After the Pentium 4 line with FSB 800 MHz was announced we looked over all chipsets offered by Intel for the new processors. But there are no chipsets we would unconditionally recommend you. The i875P (Canterwood) is certainly faster than its dual-channel brothers of the i865 family (Springdale) but its release price is noticeably higher, and the layout is much more complicated. Moreover, the mobo makers willingly believed Intel that the i875P should be positioned for the top desktop systems and even entry-level workstations, and equipped their Canterwood based models with a great deal of peripheral controllers, good accessory packs and stuck an impressive pricetags. The boards look good but too expensive.
The i865PE chipset (or i865G with an integrated video core) looks a good alternative to its elder sibling. It's not much slower, has the same functionality (except the ECC memory support) and provides a lower cost of finished solutions. But the performance level of typical i865PE based boards is comparable to the fast i845PE based levels which are officially announced for operation with the FSB 800 MHz and DDR400.
The i845PE is deprived of some features of the new generation, but it's possible to integrate external controllers to redeem the downsides of the functionality. The dual-channel memory can't be supported, but end-users value the speed rather than marketing terms. If you remember, the dual-channel SiS655 released half a year ago hardly outscored the dated single-channel SiS648, and if the latter correctly supported the Hyper Threading, the new SiS chipset would have much more somber prospects. What caused such effect if the theoretical throughput of two memory channels is twice higher than that of a single channel, and the wide FSB of the Pentium 4 is eagerly looking for an adequate memory acceleration?
The problem is that it's much more difficult to maintain stability at high frequencies, that is why the data path from the CPU to the memory has to be made longer (buffered). The memory access latency of the SiS655 and i865PE is too great making the dual-channel memory much less attractive. But Intel affirms that the Canterwood (which is principally equivalent to the Springdale) has a lower latency (thanks to the PAT), but only selected i875P dies have the required stability. However, there is an alternative consistent explanation of the reality: the performance of the i875P is normal, while the i865 series is allegedly slowed down to raise the sales of more expensive Canterwood based boards. (The only thing left to do is to find a reason to accuse SiS of, because it doesn't have yet an elite dual-channel chipset.)
The i875P is faster in the normal conditions than the i865PE, but the latter has hidden reserves, namely the internal acceleration mode. Some i865x based boards need just a new BIOS version (if the manufacturer offers such). This is actually a semiofficial overclocking of the chipset which lets you use the i865 to get a speed of the more expensive model. Now if we compare the accelerated i865PE and the i845PE with the FSB speed of 800 MHz, we will see that the former gets a decent performance gain added to its nice functionality, and becomes thus a good purchase.
In the next part of the Springdale based mainboards roundup we will touch upon several models where the PAT can be turned on as described above, and today we test the ABIT (on i845PE), DFI, EPoX and Soltek (on i865) boards that illustrate the current situation, and a specially invited star - ASUS P4P800 Deluxe. :)
OS and drivers:
The boards' specifications:
First of all, let's see what can be expected from the today's participants
and whether it makes sense to compare the old i845PE with the modern models
coupled with a dual-channel memory controller.
None of the new i865 boards has its own means to reach so high scores as those
of the ASUS with the special BIOS version that includes the Hyper Path mode. As to the other boards, the
EPoX has a bit better scores and the ABIT BH7 has comparable reading results -
the cost of the dual-channel access realization in the i865.
The MPEG4 video encoding demonstrates almost no difference between
the boards: ASUS P4P800 Deluxe takes the 4% lead,
EPoX 4PDA2+ has a 2% reserve and all the others including
the ABIT have equal results. Such results can be explained
by the stream data pumping through the buffer which masks the access delays.
The influence of these delays is well noticeable in the archiving where the ASUS P4P800 Deluxe outscores the main group by 26% (!). All the other boards take usual positions: the EPoX is a bit faster (+5%), and the others reach the finish at the same time.
Modern dynamic 3D games are another example of real applications that can be of interest for you. However, the scores of one board are noteworthy: High@1280x1024x32 in the Return to Castle Wolfenstein is the only mode where the ABIT BH7 noticeably falls behind its competitors (probably because of the lacking memory throughput when the CPU and video accelerator work simultaneously). The leader ASUS outpaces the main group by 9-12% in the low resolutions, and by 5-7% in high resolutions. The EPoX reaches +3%.
So, if such semiofficial overclocking doesn't confuse you, you have a good range of boards to choose for processors with the 800MHz FSB - starting from the i845PE to some boards on the i865PE/G (and even i865P). Otherwise, there is only the i875P, but the Springdale line has a noticeably better price/performance ratio with the same functions (except the ECC) in case you find models equal in the integrated capabilities.
Taking into consideration only "normal" boards, there is only one board
that demonstrates a bit higher performance level - EPoX 4PDA2+,
that outdoes its rivals by 3-5%. The others have equal scores. Moreover, they
have identical functions. That is why the EPoX keeps the lead
in all respects including the accessory pack. Only the tweaking and overclocking
capabilities are comparable to the Soltek boards.
Write a comment below. No registration needed!