ASUS Announces P4SDX (SiS 655) Dual Channel DDR Motherboard

[First]

Taipei, Taiwan, Oct 28, 2002 -- The world?s leading motherboard manufacturer, ASUSTeK Computer Inc., today unveiled the P4SDX socket 478 Pentium 4 motherboard based on the SiS 655 chipset. Supporting industry-leading technologies such as dual-channel DDR 333, Serial ATA and AGP 8X and an abundance of advanced features, the all-new P4SDX enables outstanding performance and value.

Dual-channel DDR 333 - The P4SDX supports dual-channel DDR design up to 4GB high-speed PC2700/2100/1600 SDRAM memory for superior performance during the most memory-intensive tasks.

The P4SDX will be available in mid-November.

Thanks goes out to Kyle Bennett from HardOCP for reporting about this first. I haven't heard a thing about the SiS 655 from any mainboard manufacturer, so this is news to me, and which is why I'm posting about it.

 

[TheCoz]

Here are some basic specs for the P4SDX...........

- Supports Intel socket 478 533MHz FSB Pentium 4 processor
- SiS 655/SiS963 chipset
- Dual channel DDR 333,4 DIMMS,up to 4GB support of PC2700 DDR memory.
- AGP 8X slot,6 PCI slots
- Up to 6 USB 2.0 ports
- UDMA 66/100/133
- C-Media CMI 9739A audio
- Optional Broadcom Gigabit LAN controller
- S/PDIF IN/OUT interface module
- Optional PROMISE S-ATA
- ATX form-factor
- 1066MB/s hub link
- SiS 963 southbridge supports 3 x Firewire ports

It'll be interesting to see if DCDDR333 support on a 533FSB gives any decent performance benefits over DCDDR266 (it shouldn't). Maybe you can provide us with some latency figures for DCDDR266 vs DCDDR333 when you get a SiS655 board Evan? The asynchronous (with the external CPU clock) DCDDR333 mode should increase latency over the synchronous DCDDR266 mode and the fact that DCDDR266 already provides more memory bandwidth than the 533FSB can accomodate (taking into account instruction fetches/prefetches) leads me to believe that DCDDR333 won't be much faster than DCDDR266 on a 533FSB.

Who will be first out the door with their DCDDR P4 chipset? Intel, VIA or SiS? They're all scheduled to appear around the same time so it'll be close!! It looks like we'll be spoilt for choice as far as DCDDR P4 chipsets go by mid-November!!

 

[Athlon4all]

Ya beat me to it stranger!!! I can't wait tosee howit performs andhow much it costs.

 

[dajo]

This is very exciting news!

I'm definitely a fan of the SIS chipset since I got my P4S533, and I can't wait for this board.

I thought I read somewhere that DCDDR would only effectively support PC2100? Guess not.

A few good reviews from the sites and I'll be getting one for sure.

Guess I'll need another Corsair Extreme stick!

 

[Hamburgerpimp]

I'm not a fan of anything buy Intel chipsets on P4 boards. But, it'll be interesting to see how this board stacks up. Obviously just the first in a line of new DCDDR memory boards.

 

[TheCoz]

Originally posted by: dajo
I thought I read somewhere that DCDDR would only effectively support PC2100? Guess not.

It depends on the chipset.

Intel E7205 (Granite Bay) only supports PC1600 and PC2100 because of it's 1:1 memory ratio. SiS 655 and VIA P4X600 support PC1600, PC2100 and PC2700 (And possibly unofficial PC3200 support) because they have the 1:1 and 4:5 ratios.

The 1:1 ratio (PC2100) is all you need though because it matches the FSB bandwidth and the memory bandwidth perfectly. It also means your memory clock and external CPU clock are running synchronously for lower latency. Running dual-channel PC2700 mode (4:5 ratio) means your clocks are running asynchronously (which is bad for latency) and your memory bandwidth is greater than the FSB bandwidth. That's like running 5 lanes of traffic into 4 lanes - you're still only getting 4 lanes of traffic out the other side!! It's a less severe version of the nForce2 syndrome - cramming 5.3GB/s (DCDDR333) memory bandwidth into 2.7GB/s (333FSB) of FSB bandwidth. Running PC2700 mode on the SiS 655 may prove a fraction faster but I wouldn't expect much of a benefit.

But I'm looking forward to the first reviews of the SiS 655 to see if the theory works in real life!!

 

[ElFenix]

heres what i would like to know... do you have to populate both channels with identical modules to get the thing to work, or is it somewhat intelligent like nforce is and really doesn't care whether the modules match in size or even if a second module exists?

 

[RanDum72]

I think its going to be like Dual RDRAM, the memory sticks have to be the same.

 

[dajo]

Originally posted by: RanDum72
I think its going to be like Dual RDRAM, the memory sticks have to be the same.

In what sense will they have to be the same? I have a stick of Corsair XMS 3200C2 and a stick of TwinMos PC2700. Do you not think that these will work together assuming there is some type of BIOS adjustment to run them within spec of the PC2700?

 

[TheCoz]

Originally posted by: ElFenix
heres what i would like to know... do you have to populate both channels with identical modules to get the thing to work, or is it somewhat intelligent like nforce is and really doesn't care whether the modules match in size or even if a second module exists?

Granite Bay requires the use of identical DIMMs for dual-channel operation but it'll work fine in single-channel mode with a single DIMM (or mixed DIMMs). Of course, single-channel mode would defeat the whole point of having a dual-channel motherboard but it would work.

SiS 655 is different and can operate in 3 different modes - 128 bit, 2 by 64 bit and single 64 bit. Here's a little info on the 655's memory controller....

The dual memory controller has independent MA, MD and control signals for each channel, and each channel supports 128 bit mode, concurrent dual 64 bit mode, and legacy single 64 bit mode. There's auto configuration built in for DIMM combinations.

This auto configuration in 128 bit mode supports precisely the same size and type of memory modules on channel 0 and channel 1, but in 2 by 64 mode can support different sizes of modules.

It will also support three modules using 2 by 64 mode, apparently.

When four DIMMs are plugged in, the modules need to be exactly the same size and type on Channel 0 and on Channel 1.

The 2x64 mode gives the 655 much greater flexibility than Granite Bay for sure but how much of a performance hit this mode takes remains to be seen. SiS 655 is certainly a very interesting product and I'm looking forward to seeing how it performs against Granite Bay.