K6-III+ works good on 133 FSB

Discussion relating to Socket 7 hardware.
User avatar
Uranium235
Senior K6'er
Posts: 183
Joined: Mon Sep 12, 2005 9:59 pm

K6-III+ works good on 133 FSB

Post by Uranium235 »

Got an Asus P5S-VM running at 600Mhz (4.5x133). Motherboard was an old HP board with a AMI BIOS. I had to hotflash the chip to take Jan Steunebrink's modded award BIOS. The Asus BIOS lies and says the RAM is running at 2-2-2 but in reality defaults to 2-3-3. I can tweak it to 2-3-2 with Wpcredit, 2-2-2 crashes the machine. Also the Asus manual lies and says the RAM is running synchronous with the CPU at the correct settings but actually defaults to asynchronous mode, again I had to use Wpcredit to set the RAM to synch. I couldn't find a SiS530 .pcr file so I've been winging it with the datasheets. Performance is inline with top VIA and Ali boards. I'll post some benchmarks later when I get time.

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

Post by DonPedro »

ah, that is nice to hear. :)

for exactly the same reason - the opportunity to run a k6 with 133fsb paired with ram clocked with 133mhz I ordered a gigabyte p5smm (from the u.s.) many weeks ago and customs auth. notified me this week that it is in process, I should get it next week. In the mean time I found a brand new soyo-5ssm and ordered it too, should also arrive next week. both boards have the sis-530 chipset you are talking about. I will happily join your quest into that area.

I also tried to locate some 530 pcr files to tweak the chipset with wpcredit but was not lucky either. anybody else out there who could help here?

I learned that the 530 chipset is not the real thing for 133 mhz. all the information I gathered said that sis considers it as overclocking (all speeds above 100 mhz). the real 133-thing did come with the 540-chipset. only problem is to locate any brand with good reputation that used that chipset in one of its boards.

I finally found out that ibm used that chipset in their aptiva series, modells 2193 and 2196. they seem to have been sold especially in the asian markets. I found many items for sale at some auctions over there but since I am not in the position to understand any asian language I did not place any bid.

So I went forth and contacted ibm-europe if they still have spare- mainboards on stock. they really do but they ask - hold your breath - about 320 euros for it (about $400,-). :(
User avatar
KachiWachi
K6'er Elite
Posts: 507
Joined: Wed Sep 21, 2005 10:53 am
Location: Pennsylvania, USA

Post by KachiWachi »

It is certainly easy enough to write your own PCR file.

Just use an existing one as your guide, and edit in the parameters (using the SiS datasheet) as you go.
Moderator - Wim's BIOS

PC #1 - DFI 586IPVG, K6-2/+ 450 (Cyrix MII 433), 128 MB EDO. BIOS patched by Jan Steunebrink.
PC #2 - Amptron PM-7900 (M520), i200 non-MMX, 128 MB EDO
PC #3 - HP8766C, PIII-667, 768 MB SDRAM
PC #4 - ASUS P3V4X, PIII-733, 256 MB SDRAM
PC #5 - Gateway 700X, P4-2.0 GHz, 768 MB PC800 RDRAM
PC #6 - COMPAQ Evo N1020v laptop, P4-2.4 GHz, 1 GB PC2700 DDR
PC #7 - Dell Dimension 4600i, P4-2.8 GHz, 512 MB PC2700 DDR
PC #8 - Acer EeePC netbook, Atom N270 @ 1.60 GHz, 1 GB RAM
PC #9 - ??? ;)
DonPedro
K6'er Elite
Posts: 578
Joined: Wed Jul 27, 2005 2:11 pm

Post by DonPedro »

how easy is easy? and what documents do I need? I locatetd a file called 0530DS10.PDF and downloaded it. does one need more information? and could you give a teach-in on that matter (favorably in a new thread since this is somehow off-topic here)?
User avatar
Uranium235
Senior K6'er
Posts: 183
Joined: Mon Sep 12, 2005 9:59 pm

Post by Uranium235 »

DonPedro, I appreciate your interest in the SiS530 and the 133 FSB. It is one of the last unexplored frontiers of Socket 7 in my opinion. I also look forward to your results when your boards arrive. :)

The SiS530 datasheet you have is the same one that I am using. It is rather easy to translate the datasheet info to Wpcredit parameters. Just look at the charts starting under "Register Descriptions" on thumbnail 81 of the .pdf file. The register numbers, like 04h or 51h, directly correspond to the Wpcredit offsets. Just disregard the "h". The bits listed under the register numbers on the datasheets directly correspond to the bits on Wpcredit 0-7 listed on the far right.

There is also a program called Tweakbios that I've been using to test different settings. You can use Wpcredit to see what data is changed by Tweakbios.
User avatar
Uranium235
Senior K6'er
Posts: 183
Joined: Mon Sep 12, 2005 9:59 pm

Post by Uranium235 »

OK here's some early benchmark results, here's the setup:
K6-3+ 400ATZ @600(4.5x133)
Asus P5S-VM Rev. 2.02, 512K L2 disabled
512MB (2x256) Crucial PC133, 2-3-2-6
Radeon 9100 64MB PCI
40GB Western Digital HDD, 7200rpm, 8MB
Windows98SE

Everest v2.20.405
Memory Read- 316 MB/s
Memory Write- 157 MB/s
Memory Latency- 203.7 ns

CPUmark 99 Version 1.0
59.2

Super PI
[1MB] 5m 16s

Hot CPU Tester Pro 4
Total Score 1170
Metamark to Mhz ratio 1.95

3DMark2001 SE
3207 3D marks
Last edited by Uranium235 on Sat Sep 09, 2006 8:46 pm, edited 1 time in total.
User avatar
KachiWachi
K6'er Elite
Posts: 507
Joined: Wed Sep 21, 2005 10:53 am
Location: Pennsylvania, USA

Post by KachiWachi »

@Uranium235 - Exactly. (Note - h = hex as to not confuse the data string with decimal.)

All you do is match the hex offsets with whatever text is displayed for them. Many are in the same place (Manufacturers ID and such), while others are "Reserved". Just edit away with your favorite text editor. If you text entry is too long, it will run over into the next space...that is the only thing you need worry about....and that is for your convienience. You can always manipulate the data (though it is easier to do so when the datasheet doesn't have to be in your hands. :P )

The "bracketing" of the numbers is also for convienience...to denote "read only" bits, etc... No worry on that either...since it is again only for display purposes.

Over at the AMD forum a young kid named Charlie22911 worked up a 530 motherboard some time ago. I think he posted some test data there as well. ( http://forums.amd.com/index.php?showtopic=12325, http://forums.amd.com/index.php?showtopic=18866 )

Also check under the username cowen1000. ( http://forums.amd.com/index.php?showtopic=16883 )
Last edited by KachiWachi on Sun Sep 10, 2006 4:55 pm, edited 4 times in total.
Moderator - Wim's BIOS

PC #1 - DFI 586IPVG, K6-2/+ 450 (Cyrix MII 433), 128 MB EDO. BIOS patched by Jan Steunebrink.
PC #2 - Amptron PM-7900 (M520), i200 non-MMX, 128 MB EDO
PC #3 - HP8766C, PIII-667, 768 MB SDRAM
PC #4 - ASUS P3V4X, PIII-733, 256 MB SDRAM
PC #5 - Gateway 700X, P4-2.0 GHz, 768 MB PC800 RDRAM
PC #6 - COMPAQ Evo N1020v laptop, P4-2.4 GHz, 1 GB PC2700 DDR
PC #7 - Dell Dimension 4600i, P4-2.8 GHz, 512 MB PC2700 DDR
PC #8 - Acer EeePC netbook, Atom N270 @ 1.60 GHz, 1 GB RAM
PC #9 - ??? ;)
DonPedro
K6'er Elite
Posts: 578
Joined: Wed Jul 27, 2005 2:11 pm

Post by DonPedro »

thank you folks! with your help we might do it!

@uranium:
did you run your benches with mainboard-cache activated? and does the P5S-VM allow to run the ram @133?

just to give you a quick comparison for your bench-data I digged out some old data. the setup is different but it helps anyway to know whether the numbers are good or bad.

GA-5ax rev. 5.2
768 RAM @ 22258 @ 100mhz
cache off
K6-3+@550 !!!!
GF-4 460MX, AGP
NT4-Svp6a

Everest v2.20.405
Memory Read- 350 MB/s
Memory Write- 167 MB/s
Memory Latency- 196.7 ns

CPUmark 99 Version 1.0
53

Super PI
[1MB] 5m 52s

Hot CPU Tester Pro 4
Total Score 1065
Metamark to Mhz ratio 1.93

if you had mainboard ON then this would explain your lead in SuperPI and CPUmark99. they both love cache, the more the better. I don't give much for the cpumark-bench, but SuperPi is acutally a real-world program that really tells whether a task can be run faster or not. but these higher numbers could also be because of the 50mhz cpu speed difference here. could you clarify (rerun if necessary) this?
User avatar
Uranium235
Senior K6'er
Posts: 183
Joined: Mon Sep 12, 2005 9:59 pm

Post by Uranium235 »

DonPedro, All those benches were run with the mainboards L2 cache off. This makes the results appear even more impressive. I couldn't get the machine to operate with the mainboard's L2 cache on above the 100 FSB anyway. I don't believe it would have mattered much because 512K L2 cache with the SiS530 can only cache 64MB of RAM. VIA and ALi boards handle L2 cache alot better than SiS boards.

The P5S-VM is supposed to be able to run the memory at 133 and there are jumper settings for that, which is how I set it up, but I was looking at the datasheets and noticed Register 52h, bit 4 controlled Asynchronous/Synchronous mode between CPU/DRAM Clock, so I looked at Wpcredit offset 52, bit 4 and found it was set to "0" or Asynchronous mode. I changed offset 52, bit 4 to "1" and this made a huge increase in performance. I would recommend that you check this setting on your boards as well when the time comes.

Thanks for including your benchmark results. Unfortunately I can't set the machine up to run at 550Mhz at the 133 FSB for a closer comparison. I can run more benchmarks at any setting if you need to see something. I'm going to leave this setup alone for a few weeks to see what else it can do.

KachiWachi, thanks for the more in-depth explanation on Wpcredit. I may write a PCR file if would be a benefit to someone but I've only found a very small amount of settings worth looking at.

:)
User avatar
Uranium235
Senior K6'er
Posts: 183
Joined: Mon Sep 12, 2005 9:59 pm

Post by Uranium235 »

Here's the jumper settings that are "supposed" to result in synchronous CPU/RAM operation according to the manual:

FS3- pins 1&2
FS2- pins 1&2
FS1- pins 1&2
FS0- pins 1&2

HOST 133MHz
SDRAM 133MHz
PCI 33MHz

I was looking to see if there were any undocumented jumper combinations that would result in actual synchronous CPU/RAM operation at 133 FSB but all combinations appear to be accounted for. The only other possibility is that a jumper has to be completely removed...
DonPedro
K6'er Elite
Posts: 578
Joined: Wed Jul 27, 2005 2:11 pm

Post by DonPedro »

thanks for the information

I still don't have the 530-boards in my hand ..... when they arrive next week I will be on holidays (ahhhh, on a greek island ....)
Super7Dude
K6'er
Posts: 77
Joined: Mon Aug 28, 2006 8:05 pm
Location: Brisbane, Queensland, Australia

ASUS P5-A Rev 1.03 and K6-III+ 450 doing 133MHz too!

Post by Super7Dude »

I can also confirm that the ASUS P5-A rev 1.03 with 1011 beta 005 BIOS is 100% stable at 133MHz FSB. I have my K6-III+ doing 4x133 for 533MHz (needs 2.1v for a stable 550 tho) :shock: Even with an old 2GB Fujitsu hard drive, it still feels fast!
DonPedro
K6'er Elite
Posts: 578
Joined: Wed Jul 27, 2005 2:11 pm

Post by DonPedro »

super7dude,

could you please get into more detail regarding your system setup (graphics card used, ram used, operating system).

I was never able to get beyond 105 fsb with my p5as and run any benchmark without system crashes. brother in arms jim reported in another thread that he successfully run the p5a with 112 (110 setting) but I just can 't believe that it is possible to get a stable system with that much overclocking: 133 fsb means pci-bus going up from 33 to 44mhz and agp from 66 to 88! did you disable the onboard cache (I guess so)?

please run some benchmarks (everest, hot cpu tester, superPi) and tell us whether you experienced any anomalies and also tell us your results.

also would like to know whether you can run 3dmark2001 in demo-loop mode and how long it runs without crashing.

edit: I am also puzzled because I am wondering on how to set the jumpers to get 133 fsb
Attachments
p5a fsb jumper settings.png
p5a fsb jumper settings.png (4.33 KiB) Viewed 14013 times
Super7Dude
K6'er
Posts: 77
Joined: Mon Aug 28, 2006 8:05 pm
Location: Brisbane, Queensland, Australia

ASUS P5A at 133MHz FSB

Post by Super7Dude »

I have three 128MB sticks of PC133 CL3 RAM
one from Amicroe with Spectec ICs,

one from Micron with unknown ICs (label says: PC133U-333-542 256MB Synch PC133 CL3"

and one from Hynix with the same specs on the label (synch 133U-333-542)

I have a HIS GeForce2 MX400 PCI card (64MB 128bit SDR) at 208/183 and a Creative PCI64V sound card.

Hard drive is a Seagate 120GB 7200.9 series

Ironically, the system just BSODd running 3DMark01SE so it might not be that stable :oops: but it gets into windows XP no probs and seems fine for light duty work.

Not sure if the PCI bus is locked... but haven't seen any data corruption yet... so fingers crossed.

All FSB jumpers covering the top two pins (1-2) gives the 133MHz FSB. The RAM also runs at timings of 3-3-5.

*edit* Everest 2.20 Home Edition RAM bandwidth score was 323MB/s... just above the other super7 boards listed and about on par with a socket 370 PIII with 100MHz FSB.

CPU has been priming for about an hour now without a problem so im guessing the 3DMark01 crash was due to the 44MHz PCI bus speed. The soundcard isn't working either. But how it even boots and (partially) works is intriguing.

*edit* sorry, that was a duron not a PIII for everest
User avatar
Stedman5040
Veteran K6'er
Posts: 271
Joined: Mon Sep 05, 2005 4:22 pm

Post by Stedman5040 »

@uranium235

I have a Gigabyte G5-SMM. Any chance of a showing a register dump on WPCREDIT after your bios tweaks? I have just set up this board as is (no tweaks) using onboard sound and video with a K6-2+550@550. Memory is running 2-3-2-6 and Everest for this aint so hot. With asynch mode

MR=127
MW=72
ML=495

With synch mode

MR=153
MW=73
ML=395

All at 100fsb with cache off.

Stedman.
Post Reply