VOGONS


First post, by tauro

User metadata
Rank Member
Rank
Member

INTRO
I have a P-III build that took me a long time to put together, lots of testing for it to be stable at top speeds, with all cards living together in peace... Later realized that it's really too fast for DOS, and that late 90's and early 00's stuff would run better on a P4 or later. Most old stuff ran fine but some DOS games weren't cooperating. Disabling L2 cache made the machine too slow, and ALL slow down utilities weren't up to the task and sometimes introduced problems with the ISA sound card. I reduced the bus speed to 66MHz and that helped, but it's still too fast for some games or software and kind of defeats the purpose of building a very fast DOS machine.

So, aiming for a more versatile build, I think the best idea is to build a Super Socket 7, Pentium MMX machine, that can effectively slow down to a snail's pace and (hopefully) run most if not all x86 software from '81 to the mid to late 90's, without sacrificing comfort (PS2 mouse, USB, AGP, PCI, ATX, etc).

THE NEW BUILD
The only SS7 board I have in a working condition is a GA-5AX Rev 5.2. I have a handful of MMX chips, so I decided to try them out and see which one was the fastest, the slowest, etc, but there were some surprises awaiting me. Take a look at this one:

sl23w.fake.jpg
Filename
sl23w.fake.jpg
File size
536.65 KiB
Views
1365 views
File license
CC-BY-4.0

It says FV80503233 / SL23W / A4 / L742003-6627 but that's supposedly a 200MHz CPU according to online sources... No matter how hard I try to clean it, it always looks dirty, text is very very dim. I think this one has been relabeled, with a rubbery substance. What do you think?
It didn't post at 3x100 3.3v. Then I found out that it doesn't even support a 3x multi! (for some reason 3 = 3.5), so it's either 2 or 3.5. It was stable at 290MHz (83x3.5 3.3v).
If this was a relabeled SL23W (200MHz), isn't it supposed to support a 3x multiplier (66x3)?

Then I tried an SL27C, which is a mobile cpu. It's core voltage is 2.45v.
It's stable at 2.5x100 2.6v. This one only supports 2x and 2.5x multipliers. I didn't know it was like this with the MMX chips, I thought their multiplier was unlocked.

sl27c.jpg
Filename
sl27c.jpg
File size
560.04 KiB
Views
1365 views
File license
CC-BY-4.0

This little boy boots and runs stable (Quake benchmark) @133Mhz with only 1.5v core voltage. It seems remarkable, but I'm not familiar with undervolting these chips.

Finally, this is the fastest one I have, an SL293, BP80503233, which runs stable at 3x100 3.0v.

sl293.jpg
Filename
sl293.jpg
File size
571.03 KiB
Views
1365 views
File license
CC-BY-4.0

Some other questions:

PCI BUS
* I have an AGP card that supports 88MHz, but I don't know how much I can push this chipset and the board's L2 cache. And also, what is the PCI bus speed at 130 and how do you calculate it? Is it always half the AGP bus? So running it at, let's say 130MHz bus, 87MHz AGP, is PCI 43.5MHz? Or does this motherboard automatically take care of that? What would the ISA clock be then...?

Here's a reference:

cpu.agp.clocks.reference.png
Filename
cpu.agp.clocks.reference.png
File size
14.76 KiB
Views
1365 views
File license
CC-BY-4.0

OVERCLOCKING AND VOLTAGES
* Are there any games or software in general that would see a considerable benefit with an overclock from 100*2.5 to 100*3? I plan to revisit mostly older titles and programs, but I like the versatility. Is it worth it to have those extra 50MHz?

* What are the safe voltages and borderline voltages for the regular and for the mobile MMX?

According to Intel's datasheet, Desktop MMX cpu's vcc2 specs are 2.8v ± 3.57% and the maximum rating is -0.5v/3.7v while Mobile MMX voltage tolerance is 2.45v +0.215 / -0.165, but it also says that the maximum VCC2 rating is −0.5V to +3.7V.

Is running the desktop CPU at 3.0v for long periods of time advisable? It's slightly above the max voltage (2.9v) but will it result in degradation?

UNDERCLOCKING
* Is there a way to achieve a 50MHz bus with an undocumented dip switch combination or some hack? The clock generator is an ICS 9148AF-75

* How does a S370 VIA C3 compare to a an S7 MMX regarding very old software? I have the impression that even if a C3 can be slowed down, it can't match a Pentium in terms of how smooth it can get at very slow speeds (feel free to correct me and provide information).

* How would you categorize the main computing eras from the early 80's to the late 90's? What computers would you say that are the most representative models for those eras and how would you recommend to best adjust the MMX configuration to give that kind performance?

I think that we can see at least three eras:

8088 / Early 286
386 / Early 486
Late 486 / Pentium

What do you think? Of course there are many nuances.

* What games or software are the best to test each era?

I'm aware of the SETMUL method, which attracted my attention in the first place, using CCD DCD BPD VPD to emulate a 386DX running at about 30MHz. Then I would probably need another software to slow it down further. Which one would you recommend? I tested some of them, but I don't know which one is better for this particular build.

To close this post, here's another fake chip from my collection:

sl27x.fake.jpg
Filename
sl27x.fake.jpg
File size
709.96 KiB
Views
1365 views
File license
CC-BY-4.0

It says "SL27X" but that's supposed to be a PPGA chip... It doesn't even do 233MHz with 2.8v, it needs 2.9v to be stable. So I think it's a relabeled 166MHz CPU. What's interesting is that it supports a 3.5x multiplier. Do all 166MHz ceramics support that multiplier? Are all the CPUs that start to "peel off" actually fakes? I think I have another one somewhere, but an early Pentium (non-MMX).

I also found that I have an authentic ceramic MMX, I think it's the fastest one there is, an SL2S9. I read that it's not that common to see a 200MHz MMX in a ceramic package. It's a lousy overclocker though and works only with 2.5x and 3x multipliers.

Feel free to comment on this kind of build and share your experience and advice 🙌

Reply 1 of 14, by buckeye

User metadata
Rank Oldbie
Rank
Oldbie

This type of build is on my "to do" list, just saving my pennies because prices are eye watering to say the least.

Asus P5N-E Intel Core 2 Duo 3.33ghz. 4GB DDR2 Geforce 470 1GB SB X-Fi Titanium 650W XP SP3
Intel SE440BX P3 450 256MB 80GB SSD Asus V7700 GF2 64mb SB 32pnp 350W 98SE
MSI x570 Gaming Pro Carbon Ryzen 3700x 32GB DDR4 Zotac RTX 3070 8GB WD Black 1TB 850W

Reply 3 of 14, by tauro

User metadata
Rank Member
Rank
Member
SETBLASTER wrote on 2023-09-18, 19:08:

ill take some of my pentium CPUs, take pictures of all of them side to side, and maybe you can tell me just by looking at them, which are fake and which are the real deal.

Sounds like fun! I'll take mine out too and see if I can show you more about what I'm talking about.

Reply 4 of 14, by pentiumspeed

User metadata
Rank l33t
Rank
l33t

That fake pentium mmx. I have plain pentium that was remarked same way, looked wrong then found out by scratching bit of top layer off revealed 150MHz instead of 166, that was back in the day from a customer's computer.

Cheers,

Great Northern aka Canada.

Reply 5 of 14, by Gmlb256

User metadata
Rank l33t
Rank
l33t
tauro wrote on 2023-07-07, 11:16:

* How does a S370 VIA C3 compare to a an S7 MMX regarding very old software? I have the impression that even if a C3 can be slowed down, it can't match a Pentium in terms of how smooth it can get at very slow speeds (feel free to correct me and provide information).

It depends, both Ezra-T and Nehemiah can outmatch a Pentium MMX in terms of flexibility when paired with a suitable motherboard that has a compatible PLL where RayeR's SMB can be used to switch the FSB on the fly in DOS. This is in addition to the slowdown capabilities available with SetMul.

I would recommend checking the following thread to see the overall pefomance of these CPUs: VIDEO BENCHMARKS (Original title: Settling the VIA C3 performance debate once and for all)

I'm aware of the SETMUL method, which attracted my attention in the first place, using CCD DCD BPD VPD to emulate a 386DX running at about 30MHz. Then I would probably need another software to slow it down further. Which one would you recommend? I tested some of them, but I don't know which one is better for this particular build.

Throttle, which another method to slow down the computer using ACPI functions if the motherboard chipset supports it.

For ALi Aladdin V chipsets, there is a similar utility: Throttle for ALi Aladdin 5 Socket 7 boards

VIA C3 Nehemiah 1.2A @ 1.46 GHz | ASUS P2-99 | 256 MB PC133 SDRAM | GeForce3 Ti 200 64 MB | Voodoo2 12 MB | SBLive! | AWE64 | SBPro2 | GUS

Reply 6 of 14, by tauro

User metadata
Rank Member
Rank
Member
Gmlb256 wrote on 2023-09-18, 20:29:

It depends, both Ezra-T and Nehemiah can outmatch a Pentium MMX in terms of flexibility when paired with a suitable motherboard that has a compatible PLL where RayeR's SMB can be used to switch the FSB on the fly in DOS. This is in addition to the slowdown capabilities available with SetMul.

I would recommend checking the following thread to see the overall pefomance of these CPUs: VIDEO BENCHMARKS (Original title: Settling the VIA C3 performance debate once and for all)

The C3s can outperform the modest P-MMX, but my main focus is if they can under-perform it, so to say. I skimmed through that thread and its centered around top performance. Still, thanks for mentioning it. I'm very interested in the C3 CPU. I think we should make a comparison about how slow can a S7 Pentium go (with a 50MHz bus, for example) vs. a S370 C3.

Gmlb256 wrote on 2023-09-18, 20:29:

Throttle, which another method to slow down the computer using ACPI functions if the motherboard chipset supports it.

For ALi Aladdin V chipsets, there is a similar utility: Throttle for ALi Aladdin 5 Socket 7 boards

I wasn't aware of this at all, thank you.
I've used throttle in the past and it felt a bit "stuttering", but I'll keep in mind this particular version for ALi V motherboards.

Reply 7 of 14, by tauro

User metadata
Rank Member
Rank
Member

So I took some pictures and these are the best tips I can give you to tell which chips are original and which are fake. First I'll upload the pictures and then I'll point out the differences which will become apparent.

SL24R
FAKE

SL24R.1.jpg
Filename
SL24R.1.jpg
File size
1.48 MiB
Views
1060 views
File license
CC-BY-4.0
SL24R.2.jpg
Filename
SL24R.2.jpg
File size
1.8 MiB
Views
1060 views
File license
CC-BY-4.0
SL24R.side.jpg
Filename
SL24R.side.jpg
File size
1.14 MiB
Views
1060 views
File license
CC-BY-4.0
Last edited by tauro on 2023-09-18, 22:23. Edited 1 time in total.

Reply 8 of 14, by tauro

User metadata
Rank Member
Rank
Member

SY022
I think this one is a FAKE but the coat of fake paint is more or less intact.

SY022.jpg
Filename
SY022.jpg
File size
1.72 MiB
Views
1049 views
File license
CC-BY-4.0
SY022.side.jpg
Filename
SY022.side.jpg
File size
832.8 KiB
Views
1049 views
File license
CC-BY-4.0
SY022.side2.jpg
Filename
SY022.side2.jpg
File size
1.12 MiB
Views
1049 views
File license
CC-BY-4.0

Reply 9 of 14, by tauro

User metadata
Rank Member
Rank
Member

SY027
ORIGINAL

SY027.1.jpg
Filename
SY027.1.jpg
File size
1.54 MiB
Views
1046 views
File license
CC-BY-4.0
SY027.2.jpg
Filename
SY027.2.jpg
File size
1.33 MiB
Views
1046 views
File license
CC-BY-4.0
SY027.side.jpg
Filename
SY027.side.jpg
File size
1 MiB
Views
1046 views
File license
CC-BY-4.0

Reply 10 of 14, by tauro

User metadata
Rank Member
Rank
Member

As you may have already noticed, the fake chips have an extra coat of paint of a slightly different color that can be noticed if you look at the processor's edges:

comparison.sides.jpg
Filename
comparison.sides.jpg
File size
641.1 KiB
Views
1026 views
File license
CC-BY-4.0

And the typography/fonts on the top of the CPUs is slightly different too. Moreover, if you touch the surface with your finger, you will feel that the letters on the fake CPUs feel depressed, as if they were etched, while the original CPUs look as if they were printed, and you can feel that the letters stick out.
That's why I think that with the fake CPUs you can't clean the thermal compound easily, as it gets stuck in the crevices of the etched letters (just an hypothesis).

all.three.jpg
Filename
all.three.jpg
File size
1.47 MiB
Views
1026 views
File license
CC-BY-4.0

The typography/letters on the back of the CPU are much more relative according to what I could see in my processors, so I couldn't find a pattern.

Reply 11 of 14, by SETBLASTER

User metadata
Rank Member
Rank
Member

Ok this is a first batch of pictures cpus,
here are the ones that came with cooler and non-ceramic,

notice anything strange?

metal1.jpg

withcooler1.jpg

in about an hour i will post the ceramic ones.
i have many fake ones that are ceramic hahaha because im trying to clean them and the paint comes out 🤣

Reply 13 of 14, by tauro

User metadata
Rank Member
Rank
Member

I have much less PPGA P-MMX processors, and only one of them is fake (SL23W).

sl23w.fake.jpg

I could tell because they marked it wrongly as 233MHz when in fact it should be 200MHz.
And more importantly, the letters are VERY DIM and you can never see them clearly.

Yours seem to look dim too. If you have other PPGA CPUs see if they have clearly readable yellow-looking letters, such as this:
sl293.jpg

All the ones I have (except the fake one) sport these yellow/whitish letters and I think they are originals.

The top of the PPGA processors seems to always be the same since that part bears no markings, so I don't think it's relevant.

You could try your PPGA chips and see if they perform well at the rated voltage and see how they overclock. Check their multipliers too. That will help you decide if they are originals or fakes.

Now about the CPGA processors, I can't tell if they are fake or original by looking at their back side. I have lots of them and I've seen different typography, sizes, placement, etc. I'd like to learn more if anybody knows how to identify them this way.

Your SL23X could be a fake, but I'm not sure, as the only way to really tell is by removing the heatsink and scraping it to see if the paint comes off. But an easier way is to get a flashlight or a strong light and take a look at the edges and see if you spot fake paint traces, like I pointed out in my previous post.

On the ones with the attached heatsink+fan, carefully remove the fan (there's a clip) and see the markings inside the cooler:
L_00013043.jpg
Maybe you could find a surprise there too.

Reply 14 of 14, by SETBLASTER

User metadata
Rank Member
Rank
Member

ok here comes the cheap fakes

Here is a picture of a pentium 100 and a pentium 75
I belive these are real, i don´t think they used to fake pentium 100s
100-75.jpg

Here are 3 pentium 120, the 2 on the left are different.
the one on the far right was a 166mhz fake, that if you scratch the surface ended up being a 120, notice how they faked the 166mhz on the back too 😀
120s.jpg

Here are 2 pentium 150, the one on the left looks strange
150.jpg

these are 3 pentium 166, and to be honest, the only one that looks real to me is the one on the left.
166.jpg

lastly this is a 133mhz and a 200mhz
while cleaning it, the paint was gone, am sad because i wanted a 133mhz, so they faked even 133mhz!
the other one 200mhz must be fake too
133-200.jpg