz
|
On a flip side of the coin, the DFI takes the advantage in obtaining much higher FSB overclock and stability even in the intensive 3D games. It outran the Abit NF7-s by approx. 47mhz FSB in 3D games. The Vdimm modification to the Abit NF7 board did not help to gain stability at higher FSBs in gaming. Abit NF7 boards in general, do not usually co-operate with the 133FSB rated chips at high FSBs. Abit boards need right combination of CPU types, memory, and special BIOS versions in order to gain a bit of FSBs. On the other hand, DFI board seems to be more independent when FSB overclocking with the 133 FSB rated chips, or any Athlon chips for that matter. The board is designed to handle high FSB overclocking and does it well right out of the box. One thing to look out for is the AGP bus frequency setting in the BIOS. The setting Auto may run the system at 50MHz PCI. Unlike the Abit NF7/s boards, the users who run the DFI system at AGP set to AUTO may experience some loss in 3D performance. After repetitive tests and observations on the DFI, manually setting the AGP to 66MHz gained performance in 3Dmark 2001 by 150 points average.
These DFI boards have issues in the BIOS department however. The board often fails to boot/ reboot when overclocking. These symptoms come usually when the system is overclocked, in combination of high FSBs of beyond 230MHz, high total clockspeeds, and high Vcore. It has not been proven yet whether it is the boards limit or CPUs limit.
Another problem is the Low CPU temperature protection on the board. If the CPU temperature reaches below 0C, the system may not function properly, such as not being able to reboot. Unable to reboot is the most common problem with this board. Once the system fails to boot, the CMOS may need to be reset several times in order to take effect. Applying the reset switch few times or so can sometimes solve the problem as well. The bright side is that the CMOS Reload feature will load all the saved BIOS setting without wasting so much time. Conclusion
The board is well equipped with features. In addition to the onboard nVidia LAN, the onboard Gigabit LAN provides some lightning fast data transfer in the network. The BIOS has lots of overclocking options such as voltages and tweaking in the memory department. The board seems better tuned in memory subsystem, which results in slightly better throughput than the previous nForce2 ultra 400 boards. Furthermore, the CMOS Reload feature comes handy especially for the overclocking gurus. The down side is the system fails to boot when it is at highly overclocked state. The system also fails to boot when it reaches sub-zero CPU temperatures. Some updated BIOS versions can resolve this issue however.
Obtaining Vdimm (Memory Voltage) modification to the board can gain stability at FSBs past 250 mhz as these boards have enough headroom in FSBs and not as sensitive to heat in the chipsets and circuitry of the board as the previous nForce2 ultra 400 boards have been. The board does not seem to require high chipset voltages at high FSBs. NorthBridge chipset seems to have no issues and massive cooling isnt necessary. It is still recommended to have sufficient air circulation in the system when highly overclocked. In overall, the DFI LanParty Ultra B is an impressive board. Even though it lacks in highest clockspeeds compare to the Abit NF7 boards, the high FSBs will make up more than enough in performance.
|