Reply 60 of 65, by VooDooMan
- Rank
- Member
Time for a small update!
After more than 2 years of "work" there is some progress at last, thanks to some amazing people 😉
At the end of 2022, some of my boards were sent to an experienced overclocker from Kazakhstan (TerraRaptor).
The purpose of his research was:
- To stabilize the motherboard at 166 MHz bus with PentiumIII-S (Tualatin).
- To check the board's maximum OC capabilities.
TerraRaptor made various modifications to the board (CPU vCore-mod, RAM volt-mod, Northbridge vold-mod and other mods) and discovered what the important element was for the board to operate properly at 166 MHz. From that moment on, the mobo is fully stable with Tualatin CPUs but The RAM compatibility problem still remains and you need to select the perfect memory for stable 166 MHZ FSB with CL2,5-2-2-6 timings.
After stabilizing the board, the next step was to check how much overclocking this ECS board could handle 😀 It was stunned to see that he had forced this "humble" motherboard to run at 203 MHz !!! - here is the video! https://www.youtube.com/watch?v=Uot4RlmNjsg .
TerraRaptor's achievements unveil the true potential of SiS635T chipset. However, that performance was not 100% stable and his results can currently be considered experimental only. Still, it showcases what P6S5AT is theoretically capable of and gives hope that the 200 MHz option will be fully accessible in the future.
The problem at that time was the LACK OF A PROPER OC BIOS! Luckily, a professional BIOS modder from Bios Mods (Antinomy) appeared. He agreed to modify the BIOS in his free time. Antinomy added some OC options and made the BIOS publicly available. He plans to make some more additions to it one day.
Finally, I could say goodby to good-old CMOSTOOL 😉 https://www.youtube.com/watch?v=UE9TADMJZPA
In this new OC BIOS by Antinomy, the following options were added: 138/138; 150/150 and 166/166. In addition, there is also an interesting alternative for people who are not friends with overclocking: 100/150 MHz - perfect for Celeron owners on this platform 😉
It's worth mentioning at this point that he is still working on it and the world should expect the final version with all possible OC options(and other improvements) including the long-awaited 200/200 MHz option any time.
Before I proceed with my current stable results on P6S5AT, it should be noted that I've managed to lower the timings at 166 MHz from CL2.5 to CL2 - Unfortunately I came across another problem.... It seems to be rather software-related. This is what it looks like: https://www.youtube.com/watch?v=HdJ53HNZnSE .
I truly hope that I will find someone who could tweak the drivers to get rid of those issues cause those artifacts appear only with graphics cards that benefit from the AGP port features. There are no such distortions with PCI and 3DFX VooDoo cards. As you probably know, I do all my testing with AGP graphics cards, so I don't really have much to compare to, but here's a chart showing the gains of CL2 at 166 MHz FSB:
Meanwhile, when we were working on the mods for the motherboards and BIOS enhancements, a young electronics engineer from the Czech Republic (TefaTronix) joined our team. He analyzed the whole motherboard thoroughly and came up with lots of cool ideas on how to implement all the new modding solutions in a practical way (for example a small self-made PCB parts with some mods). I was very fortunate to see that TefaTronix had an oscilloscope at that time. Thanks to his equipment he was able to confirm my earlier assumptions concerning the PCI/AGP dividers. He has proven once and for all that the 1/5 divider for a 166 MHz bus is there! Now we are 100% sure that this mobo operates at 33 MHz PCI bus and the AGP speed is 66 MHz when overclocked:
https://www.youtube.com/watch?v=7SluWdt4qlY
When it comes to practical mods, I like this LED light- it signals that the PSU is turned on:
There were also some visual mods like this (those wires looked really nasty):
After I got the mobos back to me with the modifications, I tried to reach 203 MHz clock speed with my own DDR RAMs myself. Unfortunately, despite many attempts, I've managed to enter BIOS at most or run MemTest for a few seconds - https://u.cubeupload.com/roksana/20241205180957.jpg . I used Pentium III-S 1133 (SL6BY) for RAM selection, which does not even require additional voltage. I tested over 150 different DDR1 sticks - so far none of them allowed me to enter Windows at 203 MHz... (TerraRaptor used his own stick by APACER)
Since 200 MHz option doesn't currently work as I expected, I was forced to lower the FSB a little bit... The BIOS for 200 MHz is not ready yet anyway and it's very difficult to find a RAM stick that will run stable at 200 MHz 😉 TerraRaptor installed a higher clocked crystal oscillator to force the 203 MHz FSB speed. Setting 166 MHz in the BIOS, would make the board actually run at 203 MHz. The 150 MHZ option in the BIOS gave 183 MHz as a result. Here, at 183 MHz it was much easier for me to select a stable memory stick. Unfortunately, none of my fastest SL6BY CPUs will go that high with air cooling, but 181 MHz was within my reach 😉 In addition, 181 MHz gives a perfect CPU clock speed of 1900 MHz and I like round numbers 😁
All I had to do was to solder a lower clocked crystal oscillator. After doing that, the board overclocks perfectly to 201 MHz (166 MHz option in BIOS) or 181 (150 MHz option)
Now the dividers look like this:
I know it's not perfect in this shape... It's not what I dreamed of... but at least at 181 MHz the PCI port has only 36.2 MHz, and the AGP port runs at around 72.4 MHz. I think that 10% overclocked bus will still allow me for the installation of most of the graphics cards I would like to test on this set 😉
The problem after replacing the crystal is that the board will not display the real frequency values on the screen... In order to see the actual values you need to use an additional clockgenerator from another board 😁 This is a bit of black magic to me (you know I suck at electronics, don't you?)... so I don't know exactly how it works... but it looks like this:
Maybe someone knows what board this is from? 😁
1900 MHZ and 181 MHZ bus!!!!
After a long search I got a Pentium III-S 1400 MHz (SL6BY) that runs stable at 182 MHz at 1.712Vcore. It is important for the processor not to exceed 45 Celsius degrees - because then I get crashes. Room temperature is around 22 degrees, computer environment ranges from 24 to 28 degrees. The processor temperature is currently 41/42 at idle and 44 under maximum load for several hours (unfortunately this Winbond sensor is going crazy on ECS and shows wrong values...):
Now, I am finally sure that this configuration is 100% STABLE AND NOTHING CAN BEAT IT (apart from extreme heat in the summertime 😜)
https://youtu.be/a_t52EwdxTw - My video after 3 hours of excessive stability tests 😉
Interestingly, this configuration heats up MUCH LESS than the same configuration with a different board - ASUS TUSL2C (MOSFETs get very high on ASUS while on ECS they do not exceed 60 degrees)... - https://youtu.be/YGDogB2ueoc ; https://youtu.be/SnS8KYfv7fg - on TUSL the MOSFETs exceded 100 degrees...
From all the DDR memories I have selected only 6 pieces run, more or less, stable at 181 MHz FSB. The timings are always the liimt... Only one stick runs at CL2.5-2-2-6 (even CL2.5-2-2-5) it is a 512 MB stick (from Corsair, surprisingly...) and it is a bit slower than the 1GB from GoodRam with CL2.5-2-3-6 😀
------------------------------------------------------------------------------------------------------------------------------------
[center]LET'S TAKE A LOOK AT THE TEST RESULTS!!!![/center]
=====================================183 MHZ i 1740 CPU CLOCK=====================================
Ti 4600 1 GB GOODRAM
2001se:
https://u.cubeupload.com/roksana/4652001se.jpg
2000mark
https://u.cubeupload.com/roksana/2002i2000.jpg
stability test:
https://u.cubeupload.com/roksana/1Goodramwielkitest.jpg
sisoft:
https://u.cubeupload.com/roksana/42fsisoft.jpg
other:
https://u.cubeupload.com/roksana/wprimeandsuperpi.jpg
Judging by those test results, one could say that SL6BX at 183 MHz and clocked at 1740 MHz (CL 2.5-2-3-6) is practically as efficient as SL6BY at 166 MHz (CL 2.5-2-2-6) with 1750 MHz CPU clock. I didn't do any more tests on this configuration since it was temporary and not effective.
=====================================181 MHZ i 1900 CPU CLOCK=====================================
Ti 4600
WIN98SE!!!! GOODRAM 1 GB NO FW:
72,2ns latency versus78,3 with 1750 MHz and 166 FSB 😀 Performance gain is 8,5 %
PC MARK 2002 gain from 5492 to 5994 points is a total of 502 points 8,375 % increase.
WIN98SE GOODRAM 512MB with FastWrites:
WIN XP: 1 GB GOODRAM NO FW:
FX 5950 ULTRA
1GB GOODRAM win98:
1 GB GOODRAM WIN XP:
GeForce 6800 GT
WIN 98 & FAST WRITES (1GB GOODRAM)
- default clocks.
- light OC
- max OC - I THINK IT IS THE WORLD RECORD WITH THIS GPU AND A TUALATIN CPU 😀
WIN 98 & FW (512 MB Corsair)
- defualt
WIN98 NO FW (1GB GOODRAM)
- 2003 mark
- 99 mark vsync finally off
- DEF 3DMARKS
- OC 3DMARKS
- max OC 3DMARKS
WIN XP 512 MB & FW:
NOW SOMETHING THAT IS NOT POSSIBLE AT ANY OTHER BOARD WITH A TUALATIN CPU AT THOSE SPEEDS!
RADEON 9800PRO
WINXP 512 MB & FW:
WIN 98 512MB and FastWrite:
3faltigkeit wrote on 2025-01-15, 14:00:Hello VooDooMan, which Memory ICs are used on your Goodram module?
Is there maybe a model number written on the ICs?
Forget about any consistency here... You can have two identical sticks - one will work, the other won't... It's just total randomness 🙁 My GoodRam sticks have no markings on the chips - just blank dark space 😉
BTW. I got a hold of an ECS clone 😉
So, this is what I have to say for now. I also wanted to let you know that I'm in talks with some well-known hardware/PCB design engineer from YouTube and we will see what the future brings 😉 May the force of ths SiS 635T chipset be with you guys 😁
Tu summarize:
- ECS P6S5AT is able to run at 183 MHz CL 2,5-2-2-6 (1900+ MHz) with proper memories.
- OC BIOS with 138/138; 100/150; 150/150 and 166/166 is ready and will get some updates.
- Mods for stable work at 166-183 MHz were developed and will be made public after this project is done.
Things to do in the future:
- To force and stabilize the board at 200 MHz.
- To add 200 MHz options and some other enhancements to the BIOS.
- To check if the 1/6 PCI divider is really there at 200 MHz.
