gerwin wrote:

Now, Feipoa wrote somewhere that the original Cyrix 6x86 had a rare 1.0x multiplier setting. Unfortunately the 6x86 is not that desirable, if only because it gets rather hot while doing nothing fancy. But I recently sourced a 6x86L PR200 split voltage model to try, and that one also supports the 1.0x multiplier reliably. At 60 MHz FSB SpeedSys says: 1.0x = 40 pts, 2.0x = 80 pts, 3.0x = 120 pts. That is easy enough to remember 😀. The point is that 40 pts equals a 486DX/4-100, at least in speedsys. It manages these 1.0x and 2.0x speeds at a mere 2,2 Volt. Disabling the L1 cache lowers the scores to around one third / one quarter IIRC.

Some 6x86L seem to mention only 3,3V on the chip, I wonder if these are still split voltage.
In addition I am still wondering if there are actual Pentium MMX chips that do 1.5x like the original Pentium, and how they are identified.

Your post got me wondering. I once read that the Pentium Overdrives, with fan detached, would go to use a 1x multiplier, but a search revealed nothing about the Socket 5 Pentium Overdrives, only the 486 upgrade chips.
Another thing I should test out, but that'll have to wait till my test setup is finally done.

I will test the chip while building my Pentium MMX retro rig (which is kinda on hold because of RL issues) and see what it does with it's fan removed.

I have one or two socket 5 boards here.. but dont going to use them, because they arent in the line of computers iam will going to use.. I should be actually stupid to have two pentium 1 systems if one can support all the games i want to play.. My biggest annoying point is, that some of the board having to Y2K bug.. I only like the boards if they are 100% solid and working..
So i leave it on the background as spare part. And yes my board(s) do have cache on the boards.

I know Socket4 PODP5V133 overdrive DOES NOT revert to a 1X multiplier with the fan removed, as I have one missing the retail fan.

If I had a Socket5 motherboard that had a killer feature (dual CPU, fast memory etc) there would be nothing stopping me from using it. You can get around CPU limitations by using a powerleap or evergreen adapter. That being said, I have no Socket5 boards simply because I've never been offered one. I guess they're far less common.

Anonymous Coward wrote:

I know Socket4 PODP5V133 overdrive DOES NOT revert to a 1X multiplier with the fan removed, as I have one missing the retail fan.

So what will happen if the fan of such a CPU is removed? Will it simply refuse to boot or will it simply sit there and overheat?

It boots at the normal speed. I didn't want to find out what happens if it overheats, so I had a 3rd party fan attached at all times.

Anonymous Coward wrote:

It boots at the normal speed. I didn't want to find out what happens if it overheats, so I had a 3rd party fan attached at all times.

That is what I expected. How can a processor know at the first seconds of booting, what is, or is not, stacked on top of him?
EDIT: oh. wait.. the overdrive fan is 5V 'socket powered'.

retrofanatic wrote:

I find that socket 7 boards are better because of what is explained in the link above and that they are more readily available.
Having the option on a socket 7 to go with a 3.3v cpu (thus reducing heat) is an advantage as well as having the option of using an MMX enabled chip (if I use windows 3.11 or other windows versions).

Is that a typo?. Socket 4 is 5 Volt. Socket 5 is 3,3V. Socket 7 is <=3,3V. Super Socket 7 defines a 3rd multiplier pin + 100MHz FSB.

We had a Compaq Presario 7170 - Pentium 90 in the past. It was great for a while (Command&Conquer!). Upgrading the video card had its limits. It turned out that it did not support 'IRQ for VGA'. So the best it could run were some cards that tolerated this, being the S3 Virge or a Riva-128. I went with the Virge. Actually tried Half-Life on it (very very low settings) just to connect to another PC for multiplayer.
Despite the initial appearance as a multimedia PC for the future; for serious things this platform was short lived, Contrary to Slot 1 for example. I still feel sorry for a friend, who got himself a brand new Pentium 60 using all his past paperboy earnings.

About the Cyrix 6x86(L) 1.0x multiplier. My 6x86MX does not have this setting, but the later Cyrix MIIv does. Unfortunately the MIIv will lock up at 1.0x, either immediately or after 10 minutes. I wonder what happens with some CPU's that they become unstable at very low clock speeds. VIA C3 has it too.

gerwin wrote:

retrofanatic wrote:

I find that socket 7 boards are better because of what is explained in the link above and that they are more readily available.
Having the option on a socket 7 to go with a 3.3v cpu (thus reducing heat) is an advantage as well as having the option of using an MMX enabled chip (if I use windows 3.11 or other windows versions).

Is that a typo?. Socket 4 is 5 Volt. Socket 5 is 3,3V. Socket 7 is <=3,3V. Super Socket 7 defines a 3rd multiplier pin + 100MHz FSB.

No, that wasn't a typo, just made an error..thanks for correcting me...I was thinking about socket 4 instead of socket 5...I apologize for the dumb mistake...it was pretty late at night...

In light of this new news (new to me), I do like socket 5 better now! 🤣

One thing I have been wanting to try is to find a Socket 5 or 7 or SS7 board that has a 1x multiplier and 75mhz FSB and run a Pentium 75 at full 75hmz FSB speed and see how it behaves. I have not gone though all of my socket 5 and 7 boards to try undocumented settings yet, I have too many projects going on right now.

Tiremaster400 wrote:

One thing I have been wanting to try is to find a Socket 5 or 7 or SS7 board that has a 1x multiplier and 75mhz FSB and run a Pentium 75 at full 75hmz FSB speed and see how it behaves. I have not gone though all of my socket 5 and 7 boards to try undocumented settings yet, I have too many projects going on right now.

The Pentium 75 doesn't have a 1x multiplier though. It's a shame it appears theres no way to get a Socket 5/7 Pentium to run on a 1x multiplier. Would've been interesting to run one @ 100MHz with a 1x multi 😜

Unknown_K wrote:

Same reason people don't love the early 486 boards, they cannot run the top of the line CPU in one. […]
Show full quote

Same reason people don't love the early 486 boards, they cannot run the top of the line CPU in one.

Personally if you want to run an early Pentium the rare Socket 4 P60/P66 are the way to go. Very unique setup that cost a mint when new.

Most of what you find in the wild (outside of a brand name system) is Super Socket 7 anyway, they made millions of those boards and you can run most anything in them.

I have a Millenium P100 system that runs an Optibase ISA video capture setup under Windows 95. Can't think of any other system I have that would be socket 5 (have a couple socket 4 and lots of later Pentium 1s).

People not being able to run the top of the line CPU in one is a good point. I think people don't view the Socket 4/5 platform as an alternative to a quick 486.

Only problem I've seen with the Socket 4 boards is that they have little BIOS settings available to the user, but I haven't tried so many chipsets/boards as they are rare and expensive. I also have a Socket 4 system as well, it performs around a DX4 - AMD X5 133.

I found this at random on google a few months back and thought I'd share. Socket-5 motherboard. With what appears to be VLB slots.

All those ram slots.. 😲 And that looks like a lot of cache on the lower right corner.

That looks old, probably the earlier socket 4. "Socket 4 VLB" quickly comes up with the Italian source, and google translate says something like this

Motherboard Unisys PW2 Premier EISA / VLB (1994/5) - This is a motherboard with CPU Pentium P54/Socket 4 at 66 MHz and particularly interesting for several reasons. First, it has a mixed architecture EISA / VLB (based on chipset Intel / ALR) which is not common and even less in a product like this, intended for use in x86 servers and in any case in high-end systems (note the number of slots EISA and the sockets for SIMM). Secondly have been very few Pentium motherboards to be equipped with bus VLB (VESA Local Bus), since the latter was closely related to the architecture of the i486 system and therefore difficult to adapt to other CPUs. Finally, this card is equipped with three VLB slots, and even this is not a common thing having electrical problems with this bus sometimes, evident already in the motherboard with only two slots. The second level cache is 256KB and is of type writeback.

Ah.. maybe so then. I thought it was socket 5, my bad. I haven't worked with anything between 486 Socket3 and Socket7 so I'm not really familiar with visually identifying boards from then.

I forgot about my Zenith Data Systems Pentium 90 Server EISA full tower, that must be a socket 5. It was made in the ERA where companies still made custom server motherboards (CPU and RAM have their own removable complexes) kind of like my IBM Model 80 and 95. Those systems seem to be collectable but are too heavy for cheap shipping.

kithylin wrote:

Ah.. maybe so then. I thought it was socket 5, my bad. I haven't worked with anything between 486 Socket3 and Socket7 so I'm not really familiar with visually identifying boards from then.

It is an interesting find anyway. 😀 I don't know how to identify the sockets either, when I can't see the CPU pin arrangement. Was just guessing by the 'premature' board layout.

I was talking about the Socket 7 Cyrix 6x86L at 1.0x multiplier. Here is that benchmark again as a screenshot. 40.00 points is like a 486DX/4-100. But I suspect Speedsys mainly benches the Floating point unit, which is the least developed aspect of the 6x86L. Strangely CPUID is disabled by default on this particular CPU, so I don't have to disable it manually in order to run speedsys.
In addition it was benched with PC-Player benchmark at stock resolution, it got 9.0 FPS. That FPS value is more like an AMD 5x86-P75 @133MHz. Here are is a nice lineup of AMD 5x86-P75 @133MHz systems (4x), on a German site.
Not having any older mainboards, like socket 5. This is the lowest I could get on Super Socket 7 without any additional tricks, such as disabling cache.

Edit:
Quake.exe +timedemo demo2 -nosound -nojoy: 13,1 FPS
PCPBench /VGAMode: 20,6 FPS

Why would you have to disable the CPUID to run speedsys?

Just with Cyrix 6x86 and MII, else the program will hang at startup.

For those who are interested in socket 5 at all start looking for boards by Gateway and there is one out there that is ATX.

The slowest speed I've managed on my Socket 5 ASUS P54SP4 with a Pentium is (no cache disabling):

According to Phil's benchmark settings:
3DBench2: 31.6
PCPBench /vgamode: 12.5
Doom: 4320 realticks (17,3 FPS)
Quake: 10.5 FPS

This was with a Matrox G200 PCI

I'm sure a Cyrix would bench even slower.

vetz wrote:

The slowest speed I've managed on my Socket 5 ASUS P54SP4 with a Pentium is (no cache disabling): […]
Show full quote

The slowest speed I've managed on my Socket 5 ASUS P54SP4 with a Pentium is (no cache disabling):

According to Phil's benchmark settings:
3DBench2: 31.6
PCPBench /vgamode: 12.5
Doom: 4320 realticks (17,3 FPS)
Quake: 10.5 FPS

This was with a Matrox G200 PCI

I'm sure a Cyrix would bench even slower.

Pentium running at what internal frequency? 1.5x40 = 60 MHz? I wouldn't mind seeing a first generation Cyrix 6x86 run at 1x40 MHz with accompanying Speedsys screenshot.