Navigate: Advice Beginners BIOS Guide CPUs Links Mainboards Memory Network Storage Video/Sound Cards Contact Forum SiteMap Sponsors WebNews Home	.	.
Prices: CPU Intel P4 2.4C-800 P4 2.6C-800 P4 2.8C-800 P4 3.0-800 P4 3.2-800 AMD AthlonXP XP 1700+ XP 2000+ XP 2400+ XP 2500+ XP 2700+ XP 3000+ XP 3200+ Athlon64 Athlon64 3200+ Athlon64 FX-51 Opteron Opteron 240 Opteron 242 Opteron 244 Opteron 246 Memory Corsair Crucial Kingston Mushkin OCZ Search Prices:

Overclocking

At stock voltage, that is 1.55 V, the system would POST at 150 MHz FSB or 3450 MHz Clock frequency but would hang shortly before loading Windows Explorer. Increasing the Core voltage to 1.65 allowed the system to run 3450 MHz without any problems, that is, there were no issues running benchmarks, e.g. Expendable. However, the gain in performance clearly showed diminishing returns, also the system overall developed a more sluggish feel.

Expendable Screenshot at 3450 MHz: Just for grins, over 200 fps average and the minimum at 143 fps (2 x 512 MB)

Conclusion

This is one of the reviews where it is difficult even to start with the conclusions. Let's start with the chipset first.

First, the E7205 appears a mature chipset, very stable and with outstanding performance. It does not appear as if any of the aforementioned bugs really impact either performance or stability of the board at all.

Second, Granite Bay as a workstation chipset can be used as a desktop chipset but this is not exactly the intended use, which is reflected in the price point for the silicon. Springdale will be more cost-effective and bring potentially better performance to the desktop, courtesy of the higher memory frequencies supported. The caveat is that we don't know how much the fifos and synchronizers will bite out of the performance if the memory bus is running in asynchronous mode to the processor bus.

Third, Dual Channel memory technology is a double edged sword, especially in conjunction with HyperThreading. Clearly, in the majority of applications thriving on memory bandwidth, Dual Channel performance blows away not only single channel performance but also whatever we have seen from Rambus, irrespective of whether it was dual or single channel. This is most likely caused by faster initial access latencies (tRAC). Applications utilizing predominantly the Level2 cache will not see much of a performance increase.

Fourth, Dual Channel operation combines two physical DIMMs to a single 128 bit wide "Über-DIMM" which is seen by the chipset as a single row of memory. This naturally cuts the number of memory addresses available in half. While the wider bus results in higher bandwidth, it also causes higher redundancy and, by extension, less effective usage of the entire memory space. Combined with HyperThreading thread level parallelism, this can cause some problems in applications demanding high amounts of system memory. The easy workaround in this case, of course, is to use more memory which is kind of self-understood in the workstation class anyway. In other words, it almost appears as if 2GB of system memory are about the minimum requirement in order to not lose performance. Keep in mind here that the P4 as a 32-bit processor can only support 4 GB of system memory.

On To The P4G8X

It looks like a great board, it feels like a great board, it overclocks great and hardly any of the BIOS settings are working. For the record, the BIOS revs used were 1.002-1003g with 1003G increasing the memory scores from 3388 to 3392 MB/sec (SiSoft Sandra Buffered) but sill not providing any differences in benchmark results at 2:2:2 or 2.5:3:3 latency settings. The latest revision posted at the time of the review was 1004, however, the page was not available for download.

Other reviews of a variety of Granite Bay-based boards have shown screenshots of original BIOS revisions and some of those revisions did not feature any memory latency options. Suddenly, the next revisions looked like everything was put back in place. We have not, however, seen anybody checking whether the new-found-again options were actually functional. In so far, we can excuse some of the broken settings, however, the recommendation to ASUS is to either fix the bugs or take out the entries. We have come across similar situations with ASUS boards in the past, the A7M266 was just one example but ASUS have always come through with BIOS revisions that addressed the issues and we are confident that the appropriate fixes are on their way here, too.

Along the same lines we are criticizing some aspects of the manual that have somehow taken a turn to the worse compared to the previous standards set by ASUS in terms of accuracy. However, this does not impact the quality of the board by any means.

We could have gone through another few weeks of testing all different sorts of application but I think the message is clear, Granite Bay is the most powerful solution for the P4 at the moment and the P4G8X is an excellent board for anybody interested in the current platform, pending the appropriate fixes for the bugs we found. If the hardware upgrade is not absolutely imminent, it still may be advisable to wait for the next chipset due early next year.

next page:    => more =>