Page 1 of 3

SiS 530 chipset question re 133Mhz FSB

Posted: Tue May 11, 2010 3:56 pm
by KenB
I am using a Gigabyte GA-5SMM mobo. I noticed in the manual that when configuring the FSB at 133Mhz, the PCI must be using a 1/4 divider (rather than typical 1/3 divider), because the PCI bus is still running at 33Mhz (which is good, for stability).

My only question is: This board features built-in AGP based graphics processor, which I am using. Doesn't the AGP have a divider of its' own, or does it use the PCI divider? I just want to make sure that if I use the 133Mhz FSB, that I'm not overclocking the graphics portion of the chipset.

My understanding is that this is one of the poorer performing chipsets, and to get max benefit, I should use 133Mhz FSB, if possible.

RE: SiS 530 chipset question re 133Mhz FSB

Posted: Tue May 11, 2010 6:20 pm
by Jim
Don Pedro uses boards with that chipset so he would best be able to help you. I don't know how one goes about setting up that type of board, because I have never had an SIS chipset board. But I would assume that if you set the jumpers correctly, and any related bios settings are set correctly, then it should work alright.

Posted: Wed May 12, 2010 10:48 am
by jsc1973
I never knew that board had a 1/4 PCI divider for the 133 FSB. If I had, I might have bought one back in the day. Hopefully they also supported a 1/2 AGP divider as well.

Even if they didn't, you might well be OK. A lot of Intel BX motherboards from that era were built with support for a 133 MHz FSB setting, even though Intel never supported this. Oddly, the BX did have an undocumented 1/4 PCI divider, but didn't support a 1/2 for the AGP. Any AGP cards run on a BX133 platform ran at an AGP speed of 89 MHz.

For whatever reason, it didn't seem to be a problem. Nearly every AGP card that was tried ran perfectly normally at the higher AGP port speed. I suspect that GPUs of that era were generally tolerant to out-of-spec AGP operation.

Posted: Wed May 12, 2010 12:58 pm
by DonPedro
hi kenb,

you will find a treasure of information in the thread

SiS530 @133MHz (GA-5SMM/Soyo 5SSM) setup, benching, tweaking

which many folks here on this forum have produced together.

it deals with sis530 chipset boards of all brands and types.
My understanding is that this is one of the poorer performing chipsets, and to get max benefit, I should use 133Mhz FSB, if possible.
you will learn also there that boards which are build around the sis530 chipset - once configured perfectly - are the foxes in the henhouse.

with regards to the bus divider I do not know with certainty what is the root where pci and agp speeds are computed from.

because: when we assume (as a rule) that the pci-bus' speed is derived from the fsb-speed of for example 100mhz, we would need a 1/3 divider to receive 33mhz regular speed and in the case of fsb of 133mhz a 1/4 divider. but from what source is the speed of agp taken? when we assume that the agp speed is always twice the pci-speed we would get 66mhz in the case of fsb=100mhz (divider 1/3) but also in the case of fsb=133mhz AND divider 1/4. so everything would be fine. but is this the case? I don't know how the speed of agp is produced. is it always double of pci-speed?

my experience with onboard vga of an sis530 chipset board with fsb at 133mhz and (thanks to the 1/4 divider) pci speed set to 33mhz does not strengthen the "2 times pci-speed" approach, because I was never in the position to run any of my sis530 boards with onboard graphics at an fsb of 133mhz, which leads me to the conclusion that the vga is then clocked higher than the expected 66 mhz.

so in order to get the best speed out of your sis530 board you have to use some pci-based graphics card.

the answer to the question of how pci and agp speeds are produced should take into account another question: what happens when fsb is set to 66mhz (which is a viable option if you use a cpu and/or ram which requires such a lower speed)? because there we would need a 1/2 divider for the pci bus. but what if the assumption of agp speed is 2 times pci-speed is wrong and agp speed is calculated in a comletely different manner? so many questions ...

Posted: Wed May 12, 2010 3:48 pm
by KenB
Thanks, DonPedro. i'll take a look at the link later tonight.

In case you read this sooner than I get to reading that link, would it be safe to say that keeping this board at 100Mhz FSB would be better for stability's sake, if I use the built-in GPU? This is part of a NAS (network attached storage) system.

The other option would be to altogether disable the GPU in the BIOS (the OS on this system can run "headless", no graphics card), but I don't think the BIOS has that option.

Another idea: Would it make sense to try 133Mhz FSB with a PCI video card installed to get the system past POST, then pull it out? The system would not have a chance to re-enable the built-in GPU.

Hhmmmm. Weird.

EDIT: Actually, I could be wrong on the "no graphics card" comment. Headless might actually mean no monitor is needed.

Posted: Wed May 12, 2010 6:24 pm
by DonPedro
KenB wrote:would it be safe to say that keeping this board at 100Mhz FSB would be better for stability's sake, if I use the built-in GPU?

Another idea: Would it make sense to try 133Mhz FSB with a PCI video card installed to get the system past POST, then pull it out? The system would not have a chance to re-enable the built-in GPU.
for sure setting fsb to 100mhz will be the the best choice regarding stability.

regarding the idea of pulling the graphics card from a running system: this will definitely not work, the system will crash immediately.

btw: never forget to disable mainboard 2nd level cache when setting fsb to 133mhz!

Posted: Thu May 13, 2010 3:32 pm
by KenB
OK, I guess I could stick a low end PCI graphics card in there, disable the mobo GPU, and then I wouldn't have to be concerned about the AGP divider.

I could take advantage of the 133Mhz FSB.

Sounds good? :idea:

Posted: Thu May 13, 2010 3:41 pm
by DonPedro
yes, thats the proven way it works without a glitch

Posted: Fri May 14, 2010 6:10 am
by KenB
"yes, thats the proven way it works without a glitch"
I'm confused. :D

I just spent about an hour going through all 16 pages of the thread listed above. Correct me if I'm wrong, but there did not seem to be any consensus that the SiS 530 chipset is truly capable of 133Mhz FSB. It seems to be hit or miss, with some people even suggesting to add variable resistors to mosfet leads or looking for unfinished jumpers (traces or solder holes) that might electrically disable the onboard video if shorted.

In the end, the only conclusion one can come to is that the SiS 530 really does not do 133Mhz reliably. But if you can get it to do so, it's very best performance only about meets that of the basic standard performance levels of other chipsets (ALI, VIA, etc). In other words, it totally sucks.

OK, so, unless I'm wrong about the above, I think I will just move on to a different chipset altogether for this system I'm working with. Doesn't seem worth the hassle anymore to deal with this SiS 530. :lol: Just a neutered low end chipset with no potential, me thinks.

Posted: Fri May 14, 2010 12:49 pm
by Jim
A lot of the variation will arise from the quality of RAM the individual user is using. Never having used an SIS chipset board, I can't comment more than that. In my experience the best Super 7 board to use would be the ASUS P5A. Rev 104. (NOTE : The Rev 105 & 106 do NOT support K6+ processors). With that one you really don't need WPCredit to achieve very good results. I used Powertweak, (which may or may not work with the OS you are using), to achieve comparable results with a P5A-B. The P5A is preferable though because the available PCR files for WPCredit don't work on a P5A-B.

Posted: Fri May 14, 2010 2:06 pm
by KenB
Well, the RAM I would be using is already Micron PC133 at CAS 3 from Dell, so I'm pretty sure that RAM can handle PC133 settings. It just seemed to me, after reading through that whole 16 page thread, that some were getting desperate to get the SiS 530 just to perform decently with such sophisticated ideas as electrical modification. At that point, it's no longer worth it to me. Although I totally understand the motivation behind it all, so don't get me wrong, I love to tinker too, but I don't want to get too tinkery with a PC being used for backup and data storage, which is the kind of PC this SiS 530 mobo is currently in.

The P5A board would be awesome, but I do not have one of those. Not about to spend more money for another old socket 7 board, either. I have two, already. :)

Posted: Fri May 14, 2010 5:03 pm
by DonPedro
kenb,

I would hesitate to say that the soyo board does the 133mhz or not. but i definitely would say that the gigabyte as well as the asus sis530 boards are capable of working at 133mhz speed without problems and I don't know where you read that a) that one has to fiddle around with electric circuitry and b) that even when it works
it's very best performance only about meets that of the basic standard performance levels of other chipsets (ALI, VIA, etc).

you won't get any via mvp3 or ali-5 chipset board reaching 400+mb/s ram speed. which translates to 278 seconds for superpi 1mb. if you can beat that with an via or ali board come back and let us know (screenshot).

the sis530 is truly capable of operating fsb and and ram at 133 mhz, which you can verify when you read through the spec-sheets for that chipset.

what it needs:
- deactivating mainboard cache
- setting ram parameters in bios to what your ram is capable of with the exception that RAS To CAS Delay (tRCD) MUST be set to 3T
- no use of onboard graphics device (the asus p5s-vm has a jumper to electrically cut it off): that means: no installation of drivers for the onboard solution, booting from external device


In other words, it totally sucks.
oh dear!

Posted: Fri May 14, 2010 8:13 pm
by Jim
One other thing I would mention : In my experience "Micron" is LOUSY RAM, the only board I could get it to run OK in was the ASUS P5A-B, and be it noted that ASUS overvolts EVERYTHING. In DFI boards, and FIC boards, Micron caused Problems, which is why I switched to Hynix. Further, Crucial is not a manufacturer, but rather the brand name used by Micron for their best RAM. Their lower grade stuff was sold as MT which stood for Micron Technology. Later they switched their name to "Buffalo Select", - which is no longer sold in North America.

Posted: Fri May 14, 2010 11:02 pm
by KenB
Hi DonPedro and Jim,

Would I be wrong in saying these 400MBps+ performance ratings are only if I use WPCredit and Windows? Even in some cases, certain graphics card combinations and certain multipliers? you mentioned this in a chart on page 5 of the thread you referenced.

I guess it would have helped if I had mentioned that my K6 system is a FreeBSD-based (FreeNAS, specifically) PC, not Windows. In such a case, some tweaks will not work for me (WPCredit). even so, your advice on not using the onboard graphics or onboard L2 cache chip is helpful for best performance of the hardware itself, OS aside. Thank you.

I'll have to see if FreeNAS loads a graphics driver for the onboard GPU.

In my quick reading of the 16 pages, it may be that I misunderstood something (understatement)
:oops:

Jim, the RAM stick I have is from Dell, using MT chips. Are you aware if the Dell modules are purchased through Crucial, or are they the lower quality units you speak of? Even so, the modules themselves are labeled to operate at 133Mhz FSB at CAS2 with ECC. I also have a PC133 MT module at CAS3, no ECC.

Posted: Sat May 15, 2010 9:30 am
by Jim
I very much doubt if Dell used Crucial. That stick is probably MT aka Mouse Turd. You can try it; but if the machine has problems, there is your number 1 suspect.