Thanks, man!

CPU-Z Vintage Edition under Windows ME, also here completely stable with 2,8V VCore (GLQuake with Riva 128ZX/8MB SGRAM and Prime tested):

feipoa wrote on 2020-12-06, 03:46:

Does it have SMP enabled? Or disabled like the Tillamooks?

someone reported being able to get two tillamooks working without cache I'm not sure where that post went it may be in the "tillamook working with cache thread" or the "dual socket 5/7 motherboard thread". when asked they didn't do any more tests though.

I think two tillamooks would work on upgrade adapters and with pin mods on a old socket 5 motherboard that used non intel smp

I personally have 1 tillamook and a dual socket 7 tyan s1564 board based on a 430hx if someone wants to donate a tillamook I can do some tests but I have already tried pairing a 233mmx and a tillamook 266 on it. it detected only a single tillamook (which it actually did call tillamook in bios)

so in skimming this thread it seems no one has figured out how to do 4x on a p55 yet? I'm really hoping to be able to somehow overclock these 2 233mmx chips

edit here it is Re: Tillamook 266MHz and working L2 cache?

There simply isn't a register for 4x on P55C cores.
The first stepping didn't even have the 3.5x register.
The chip had great yields and many run at 262MHz with 75MHz FSB and stock voltage but at official ~266 MHz many people wouldn't have understood why they should buy a 266MHz Pentium II. Because of that I suspect Intel actively chose not to implement multipliers >3.5x on P55C chips for marketing reasons.

Doornkaat wrote on 2021-03-02, 12:15:

The chip had great yields and many run at 262MHz with 75MHz FSB and stock voltage but at official ~266 MHz

I'm actually playing right now with a Pentium MMX 233 that runs at 290 MHz (FSB 83) semi-stable at 2.8V and 100% stable at 2.9V 😀 It's quite impressive how fast it can be... A Voodoo 2 is bottlenecking the CPU in GLQuake at 640 x 480. Even Unreal Gold is actually playable with an average of 39 FPS (13 FPS min) during the first two levels. My second PC from two decades ago (an AMD K6-2 @ 500 MHz with a Riva TNT 2 M64) was a LOT slower in games... if only I had known at the time 😀

Some beautiful scores and overclocks - the 333mhz is pretty beefy.
I have one running @124x2.5(~310mhz) on an ASUS P5A-B. It does take only 450 realtics to run demo3 of SW Doom which translates to 165.9fps. It's kinda "rendered" obsolete by the existence of the Tillamook though.

feipoa wrote on 2012-09-10, 04:47:

I spent about 2 weeks reading this whole tread when the time permitted. It did not seem as if there was a clear answer on setting a 4x multiplier on P233 MMX chips. It was noted that the BF0 and BF1 pins should be shorted to set 4x. Another comment was to short BF0 and NC for 4x. RG100 tried both of these but did not check the actual CPU speed, for example, with Chkcpu or in Windows. He relied soley on the BIOS table look-up information at boot, which may not display the correct CPU speed if that speed is not in the table.

On the vintage-computer.com thread, there was much talk of running this 4x experiment, but nobody posted any photo evidence of 4x working on a non-Tillamook chip. Tetrium mentioned he was going to try this, but did not follow through in the vintage-computer thread.

I don't have any P233 MMX chips or S7 motherboards to kill for this test so I will not be trying this myself. Has anyone successfully run a P233 MMX at 4x and have photo evidence of it, for example, a CPU photo, chkcpu, and a benchmark? The best bet for this test is to try something modest, like 66x4 (266 MHz) at the rated voltage. It seems a lot of people who tried this 4x test were gunning for 300 MHz and above. Even 66x4 would might be of moderate benefit since you don't need to overclock the PCI bus for running the chip at 75x3.5.

I don't know if that's already covered in another thread since 10 years ago or so, but I definitely searched about such thread and couldn't find one.
Well, I happen to have such CPU I think. I also have a QDi Titanium IIIB motherboard (yes, it's IIIB not IIB that everyone seems to have) that has the Speedeasy menu. I just plugged it in, set 4x with the default FSB of 66MHz (can't overclock it at all on this mobo) and appeared as 266MHz on the POST screen. I thought it was definitely wrong, but benchmark numbers were improved. That was months ago. If this is not documented yet somewhere I can try again and take some photos.
But that motherboard definitely needs the patched BIOS, which is here [http://www.steunebrink.info/k6plus.htm] listed as "coming" in order to get 75MHz FSB speed and HDD size limit increased from 8.4GB to 64-128GB, else I constantly have to use DDO software...

I was able to get several Pentium 233 MMX chips to run at 3x100MHz on MVP3 boards, seemingly stable, but I haven't used any of them at that frequency for an extended period of time so I have no clue what the long term effects will be. I have one sample that ran at 300MHz stock voltage - played with it for a few hours - but the rest of my chips need a voltage bump. 2.9V should be safe. Thing is, most chips I tested are windows stable only at 3v or more, and they will get hot. I used a Titan brand copper core socket A cooler for testing.

elfuego wrote on 2009-03-27, 12:52:

5u3 wrote:

Why not take a K6(-2/-3)?

Because they suck? 😜 Sorry, but now that I've tested K6, K6-2 and recently K6-3 ranging from 266Mhz to 600Mhz I can tell that those CPUs are nuisance when compared to genuine Intel of that time. The only good thing about K6 is, as you already mentioned, ability to run from 120Mhz to 600Mhz, in a setting you choose.

But on the performance side - there is only Intel.

The only game I've seen a P55C really smack down a K6-2 is Quake, and that due to the game engine being highly optimized for the P55 architecture - specifically the P55's FPU - but once you bump up clock speeds and add on chip L2 cache, the situation turns around. Even at 300MHz a P55C will be slower than a 500Mhz K6-2+. Not proportionally slower, but slower, nevertheless.

IF you could get a P55C to run at the speeds a K6-2+ / K6-3 is capable off, it might be a different story - but again, only in quake 1. Quake 2 has a 3dnow patch and with that a K6-2+ performs very well - and in other games it's no longer a CPU speed issue, but a platform performance issue - for example Half Life 1. It stutters and lags on a super socket 7 platform regardless of what CPU and video card I tried, but it's nice and smooth on a 350MHz pentium II using the same video card I used in the super 7 build (Geforce 2 MX400) - that was my experience with the super 7 platform. Half Life 1 is still quite playable on a super 7 PC, but it's much smoother on a pentium 2.

Socket3 wrote on 2023-07-28, 07:37:

Half Life 1 is still quite playable on a super 7 PC, but it's much smoother on a pentium 2.

It's really not. I mean, it depends on the actual CPUs (and motherboards) being compared and the video card that's being used.
With a Voodoo 3, an AMD K6-3+ @ 550 MHz (or more), running on a good motherboard like the Asus P5A/P5A-B/Gigabyte GA-5AX/GA-5AA will definitely smoke a Pentium 2 350 running on a 440BX in... anything, including Half Life.

If, on the other hand, we are comparing a Pentium 2 350 on a 440BX motherboard with a normal K6-2 @ 500 MHz running on a crappy SS7 motherboard... yeah, the Pentium 2 will definitely be faster/smoother.

bloodem wrote on 2023-07-28, 07:49:

It's really not. I mean, it depends on the actual CPUs (and motherboards) being compared and the video card that's being used. W […]
Show full quote

Socket3 wrote on 2023-07-28, 07:37:

Half Life 1 is still quite playable on a super 7 PC, but it's much smoother on a pentium 2.

It's really not. I mean, it depends on the actual CPUs (and motherboards) being compared and the video card that's being used.
With a Voodoo 3, an AMD K6-3+ @ 550 MHz (or more), running on a good motherboard like the Asus P5A/P5A-B/Gigabyte GA-5AX/GA-5AA will definitely smoke a Pentium 2 350 running on a 440BX in... anything, including Half Life.

If, on the other hand, we are comparing a Pentium 2 350 on a 440BX motherboard with a normal K6-2 @ 500 MHz running on a crappy SS7 motherboard... yeah, the Pentium 2 will definitely be faster/smoother.

....mmmm yeah, I was being generous when I called it "playable". I can play it on my super 7 build, I could complete the game, but it's not exactly enjoyable. Average framerate is around 40 fps with a Geforce 2 MX, but the game stutters and there's even micro-freezing from time to time. Off the top of my head, minimum FPS on the super 7 build at 640x480 was single digits, while on a 350mhz Compaq Deskpro EN with the same exact geforce 2 MX minimum FPS was 17 or 18 FPS. Still low, but overall, a much better experience. Stuttering is not completely gone on the p2, but the occasional freezing is not present on the p2. Oddly enough the super 7 manages higher max fps, but that's irrelevant with stuttering and micro-freezes.

bloodem wrote on 2023-07-28, 07:49:

It's really not. I mean, it depends on the actual CPUs (and motherboards) being compared and the video card that's being used. W […]
Show full quote

Socket3 wrote on 2023-07-28, 07:37:

Half Life 1 is still quite playable on a super 7 PC, but it's much smoother on a pentium 2.

It's really not. I mean, it depends on the actual CPUs (and motherboards) being compared and the video card that's being used.
With a Voodoo 3, an AMD K6-3+ @ 550 MHz (or more), running on a good motherboard like the Asus P5A/P5A-B/Gigabyte GA-5AX/GA-5AA will definitely smoke a Pentium 2 350 running on a 440BX in... anything, including Half Life.

If, on the other hand, we are comparing a Pentium 2 350 on a 440BX motherboard with a normal K6-2 @ 500 MHz running on a crappy SS7 motherboard... yeah, the Pentium 2 will definitely be faster/smoother.

This is one of the few times I will disagree with you bloodem. K6-III+ 550 with a Voodoo 3 trades punches with a PII 350 on 440BX in 3D games. Sometimes the K6-III will be faster, other times the PII 350 will be a bit faster on average but offer more consistent performance and sometimes they'll be even. K6 is all over the place really, depends on what bottleneck given game is hitting.
There are some benchmarks here:
Pentium II/III VS K6-III+

GeorgeMan wrote on 2023-07-24, 13:50:

I don't know if that's already covered in another thread since 10 years ago or so, but I definitely searched about such thread a […]
Show full quote

feipoa wrote on 2012-09-10, 04:47:

I spent about 2 weeks reading this whole tread when the time permitted. It did not seem as if there was a clear answer on setting a 4x multiplier on P233 MMX chips. It was noted that the BF0 and BF1 pins should be shorted to set 4x. Another comment was to short BF0 and NC for 4x. RG100 tried both of these but did not check the actual CPU speed, for example, with Chkcpu or in Windows. He relied soley on the BIOS table look-up information at boot, which may not display the correct CPU speed if that speed is not in the table.

On the vintage-computer.com thread, there was much talk of running this 4x experiment, but nobody posted any photo evidence of 4x working on a non-Tillamook chip. Tetrium mentioned he was going to try this, but did not follow through in the vintage-computer thread.

I don't have any P233 MMX chips or S7 motherboards to kill for this test so I will not be trying this myself. Has anyone successfully run a P233 MMX at 4x and have photo evidence of it, for example, a CPU photo, chkcpu, and a benchmark? The best bet for this test is to try something modest, like 66x4 (266 MHz) at the rated voltage. It seems a lot of people who tried this 4x test were gunning for 300 MHz and above. Even 66x4 would might be of moderate benefit since you don't need to overclock the PCI bus for running the chip at 75x3.5.

I don't know if that's already covered in another thread since 10 years ago or so, but I definitely searched about such thread and couldn't find one.
Well, I happen to have such CPU I think. I also have a QDi Titanium IIIB motherboard (yes, it's IIIB not IIB that everyone seems to have) that has the Speedeasy menu. I just plugged it in, set 4x with the default FSB of 66MHz (can't overclock it at all on this mobo) and appeared as 266MHz on the POST screen. I thought it was definitely wrong, but benchmark numbers were improved. That was months ago. If this is not documented yet somewhere I can try again and take some photos.
But that motherboard definitely needs the patched BIOS, which is here [http://www.steunebrink.info/k6plus.htm] listed as "coming" in order to get 75MHz FSB speed and HDD size limit increased from 8.4GB to 64-128GB, else I constantly have to use DDO software...

With the years since my last reply here I have come to the conclusion that the p55c does not have a 4x miltiplier on the silicon.

See, there were some p55c chips that were fully unlocked, and able to overclock far beyond their rating. Example, there is a 133, 120, 150, 166, 200, and 233 mmx and I have encountered unlocked versions of most.

None of these had a 4x multiplier that I could set.

But, if I am wrong, here is how you would find out.

On intel they set 4x through a different bf2 pin than on amd/cyrix

If you used one of my tweaker devices (in my sig, go to pcb projects) you can program this register.

I may one day go through my collection and find unlocked examples and try setting intel bf2 pin and see what happens.

But I don’t believe it exists, I think the only intel cpu to have it was tillamook.

Tillamook is supposed to have a 4.5x multi btw, but I have yet to be able to verify its existence

Sphere478 wrote on 2023-07-28, 14:12:

With the years since my last reply here I have come to the conclusion that the p55c does not have a 4x miltiplier on the silicon […]
Show full quote

With the years since my last reply here I have come to the conclusion that the p55c does not have a 4x miltiplier on the silicon.

See, there were some p55c chips that were fully unlocked, and able to overclock far beyond their rating. Example, there is a 133, 120, 150, 166, 200, and 233 mmx and I have encountered unlocked versions of most.

None of these had a 4x multiplier that I could set.

But, if I am wrong, here is how you would find out.

On intel they set 4x through a different bf2 pin than on amd/cyrix

My old 2014 test of six variants of normal non-tillamook P55C / Pentium MMX is at the bottom here.
Pentium MMX Multipliers
Only tried BF0/BF1 jumpers, and probably also the AMD BF2 jumper which has no effect for intel processors.
Never seen 1.5x or 4.0x on such.

Socket3 wrote on 2023-07-28, 13:11:

Average framerate is around 40 fps with a Geforce 2 MX, but the game stutters and there's even micro-freezing from time to time. Off the top of my head, minimum FPS on the super 7 build at 640x480 was single digits, while on a 350mhz Compaq Deskpro EN with the same exact geforce 2 MX minimum FPS was 17 or 18 FPS.

Strange, that's not my experience at all. Were you using proper, older drivers (like 7.76)? If not, that would explain it. Also, what specific motherboard? Even the sound card is very important - some sound cards can completely kill the performance on SS7.
Anyway, when it comes to SS7 and Half Life, I think the Voodoo 3 (with the MiniGL driver), is the only option to have a somewhat playable experience (by more... modern standards).

Garrett W wrote on 2023-07-28, 13:23:

This is one of the few times I will disagree with you bloodem. K6-III+ 550 with a Voodoo 3 trades punches with a PII 350 on 440BX in 3D games. Sometimes the K6-III will be faster, other times the PII 350 will be a bit faster on average but offer more consistent performance and sometimes they'll be even. K6 is all over the place really, depends on what bottleneck given game is hitting.
There are some benchmarks here:
Pentium II/III VS K6-III+

Don't get me wrong, I'm not pretending that the SS7 platform is something that's not. We all know that it's a slow (and many times buggy) platform, something that I've been saying many times before. 😀
But, when it comes to the "stock" Pentium 2 350 vs a faster and properly tuned SS7 platform (which, granted, it's quite hard to do, many stars must be aligned), the Pentium 2 350 is definitely slower in my experience (at least in the majority of cases).
Of course, also true is the fact that the Pentium 2 350 can usually be overclocked just fine to 466 MHz by increasing the FSB to 133 MHz (I have two P2 350 CPUs that are perfectly stable at this frequency without a voltage bump), at which point it will destroy any K6-2/3(+) CPU. It's normal/expected; not only does it have a stronger FPU, but it also has the 440BX chipset going for it, with its incredible memory performance.

Regarding those benchmarks, there are definitely some red flags that I see (different testing conditions & components between the platforms, the Voodoo 5 being used on SS7 - this card being slower in heavily CPU bottlenecked games compared to the Voodoo 3 because of more driver overhead, the fact that 256 MB of RAM were used on the Asus P5A-B rev 1.04 - which right out of the gate will typically decrease the performance by 5 - 10%, even with CPUs that have on-die cache). And don't even get me started with the "Unofficial service packs"... a complete no-no, especially for slower platforms like the SS7.

What I do know for sure is that I played Half Life many times on my SS7 systems, and I've never seen single digits. Typical framerates are ~ 70 - 75 FPS for less cluttered environments, 30 - 45 FPS for more demanding areas, and there will also be occasional drops to ~ 15 FPS during very intensive scenes. Overall, the experience is surprisingly enjoyable on these specific systems, but this is far from being the norm.

bloodem wrote on 2023-07-28, 16:54:

Regarding those benchmarks, there are definitely some red flags that I see (different testing conditions & components between the platforms, the Voodoo 5 being used on SS7 - this card being slower in heavily CPU bottlenecked games compared to the Voodoo 3 because of more driver overhead,

I don't think that's the case. I've tried both cards on slower systems such as these, they perform identically in CPU limited scenarios and the Voodoo 5 pulls ahead on the off-chance that the GPU load increases. I remember specifically testing for that possibility, do you happen to have uploaded anything showing this?

bloodem wrote on 2023-07-28, 16:54:

the fact that 256 MB of RAM were used on the Asus P5A-B rev 1.04 - which right out of the gate will typically decrease the performance by 5 - 10%, even with CPUs that have on-die cache).

On that we agree, 128MB would have yielded slightly better results, but I'd wager closer to the 5% mark rather than 10%.

bloodem wrote on 2023-07-28, 16:54:

And don't even get me started with the "Unofficial service packs"... a complete no-no, especially for slower platforms like the SS7.

I have to fight you on this one too. SP2.1a is pretty much tried and tested, it includes only official, Microsoft updates for Win98SE. I have always had a great experience with it, none of that unofficial SP3 bloat.

bloodem wrote on 2023-07-28, 16:54:

What I do know for sure is that I played Half Life many times on my SS7 systems, and I've never seen single digits. Typical framerates are ~ 70 - 75 FPS for less cluttered environments, 30 - 45 FPS for more demanding areas, and there will also be occasional drops to ~ 15 FPS during very intensive scenes. Overall, the experience is surprisingly enjoyable on these specific systems, but this is far from being the norm.

This has not been my experience unfortunately, but maybe I've been doing something wrong here. In my experience, Half-Life, even on a K6-III+ 550 will drop framerate in the teens as soon as a couple of enemies show up on screen and the player happens to be in an outdoor area.

Garrett W wrote on 2023-07-28, 17:40:

I don't think that's the case. I've tried both cards on slower systems such as these, they perform identically in CPU limited scenarios and the Voodoo 5 pulls ahead on the off-chance that the GPU load increases. I remember specifically testing for that possibility, do you happen to have uploaded anything showing this?

It's possible you're right. I only tried the Voodoo 5 on a K6-2 build once and I do remember seeing a noticeable difference between it and a Voodoo 3, but that's definitely not enough data to draw a pertinent conclusion.

Garrett W wrote on 2023-07-28, 17:40:

On that we agree, 128MB would have yielded slightly better results, but I'd wager closer to the 5% mark rather than 10%.

Well, you are wrong. But I was more wrong than you... 😀 I just did a few tests on one of my SS7 builds and the performance difference with external cache on or completely off is actually 2 - 3%.

With external cache ON:
GLQuake 640 x 480: 186.3 FPS
Quake 2 3DNow! 640 x 480: 113.4 FPS

With external cache OFF:
GLQuake 640 x 480: 182.1 FPS
Quake 2 3DNow! 640 x 480: 110.7 FPS

I think this needs further investigation, because I distinctly remember seeing larger differences, but it's very likely that those differences shrink on more powerful builds with higher FSB frequencies (this particular system is running a K6-3+ at FSB 112 / 616 MHz / Voodoo 3 3000).

Garrett W wrote on 2023-07-28, 17:40:

I have to fight you on this one too. SP2.1a is pretty much tried and tested, it includes only official, Microsoft updates for Win98SE. I have always had a great experience with it, none of that unofficial SP3 bloat.

My experience with service packs, including 2.1a, has always been abysmal. I've never seen any improvements in games, but I've encountered all sorts of weird issues after installing them on certain platforms... so I've stopped doing that a long time ago.

Garrett W wrote on 2023-07-28, 17:40:

This has not been my experience unfortunately, but maybe I've been doing something wrong here. In my experience, Half-Life, even on a K6-III+ 550 will drop framerate in the teens as soon as a couple of enemies show up on screen and the player happens to be in an outdoor area.

Do you have a save game from a particular section of the game that you say runs poorly? I'd like to test it out as well.
Last time I've played and finished Half Life on a SS7 build was... 23 years ago. On my newer SS7 builds I've played it occasionally and I've only reached Chapter 7.

Try Half-Life Uplink, it's a fun vertical slice of the game that contains a cut chapter from the game. It's also a great stress test, there's an outdoor area where you fight the marines where you can notice what I was referring to in my earlier post.

bloodem wrote on 2023-07-28, 16:54:

Strange, that's not my experience at all. Were you using proper, older drivers (like 7.76)? If not, that would explain it. Also, […]
Show full quote

Socket3 wrote on 2023-07-28, 13:11:

Average framerate is around 40 fps with a Geforce 2 MX, but the game stutters and there's even micro-freezing from time to time. Off the top of my head, minimum FPS on the super 7 build at 640x480 was single digits, while on a 350mhz Compaq Deskpro EN with the same exact geforce 2 MX minimum FPS was 17 or 18 FPS.

Strange, that's not my experience at all. Were you using proper, older drivers (like 7.76)? If not, that would explain it. Also, what specific motherboard? Even the sound card is very important - some sound cards can completely kill the performance on SS7.
Anyway, when it comes to SS7 and Half Life, I think the Voodoo 3 (with the MiniGL driver), is the only option to have a somewhat playable experience (by more... modern standards).

Garrett W wrote on 2023-07-28, 13:23:

This is one of the few times I will disagree with you bloodem. K6-III+ 550 with a Voodoo 3 trades punches with a PII 350 on 440BX in 3D games. Sometimes the K6-III will be faster, other times the PII 350 will be a bit faster on average but offer more consistent performance and sometimes they'll be even. K6 is all over the place really, depends on what bottleneck given game is hitting.
There are some benchmarks here:
Pentium II/III VS K6-III+

Don't get me wrong, I'm not pretending that the SS7 platform is something that's not. We all know that it's a slow (and many times buggy) platform, something that I've been saying many times before. 😀
But, when it comes to the "stock" Pentium 2 350 vs a faster and properly tuned SS7 platform (which, granted, it's quite hard to do, many stars must be aligned), the Pentium 2 350 is definitely slower in my experience (at least in the majority of cases).
Of course, also true is the fact that the Pentium 2 350 can usually be overclocked just fine to 466 MHz by increasing the FSB to 133 MHz (I have two P2 350 CPUs that are perfectly stable at this frequency without a voltage bump), at which point it will destroy any K6-2/3(+) CPU. It's normal/expected; not only does it have a stronger FPU, but it also has the 440BX chipset going for it, with its incredible memory performance.

Regarding those benchmarks, there are definitely some red flags that I see (different testing conditions & components between the platforms, the Voodoo 5 being used on SS7 - this card being slower in heavily CPU bottlenecked games compared to the Voodoo 3 because of more driver overhead, the fact that 256 MB of RAM were used on the Asus P5A-B rev 1.04 - which right out of the gate will typically decrease the performance by 5 - 10%, even with CPUs that have on-die cache). And don't even get me started with the "Unofficial service packs"... a complete no-no, especially for slower platforms like the SS7.

What I do know for sure is that I played Half Life many times on my SS7 systems, and I've never seen single digits. Typical framerates are ~ 70 - 75 FPS for less cluttered environments, 30 - 45 FPS for more demanding areas, and there will also be occasional drops to ~ 15 FPS during very intensive scenes. Overall, the experience is surprisingly enjoyable on these specific systems, but this is far from being the norm.

Awe64 value an forceware 12.xx