K7S5A Owners: Don't overlook that 166/166 setting!

[UNCjigga]

A quick trip to the SiS page for driver updates made me notice their ad for the newly announced SiS645 chipset. It seems like this is their Pentium 4 DDR chipset with MultIOL support as well. A glance at the specs reveals 'DDR333 support'. Hmm, I'm ASSUMING that would mean an FSB of 166MHz running some PC2600 DDR RAM!

So, if PC2600 is the official successor to PC2100 (heh, that puts all those PC2400 adopters outta luck!) then can we expect to slap some of this good sh!t in, crank it up to 166/166 and let the games begin? Does anyone know what PCI/AGP dividers this thing is using at this setting? Like the cheap board itself, this is almost to good to be true...

 

[GoldenLion]

Hmmm...

 

[UNCjigga]

Hehe, I forgot about the actual CPU in all this FSB madness. I always thought the limiting factors in FSB overclock were always the RAM and the chipset...how does CPU fare in the equation? I have a sneaking suspicion that the SiS735 really does internally support a 150 or 166 MHz FSB...and I'm assuming the release of the SiS645 chipset means memory makers are gonna be on the PC2600/PC2700 bandwagon soon.

Nevertheless, even if CPU is the limiting factor, you can always run the memory asynchronous at 133/166 or 150/166, or else just put it all at 150/150. Now we just gotta wait for the memory to appear...

 

[L3Guy]

<< I'm ASSUMING that would mean an FSB of 166MHz running some PC2600 DDR RAM! >>



The current FSB for a P4 is 400 MHz, ( 100 Mhz quad data rate). With PC2600, it would not be synchronized with the memory bus.
The AMD ran the same way with PC133 and a 200 Mhz FSB, so there is precedent.

What is BAD is having a FSB slower than the memory bus. The Intel 820 chipset had that problem on the P3.

Just babbling.

Doug

 

[Athlon4all]

First I just thought I'd mention, here's a quote from OCWorkbench: "At the AMD forum in Japan on 23/8/2001, SiS demonstrated a 735 system running 166Mhz DDR FSB [equal 333mhz fsb]. The CPU (which I would like to know what CPU did they use) was running at 166Mhz FSB, RAM is also running at 166Mhz DDR. It was paired up with a SiS315 graphics card to demonstrate a full SiS solution." So it is coming.


<< The current FSB for a P4 is 400 MHz, ( 100 Mhz quad data rate). With PC2600, it would not be synchronized with the memory bus. The AMD ran the same way with PC133 and a 200 Mhz FSB, so there is precedent. >>


But, SiS 645 will be using the extremely low latency memory controller of SiS 735 (Which does support DDR333 btw), and that should minimize the latency issue because of the fact that the fsb will be 100 and the Mem bus at 166. SiS 645 should be a very good chipset.


<< that is BAD is having a FSB slower than the memory bus. The Intel 820 chipset had that problem on the P3. >>


Whole different situation, the PIII's fsb did not provide as much bandwidth as PC800 RDRAM (PIII's fsb is able to provide 1.064GB/ps of bandwidth, PC800 RDRAM provides 1.6GB/ps in the Single Channel configuration of 820), but with the P4, despite the fact that the fsb is running at 100MHz, and PC800 runs at 400MHz, they actually provide the same amount of bandwidth, because the P4's bus is as u pointed out Quad pumped, it actually provides 3.2GB/ps of bandwidth, and PC800 in Dual Channel config of 850 provides 3.2GB/ps also so this is a whole different situation. I will admit that running the fsb asynchorusly does add latency, but SiS and VIA have plenty experience making chipsets run asynchorously. Look at VIA KX133, it despite running asynchorously out performed the synchorous AMD 750, 850 performs very good compared to P4X266 despite it's asynchorous fsb. So I am very confident about 645's chances.

 

[DAPUNISHER]

Now let's all hope they write some software code that will take advantage of this "BIG DOG" chipset:disgust:

 

[SexualChocolate]

Actually, the SIS Chipset supports a 1/5 PCI Divider , so at 166 your FSB is still normal.

 

[SiS735POWAH]

But what about the AGP?

 

[Athlon4all]

AGP bus is supposed to run at 66MHz, and because The PCI bus would run at a 166fsb 1/5 of the fsb, and then the AGP runs at 2X the PCI bus (PCI bus in spec is 33MHz), so if the PCI bus is in spec, automatically so is AGP.

 

[Rand]

<< AGP bus is supposed to run at 66MHz, and because The PCI bus would run at a 166fsb 1/5 of the fsb, and then the AGP runs at 2X the PCI bus (PCI bus in spec is 33MHz), so if the PCI bus is in spec, automatically so is AGP. >>



Just because the PCI bus has the appropriate divider doesnt mean the AGP bus will. The AGP bus does not always run at 2X the speed of the PCI bus, it is speced to run in that manner however depending upon the divider of the AGP bus it may not necessarily.
An example of this situation would be the old BX chipset in which many boards supported the 1/4 PCI divider to allow for in spec operation at 133MHz FSB, but only has a 2/3 divider for the AGP bus, and so the AGP bus was forced to run at 87MHz while the PCI bus was operating at 33MHz. The PCI and AGP buses are not tied together and operate entirely independently of one another.
At a 166MHz FSB you would need a 2/5AGP bus divider.


As for the SiS645 chipset, I'm inclined to believe it will only support the conventional quad pumped 100MHz FSB of the P4, or possibly a quad pumped 133MHz FSB to cate for future P4's.
I sincerely doubt it will natively run a quad pumped 166MHz FSB for a data rate of 664MHz.
In all likelyhood it will just run the usual quad pumped 100MHz FSB/400MHz data rate and run the memory asynchrounously in order to run the PC333 DDR SDRAM at 166 to provide just a little more memory bandwidth then current DDR P4 mobo's.



<< Actually, the SIS Chipset supports a 1/5 PCI Divider , so at 166 your FSB is still normal. >>


I was unaware of this, could you provide the URL to some tech docs that could confirm this fact, or perhaps inform us as to where you hathered this information?
I strikes me as rather surprising if this is indeed factual.