It is possible that it depends on the card or Phil changed the conf. I don't know. I demonstrated this pneumonia in another thread, but I cannot find that thread. The Serpent Rider was in the discussion - perhaps he can recall the thread. I thought it was the Proserva or a PPRO build, but when I searched for "phil" in those threads, I couldn't find it.

Maybe you can get a different version of Quake binary or tweak arguments to get slightly better results. But the beauty of the Phil's pack is it allows for a easy direct comparison. So whenever I share Quake/DOOM/PCBench/... results I always use Phils pack and urging others to do the same. The point is not about achieving a higher number, it is about measuring if the same thing is running faster.

As a big fan of the channel I enjoyed every second of CPU Galaxy's video and congrats to the 180 MHz CPU and amazing result with the fast PCI bus with that cool hot switch trick. But still believe excluding other 5x86 Socket 3 CPUs is like holding a tuning car competition, but restricted to 4-cylinder engines only 😀

The hot swap method of getting the PCI bus at 40 MHz on the M919 has been around for some time. Since the 33/40 MHz is the difference between a jumper on or off, people usually use the turbo button switch to switch to 40 MHz once past POST.

I personally don't think Cyrix 5x86 chips should be excluded from this competition. They are more a 486 than a Pentium in my opinion. I bet I can get one of my IBM's to run at 150 MHz long enough to complete the Quake timedemo.

I kinda want to try that as well. I was messing with one of my boards months ago and accidentally set it to 50mhz fsb with one of my Cyrix 120mhz CPUs and it posted.. pretty sure it was at default voltage as well.

I had a IBM 5x86 running at 150 MHz in DOS pretty well, but it died suddenly. Point being that if I want to risk another CPU, I can probably try for Quake. I'd need to setup proper cooling though. Photos from my archives:

I think you can forget about any stability at all with 150MHz. Cyrix had to switch to a different manufacturing process to do the 6x86-P200+, and even then it was too hot and they had to try again with the 6x86L.

I thought it was the Proserva or a PPRO build, but when I searched for "phil" in those threads, I couldn't find it.

Re: Improving Quake performance on POD83.

But still believe excluding other 5x86 Socket 3 CPUs is like holding a tuning car competition, but restricted to 4-cylinder engines only

You can't pick a batch of Cyrix CPUs and test the best one for overclocking. Arguably, even vanilla DX4 from Intel aren't as available.
So chances of achieving good results are quite slim, at least without below zero Celsius setup.

Wow, you are good. I don't know how you found that amongst the mess of threads. Here's the photo showing the discrepancy in results, Re: Improving Quake performance on POD83.

mpe wrote on 2021-04-01, 07:26:

But still believe excluding other 5x86 Socket 3 CPUs is like holding a tuning car competition, but restricted to 4-cylinder engines only 😀

I mean, there is a reason why they do that in professional racing :p

this is about the fastest 486 after all, and not the fastest socket 3 setup, right? ^^

doesn't mean it's not interesting to see what's possible.

Both are good sport.

My only concern (and maybe complain) is the lack of clarity around "how real the computer used in the test is".
Is it a one trick pony - minimal configuration and potentially not completely / long term stable, or is the real deal - fully maxed out system running with full gear.
To use the sport car analogy - can it do more than one lap ?
The first case is kind of meh, if you ask me.

Still.
It looks like cpu galaxy achieved highest score with 486 cpu atm.
But not clear what is the highest score with a socket 3 cpu of any kind, considering MPE's point about staying true to Phil's vanilla setup.
Do we know ?

obviously hunting the highest possible number isn't going to be long term stable ^^

But not clear what is the highest score with a socket 3 cpu of any kind

Obviously that would be POD. With some modding, 125 Mhz is probably possible.

To use the sport car analogy - can it do more than one lap ?

Well, I've tested even Quake 3 on 180 Mhz 486, so it's possible to achieve.

The Serpent Rider wrote on 2021-04-03, 16:12:

Obviously that would be POD. With some modding, 125 Mhz is probably possible. […]
Show full quote

But not clear what is the highest score with a socket 3 cpu of any kind

Obviously that would be POD. With some modding, 125 Mhz is probably possible.

To use the sport car analogy - can it do more than one lap ?

Well, I've tested even Quake 3 on 180 Mhz 486, so it's possible to achieve.

Thanks. I have seen your post - great stuff. Looks pretty stable - the real deal ! 😁

CPU Galaxy wrote on 2021-03-27, 14:44:

My new video is online and I did beat your scores. 😀

https://youtu.be/qPRSHI4JoWI

Damn.

I tried to reach the same heights with my M919 but my cache module just didn't want to play nice. It isn't completely stable at 3-1-3 and at 2-1-2 it completely gives up.
The highest score I could get in 3DBench 1.0c was 111.1 with an Asus V3000 which, at least, took the 60MHz PCI without complaining. The VGA transfer rate wasn't stellar, though, it hovered around (~27000 chr/ms in Landmark) what my slower Abit PB4 board gets at 40MHz FSB=PCI. The faster one at 40MHz completely blows it away. It's also faster in Wolfenstein at 160MHz than what this M919 could do at 180MHz, having the cache at 3-1-3 and the RAM at 1WS. The PB4 can achieve 103fps in 3DBench 1.0c, btw.

Cyrix/IBM PQFP 5x86 came with aluminium heat spreaders buried in molded epoxy while AMD chose copper heat spreaders. Makes some difference.

I'm still content with running my IBM 100MHz at 120MHz. It's not the fastest but I feel it's safe and quite fast enough for a 486 😁

120MHz is easy enough. 133MHz is where the fun starts.

CPU Galaxy is raising the bar -- 200 Mhz 486, 66Mhz PCI, Peltier cooling, 21.6 FPS

https://www.youtube.com/watch?v=LI1_RlVLhu8&t=1333s

Then they do a 29.3 FPS GL Quake @ 640 x 480 just because they can.

cyclone3d wrote on 2021-04-01, 08:49:

I kinda want to try that as well. I was messing with one of my boards months ago and accidentally set it to 50mhz fsb with one of my Cyrix 120mhz CPUs and it posted.. pretty sure it was at default voltage as well.

It's a couple of decades since it happened, but when I was trying to get the most of out my Cyrix, I tried it at 50Mhz bus speed, and what seemed to happen is that it ran, but automatically dropped to a 2.5x multiplier, so it was running at ~125, also the motherboard seemed to have kicked in wait states for 50mhz operation, so it actually seemed to be 3-5% slower than running at 3x40. In no other way could I seem to get this 2.5 multiplier to kick in though, other than trying to boot it at 50mhz bus, I don't know if it was the board doing it, or the CPU doing it. Now I did discover the undocumented 60Mhz, this is on an Amptron board that is a close cousin to the M919, and had it at 2x60, which benched higher. Funny thing is I got into multiple arguments about that over the years, people telling me it wasn't possible, but now I can point them to ppl who've done it on Vogons and shut them up 🤣. I found the 66Mhz setting also, but stability was extremely poor on that, could get a DOS prompt but nothing would run more than a few seconds. I was unable to determine whether it was CPU heat, chipset, graphics or I/O crapping out because all of them were on the raggedy edge really.

The Cyrix 5x86 series did not have any fractional multipliers. If you saw the BIOS screen or some other software say "125 MHz", it was mistaken.

Amptron was a US branding for PC Chips. I believe they are one in the same.

I don't think any M919 can do 66 MHz reliably. 60 MHz FSB was max if you want stability from what I could determine.