sis655

[jkapa]

It sounds like it´s pretty good. At least after seeing the disapointing review on the e7205. Thinkin of getting a Gigabyte board with this chipset GA-SINXP1394 (http://w.giga-byte.com/products/8sq800ultra2_spec.htm) with a P4 2.0@2.7.

There´s a review (http://www.ocworkbench.com/2003/gigabyte/sinxp1394/sinxp1394-1.htm) on ocworkbench (never heard of it before).

Tell me what you think. I´m really open to suggestions at this point. Want to assemble my new system ASAP :)

Joao

[BipolarBill]

For Intel CPUs, nothing beats the PE chipset.

[Beeblequix]

For Intel CPUs, nothing beats the PE chipset.

Does that also include the 850E? I *personally* would not get involved with anything Rambus myself (I have an extreme bias against them). I'm interested to hear your view, regardless whether or not Rambus' policies also turn your stomach.

regards,
ß

[BipolarBill]

The PE chipset is strictly DDR.

[jkapa]

which do you think is the best 845PE motherboard?

[BipolarBill]

The Asus P4PE.

[jkapa]

had my eye on that one myself.... looks good

[Beeblequix]

The PE chipset is strictly DDR.

That really doesn't answer the question. I realize it's DDR, but I wanted your insights as to why DDR for P4? I have my own insights, and they're mostly based on my distaste for Rambus, Inc., not the technology per se. But then, not everyone esteems my opinions as highly as I do.

ßeeßle, Inc.

[BipolarBill]

My reasoning for DDR? Simply that Intel is finally brushing Rambus off.

[causticVapor]

Well, the reason is twofold -- first, the legal issues, second, the fact that RDRAM runs very hot, consumes a lot of power, is difficult to manufacture, and is also difficult to build a motherboard around. You can always rearrange circuitry to make room for wider bus paths. Higher clock speeds, however, are a bit more complicated. Which goes into another arena.... RDRAM is very difficult to overclock; the high-end can only do 600 (1200) MHz, which tells you something about the technological headroom. 1333 MHz and 1400 MHz RDRAM were on a roadmap somewhere, but it is impossible how this could be achieved without a miracle or 90-nanometer technology. In the past there was also latency issue; now since the manufacturing process has been refined that is no longer true.

Even Rambus seems to have few future plans for RDRAM at the moment... I guess they figure they'll make enough money off of lawsuits... :p

DDR is easier to overclock, and is more cost-effective because it relies on higher bus widths and lower latencies. It also shows more promise and achieves nearly the bandwidth of RDRAM, at much less heat. A "PC3700" module running at 233 (466MHz) allegedly can meet PC1066/PC4200 in bandwidth terms. And DDR used to cost less... but premium modules like Corsair and the 466MHz GEiL are almost as expensive as thier RIMM counterparts.


Also, the new Springdale chipset, while Dual DDR, seems to suffer from similar perils as that of the original nForce did for AMD. There is almost no performance gain from theoretically doubling the bandwidth by doubling the bus width, and it is suspected that this is due to the fact that the chipset is not tweaked. There should be a successor, Canterwood, that will fix this. High-end versions will offer a "turbo" mode that makes the memory bus throughput actually 1.5times that of single channel, instead of 1.1-0.8 times. And it'll support the future 800MHz FSB Northwoods and Prescotts (supposedly.)