Tech Support Forum - View Single Post

**numby** · 07-27-2004, 04:55 AM

Hi Adrian,
I found this in a review.

Quote:

Now a few words regarding the memory configuration. On board the Gigabyte 7NNXP there are four DIMM slots which unlike other boards based on the nForce II chipset are numbered in the traditional way: 1-4 starting from the socket. The first and second slots relate to the first memory channel, with the third and fourth to the second channel.
So it's better to fill the slots in the order they are colored. Reminding it that the nForce II is a dual-channel chipset, and to make the most of the performance the modules in the first and second channels must be there.
Now some bad points: it is about incompatibility of Gigabyte 7NNXP to Kingston's HyperX PC2700 memory modules. I started testing the board with the XP3200+ processor that runs at FSB=200 MHz. Of course, in so doing the memory was running in the dual-channel mode. Upon successful completion of the tests (stable operation both at 2-3-6-3 timings and at much lower 2-2-5-2 timings), I switched the system to FSB=166 MHz and tried to determine the available set of multipliers. But the system started running very unstably - I even was unable to boot Windows XP. As a result, having spent a lot of time searching for the source of the problem I replaced the HyperX memory with Kingston's PC2700 ValueRAM modules. As a result, the system started running stably. For the purity of experiment, I installed Corsair XMS PC3200 Platinum Series modules and set FSB to 166 MHz. The result is the same - a stable running system.
To all appearances, this problem is rooted in the board's BIOS (during the tests we used its latest version - F10) and may be eliminated in future. If you already have five HyperX PC2700 on hands, don't worry - set the FSB/MEM = 200/200 and get a stable running.
By the way, I did it on purpose mentioning that the problem rooted in PC2700 modules. It's quite probable with other HyperX memory modules (e.g., PC3200 or PC3500) there won't be any problems at all.

So it appears to me the problem steams from the memory...