KT400A vs nForce2 vs SiS746

Motherboards/AMD S462 by piotke @ 2003-05-15

A couple of months ago nVidia introduced the nForce? chipset for AMD. It outperformed VIA?s KT400 chipset by a considerably margin. Now VIA introduced a newer version, the KT400A. Is it a decent alternative for the Nforce?? And what about Sis ? They have the Sis 746 that boasts 200Mhz FSB. I compared three motherboards based on these chipsets, and of course, overclocked them. Tested motherboards: Abit NF7-S v2.0 (nForce2) , DFI Lanparty (KT400A) and the ECS L7S7A2L (SiS746)

  • prev
  • next

Overclocking Remarks

Overclocking:

Since I used on overclocked CPU during almost all the tests I'm going to spend some time on this issue. Which board has the most overclocking options and how can they handle aggressive timings, high fsb's and high voltages?


In the table below you can find the maximum voltages that can be set.

Voltage / MainboardECSABITDFI
Vcore+ 3.8 % 2.32.000
AGP/1.851.8
Chipset/1.72.8
Vdimm+ 5 % 2.92.8



ABIT NF7-S v2.0
Madshrimps (c)

As you can see, the ABIT NF7-s has the highest possible voltages, except for the chipset, but that’s because it has another type of chipset than the DFI.

By raising the Vdimm to 2.8 I could get the reach 215Mhz on the FSB stable, this was with the FSB running in sync with the memory in dual channel and the VDD (voltage delivered to the chipset) at 1.6v (default).

By raising both the voltages to 2.9 Vdimm and 1.7 VDD, I could raise the fsb to 220Mzh but this was not stable so I reverted back to 215Mhz, which is quite nice already.

The CPU voltage can be set to a maximum of 2.3Vcore. Too bad that the board can't handle this as I tried 2 different cpu's (two times an 1800+ JIUHB, but one of these was a DTL3C, the one with the 1.5Vcore) and both times the board maxed out at 2.125 Vcore.

A great advantage of this board compared to the ECS is that it can handle fast ram timings. I was able to run my ram at 4-3-2 CL 2. In the benchmarks you can see that this gives a big performance boost.
If I put the timings on 5-2-2, or 6-2-2, that resulted in a no boot. The maximum memory bandwidth I got was over 3.3 Gb/s (Sandra). But this was with the unstable 220 fsb.

Another nice feature is the ability to set the AGP speed manually somewhere between 66 and 99 Mhz. Most PCI raid controllers and some soundcards can't handle high PCI speeds while using a high fsb. Well, this problem is now solved as the PCI is running on a locked bus, upping the FSB doesn’t increase the PCI or AGP (as seen on older motherboards)

The Nforce² chipset itself it active cooled with a small heat sink and fan (which produces a terrible high pitched sound).

Ram timings used for benchmarking the Abit
On the Abit there wasn't a "Fast Mode" option, so I let the motherboard read the settings from the ram, by choosing "SPD".



ECS L7S7A2L
Madshrimps (c)

This is one of the first ECS main boards that supports overclocking. You can set the fsb to a maximum of 200. My board had a passive cooled northbridge. And even with an fsb of only 166 it got extremely hot. This couldn't hold me back though. I pointed a 80 mm fan at the small heatsink and started raising the fsb, stopping at 197Mhz as this was the maximum I could achieve.

Not that bad you will think, but I got disappointed several times...
The chipset has severe problems with the ram on aggressive timings. I was able to use cas latency 2 on the ram, but that was only stable enough for some windows applications, and sometimes stable enough to benchmark with SiSoft Sandra.

Sometimes yes, the board had a lot of troubles of booting with aggressive timings. Once I got it to boot it was quite stable (But certainly not stable for more intensive benchmarks such as 3d mark.), but unfortunately booting at these settings was a 50/50 deal.

Like I said it is one of the first overclocking boards of ECS. You can change some voltage settings, not extremely high, but you change it. Normally you just can choose the desired voltage in BIOS. With this board you have to start calculating... You can for example add an extra 3.8 % Vcore to the CPU...

You can also change the multiplier of your CPU, but as you maybe could have expected by changing some jumpers on the mainboard itself !

Ram timings used for benchmarking the ECS
On the ECS I wasn't able to run the memory on max timings. The best stable timings I could choose were "Fast Mode" and 2.5 T ram Command



DFI KT400A
Madshrimps (c)

I had some issues getting 200 fsb on this board.

Overclocking works kind of weird on this board. First you have to set the fsb higher, by using dipswitches on the board. When you've raised the fsb this way to 166, then you can select in bios an fsb up to 250. “WOW” was my first reaction, but I got disappointed quite quickly.

I set the fsb to 200, and lowered the multiplier, both in BIOS, this resulted in a no-boot.
No big deal, I cleared the CMOS, entered BIOS, lowered mulitplier to 7 and then I changed the fsb to 200 by using the dipswitches. Again no-boot.

Then I set back the FSB to 166 by using dipswitches and started to up the fsb in bios. The max stable I could get that way was somewhere between 190 and 195. In Windows I opened the Sandra mainboard information module, and I saw that my memory was running at more that 220 Mhz. That was the problem I thought, because I set in BIOS the Memory speed "By SPD". In the Genie BIOS part I saw that it was running @ 400 Mhz...
Seemed normal to me, but it wasn't that way.
In the BIOS I lowered Ram speed to 333, and that way I could raise the fsb to somewhere between 195 and 200. And ram was running at 2*200 Mhz in Sandra.
200 fsb wasn't possible. I think it was because the fsb also overclocked the PCI bus.? but I'm still not sure about that. 195 fsb was possible with aggressive timings...

Ram timings used for benchmarking the DFI
The best stable result I got in the end was a fsb of 190 and the memory running at 224 Mhz, so I used these settings for benchmarking.
  • prev
  • next