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

Discussion relating to Socket 7 hardware.
DonPedro
K6'er Elite
Posts: 578
Joined: Wed Jul 27, 2005 2:11 pm

Post by DonPedro »

kachi and all,

the bios routine at startup is obviously not in the position to set register 60h accordingly, be it that the programmers back then did not even think of implementing ram modules of 512mb ram (because they were not available at that time) or be it that the lookup-table is faulty.

here is the screenshot of what wpcredit shows at reg 60h with an 512mb ram module installed. I combined it with an excerpt of the sis530 datasheet.

one sees that bits 3..0 should be set to 1-0-1-1 when an 32Mx8 module is used (organisation 2x13x10), but instead it is set to 1-0-1-0 which is the organisation of 256mb ram modules (2x12x10). so what it needs is to set bit 0 to 1.

the first table provides an overview of internal organisation of chips used and organisation of dimm-modules. although the table is from a german pc-magazin I think that it is still readable to someone with technical background. cells in bold mean "standard" configuration.

edit: changed "set bit 1 to 1" to "set bit 0 to 1"
Attachments
sis530 register 60h.png
sis530 register 60h.png (36.88 KiB) Viewed 25133 times
ct ram chip dimm modul explanation.png
ct ram chip dimm modul explanation.png (68.62 KiB) Viewed 25133 times
Last edited by DonPedro on Wed Apr 09, 2008 3:21 pm, edited 1 time in total.
DonPedro
K6'er Elite
Posts: 578
Joined: Wed Jul 27, 2005 2:11 pm

Post by DonPedro »

@uranium,

thanks for the flowers! :)
to be honest it is the same setup and uses the same wpcredit tweaks as we already found. the only thing was that I populated all 3 dimm-slots with 256mb ram each. second I tested an addon-card (usb, firewire, sata) at the very same time. I was/am happy that the performance did not drop but went up instead! :) (I will report on my add-on card experience I announced before x-mas soon).

anyway, here is again a summary of the bits I set via wpcredit:

=========================================
50: bit 6 set to 1 (mem single write cycle) or 0 in w2k
52: bit 7 set to 0 (dram time delay set to 0)
52: bit 6 set to 1 (single read allocation)
52: bit 4 set to 1 (sync cpu-dram mode)
55: bit 2 set to 0 (disable res bit for data read dram)
56: bit 7..6 set to 11 (6 clocks)
58: bit 7..6 set to 00 (2 clocks)
58: bit 5..4 setting from 01 to 00 crashes (from 3 clocks to 2 clocks) !!!! never succeeded once here
5D: bit 4 set to 1 (dram write retire rate = x-1-1-1)
5D: bit 3 set to 0 (cas latency = 2 clocks)
===========================================


@kalabok,

according to the table that c't magazin provides 32x64 modules don't exist.

EDIT:
oh, in case you are talking about the organisation of the whole modul, yes, modules with 32x64 org. do exist, but are of 256mb size not 512mb. single chips with 32x64 org. don't exist.
DonPedro
K6'er Elite
Posts: 578
Joined: Wed Jul 27, 2005 2:11 pm

Post by DonPedro »

in the meantime some news on getting 512mb ram sticks working, from jan steunebrink, an excerpt from an e-mail I received yesterday:

"Last time I worked on the P5S-VM BIOS I concluded that the the memory
detection logic is not in the main BIOS module, which surprised me
greatly... The only other piece of BIOS code left to examine now is the
bootblock. Normally this block only performs some initial setup and
decompresses the other modules in shadow-RAM and continue the POST from
there. But maybe on this BIOS is does some more...

So I've started examining the bootblock code to see if the the the memory
detection logic is there. I never had to patch the bootblock before, so this
piece of code is new to me. ;-)

I expect to have some time to spend on this project in a week from now.
I hope I will find the the memory detection logic then. Keep your fingers
crossed! :-)

Greetings, Jan.
User avatar
Stedman5040
Veteran K6'er
Posts: 271
Joined: Mon Sep 05, 2005 4:22 pm

Post by Stedman5040 »

Has anyone yet managed to get a stick of 512mb to run in a SIS530 chipset board?

I have my set up running now as follows

AMD K6-III+450 @ 600 (133 x 4.5 ) external cache disabled
Gigabyte GA5SMM v1.2 bios F5
Radeon PCI 128mb 9250
2x256mb Twinmos PC133 CL2 @2-3-2-6

I have a large copper cooler on the CPU but with the fan going at about 3500rpm which keeps the noise levels down.

Running this set up with the WPCredit modifications as before I get

Everest Home 2.20

MR/MW/ML 402/156/166

and

Superpi 1M 4m 50s

Sandra 2004 memory bench

214/211

Not bad results for a SS7 system, although not quite to the speed of the Asus set up quoted above.

I am hoping to progress to a water cooled set up to see if I can manage to squeeze another 66Mhz out of the CPU or try for 124x5.5. I have the cpu heatsink already and now have to just get the rest of the set up.


Stedman
DonPedro
K6'er Elite
Posts: 578
Joined: Wed Jul 27, 2005 2:11 pm

Post by DonPedro »

stedman,

it is a long time I asked steunebrink to look into this matter, I will contact him again, the 512mb ram question is really something I would like to have available on this board ....

I also have the water cooling thing in my mind to go for 666mhz. I found a set that fits on socketA/socket370/socket7 sockets, but the price up to now is not very inviting ....


innovatek i-COOL rev3.0 for about 200 euros (complete set, not just cooler)
User avatar
Witte
Newbie K6'er
Posts: 18
Joined: Sun Sep 28, 2008 11:30 am

Post by Witte »

Hi all,

I'm new to the forums but I'm glad I've ever found it. At this moment I'm experimenting with the Gigabyte GA-5SMM mainboard. It's a rev1.3 with no onboard SRAM. I was overclocking this board but I failed in having the system boot at anything higher then 112MHz FSB.

Next I modified an old PSU to have the 3.3V line regulated up to 3.7V but unfortunately this did not increase the overclocking headroom, I still couldn't log into windows at 124/133MHz. Then I found out the VCC3 notification inside the BIOS. It monitors the chipset core voltage I suppose, next what I did was finding and modifying the regulator circuit and few minutes later I found myself booting into Windows with the 133MHz bus speed selected!

I think this may be very handy for all these who are having problems with overclocking their GA-5SMM board, I'll upload pictures ASAP.
User avatar
Witte
Newbie K6'er
Posts: 18
Joined: Sun Sep 28, 2008 11:30 am

Post by Witte »

VCC3 modification:
Attachments
VCC3mod.jpg
VCC3mod.jpg (122.46 KiB) Viewed 24698 times
Last edited by Witte on Mon Sep 29, 2008 9:46 am, edited 1 time in total.
DonPedro
K6'er Elite
Posts: 578
Joined: Wed Jul 27, 2005 2:11 pm

Post by DonPedro »

witte,

thank you for your investigation.

your success-setup at 133mhz does it still include the modified psu or is the variable resistor alone able to do the job?

also, the picture does not show pin 1 and 3 connected via the variable resistor but only a wire connected to pin 1. does the picture actually show the final setup? if this question is somehow stupid please forgive me, I am not very familiar with soldering etc.

also, by "setting the variable resistor to 50kohm at start ..... " am I right that this means that one should start with full 50kohm and then gradually should decrease the resistance, try to boot, if it fails again should decrease the resistance and so on until the machine boots and runs(!!) with enough stability?

since you seem to be familiar with soldering, exchanging and replacing parts of the electric circuitry could you please look into the following and tell us the outcome:

all of whom I know (including me) were not able to run the system at 133mhz fsb/ram when onboard-cache is ENabled. I think the reason is that the problem is with the clock generator / tagram for the cache-ram that is not built to run at 133mhz, but the cache-ram is (5ns should make 200mhz). if it is possible to replace that unit with one that does 133mhz we would be able to reach new speed records regarding ram access/read/write as long we are playing within the limits of ram cached (which in case of 512k cache translates into 64mb of ram).
User avatar
Witte
Newbie K6'er
Posts: 18
Joined: Sun Sep 28, 2008 11:30 am

Post by Witte »

your success-setup at 133mhz does it still include the modified psu or is the variable resistor alone able to do the job?
VCC3 is taken from the 5 Volt line, anybody who copies the modification should be able to get VCC3 up to roughly 4,1V irregardless to the 3,3V line voltage. Unfortunately 133MHz is no where near stable and I feel I need just a bit extra. I'm thinking of using my own build circuit and bypass the LX8383A.

also, the picture does not show pin 1 and 3 connected via the variable resistor but only a wire connected to pin 1. does the picture actually show the final setup? if this question is somehow stupid please forgive me, I am not very familiar with soldering etc.
What you see in the picture is how I did it during tests, I solder wires onto the mainboard in order to have the variable resistor right next to me instead of on the mainboard where it is hard to adjust (due to space). You don't see a wire to the right leg of the IC because I took the 5V line as reference. It should be no difference then what I showed on the picture as the 5V line is used as Vin for the LX8383A (aka pin3). In the end I did make the modification where I had the resistor right on top of the LX8383A because 4,1V is the maximum and I needed no lower then that :)
also, by "setting the variable resistor to 50kohm at start ..... " am I right that this means that one should start with full 50kohm and then gradually should decrease the resistance, try to boot, if it fails again should decrease the resistance and so on until the machine boots and runs(!!) with enough stability?
That's correct, at 0 ohm, that's when you have the resistor completely turned over you'll get VCC3 at around 4,1V, it was no problem for me but just to be sure I added a fan on top of the North Bridge to prevent it from overheating. It did fell warm.
all of whom I know (including me) were not able to run the system at 133mhz fsb/ram when onboard-cache is ENabled. I think the reason is that the problem is with the clock generator / tagram for the cache-ram that is not built to run at 133mhz, but the cache-ram is (5ns should make 200mhz). if it is possible to replace that unit with one that does 133mhz we would be able to reach new speed records regarding ram access/read/write as long we are playing within the limits of ram cached (which in case of 512k cache translates into 64mb of ram).
I've haven't had play with enabling/disabling onboard cache since I'm running the board with an P166MMX, makes no sense to disable it. I'll soon try it just to know if it would mean any difference in terms of overclocking but from what I can make up is that it is probable the onboard VGA which may hold back the overclock (wasn't it mentioned somewhere in this thread?). I'm using a PCI card though.

Your guess is that the TAGRAM may hold back the overlocking? What chip (I don't have the mainboard around at this moment) is it?

Concerning the L2 cache, the chipset support up to 2MB L2 cache, unfortunately there are no 2MB chips (as far as I know) so if we were to replace it we would be stuck with 1MB maximum. Replacing the cache is very hard to do, don't think I'll succeed in doing so :)
Attachments
cacheconfig.jpg
cacheconfig.jpg (62.92 KiB) Viewed 24694 times
DonPedro
K6'er Elite
Posts: 578
Joined: Wed Jul 27, 2005 2:11 pm

Post by DonPedro »

re replacing / upgrading the cache:

there are 1mb cache chips available, so for example the epox mvp3-g5 mainboard features 2mb of cache (two 1mb chips). of course, replacing or adding another chip needs special machinery I guess ...

re mainboard cache enabling / disabling:
well, all parties who participated in this thread failed to run any sis530 mainboard at 133mhz fsb/ram speed with cache ENabled, but succeeded when DISabled. on the other hand I own a ga-5mms and a soyo sy-5ssm; both are not capable to run at 133mhz even when cache is disabled.

the only board I own that is able to the job is the asus p5s-vm. this board features pins to actually completely disable the onboard vga. but others here in this thread run their ga-5mms at 133mhz without trouble.

yes, I think it is the tagram which is not capable of doing 133mhz. this chip can be found in the neighbourhood of an about 7mm by 2-3mm silver shining oval device (which has 14.318 printed on it). the tagram chip differs from mainboard to mainboard. my soyo sy-5ssm has two chips here: an sdt sb61l256bs-8 chip and an ics 9248AF-128. Idon't know which one is the tagram, but I think it is the sdt-chip.

here is a link to an example of a tagram chip marked "2". the oval silver device can be seen in this photo just above the left of the 2 orange devices. click here
User avatar
Witte
Newbie K6'er
Posts: 18
Joined: Sun Sep 28, 2008 11:30 am

Post by Witte »

Hey DonPedro,

When you say fail, did your system not boot at all? Because mine does but it crashes during WinXP boot. I does boot all way op to the dualboot menu, that's why I think some extra voltage might be enough.

About the chips, the 7mm by 2-3mm silver shining oval device you talk about is in fact a crystal which oscillates at an harmonic frequency of 14,318MHz. It's the base frequency of any todays mainboard. The PLL chip aka clock generator divides and multiplies that frequency until you get the various bus speeds found in an existing pc system (100MHz/33MHz). It is the PLL which builds the CPU host clock, inside the CPU the host clock is being multiplied so that you get a much higher CPU speed (200MHz for example).

On your Soyo board the ICS 9248AF-128 chip you talk about is the clock generator. On the Gigabyte GA-5SMM I found the ICS9248-BF-81 clock generator. Even though the chipset may support 133 mode, and even though you have a BIOS which can set 133 mode, if the PLL does not support it you won't ever be able to set your system to 133 mode. You can off course desolder the original 14MHz crystal and replace it with an 15MHz one, just don't forget that you will be overclocking your entire board that way because the IDE/PCI/USB/and various other bus speeds are also being generated by that ICS chip. You may encounter data loss and such, most people don't ever touch the base crystal because it is not of much use. In theory its possible but it is not very practical so better leave it is it where. For more information you should 'google' for TurboPLL.

TagRAM, I know that some boards come with TAGRAM sockets, unfortunately that is not the case with the GA-5SMM so we can not quickly go change it. The sdt sb61l256bs-8 you talk about could be the TAGRAM chips on your SOYO board, can't find any datasheet confirming your thoughts. But if it looks like the one you gave as picture you can be pritty sure it's a TAGRAM IC. On the GA-5SMM, can you find something alike, or put some names which you find on chips on your board? Could you maybe shoot me a high-resolution picture of the board so that I could have a look at it myself? From a low res picture I'm asuming the TAGRAM chip might be next to the NB, on its left side.

Oh, I did manage to modify the CPU Vcore regulator but at +0,1V it shuts itself down so it is of not much use.
User avatar
Witte
Newbie K6'er
Posts: 18
Joined: Sun Sep 28, 2008 11:30 am

Post by Witte »

Hey

I found the SDT sb61l256bs-8 chip on my GA-5SMM rev1.3 too, it's TAGRAM for sure. The "-8" at the end makes me believe it is 8ns static RAM which is only up for 125MHz. I googled for TAGRAM datasheets and found a chip which is roughly as old as the SDT chip, it should be pin compatible. In the datasheet I found that pin 28 is the VCC pin of the TAGRAM chip and after measuring around on my board I found that my thoughts were correct and that it is taking its power from VCC3: so altering VCC3 should make it possible to get higher clocks on the TAGRAM chip.

What I did next was completely get rid of the LX8383A power regulator and replace it with an switching power circuit which I made myself. It allows me to go up to 12V, no I'm not that crazy but I still went up to 4,5V for VCC3. Unfortunately no success, at 133MHz mode the system is still crashing during WinXP boot.

Just checked another board, it too comes with 8ns TAGRAM. Anyone found faster chips?
Attachments
tagram.JPG
tagram.JPG (16.43 KiB) Viewed 24665 times
DonPedro
K6'er Elite
Posts: 578
Joined: Wed Jul 27, 2005 2:11 pm

Post by DonPedro »

witte,

so what we need is a substitute tagram with 7.5 ns or better. one we find the correct chip it then only needs a skillfull and experienced solderer, the pins are small but still big enough for a successful operation.

have you tried to boot without onboard cache yet?
User avatar
Witte
Newbie K6'er
Posts: 18
Joined: Sun Sep 28, 2008 11:30 am

Post by Witte »

Yes but it took ages, didn't know if the system actually hanged during boot or if it was still working. Did a reboot then and tried some other stuff, didn't really had enough time with the board yesterday :)
Soon I'll pick up an P5A and have it compared with the GA-5SMM, maybe we learn something :)
Jim
K6'er Elite
Posts: 1745
Joined: Wed Jan 21, 2004 7:10 pm
Location: Toronto

Post by Jim »

Keep sharing what you learn please. You know far more about electronics than most people here, as far as I can tell..
Superpuppy 3
K6-3+ 450 ACZ (6x100)
DFI K6BV3+/66 Rev B2 (2 Meg) w/ 2x28mm Chipset Fans
2x256 Meg PC 133 Hynix SDRAM
1x 20G Maxtor (7200)
2x 80G Maxtor (7200) Ducted w/ 2x486 Fans Mount
52/24/52/16 LG CDR/RW/DVD
8/4/3/12/24/16/32 LG Super Multi
ATI 9000 aiw Radeon AGP
SB Audigy 1 MP3 Sound
CMD 649 IDE Controller
NEC USB 2 Card
Post Reply