I've a Slot 1 P3 600 that runs just about all my dos/win9x games just fine as is, no need for slow down utils at all. stock 100Mhz FSB

Out of your list I can confirm Raptor Dune2 Warcraft 1 work fine Jazz will need the pascal patch.
Latest game it runs is Need for Speed 4 Next game in my list is GTA3 which does struggle but better off on XP era rigs anyway.

So if it was me I wouldn't worry about your S370 boards limitations, your already doing better then me at 133Mhz FSB.

I would think the V3 is fine for graphics , see how it goes and upgrade only if you feel like its holding back some of the D3D titles.
Not sure off the top of my head about your dos 3dfx games and v3 compatibility though? someoen else will need to confirm that side of things

Right so I had a look at my stock of VT82C693A mainboards and tested them with the C3 1200. I also went into bios on each of them looking to see if any have the ability to set FSB from bios. While doing this, I discovered a few things...

- Gigabyte GA-6VXE7+ - this board is REV 1.0 and does not officially support any CPU with a multiplier higher then 8x - that's fine for most coppermine chips, since they top out at 1100MHz, but useless for my C3 witch requires a multipier of 9x. Multiplier issue aside, the board posts with the VIA chip but will not correctly identify it. Post screen sais "Intel Pentium III ;86 mhz". I didn't try booting into dos and checking with speedsys or HWinfo, but it is possible the CPU would work regardless. No FSB option in bios.

- Asus CUV4X-M - for some reason I remembered this board had an ISA slot. It does not. In fact it doesn't even use the VIA 693A chipset like I thought it did - it uses the newer 694X chipset. It's a decent enough board and the 1200MHz C3 is correctly identified and functions as it should on it, but lack of ISA puts it out of the race. Might as well try a socket A Geode build if ISA is off the table. This board is part of a tiny SFF mATX build I made a while ago, witch used to run a Voodoo 4 and a 1100MHz Coppermine Pentium 3. The CPU is still in there, and the PC is complete apart from the V4 witch I sold a while ago. CMOS battery was dead as well since I haven't used this PC in over a year. I cleaned the PC up after testing and installed a Geforce 3 Ti200 in place of the voodoo 4.

- ECS P6VXA - again, no ISA slot and VIA 694X chipset. This board does however allow setting the FSB from BIOS, but lack of ISA disqualifies it.

- Iwill VA133 Plus - this is a great little board. It's one of the few boards I own with jumpers for setting v-core and suports a wide range of voltages, down to 1.2v, something really useful if you want to run a pin modded tualatin celeron. I've tried tualerons up to 1400MHz in it and it ran perfectly stable. It was part of my main socket 370 win98 PC for over a year, running a 1000MHz tualeron at 1333MHz (133x10). This was 4-5 years ago. It was superseded by an ABIT ST6 and a 1400MHz 512KB pentium 3. Unfortunately, I managed to damage the board w/o even realising.... the thing has a couple of tiny SMD capacitors mounted right under the CPU socket cooler clips - SMDs witch I managed to tear off at one point, probably when I took it out of use. The board no longer posts. I'll get around to fixing it when I'm finished remodeling. I also remember it has an AUTO jumper setting for FSB which did allow me to set 66/100/133 from bios.

- Shuttle HOT-687V2 - this is what I've been running my 1200MHz C3 in for the past two months. It runs great, it's perfectly stable, but lacks the ability to change FSB from bios* and fitting a longer ISA card is problematic, since the back of the card runs on top of the front panel header. I worked around this issue by bending the pins 45 degrees up and down, making room for my sound blaster 16. Now about the asterisk above - the manual states that if you set the jumpers for 133Mhz, you should be able to set FSB from bios - but the selection is very limited and likely implemented only for light overclocking as it does not allow setting anything lower then 112MHz. Despite this, this board is probably my best candidate going forward since it has physical jumpers which I could route to a set of switches installed on the front of the case...

- Lucky Star 6VA693AM - this is another interesting board, but it doesn't properly support the C3 Nemiah. It allows FSB setting by either jumpers (3 jumpers, very very easy to route to some switches on the front of the case) as well as an AUTO setting witch lets you pick FSB from bios. It also supports the 1.45v - selectable by jumpers - that the 1200MHz C3 requires. The board works perfectly fine with my 800MHz C3 Samuel, but doesn't correctly identify the 1200MHz Nemiah. Post screen states "VIA C3 300MHz-A", and the system locks up after displaying the POST table. It's possible there are BIOS updates that support the Nemiah core for this board, but considering it's made by a manufacturer only popular in eastern europe and russia, it's rather unlikely that someone bothered.

This is every VIA socket 370 board I remembered owning, apart from some oddball AT form factor motherboards form Acorp and Zida. The Zida boards run the intel 440FX and LX chipsets respectively, and the Acorp boards only support older VIA C3 processors - not the Nemiah core. I also have a slot 1 / socket 370 combo ECS board but that's using the older apollo pro chipset with no support for the Nemiah or FSB 133 chips.

Now on to the slot 1 platform!

For my slot 1 build I wanted to use a board that supports the 2x multiplier and has decent CPU settings in bios. This limits me to my ABIT BH6, as it its the only one of my slot 1 boards that ticks all the boxes. I also have a couple of BE6-II boards, but for some reason the lowest multiplier these allow for is 3x. I also have a ZIDA AT form factor slot 1 board - currently running a 333Mhz pentium II, but it's an LX chipset board which means overclocking a pentium II to 500MHz is extremely unlikely (although it does support multipliers as high as 11x).

Here's the BH6:

Unfortunately the RTC battery holder had the metal arm that keeps the battery in place snapped off, but after spending 20-30 minutes sifting through boxes looking for my soldering equipment, I managed to install a replacement battery holder.

My selection of 333Mhz pentium II chips includes a SL2VT and SL2KA. I also have a 266Mhz SL2HE chip witch is unlocked and will happily run at 333Mhz, but requires 2.85v for that. 500MHz on this chip is not possible, although I did manage to get it up to 400MHz (4x100) speedsys stable by feeding it 2.95v. Don't know if it's stable in windows or under gaming at 400Mhz, but that matters little since for this build I'd like to run stock voltage or a maximum of +0.5v.

The SL2VT is stable at 500Mhz, but is unfortunately multiplier locked. By default it runs at 2.1V and for 500Mhz operation all it takes is 2.15v. Can't find any info on it on CPUworld, only an e(vil)bay link for an auction.

The SL2KA runs at 2V, is unlocked and can run at 133Mhz, but I have difficulty taking it to 500MHz. For 500Mhz operation I had to increase voltage to 2.1V and increase L2 cache latency. It is speedsys stable, but I'm not happy about the voltage requirements or the cache latency increase.

There's still hope yet. I know I have a 333Mhz pentium 2 someware, which is not only unlocked, but also runs at 500MHz with default latency and stock voltage (on the BE6-II at least). Don't remember where I put it - it's likely hiding in my dell optiplex or compaq deskpro en.

Util then here are some pics of the SL2KA posting at 133 and 500MHz. Speedsys and other benchmarks will follow, as soon as I find some more time.

chinny22 wrote on 2023-01-05, 17:53:

I've a Slot 1 P3 600 that runs just about all my dos/win9x games just fine as is, no need for slow down utils at all. stock 100M […]
Show full quote

I've a Slot 1 P3 600 that runs just about all my dos/win9x games just fine as is, no need for slow down utils at all. stock 100Mhz FSB

Out of your list I can confirm Raptor Dune2 Warcraft 1 work fine Jazz will need the pascal patch.
Latest game it runs is Need for Speed 4 Next game in my list is GTA3 which does struggle but better off on XP era rigs anyway.

So if it was me I wouldn't worry about your S370 boards limitations, your already doing better then me at 133Mhz FSB.

I would think the V3 is fine for graphics , see how it goes and upgrade only if you feel like its holding back some of the D3D titles.
Not sure off the top of my head about your dos 3dfx games and v3 compatibility though? someoen else will need to confirm that side of things

Thanks for your input!

Jazz is a little iffy on anything faster than 400mhz... on my K6-3 build, at 200Mhz with L2 cache off, the game runs perfectly, but if I forget to launch setmul the game has occasional stuttering - like there's an issue with scrolling. Same with commander keen. At 500Mhz warcraft 1 will not correctly detect my sound card (An ESS based Guillemot Maxi Sound 64) and scrolls way to fast.

I don't plan on running DOS 3dfx games... in dos. The glide games I mentioned have unofficial glide patches that require win9x and run fine on a voodoo 3, voodoo 2 and voodoo bansee. Carmageddon has an official dos glide port/patch but it will only run on a voodoo 1 / pentium 200 or voodoo 2 + pentium 133 (providing you replace glide.ovl in the game folder with the one found in the voodoo 2 driver folder). The windows glide version runs on any 3dfx card I tested it on. Uprising is a win9x game, and usually requires a voodoo 1, but will run on a voodoo 2 with patch 1.05. I haven't tested it on a voodoo 3 tough... Pandemonium on the other hand... I haven't been able to run pandy on anything other than a voodoo 1.

I'll take your word on Jazz, It's not a game I play often so not have played it since upgrading from P2 400 to P3 600.
Never really got into Keen but Warcraft I do know about at least!
Haven't experienced the sound issue? What version are you running? I've got 1.22 (CD-ROM version) but patches exist to bring you up to 1.21
https://www.blizzardarchive.com/pub/index.php?id=war1
It could be all my P2 and faster rigs have a SB16 or AWE so could also be a bug in setup related to ESS and not creative cards.
Scrolling is fast War2 suffers the same problem I got round this by clicking on the minimap, you do get used to it.

chinny22 wrote on 2023-01-05, 17:53:

Out of your list I can confirm Raptor Dune2 Warcraft 1 work fine Jazz will need the pascal patch.

Raptor works fine if you copy the install folder from somewhere but if you try to install it with original disks you have to slow down your system or it will fail. Haven't tried the CD version.

Ok, I'm back. The slot 1 system is still under experimentation. I managed to find a 333MHz CPU with an unlocked multipier but it's not stable at 500MHz unless I pump an extra 0.15v in it and drastically increase cache latency. Even then I'm not sure it's 100% stable. The CPU that can do 500MHz at stock voltage is multipier locked.... so it's either 333MHz or 500MHz.

For now I put my Cyrix all in one build together using the Shuttle HOT-687 Ver 3.1 motherboard which seems to support the C3 Nemiah. I still have some problems with the build - first and foremost being very poor memory performance. In Aida64 the system gets 300-330MB/sec memory read, witch is very low for a VT693A. This chipset should be getting someware between 650 and 850MB/second. Maybe the shuttle board does not fully support the Nemiah CPU. On POST, the board is identifying the CPU as VIA Nemiah 1193MHz - which is exactly the frequency the chips runs at, and I'd expect the POST screen to read VIA C3 1200, not a string followed by the CPUs current clock speed.

On the bright side, I discovered the CPU will happily run at 1333MHz with no increase in voltage - just by setting the multiplier to 10x with setmul. It works in DOS as well as Windows 98. I tested it at this speed for 5-6 UT99 matches, seems perfectly stable, if slow.

Thread update.

Still looking for that "one motherboard" that will play nice with my 1200MHz C3 nemiah - no luck so far. Looking up VIA 694X and 694T boards on theretroweb, there were quite a few made, but VIA CPU compatibility is not mentioned in the manuals half the time, let alone Nemiah core compatibility.

A recent lot of retro hardware from a recycling center got me a MSI MS-6309 rev 1.1 witch I had very high hopes for - unfortunately it turns out to be an incomplete / budget implementation of the VIA 694X chipset. This is because the 6309 does not support the Nemiah core whatsoever, and to top it off, it only supports Vcore down to 1.75 - in theory. I say in theory because bios states that at 1.75V, an actual 1.83-1.84v are going to the CPU, and my multimeter read close to that at 1.82v. This means this board is incompatible with the 1.45v nemiah from the get-go, and it's also not a good candidate for running pin modded Tualatin CPUs. I had a go with a 1000MHz Celeron-S witch is fully stable at 1333MHz / 1.6v on most boards - it gets to 50C very quicly on the MSI, is unstable even at 1000MHz because of the very high Vcore this board provides.

To top it all off, the MS-6309 has a really annoying clicking bug when paired with a SB16 or SB32 (an AWE64 seems to work OK, as well as other ISA cards I tested) - consisting of an audible "pop" or "click" showing up on top of whenever sound I'd try to play. This "pop" is ritmic, independent of system load or HDD activity (so it's not interference) and plays about once every second.

This is a real shame, because the MS-6309 is one of the fastest and most stable implementations of the 694x chipset I've come across so far:

===================================
P3 1100MHz FSB100 - msi via 694x - 256mb PC100 CL3-3-3-6 - Voodoo 3 3000 AGP
===================================

ut99 800x600 - dm codex - 81.29 avg
Quake 2 timedemo 1 - 118.3 fps

aida64 2.5.2:
memory read cl 3-3-3-6 100mhz - 767mb/s
fpu julia - 240 pts
fpu vp8 - 181 pts

===================================
P3 1000MHz FSB133 - msi via 694x - 256mb PC133 CL2-3-3-6 - Voodoo 3 3000 AGP
===================================

ut99 800x600 - dm codex - 80.52 avg
Quake 2 timedemo 1 - 118.9 fps

aida64 2.5.2:
memory read cl 2-3-3-6 133mhz - 833mb/s
fpu julia - 238 pts
fpu vp8 - 187 pts

Those results are not bad at all, in fact they are very close to my 440BX baset Abit BH6:

========================
P3 1100MHz (1133mhz fsb 103) - Abit BH6 + Slotket - 256MB PC103 CL 2-2-2-? - Voodoo 3 3000 AGP
========================

ut99 800x600 - dm codex - 80.52 avg
Quake 2 timedemo 1 - 117.8 fps

aida64 2.5.2:
mem read cl 2-2-2-? 103mhz - 813mb/s
fpu julia - 248 pts
fpu vp8 - 192 pts

The thing is, while checking MSI MS-6309 compatibility with VIA chips, I ran some tests using a VIA C3 800A based on the Samuel core, witch needs 1.75v, and decided to compare the results with the Shuttle HOT-687V2 I had in my office retro PC, and found that the Shuttle mainboard's performance was lacking to say the least. Using a slower CPU, the MSI managed to score over 700MB/sec memory read with the 800A, and very very respectable FPU scores in AIda64, close to a 600MHz Celeron! UT99 was very playable on the MSI + C3 800 combo, while on the Shuttle + C3 1200 it's a stuttery mess. Quake 2 managed to score a respectable 78 fps on the MSI + C3 800 at 800x600, witch is a little higher then the Shuttle + 1200MHz C3.... Unfortunately I did not think to write down the results for the MSI + VIA C3 800 combo, but I will be re-running the tests in the next couple of days. Since I already had both systems apart and did not have the 1200MHz VIA chip on hand, I quickly assembled a test rig using the shuttle board and a pin-modded 1000MHz Celeron-S (tualatin). Here are the results:

================
tualeron 1000mhz fsb100 - Shuttle HOT-675v - 256mb pc100 cl 3-3-3-6
================

ut99 - 51.29 fps
quake 2 = 79.6

aida64 2.5.2
mem read - 264mb / sec
fpu julia - 224 pts
fpu vp8 - 162 pts

tualeron 1000mhz fsb100 - shuttle - 256mb pc133 cl 2-2-2-6
aida read - 266mb/sec

These results lead me to believe that the Shuttle board is a very, very bad implementation of the 693A chipset, as it scores much slower then it should with a 1000MHz Celeron-S. Quake 2 should score over 100 fps at 800x600 and UT99 should score around 70-80 fps. Memory read is half of what it should be on the 693A chipset, when comparing the shuttle board to my iwill witch uses the exact same chipset, and managed to score 533 Mb/sec in aida64 2.5 witch PC100 SDRAM.

All of these tests have finally made me understand why some users simply despise VIA chipsets - a lot of implementations are just garbage, and that is purely the manufacturer's fault. I can't get around how Shuttle managed to f#(% up the 693A's performance so badly that it scores half of what it's supposed to in memory tests.

693A always had terrible memory performance and is a poor choice for a retro rig. It was possible to reconfigure chipset via wpcredit to fix it and get performance closer to 694X. 694X at 133FSB ~ 440BX at 100FSB. The wpcredit fix made it viable for Windows but not DOS. I'm surprised you have collected so many boards with this chipset instead of 694X or 694T (these are valuable).

AlexZ wrote on 2023-05-06, 07:48:

693A always had terrible memory performance and is a poor choice for a retro rig. It was possible to reconfigure chipset via wpcredit to fix it and get performance closer to 694X. 694X at 133FSB ~ 440BX at 100FSB. The wpcredit fix made it viable for Windows but not DOS. I'm surprised you have collected so many boards with this chipset instead of 694X or 694T (these are valuable).

I'm starting to notice that, yes.

The thing is, I found not all 693A boards are this bad (but most are). For example the Iwill VA133 Plus gets close to the 694x in performance - very close if you use a 1000MHz/133fsb pentium 3, and the Iwill can go down to 1.4v in bios allowing for hassle free pin-modded tualatin use. I've had chips as fast as 1400MHz in there, and at that speed it blasts most other Pentium 3 builds out of the water. Off the top of my head, using 133MHz CL 2-3-3-6 memory it would score a bit over 650mb/sec in aida64 memory read speed, which is not wow, but respectable. I even got it to run a 1200MHz Celeron-S at 1600MHz (12x133) / 1.6v at one point, but it wasn't fully stable in the long run. Too bad I damaged it - and have no idea what values the SMD caps near the CPU socket would have....

I found a page describing WPCREDIT here: https://www.overclockers.com/wpcredit/ - the guide seems straight forward enough. I imagine the Iwill already has 4 way interleave enabled by default -witch would also explain the performance and why it's rather picky about ram compared to other 693A boards that seem to post with anything I put in the memory slots.

As for collecting the boards - some were free, others were under 10$/board... and up until a few days ago I had no idea there was such a difference in performance between the 693a and the 694x.

694x boards are also quite a bit harder to find - out of 10 skt 370 via boards I stumble across only 2 are 694x and only one of those have an ISA slot...

AlexZ wrote on 2023-05-06, 07:48:

The wpcredit fix made it viable for Windows but not DOS.

nothing a TweakPCI - A DOS utility to view/modify PCI configuration registers couldnt fix I imagine

That guide doesn't list many optimal settings to enable on 693A to get decent memory performance. You can view 693A as a beta version of 694X and it may not be fully stable with all optimizations enabled. I'm only listing those that should be stable:

The best known optimization is: 1-0 Bank Interleave - 4-way
but there are other important settings as well:

Device 0, offset 50:

• CPU Hardwired IOQ (In Order Queue) Size - 4-Level (bit 7 -> 1)
• Read-Around-Write - Enable (bit 6 -> 1)
• CPU / PCI Master Read DRAM Timing - Start DRAM read before snoop complete (bit 0 -> 1)

There are also some settings in device 0, offset 51 that improve performance slightly:

• CPU Read DRAM 0ws for Back-to-Back Read - Enable (bit 7 -> 1)
• CPU Write DRAM 0ws for Back-to-Back Write - Enable (bit 6 -> 1)

with these changes 693A should have almost the performance of 694X. If you use TweakPCI as rasz_pl suggested you should be able to fix performance in DOS as well.

AlexZ wrote on 2023-05-08, 08:47:

That guide doesn't list many optimal settings to enable on 693A to get decent memory performance. You can view 693A as a beta ve […]
Show full quote

That guide doesn't list many optimal settings to enable on 693A to get decent memory performance. You can view 693A as a beta version of 694X and it may not be fully stable with all optimizations enabled. I'm only listing those that should be stable:

The best known optimization is: 1-0 Bank Interleave - 4-way
but there are other important settings as well:

Device 0, offset 50:

• CPU Hardwired IOQ (In Order Queue) Size - 4-Level (bit 7 -> 1)
• Read-Around-Write - Enable (bit 6 -> 1)
• CPU / PCI Master Read DRAM Timing - Start DRAM read before snoop complete (bit 0 -> 1)

There are also some settings in device 0, offset 51 that improve performance slightly:

• CPU Read DRAM 0ws for Back-to-Back Read - Enable (bit 7 -> 1)
• CPU Write DRAM 0ws for Back-to-Back Write - Enable (bit 6 -> 1)

with these changes 693A should have almost the performance of 694X. If you use TweakPCI as rasz_pl suggested you should be able to fix performance in DOS as well.

Would it be possible to edit the BIOS and unhide or enable these tweaks there?