You know, this thread made me check my main PC (Zen 3) for MMX support, and it's indeed still there! Along with MMX+, whatever the heck that is. My work laptop (11th gen Intel) also supports it.

I could've sworn I read that one of the two big CPU makers had axed MMX support because of how irrelevant it was, even back then. That would've been what, 10+ years ago? Not that it matters, because obviously I was dreaming. It wouldn't surprise me, though, if MMX was rolled into SSE when x64 took over. Maybe that's what I'm misremembering.

Anyway. You now have my permission to return to the original topic of MMX on the actual MMX. Pentium. 😜

Edit: It was 3DNow! AMD dropped it back in 2010. https://bit-tech.net/news/tech/cpus/amd-retires-3dnow/1/

the3dfxdude wrote on 2023-05-26, 18:04:

Strange, I remember using a P2 on a thinkpad. But they weren't necessarily good for battery life! But it's quite typical that the mobile processors were the previous gen at low voltage and die shrunk. They didn't push them for desktop systems, which is why I didn't know anyone really making builds out of them back in the day.

Yes, they do exist... "Mobile Pentium II"... two variations (Tonga and Dixon core): https://en.wikipedia.org/wiki/Pentium_II#Mobile
Pentium 4 Mobile also exists. But in both cases they were indeed poor mobile processors because of the high power consumption, leading to poor battery life.
So they wouldn't have been a very popular option.

awgamer wrote on 2023-05-26, 17:10:

The Serpent Rider wrote on 2023-05-26, 12:00:

300MHz.

Tourist!
https://www.youtube.com/watch?v=fqB4yCssUCM

Also 500 MHz is possible with a good Tillamook: Retroclocking: Worlds Fastest Pentium MMX

The Serpent Rider wrote on 2023-05-26, 17:56:

Yeah, that will tickle the "almighty" K6-III too =P

Apparently it takes a 500 mhz mmx to get around a k6-iii orbit.

Last time I tried the Pentium MMX 233 maybe I had too high expectations on the multimedia side but it could not (just like the K6 233) help a lot for MPEG2 high bitrate real time decoding, even with some card supposed to offload a bit of that work like the Rage Pro PCI with Motion Compensation which can only help using higher clocked cpus even using the fastest sw possible it could reach 20fps frame rate with lowest details.

Anyway on the Windows usage side, it was a good cpu I wish I had instead of the brand new computer I bought later in those years with the K6-2 but with the Pentium II ready to sell a lot I remember many were only talking about the new generation cpus back then.

The joy of disabling the FPU.

I still see MMX as largely as a 1997 consumer focus thing. Intel's attempt at vendor lock-in (before K6 and 6x86MX came , anyway). Some of the worst games of 1997 required MMX. 😀

kolderman wrote on 2023-05-27, 08:15:

The joy of disabling the FPU.

Yes, sadly. x87 and MMX would have been a nice duo if they could have worked in tandem.

Too bad dual-CPU systems didn't support using X87 on one CPU and MMX on the other, simultaneously.
That would have been interesting.

auron wrote on 2023-05-26, 12:33:

pentium MMX was intel's attempt to sell the p5 architecture for another year with overhyped MMX marketing

agreed, MMX was useless, additional cache was doing all the work

auron wrote on 2023-05-26, 12:33:

tillamook is interesting but did people actually use them in desktop systems back in the day?

Re: Pentium 1 366mhz The Infamous Golden Tiger
https://akiba--pc-watch-impress-co-jp.transla … en&_x_tr_pto=sc
June 6, 1998 you could buy "Golden Warrior" with 3x multiplier in Taiwan. 12,000 to 15,000 yen = ~$120-150. Same price as ordinary MMX Pentium 200MHz, Cyrix 6x86 MX PR266, between K6/233 and K6/266.
Ofcourse two months later Intel released 300A at $150 killing viability of Socket7 for anyone with a little bit of common sense. By December 1998 300A was ~$90 https://akiba-pc.watch.impress.co.jp/hotline/ … 1226/p_cpu.html same price AMD was selling hilariously slower K6-2/300. You could argue oh but expensive motherboards https://akiba-pc.watch.impress.co.jp/hotline/ … 1226/price.html doesnt look al that bad for 440BX, SS7 100MHz AGP boards were >$100
FIC PA2013 (Socket 7M/B, MVP3, ATX, 1MB) 1MB cache, old revision 9,980 ~$90
FIC PA2012 (Socket 7M/B, VP3, ATX, 1MB) 5,800 ~$55
EPoX EP-61BXA-M (Slot 1M/B, 440BX, ATX) Non-returnable 9,800 ~$90
Abit BH6 (Slot 1M/B, 440BX, ATX) no warranty 12,470 ~$110
Abit BH6 (Slot 1M/B, 440BX, ATX) Only for initial failure 12,750 ~$115

so yes, you could buy them in Asia. No, it made no financial sense beyond short period pre Mendocinos.

auron wrote on 2023-05-26, 12:33:

compared to K6-2/III it seems like a pretty irrelevant upgrade option.

People seem to not realize what a ripoff k6-3 was when released new, they actually tried to price them at Pentium 2/3 level.
"First K6-3 to show up in Japan was K6-III/400 at hilarious 35,500 yen aka $295!! in March 1999 https://akiba-pc.watch.impress.co.jp/hotline/ … 0313/p_cpu.html Price of full Pentium II 400MHz or over four Celerons 300A"
K6-2 came down after Mendocino Celeron release, but always stayed above its price points. Speed comparison https://www.cnews.cz/test-historickych-proces … il-i-1995-1999/

kolderman wrote on 2023-05-27, 08:15:

The joy of disabling the FPU.

Jo22 wrote on 2023-05-27, 08:39:

Yes, sadly. x87 and MMX would have been a nice duo if they could have worked in tandem. […]
Show full quote

kolderman wrote on 2023-05-27, 08:15:

The joy of disabling the FPU.

Yes, sadly. x87 and MMX would have been a nice duo if they could have worked in tandem.

Too bad dual-CPU systems didn't support using X87 on one CPU and MMX on the other, simultaneously.
That would have been interesting.

The aliasing of MMX registers to the x87 FPU ones was deliberately done for compatibility reasons with early multitasking OSes. AMD did the same trick with 3DNow! but at least it was more helpful than MMX in 3D rendering since it supported floating-point values.

MMX as a Matrix Math eXtension was useful for all sorts of DSP applications, I think.
Unlike x87, it was using Integer. It was a fine complementary counterpart to x87.

Both working together would have been awesome (I'm repeating myself, I know).
Being a SIMD, MMX could multiply a set of values through a single command.

That's why it had potential for implementing dial-up modems in software, for video decoding, Mandelbrot generators, audio effects for soundcards etc.

If done right, it could have been useful, indeed.
Modem standards like V.90 or V.92 were always changing in the 90s, so it wasn't silly to consider a software solutions.

In ham radio, at the time, soundcards and Windows programs were replacing conventional filter converters.

They essentially were an early form of a software-defined radio (SDR).
By the help of MMX, many things could have been implemented earlier.

Software-based DSPs require a lot of processing power, but are flexible, too. They stand and fall by the quality of the soundcards, DACs/ADCs.

The Windows 95 version of Xing MPEG Player supports MMX, I vaguely remember.

DirectX 6.1 of Windows 98SE used it for certain tasks. That makes sense, because it's from the time when MMX was new.

Of all the SIMDs we have now, MMX looks the most humble to me. It's not all over the top yet.

Edit: Impluse Tracker (IT) had separate MMX-enabled sound drivers, I vaguely remember..
MODPlug Player supports MMX and DirectSound, too, I think.

MMX was great for image processing. Scaling/filtering bitmaps, texturing triangles and whatnot. It was also useful for certain parts of video decoding.
But it came a bit too late, as videocards were taking over these tasks.

That’s when Intel Started the “Bunny Suit People” commercials.
These Bunny Suit dolls lasted all thru the 1990’s and Pentium ll MMX Processor duration.
You could buy these Bunny Peoples at Intel Santa Clara Head quarters Employee and Visitor store in Silicon Valley which was open to the Public. I think it was part of the Intel museum experience open to School field trips and Public.

Jo22 wrote on 2023-05-27, 14:12:

MMX as a Matrix Math eXtension was useful for all sorts of DSP applications, ... dial-up modems in software, for video decoding, Mandelbrot generators, audio effects for soundcards etc.

Yes, It was best used as audio dsp. On games front we got 1997 Ubisoft POD with a huge "Designed for Intel MMX" banner on all boxes https://www.mobygames.com/game/644/pod/cover/ … 90/cover-51358/ while MMX was used by one optional audio filter 😀. Amazingly someone working in Intel "developer relations group" at the time is on HN and chimed in https://news.ycombinator.com/item?id=28237085

"I can tell you that Intel gave companies $1 million for "Optimized" games for marketing such."

$1 million for one optional MMX optimized sound effect. And this scammy marketing worked! Multiple youtube reviewers like LGR or AkBKukU remember vividly to this day how POD "runs best/fastest on MMX".

Jo22 wrote on 2023-05-27, 14:12:

They essentially were an early form of a software-defined radio (SDR).
By the help of MMX, many things could have been implemented earlier.

and thats why Microsoft wasnt really in love with it, there can be only one monopoly 😀
MMX was one of Intel's many Native Signal Processing (NSP) initiatives. They had plenty of ideas for making PCs dependent on Intel hardware, something Nvidia is really good at these days (physx, cuda, hairworks, gameworks). Thankfully Microsoft was quick to kill their other fancy plans https://www.theregister.co.uk/1998/11/11/micr … _said_drop_nsp/ Microsoft did the same thing to Creative with Vista killing DirectAudio, out of fear that one company was gripping positional audio monopoly on their platform.

rasz_pl wrote on 2023-05-27, 20:50:

Microsoft did the same thing to Creative with Vista killing DirectAudio, out of fear that one company was gripping positional audio monopoly on their platform.

This has probably more to do with AC97/HD audio codec domination of the market.

rasz_pl wrote on 2023-05-27, 20:50:

Yes, It was best used as audio dsp. On games front we got 1997 Ubisoft POD with a huge "Designed for Intel MMX" banner on all boxes https://www.mobygames.com/game/644/pod/cover/ … 90/cover-51358/ while MMX was used by one optional audio filter 😀. Amazingly someone working in Intel "developer relations group" at the time is on HN and chimed in https://news.ycombinator.com/item?id=28237085

"I can tell you that Intel gave companies $1 million for "Optimized" games for marketing such."

$1 million for one optional MMX optimized sound effect. And this scammy marketing worked! Multiple youtube reviewers like LGR or AkBKukU remember vividly to this day how POD "runs best/fastest on MMX".

I disagree that MMX was a "scam", there is third-party software that used these instructions with zero involvement from Intel. Example: Impulse Tracker does have some drivers with MMX version for the "resonant filter" feature.

The Serpent Rider wrote on 2023-05-27, 21:36:

This has probably more to do with AC97/HD audio codec domination of the market.

Exactly, onboard audio became mature and the advantage of having hardware accelerated audio were disappearing. It also helps that the questionable quality of Creative's drivers accelerated that dominance.

Gmlb256 wrote on 2023-05-27, 21:45:

I disagree that MMX was a "scam",

In case of gaming it totally was. Intel repeated this trick when introducing P3 with SSE. Intel paid Rage Software to develop "demo" for Comdex fall'99 and pretend its an upcoming game, demo showcasing how great SSE is. That one non public demo, Disptched, was used for two years by Anandtech to prove how fast Pentium3 is Re: Athlon vs Duron 😀 Exactly one journalist managed to experience this demo at Comdex and wrote about it for French hardware portal https://www.hardware.fr/articles/95-5/jeux-optimises.html archive.org https://web.archive.org/web/20210807020608/ht … -optimises.html with pictures
"looking closely at the demo, we can see - as you can see on the screenshots - that the SSE version is less detailed than the non-SSE version (see the floor). Intel would it try to roll journalists in the flour?"

Gmlb256 wrote on 2023-05-27, 21:45:

The Serpent Rider wrote on 2023-05-27, 21:36:

This has probably more to do with AC97/HD audio codec domination of the market.

Exactly, onboard audio became mature and the advantage of having hardware accelerated audio were disappearing. It also helps that the questionable quality of Creative's drivers accelerated that dominance.

exactly the opposite, if "onboard audio became mature and the advantage of having hardware accelerated audio were disappearing" then Creative would be losing market share and Microsoft wouldnt have to do anything. Instead they unliterary ruined Creative Technology, stock went from $22 pre Vista to ~$5 in one year.

Correlation =/= causation.

I think there are two things here:
1) Intel may have used shady marketing for MMX/SSE
2) MMX/SSE may or may not be valid technologies of themselves

As for 2), as I already said, MMX had its applications. I've written my share of hand-optimized MMX back in the day.
It's just that around the same time, dedicated hardware (video/soundcards) became available which could do some of the tasks in hardware, faster and better than MMX could.
On the other hand, they were not fully programmable generic CPUs, like the Pentium MMX.

Considering how many features both cpus and video chips were supposed to "change the world" back in those days when personal computers became the must have home device, I'm not surprised that for the generic consumer none was realistically really needed to upgrade any previous config. When I bought the brand new K6-2 350 config, I could have totally lived with any Pentium MMX/K6 config for Office task and gaming. Same thing for video cards where a Riva or Savage4 card was mostly enough for those games and monitor resolutions.