wrote:A 486 won't crack 99 fps, ...
Hehe, on mine it does. 118.1 FPS, using 3DBench version 1.0c. Which leads me to ask, is overclocking allowed? 😉 You could "flag" this using an additional "Overclocked?" column in a table.
Nope, no overclocking 🤣
wrote:Yes, but:
SuperPI requires Win32s and a RAMdisk would take RAM that would be useful for SuperPI.
Good question (even though it technically isn't one 😜), how much RAM does SuperPi use anyway?
Another way "could" be to use other removable media, like ZIP or 2.88MB floppy.
And at any rate, theres not much stopping us from putting 64 megs of memory in a 486 nowdays (or even 32MB in a 386, IF we have the RAM and board that supports them).
Edit:
wrote:Nope, no overclocking 🤣
I was thinking about this. Even though I like to have overclocked results not included in the list of "standard" results, this does pose 2 thingies:
*It's a shame to simply discard these results as from a retro-perspective I find it interesting how very high clocked older gen stuff compares to standard newer gen hardware (like:How does a 486 at 200Mhz compare to a Pentium?)
*Some of the highest clocked CPU's for a given socket are very scarce, but do exist. One example:Someone on this forum recently posted results for his Cyrix Socket 3 5x86-100 clocked at 133Mhz. Technically this is an overclock. However, 133Mhz Cyrix parts do exist but are next to impossible to find. If I were to have the 133Mhz part and post a result, it would be included, but someone overclocking his CPU from 100Mhz to 133Mhz and posting results would not be included, which imo is somewhat unfair.
Same goes for the AMD 5x86-150Mhz and 160Mhz and many other parts I can think of.
To a lesser extend this also goes for underclocked results. It would be a shame not to include, for instance, a Pentium 2 366Mhz.
My suggestion would be to not discard the overclocked results entirely, but put them in a separate category. Example:
Have a Socket 3 category, like Maulwurf already done in his 3dbench thread. This category could also include non-standard results that are not clocked higher then the fastest available processor for that given socket.
Examples:
*486SX @ 40Mhz or 50Mhz.
*Pentium Overdrive 83 running at 66Mhz (1x multi, 1/2PCI divider and 66Mhz FSB)
*5x86 @ 100Mhz with FSB=25Mhz and 4x multi.
*Pentium 2 366Mhz (semi-unlocked P2-400 with 66Mhz FSB and 5.5x multi)
*Cyrix MII PR400 (standard 95Mhz FSB 3x multi) running at 3x100Mhz (very slight overclock, but the PR433 is QUITE hard to find, but exists).
*K6-III+'s: These CPU's were branded with different speeds and voltages (400Mhz 1.6v, 450Mhz 1.7v, 500Mhz 1.8v/2.0v, 550Mhz 2.0v). These processors are basically exactly the same, but underclocked for laptop and embedded use. One seller is selling the 400Mhz ATZ 1.6v ones for cheap, but the higher clocked ones are MUCH harder to find (and impossible for the same price).
Next would be a Socket 3 non-standard category with results clocked higher then was normally available (a good boundary would be anything clocked over 133Mhz, even though very few 5x86's came in 150Mhz and 160Mhz examples).
And personally I've always wanted what a Pentium MMX at 300Mhz (3x100) performs like compared to the Cyrix MII 300Mhz, K6-X/300, Pentium 2-300, Celeron-300, Celeron-300A and maybe a couple others 😀
Re: Tetrium's OC'ing thoughts = Yes. And also, a lot of folks seem to have their 486s set to OC'd status. Eg an Intel DX4-100 seems to be commonly run at 120 MHz, and an AMD 5x86 P75 seems to be commonly run at 160 MHz.
IMHO, these commonly run configs would be useful to include. Just flag them up as "overclocked", and they can happily coexist side-by-side next to the "normal" results in any table(s), so that everyone can see at a glance what is being overclocked, and what isn't.
i'm almost inspired to strip a DOS quake engine to the minimals just for running some dynamic lights on a textured bsp to benchmark with. cant get around the fpu requirement though
Simulating a CPU one doesn't have is totally fine. I'm assuming the FSB won't be touched, simply the multi changed. E.g. running a 200MMX at 233.
I would do the same thing with a K6 to be honest. Pick the highest model one has, and simulate all the lower models.
But I don't want results past their specifications. I would like the database to be for reference use only and as compact as possible.
Plenty of existing threads covering overclocking, RG100 you have a ton of these and people can go there and check out your settings and your results.
IMO overclocking old parts doesn't make much sense, as you can just build something faster. E.g. why overclock an AMD 5x86 when you can get a Pentium? I get that it's fun and interesting and there is a place for this, but not in a reference database.
Why not make a overclocking list RG100? All the overclockers can go nuts and try to beat each others scores.
Anyway back on track: Anyone had any luck with Landmark Speed Test or the other benchmarks?
The FSB will be touched however, especially on pre-P3 CPU's.
The P2-366 isn't official, it's just a semi-unlocked P2-400 that runs at 4x100 but made to run at 5.5x66.
Any of the K6-X's can be run at 66Mhz or 100Mhz FSB. Many chips even support 2 different FSB's officially.
To me it's not about overclocking for top position, it's also about "what if?". I just like to fill in the gaps, so to say.
And it's also part of history. People overclocked their CPU's so they could keep up with the increase in CPU performance, or to use more modern software.
Well, at least I'll have the top position for the Cyrix 5x86, since I'm the only one who owns a 133Mhz part 😁
I might make a private database however, maybe the idea of someone else maintaining the unofficial database isn't such a bad idea anyway 😉
>Norton SI 8.0 wouldn't fit on the second disk (3dbench2 is larger)
--8086 Create an executable that works on any 8086 CPU.
What about Trixter's TOPBENCH?
