VOGONS


K6-2/3+ Success on Asus P5A 1.06 - simple mod

Topic actions

First post, by FesterBlatz

User metadata
Rank Member
Rank
Member

The title says it all! Following a vague post on k6plus.com regarding a simple modification to correct the "slowness" issue when paring K6-2/3 "plus" variants with Aladdin V rev G Northbridge equipped P5As (rev 1.05 & 1.06), I went ahead and tried this on my P5A rev 1.06 -- so far the results look promising!

The post I saw is here: http://www.k6plus.com/phpBB3/viewtopic.php?f=3&t=2340

To attempt this myself, all I did was use a 1K resistor to pull WB/WT# (pin AA5) up to Vcore 3.3v at VCC3 pin AJ19.

So far the only test I've done is run Quake 2 using 3DNow Software 800x600 mode with a K6-2+ @550MHz installed, and compare to my MVP3 based K6-3+ 550MHz machine with the exact same Quake settings, 98SE, identical video cards and driver version. I get 19.8 FPS on the modded P5A K6-2+ and 16.1 FPS on the MVP3 K6-3+. Considering the K6-3+ on the MVP3 board has double the on-die L2 cache, my P5A outperforming it even slightly seems to be a good indication that it isn't being hampered by the "slowness" hardware bug that effects rev 1.05 and newer boards.

I have no idea what the actual "bug" is, but this seems to indicate the ALi M1541 rev G Northbridge may not be asserting WB/WT# correctly--either poor timing or not at all. Somehow simply forcing that input high seems to work around the issue... I'm not sure if there are any side-effects yet, but so far the system seems perfectly stable.

Having just got back into SS7 since the 90s, I don't have a collection of benchmark utilities installed yet to perform a more detailed comparison. But considering Quake2 seems to play just fine the bug appears to be a non-issue now. Prior to the mod I was getting about half that frame-rate.

I'll continue to mess around with this and be sure to update with pictures and whatever benchmarks I can provide, so feel free to request.

Hopefully some folks find this useful!

UPDATE: Here's a picture of the mod. I used a SMD resistor but an axial resistor would work fine too. Just be careful not to accidentally bridge to any adjacent pins or vias. It's a little difficult to see in the picture, but I used Kapton tape to help prevent any shorts.

IMG-2656.JPG
Filename
IMG-2656.JPG
File size
1.87 MiB
Views
9415 views
File license
Fair use/fair dealing exception
Last edited by FesterBlatz on 2017-11-29, 00:42. Edited 4 times in total.

Reply 2 of 109, by Stedman5040

User metadata
Rank Newbie
Rank
Newbie

Hi,
I am no electronics whizz, but are you saying that you just have to solder a 1 k ohm resistor to bridge the 2 cpu socket pins you have outlined above. If so that would be great news. I have P5A versions 1.05, 1.06 and 1.04. So I could test the relevance of activating the defective chipset tagram with a K6plus cpu.

A pic of the two pins on the back of the circuit board would be useful with orientation shown.

K6 III+ 500
Epox MVPG2
512Mb Hynix CL2 SDRAM
40MB WD HDD
Creative GeForce2 Ti
CMI8738 Sound card

Reply 4 of 109, by FesterBlatz

User metadata
Rank Member
Rank
Member

Exactly, just tying two pins on the back side of the motherboard with a 1K resistor is all that's needed.

It was after midnight and I've been struggling with a head cold so I didn't bother getting any pictures yet. I'll post some later tonight to help illustrate.

Stedman5040 wrote:

Hi,
I am no electronics whizz, but are you saying that you just have to solder a 1 k ohm resistor to bridge the 2 cpu socket pins you have outlined above. If so that would be great news. I have P5A versions 1.05, 1.06 and 1.04. So I could test the relevance of activating the defective chipset tagram with a K6plus cpu.

A pic of the two pins on the back of the circuit board would be useful with orientation shown.

Reply 5 of 109, by Stedman5040

User metadata
Rank Newbie
Rank
Newbie

Question

After looking at the spec sheet for the cpu I presume that you can use any of the many vcc3 pins that are available on the cpu pin layout.

This is really amazing and I would imagine that the same mod would work for the higher versions of the baby AT board the Asus P5A-B.

I have already done some testing on the P5A board Rev 1.06 with a K6-III standard cpu at 500Mhz and it does work faster than the Rev 1.04 board. So using the plus variant at 5.5 multiplier should work really well with regards to memory transfer.

K6 III+ 500
Epox MVPG2
512Mb Hynix CL2 SDRAM
40MB WD HDD
Creative GeForce2 Ti
CMI8738 Sound card

Reply 8 of 109, by FesterBlatz

User metadata
Rank Member
Rank
Member

Nope, I'm not certain. But since the CPU's I/O is in the 3.3v voltage domain, and the WB/WT# is an INPUT pin, I assumed VCC3 to be the correct power plane to use.

I couldn't say about the BIOS mod, but that's an interesting prospect. However, assuming Asus was aware of the issue and blamed it on a hardware bug, if it were fixable in the BIOS wouldn't they have offered a BIOS update to fix the issue?

Last edited by FesterBlatz on 2017-11-29, 12:03. Edited 1 time in total.

Reply 9 of 109, by meljor

User metadata
Rank Oldbie
Rank
Oldbie

Wow! Amazing news! I have most revisions but also the 1.05 P5A/B and 1.06 P5A, will certainly test this! Can't i simply use a wire for this or will it destroy something (sorry, don't know anything about those components). Would be nice to have the latest and fastest revision with a ''+'' cpu and also 512mb fully cacheable ram.

About the k6plus website: YES absolutely. Visiting often and also find the modified bios section very helpful.

lazibayer wrote:

Are you sure it's VCC3? Because WB/WT# pin is located on the VCC2 plane.
I wonder if it's possible to mod the BIOS so the board will constantly pull high WB/WT#.

If it was that simple Asus would have probably fixed it quickly? As far as i know other brands didn't have these problems. Most boards/brands were sold with older rev. chipsets but i've heard of GA-5AX boards that did have the G chipset and no problems with ''+'' cpu's. Some late GA-5AX rev. 5.2 boards even came with H rev. chipset.

Maybe Gigabyte DID read the memo from ALi about a change in pinout 🤣

asus tx97-e, 233mmx, voodoo1, s3 virge ,sb16
asus p5a, k6-3+ @ 550mhz, voodoo2 12mb sli, gf2 gts, awe32
asus p3b-f, p3-700, voodoo3 3500TV agp, awe64
asus tusl2-c, p3-S 1,4ghz, voodoo5 5500, live!
asus a7n8x DL, barton cpu, 6800ultra, Voodoo3 pci, audigy1

Reply 10 of 109, by FesterBlatz

User metadata
Rank Member
Rank
Member

I'd not recommend directly wiring without the use of a resistor. I haven't looked at the M1541 datasheet (assuming one is even available), but if the chip's I/O pins are of the open-collector or MOS family (likely) as soon as that pin leaves a high-impedance "off" state and tries to assert a "low", the M1541 would almost certainly be damaged.

The 1K resistor is used as a current limiting device that allows the M1541 to still pull WB/WT# low without damaging anything.

Reply 11 of 109, by ElementalChaos

User metadata
Rank Member
Rank
Member

Talk about perfect timing! I was hunting for a SS7 board the past few days and landed on the P5A-B, found a good deal on one but unfortunately was a Rev 1.05. I just went ahead and bought it and will absolutely be trying this out and updating with results.

Pluto, the maxed out Dell Dimension 4100: Pentium III 1400S | 256MB | GeForce4 Ti4200 + Voodoo4 4500 | SB Live! 5.1
Charon, the DOS and early Windows time machine: K6-III+ 600 | 256MB | TNT2 Ultra + Voodoo3 2000 | Audician 32 Plus

Reply 12 of 109, by FesterBlatz

User metadata
Rank Member
Rank
Member

I just updated my original post with a picture of the mod.

This may be the wrong place to ask this, but I'm curious if anyone could point me in the right direction for a copy of CPU-Z that won't crash on a SS7 machine running Windows 98. I've tried several versions that all claim to be 98 compatible, but they all either refuse to run or blue screen....

As for benchmarking the modded motherboard, aside from Quake 2 time demos, what would be a good toolset for comparing the performance of this ALi Aladdin V K6-2+ setup to my MVP3+ K6-3+?

Reply 14 of 109, by mrau

User metadata
Rank Oldbie
Rank
Oldbie

this is also quake 3 territory; sandra for some synthetic stuff also (memory, disk performance and so on) - in my memory the k6 machine was always a little bit power depraved in that regard
edit: also do phils benchmark suite if relevant

Reply 15 of 109, by Stedman5040

User metadata
Rank Newbie
Rank
Newbie

For benchmarks try superpi 1M. Also look at the Everest MR/MW/ML for the memory subsystem. Bios settings and WPCREDIT adjustments can help with getting the best performance out of your system. You should find some help on the K6plus site. For the bog standard K6-III+ cpu running at 550MHz on the P5A v 1.04 I can get a superpi 1M score of about 4m57s under Win98SE with 512mb memory at 2-2-2-5. Using the same multiplier and increasing fsb to 112MHz with external cache still on(cpu at 618MHz) it comes down to 4m33s which is quite fast for a SS7 set up. Hoping the G rev chipset might be faster. Although the manual states the available overclocked fsb as being 110MHz my board reported it as being 112MHz.

K6 III+ 500
Epox MVPG2
512Mb Hynix CL2 SDRAM
40MB WD HDD
Creative GeForce2 Ti
CMI8738 Sound card

Reply 16 of 109, by meljor

User metadata
Rank Oldbie
Rank
Oldbie

Thanx for adding the picture, makes it a lot easier.

asus tx97-e, 233mmx, voodoo1, s3 virge ,sb16
asus p5a, k6-3+ @ 550mhz, voodoo2 12mb sli, gf2 gts, awe32
asus p3b-f, p3-700, voodoo3 3500TV agp, awe64
asus tusl2-c, p3-S 1,4ghz, voodoo5 5500, live!
asus a7n8x DL, barton cpu, 6800ultra, Voodoo3 pci, audigy1

Reply 17 of 109, by zyga64

User metadata
Rank Oldbie
Rank
Oldbie
FesterBlatz wrote:

This may be the wrong place to ask this, but I'm curious if anyone could point me in the right direction for a copy of CPU-Z that won't crash on a SS7 machine running Windows 98. I've tried several versions that all claim to be 98 compatible, but they all either refuse to run or blue screen....

I've the same experience with K6-2, CPU-Z and Windows 98 (blue screen). But there is also similiar utility WCPUID (from the autor of WPCREDIT)
http://hp.vector.co.jp/authors/VA002374/src/download.html which works in this environment quite well.

1) VLSI SCAMP /286@20 /4M /CL-GD5422 /CMI8330
2) i420EX /486DX33 /16M /TGUI9440 /GUS+ALS100+MT32PI
3) i430FX /K6-2@400 /64M /Rage Pro PCI /ES1370+YMF718
4) i440BX /P!!!750 /256M /MX440 /SBLive!
5) iB75 /3470s /4G /HD7750 /HDA

Reply 18 of 109, by Tetrium

User metadata
Rank l33t++
Rank
l33t++

A lil bit late, but this is a fantastic find! 😁
Another mystery solved! 🤣

Whats missing in your collections?
My retro rigs (old topic)
Interesting Vogons threads (links to Vogonswiki)
Report spammers here!