486DX4-100 MHz Benchmark Battle [ Intel, Cyrix, AMD, TI, ST]

#vintagecomputer
In this video we have a benchmark battle of 5 different 486DX4 100 MHz CPUs. Intel, Cyrix, AMD, TI, ST...
Who will win the race?
At this benchmark we will also test Intel and AMD with WT (Write Trough) Level 1 Cache and WB (Write Back) Cache to see how big the difference is.
We are combining here synthetic benchmarks as well as 3D benchmarks and games. All based on DOS 6.2.
The board I am using is a Asus PVI486-SP3. This great board is supporting all different brands of 486 cpus.
Enjoy the video and please don't forget to subscribe. :-)
Cheers,
Peter
Music provided by WidthSound-NC.
/ widthsoundnc
Music composed entirely by Widthsound-NC
►Download music:
Itunes: / album .
Deezer: www.deezer.com....
Spotify: open.spotify.c....
Google Play: play.google.co....
Tidal: listen.tidal.c...

[@EgonOlsen71]

Actually, Quake isn't heavily using the FPU. At least not in the way as we would think of it today. What it does is more of a trick called overlapping FDIV. For texturing, you have to apply perspective correction (or otherwise, textures would wobble like on the PlayStation 1). Doing that every pixel is too expensive in software on these old CPUs, so Quake does it every 16 pixel with linear interpolation in between. This perspective correction requires a division and divisions are expensive. So Quake uses the fact that the Pentium can execute FPU and integer operations in parallel. While doing the linear interpolation between pixels x and y using the integer units, the FPU is doing the perspective correction for the next chunk of pixels in parallel. A 486 (or K5 or 6x86 or...) can't do this, because on it, FPU and CPU won't run in parallel. So this FPU division actually stalls the CPU and the code only continues once it's done, making them run Quake much slower. Long story short: It's not really heavy usage of the FPU, just a quite tricky one.

[@mtunayucer]

Very nice explanation, thanks

[@AncapDude]

This explains why my friend with a P90 outperformed me so much in Quake with my 5x86-133. Thanks.

[@trajanaugustus8783]

My first Intel CPU was an i486SX-33 pretty good CPU for what I needed back then except in Aces of the Pacific PC game. Had lots of stuttering when numerous aircraft appeared on the screen. Later I upgraded my system to a iDX4-100mhz CPU. It was like night and day, no more stuttering in AOP or any other high density game. I learned a lot from my earlier experiences with the intel and later AMD CPU's. This is true nostalgia, thank you for doing this, brings back great memories from back then.

[@PROSTO4Tabal]

I just discover this great channel, I didn't even know it exists, you been recommended by RetroSpector78 and his 486 build. Nice

[@sweetkaratepunch]

Love your channel. Thanks for answering that DX4 question I've had rolling around in the back of my head for the last few years!

[@CPUGalaxy]

Thanks for your feedback. I am very happy that you like it. Thanks for visiting my channel. ☺️

[@phreapersoonlijk]

I love those self made front i/o connectors/leds/button ! Brilliant !

[@CPUGalaxy]

yeah, they are very handy, especially when you test a lot of hardware on the bench :)

[@Rotwold]

@@CPUGalaxy A video on those alone would be interesting :)

[@TheMovieCreator]

Very nice! These tests show an important factor of Write-Through vs Write-Back cache. WB adds extra complexity and some overhead, but as long as you don't completely fill the cache all the time it gives an overall improvement over WT.

[@RetroSpector78]

Really great video ! I really enjoyed that a lot ... always nice to see the family getting back together for a little reunion.

[@CPUGalaxy]

Hi Davy, thanks for your feedback! Glad to see you around on my channel 😉. Yeah, this comparison I wanted to do long time already coz I‘ve been curious too how this battle would end up. Sending you greetings from Austria 🇦🇹

[@philscomputerlab]

This is a great video! Contrast to the 386, here the Intel is indeed the strongest one. But AMD had the 133 MHz model of course. I like the IntelDX4 because it has good support with motherboards, it is usually mentioned in the manual. With some of the other CPUs it can be guesswork.

[@MrToldi]

Top work! I really thought that there would be just a very slight edge in favour of the WB chips, maybe 5%. Good to know of the difference in performance. Btw. that mobo looks very smart, would be awesome to see a comparison of some different socket 3 motherboards and their chipset performance with the winner of the DX 4 cpu battle. Keep up the good work!

[@warrax111]

Problem is, in this age, that you may be considered pretty lucky, if you own at least one 486 motherboard. :) Who owns 5 or 6, for test, is eighter very lucky, or rich. But maybe someone could borrow for tests.

[@CPUGalaxy]

I will do this comparison and benchmark with some boards. Will cover following chupsets: VIA, SIS, Intel and UMC ☺️

[@clintthompson4100]

All the 486 chips are great especially for day to day applications, but seeing the WB is clearly better choice I would still pick the AMD chip. Being an AMD fan myself and for the price to performance I still feel it would be the better choice and with the 133mhz version and the 133ADZ model wich is actually a 160Mhz 40Mhz X4 it can be a real beast in the 486 chips. Also the Cyrix 5x86 133 is a beast but a rare and hard to find chip. The 486 was a great cpu. Such a great time when hardware was changing so fast and in leaps and bounds.

[@rodrigoramirez8627]

Don't believe it, I was lucky enough to have the intel dx4, the am486 dx4 and the cyrix dx4 since all of these were used by IBM (my dad worked at IBM in the 90's) and at some point when the equipment changed, they threw them whole back then Anyway ... the am486 at 133mhz is low in performance compared to an intel dx4 at 99mhz (100mhz) and the difference is very noticeable I do not intend to put together a discussion, just share my experience with these processors It goes without saying that I am not a fan of Intel processors since I have a preference for AMD from the Athlon 64 X2

[@w00tDr]

Thanks for this study. I have been enjoying your channel. The results were close to my expectations I have one minor critique concerning the bar charts: when the bars are zoomed in, it appears that the differences between each processor are much greater than they actually are. The viewer should have a better visual sense of the differences if each bar was plotted from the zero point. Other than that, great job, and I hope to see more.

[@CPUGalaxy]

Thank you very much for your FB. You are right with your critique and I will consider that with my next bar charts. 😉

[@anomaly95]

I would say display both. Graphs from the zero point are going to show just a tiny difference at the end and not show the differences very well between these CPUs. Zero point graphs would be good for comparing all of these, lower clocked 486s and 60 or 66MHz Pentiums though.

[@grapsorz]

the moste important test i was waiting for was the OC test. and yes we had started to OC the CPU's at this time in the game ;) i hope for a follow up video here to make it complete ;)

[@CPUGalaxy]

I planned an OC video but not with DX4s. I want to max out the Asus board with OC PoDP to 100 MHz and AMD-x5- to 160 MHz to compare... Maybe I will add here then a DX4 then too.. 🙃

[@grapsorz]

@@CPUGalaxy what is the max speed you can get the DX4 100 to? i remember 140+ speed's.

[@xenuno]

Should be able to handle 3 GHz on liquid nitrogen

[@MrJorgalan]

I love your videos about those old CPUs. My first home CPU was an Intel 486 33 MHz; it was the preferred brand, although in my professional training, I used a 286 to program in COBOL. Later, I got to know AMD with a friend’s DX4, which was quite fast. As for Cyrix, they were always listed in computer magazine price lists as the cheapest, even below AMD.

[@noth606]

Nice! I had expected it to be a bit of a closer race than it was in the end, but nice to see someone with all the gear put in all this effort to bring us this info. Thank you!

[@CPUGalaxy]

Thanks. A follow up is planned. I found out that there are a lot of functions on the cyrix which needs to get activated bit a bit setting tool. That is boosting than the cyrix extremely. It might change at the end the rating. 😉

[@noth606]

@@CPUGalaxy Cool, will be interesting to see what kind of difference it makes. Thanks again :-).

[@Blade2086]

Hello, great review! Thanks a million! :-) Can you also test the AMD DX4 with WB and 16kb cache? It will be a more fair comparison vs Intel’s 16k cache... AMD Am486 DX4-100V16BGC

[@Tigrou7777]

Great video. You show Duke Nukem 3D at the end but did not benchmarked it. My guess is it will give something similar to Doom (since that game is not using FPU at all).

[@AjinkyaMahajan]

Wow I was surprised to know ST manufactured 486 🧐 Nice Video Cheers ✨✌

[@CPUGalaxy]

Thank you 🙏🏻

[@NielsHeusinkveld]

If only we had such reviews available back in the day! Personally I would prefer if your charts always have the scale from zero instead of sometimes being zoomed in, like SpeedSys going from 34 to 44, but then others starting at 0. I think most humans judge the performance by the length of the bar, and that only 'works' when you start from zero.. In my opinion it gives a clearer picture of the differences, even when they sometimes are very small, I prefer to see similar bars rather than a zoomed in bar that seems to show a 'big' difference.

[@WR3ND]

Very nice. That Intel 486 DX4 100 with WB is the one I got for my pre-Windows 95 DOS6.22/Windows 3.11 build. (Yes, I'm aware there are faster 486 CPUs). Cheers.

[@ted-b]

Great stuff and well made too, thanks!

[@luckyluckydog123]

Great video! any chance we might get to see the results for those 30 vesa video cards?

[@CPUGalaxy]

nice idea. Could be an interesting video to benchmark some vlb cards. I will keep this in my mind. 👍🏻

[@arthurmann578]

That would definitely be a cool idea as I still have quite a number of the "oldie but goodie" (and some NOT so good) video cards lying around! 👍👍

[@intrinia]

Ich bin so froh das ich diesen Kanal gefunden habe! :D

[@mvickers03]

I’m currently building a 486 machine and have chosen the Overdrive dx4 100. Can’t wait to have a play with it.

[@HeavyD6600]

Great video, dude. Really unexpected results. Now on to 5x86s!

[@CPUGalaxy]

yeah, maybe a battle between Intel Pentium Overdrive for Socket 3 overclocked to 100 MHz against a 120 MHz Cyrix 5x86 ... could be interesting 🤔 😃

[@HeavyD6600]

@@CPUGalaxy Good idea. I'd like to compare with my AMD 5X86 133 and Cryrix 5x86 100GP that runs at 120 no problem!

[@mihaigabrielbabutia4595]

Just discovered this lovely channel with awesome content! Keep up the good work, I love it! My very first PC was a AM486DX4@100 and watching your videos brings back pleasant memories.

[@CPUGalaxy]

Thank you!

[@crusader2.0_loading89]

You bring back so many memories, thank you

[@bt3779]

Great video, but I think it would be interesting to see a comparison of the prices from a point in time when they were all out, and if possible sales figures or market share to round out the picture.

[@1337Shockwav3]

Can confirm that WT is faster than WB in Doom as I just tested that yesterday with Intel DX2-66 processors on my ASUS VL/I-486SV2GX4 1.7 board. Due to the lower clockrate I had a performance difference of roughly 5% How's the PCI vs. VLB performance on the Asus 486SP3? I tested a CL5428 VLB against an ET6000 PCI on one of my VIP boards yesterday and was shocked how the PCI card performed roughly 40% slower - albeit the ET6000 being a superior design and the VLB running at "just" 33MHz On a sidenote: A friend of mine happens to have an AMD DX4-100 with 16K WB cache (Am486DX4-100V16B) - would be interesting to see how those fare against their Intel counterparts. Some boards seem to be extremely picky about AMD DX4 CPUs and usually seem to assume 8T models.

[@Xaltar_]

Great video, brings back a lot of memories. I remember frying my SX 25 by overclocking it to 40mhz 😂 It ran great for about 5 mins before smoke started pouring out of it, even with a cooler on it. I already had a DX4 100 waiting to go into the system and it had worked fine @33mhz so I threw caution to the wind and tried to see how fast I could make the SX 25 😁

[@mlodzin90]

Smoke? How you overclocked it? What method? Raising FSB could only make your system unable to start, not frying a CPU...

[@gentuxable]

@@mlodzin90 surely he overvolted to go along with it.

[@aldergas01]

Very nice work, thanks you. For you hard testing we know how long it can take.

[@CPUGalaxy]

thank you. indeed. that testing takes it time. 😉

[@_f355]

6:35 that's interesting, if the memory bandwidth is 110 MBps and the L1 cache is 94 MBps, what's the point in having the cache? am I missing something? that said, the NVMe SSD in the computer I'm writing this on does about 1.8 GBps at writing. we've come a long way since DX4 days :)

[@vulturius7664]

Great channel ... Arnold ;-)

[@Frogmelon]

I remember overclocking my Cyrix DX4-100 to 120Mhz. :)

[@ferrumignis]

40MHz FSB? That should give a really useful speed improvement

[@M0F0750il]

I love your videos! These benchmarks are what I would have and did but on a smaller scale back in the 90s. Hope you can find the time to make many more benchmark comparisons like this one. Just wish I'd found your channel sooner, I must thank EEVblog for sending me there. Happy New Year :)

[@CPUGalaxy]

Thank you very much. Happy new year too 🎉😃

[@HerecomestheCalavera]

Great video, a very thorough comparison.

[@casualretrocollector]

Excellent video! I found recently an old laptop that wouldn’t boot. Inside it I found an Intel dx4 75 mhz. I installed it on my old pc set the bus speed to 33mhz and it runs at 100 mhz no problem.

[@movax20h]

They are pretty identical in performance. Max 10% difference. The graphs are stretched and not all start at 0, so differences look bigger than they are.

[@Protosszergling]

Quake, Need for Speed 1, and Starcraft were the "heaviest" games I used to play on my PC with a 486DX4 processor. Good times.

[@diegodonofrio]

Great comparison 👏👍

[@RuruFIN]

I've always thought that they were pretty identical in performance, pretty huge differences between the "same" CPU.

[@mattj5155]

The differences aren't "huge" though. The difference between the fastest and the slowest is only around 10%.

[@Xaltar_]

It's the way his graphs are scaled that makes the differences more pronounced seeming. For the 5 - 10% performance deficit (task dependent) between the intel and the other CPUs you would have paid anywhere up to 40% less for the slower Cyrix and AMD models. It was a case of pay through the nose for the best or get the best value and sacrifice a bit of performance. Cyrix and it's derivatives were by far the cheapest options and offered the best value for money of the group, every CPU shown here was a significant upgrade from the DX2 66 and slower 486s. I remember upgrading from an intel SX 25 (overclocked to 33mhz) to an intel DX4 100 (WB) and being blown away by how much faster it was. Doom, Duke 3d and Rise of the Triad got a lot of screen time on my 486 DX4 100 :)

[@ionstorm66]

@@Xaltar_ yeah back in the day it wouldn't be a mhz to mhz comparison. The cost difference would of made the AMD compare to a 75mhz DX4 and the Cyrix would of been about the same as a DX2.

[@MichaelClark-uw7ex]

I built several machines with Cyrix CPUs back in the dark ages of computing. The Cyrix MII 250Mhz was a great chip in its day.

[@GabrielZ666]

Amazing video, I love Benchmarks! I have an Intel Overdrive DX4ODPR100, will it perform like a regular DX4 with WT?

[@CPUGalaxy]

Thanks, yes, the ODPR100 will perform like a regular DX4

[@GabrielZ666]

@@CPUGalaxy Awesome, thank you!

[@sedrosken831]

That's actually a complex question, dependent on more than *just* the CPU. Your chipset, memory configuration, l2 cache/lack thereof (my ODPR actually disables the L2 on my board, can't figure out if it's the CPU itself or my BIOS) can have a big impact. Personally I gained much more from moving from my RAM configured at 2WS to 0 than from the DX2-66 to DX4-100 upgrade.

[@Trancelistic]

Just how I remember Cyrix, in all generations. It sucked at games but the rest was quite fine.

[@RodBeauvex]

Come on now, treat that Tseng Labs card to a second megabyte of RAM. :D I have one of those voltage adapters for old motherboards and once compared the DX4 to the Am5x86, and remember being kind of shocked how close they were. The AMD won, but not by a huge margin.

[@CptJistuce]

The TI chip still greatly surpasses all the others in the key "number of colors on the package" metric. Thing's downright festive.

[@arjovenzia]

The AMD DX4-100 has a very soft spot in my heart, being the first PC I built, and was my own, so free to fiddle and tinker with, unlike the family computer. Modified n tweaked just about everything on that rig. And I'm still an AMD fanboy to this day.

[@anomaly95]

I sold the vast majority of these CPUs retail back in the day. Unfortunately, the WB and WT aspects were lost on 99% of people. Of course, it was the very early internet days and most people got their computer info from magazines. Not many people cherry picked s-spec or part numbers. Also, a lot of people were just happy to be upgrading from 386s and early 486DX/SX computers.

[@AncapDude]

True, never heard from WB/WT until now 😂

[@RetroSpector78]

Great video ... my AMD DX4 (8kb WT) with same SiS chipset performs a lot slower (with PCI card I only get 50.9 on 3DBench 1.0c, 30fps on doom, .....). Now I'll need to find where the bottleneck lies :)

[@jacq0272]

Nice video!

[@lordelectron6591]

This is the first time I've known cpus other than intel or amd

[@misterspock1]

Good Cyrix video: kzbin.info/www/bejne/n4iqcpelgt9kmZI

[@KJohansson]

Like the clean and proper parts you use, no dirty scuffed parts. They all look like they where bought new today. Regarding the lower score for the WB caches in some tests, could they be related to less sequential excecution than the more synthetic benchmarks?

[@rodrigoramirez8627]

What memories ... one of my first PCs that I built, an Intel DX4 at 99mhz on a 486 compact presario motherboard, with 32MB of RAM, with a 512k Trident GPU that I don't remember the model

[@86smoke]

Probably 9000 series

[@dizzywow]

100MHz. 33.3 x 3.

[@catsspat]

Brings back so many memories.... Oh crap, I just revealed my age range.

[@the_kombinator]

My first 486 was a DX4-100 and I was about 14 when I got it (maybe 15?) - yeah I put the CPU in backwards. CPU survived, the motherboard did not. It was 3 days old so back to the store it went for a "uh, this didn't work" (with the CPU replaced correctly) - 15 mins later tech comes out with a new one, warranty claimed. Nice board, 3xISA 3xPCI, eventually I had a 5X86 P75 in there ;) I learned a valuable lesson, but it was useless on all subsequent keyed CPUs.

[@B24Fox]

Too bad you couldn't get in there the AMD DX4-100 with WB 16kb cache. Maybe a short video just for it? :D

[@amedvedev]

Cool comparsion! Why no amd486dx4-100 16kb write back? difference in speed is pretty interesting

[@MadRat70]

486DX2-100 on non-standard 50 MHz FSB was great for video encoding in its day.

[@fradd182]

Just as i predicted prior to watching benchmark results, intel pushes ahead a bit, the rest of the pack lags behind. But still, AMD had dx5 version that was overclockable to 160Mhz, i think.

[@CosmoRiderDE]

Some made it to even 180 or 200. There are boards that produce whooping 60mhz bus clock hehe

[@clintthompson4100]

The AMD 5x86133ADW model would run at 133Mhz normally setting the motherboard to a 2x setting then the chip would double that for the 133 speeds but the ADZ Model was underclocked to 133 but its true speed was 40Mhz bus X4 for a 160 Mhz speeds and they even made a 50Mhz bus X3 for a 150Mhz also under clocked to 133Mhz but can't remember the model number of that chip and only a few later 486 MB supported that kind of bus speeds but they had stability issues since most PCI or ISA cards did not enjoy a 50 megahertz bus and having to go into the bois and doing a PCI/ISA bus divide just to try to get the cards to run stable. ISA card ran happy at 8 Mhz and most would deal with 10MHZ but anything after that they were not happy and ran with stability issues or just killed the cards after a short while.

[@CosmoRiderDE]

@@clintthompson4100 as far as i know the ADZ was only different to thermal limits. ADY and ADW = 55°C limit and the ADZ had around 80°C limit. Whats different inside i dont know. Maybe a smaller lithography. However, these were all beasts. And yes, stability was a problem at times. The only mainboard i have played thoroughly on with this CPU is the legendary fake cache PCChips M919. But did not go beyond 160Mhz. I made a little mistake above, what i meant is that some boards do support 60mhz (undocumented) bus clock. And the M919 is one of these. so there would be theoretical even 240mhz posible...

[@GadgetUK164]

Fanastic video =D

[@spacemanbowkonami]

Thanks for reminding me of fhe Tsenglabs ET4000! I used to have that one and forgot all about it. Cyrix was always the underdog CPU that we all hoped would keep the competition going but just didn't make it. And as far as AMD and Intel go, realistically, up until Ryzen there was never a compelling enough reason to go with AMD. Sure, they were the first to push for 64 bit, and Intel had their integer flaws on the Pentium. But it pales versus the AMD's that went up in smoke from having no thermal protection and a poorly attached cooler.

[@86smoke]

Well, it was worth to switch to AMD when late PIII and early PIV were realeased. Athlons (Barton core especially) were an awesome machines back then.

[@dizzywow]

@@86smoke Right, the Athlons were the best, for a while.

[@Paar86]

An user on Vogons forums gave a tip to run L1 cache in WB and L2 cache in WT with IntelDX4. Tried it and this combination gave me the best results in all benchmarks (used the ones from Phils package). Could be interesting to compare. Still, I'm trying to see comparison of A486DX4 with 16kB L1 WB and IntelDX4 with L1 WB cache. I heard that IntelDX4 has some integer optimizations/enhancements in comparison to DX2 so I was wondering how they compare to CPUs from AMD.

[@thomasmica2856]

Where is UMC ?

[@AiOinc1]

Out of curiosity, what is the state like with DX2 / 66MHz chips? I finally managed to find a socket 3 board that did not have damage from a barrel battery and over the years I've collected 4 different socket 3 CPUs to choose from. An Intel 486DX2-50, which I've discounted because of it's lower FSB and lower clock, an Intel 486DX2-66, the ST 486DX2-66, and the OEM use TI 486DX2-66. Is there any worthwhile difference between these? I noticed the ST chip has "Heatsink required" and the DX2-50 has one permanently attached. I would love any insight or suggestions you might have. I've only got the chips for 128K of L2 cache, if it helps any. EDIT: I hope this helps someone, I did some testing between all of my chips now, and of course the DX2/50 performed by far the worst and could not clock up (Was to unstable at 66MHz to do anything) so it wasn't worth it. The ST performed identically to the TI chip (I think they are the same inside) but the ST ran much, much hotter and needed higher voltage to stay stable. At 80MHz it was already at 5v and extremely hot even with a heat sink. The TI runs the coolest for sure, and was stable at 80MHz at 3.3v! I was surprised by this for sure. The Intel DX2/66 seems to be right in the middle of the pack, average overall. I'm going to do some further testing with overclocking as the TI seems to take to this really, really well. Interestingly, the TI lost performance in Topbench and 3Dbench for some reason, though. I'll report back again with findings! EDIT2: The Intel chips really really hate to be overclocked, at least both of mine do. The DX2/50 will not work at 66 and the DX2/66 will not work at 80. Can't get a difference no matter what I do. Think the TI will be my winner here. I hope this helps someone out! Of the 4 I tested, the best overall was "TI486DX2-G66-GA" and a good runner up was "ST486DX2-66GS" The ST says "heatsink required" and if you're running it >3.45V and >50MHz you absolutely need one. Below that it runs generally cool to the touch. It has similar performance to the TI but runs very hot. The TI chip runs really cool and low voltage, and performs slightly better while also being more stable. How interesting! I guess they're not exactly the same inside after all.

[@jasejj]

One nice result from this is it confirms once and for all an argument I had with a PC builder at the time, who insisted that the Cyrix-branded chips were worth the premium they had on the open market at the time (typically 20%more expensive) than the SGS and TI chips. Complete tosh, as it turns out 😁

[@wuuduu609]

my dream hardware, from beginning of 90s :-)

[@brainthesizeofplanet]

Ohh what times, 1,2m transistors... Wow look how far we have come

[@AaaAaa-ly3on]

-I don't believe a single thing you said - cause my first PC had Cyrix CPU and I loved that Computer!.. ;))

[@CPUGalaxy]

Then watch this here please. ☺️ kzbin.info/www/bejne/faGVf62tZcl7Zrc

[@AlyxSharkBite-2000]

I know I’m late to the party but on the ibm chip they had 2 486s. The Blue Lightning which was off the Intel Core and then the “normal” one which was the Cyrix core

[@AG-jj3lx]

It is a shame Cyrix could not go down the ARM route with licensing out their designs. Would have loved having multiple competitors for AM4/AM5 or Intel socket types :)

[@johng.1703]

I'm trying to remember how I had my AMD 486 set up back in the day, I remember it was on a writeback board with 256 or 512k of stick cache, and was clocked on a 50MHz bus speed. I remember it was close to a Pentium 133 in benchmark results, but I did have it overclocked. think it was like 2.5x 50 for a clock of 125Mhz, I might have even pushed it to 3x 50, I remember it being fast for the time.

[@ralcool5932]

Posting this, as I haven't seen any one else try. Intel DX4 OEM CPUs also had a 2x, 2.5x, and 3x multiplier modes. i was building systems back in the day. You can see in the jumper charts 2x ,3x Have you tried this?, Mr CPU Galaxy. Thus a 50Mhz cpu bus on 2x multi for 100mhz. Best vesa speed, but difficult to make stable.

[@wishusknight3009]

There was a run of amd 100s with 16k of cache as well, but the do perform identically to the intel counterparts. They are essentially the same as the x5 but for some reason mine has no 4x multi.

[@todprog]

Cool test, it would be interesting if there is info to compare also the prices. I guess Intel ones were way more expensive than the 5-10%? performance difference to the AMD's, and even more to the Cyrix's.

[@the_kombinator]

Oh man, I've got 20 or so different 486 CPUs (from SX-20 to Overdrive) and the only one I am missing is a DX50 (I have an SX2-50, but not the same - It would have been nice to see if my Mach32 VESA could handle a 50 MHz bus speed). Apparently my board will support an i487 as well, so that'll be interesting as I have one. Looks like my benchmarking plan was quite similar to yours.

[@krz8888888]

What a beast!

[@VK2FVAX]

Enjoyed this greatly. Thank you. Would have loved to see cachechk.exe for them. Does the Cyrix or Ti have any cache tuning utilities and did you run them? I know with the DrX/2's the default performance is noticably much less and Cyrix was still doing this in the PR200+ era (pre-MMX at least). Some of these features were auto-disabled for older boards that had difficulty with things such as write-back cache and DMA regions.

[@Crazy80ivan]

I wonder if the tool u used voor the Cyrix 5x86 also works with the DX4 from Cyrix, because it seems way to slow.

[@WooShell]

I think the "wrong" CPU speed in CheckIt is some normalized number based on the Dhrystone measurement (comparing to a 286-6 AT was commonly done). I see the same figures when benchmarking with Landmark, and they explicitly state this in their documentation.

[@AdamoReggiani]

Ciao all, as an old fan and user of this kind of computer back in the 90's I really appreciate your channel and videos. Please next time try to use a geometric mean as a way to compare different results instead of fudging the scales. The geometric mean gives much better and realist result over numbers of different scales. en.wikipedia.org/wiki/Geometric_mean

[@TheSwillMan]

very shocked there is so much difference between AMD and Intel versions! i thought AMD was producing exact clones of Intel CPUs during this era.

[@davidca96]

I had an AMD 486DX4-100, it was a beast back then.

[@OhFishyFish]

The only benchmark that matters in those is Duke Nukem 3D. 👍

[@VioletGiraffe]

Thanks for the benchmarks, obviously it was a ton of work. But man, I hate bar graphs that don't start at zero! They exaggerate the differences! @7:30

[@jasejj]

I always compared these CPUs by cost, not advertised speed (still do to this day). As I recall, by 1996 the Cyrix chips were very much cheaper than the AMD and especially Intel. For example an SGS-branded DX4-100 was around a third the price of the Intel, and the DX2-80 was a real sweet spot when it came to price-performance when putting together budget systems. If the Intel was £100 and the SGS maybe £30-35, it wasn't fair to compare the two. If the SGS was compared to the like of the Intel DX-33 or DX2-66 (which were still more expensive as I recall), the Cyrix line absolutely blew Intel away. The chips do the same job, so that's really the best way to compare.

[@TheLemminkainen]

il bet Cyrix was much cheaper than Intel :)

[@MichaelClark-uw7ex]

It was, I built several machines with Cyrix CPUs they were even cheaper than AMD at the time and had comparable performance IMO.

[@TheLemminkainen]

@@MichaelClark-uw7ex I had 6x86 120+ back in the days. Sadly it went to junk yard 10y ago.

[@retrocomputinggrotto]

Some proof that 'back in the day' Intel really was the king of the CPUs - with recent stock shortages of Intel chips (and their increasing costs) I wonder if AMD now has that crown?!

[@protonjinx]

I read some post somewhere that some DX4-100 chips can run as DX2-100 (50MHz FSB). Is that something you have info on?

[@CPUGalaxy]

Interesting that you ask that coz some days ago I was trying that. The board was posting but crashed always during startup. Actually I guess a performance issue of some peripherals. But i am in that topic and if I can get it to work I will make for sure a follow up to see how much performance we can gain out of that.

[@protonjinx]

@@CPUGalaxy Test the mobo/gfx with a DX-50 to see if it can hack it.. I was told only a few vesa cards can run stable at 50.

[@CPUGalaxy]

yeah, but to keep the performance I need to use the ET4000WP32 or the ARC1000 video card. Those are far the best ones. If I would use a cirrus logic or others I would definitely lose the performance I could gain by the 50Mhz bus clock.

[@vapourmile]

How about an FPU benchmark? The 8087 80287 and the Weitek, AMD and equivalents.

[@CPUGalaxy]

Thanks for your input. Indeed I was thinking about that already. I might do one for 387 FPUs.

[@vapourmile]

@@CPUGalaxy Another I'd like to see is cross-architecture benchmarks, pitting the Motorola 680x0 series against the Intel 80x86 series with various integer and floating point benchmarks.

[@zosxavius]

There is a FPU benchmark in this video though its not very extensive. Quake hits the FPU pretty hard so the 10% loss to AMD mostly matches benchmarks pitting the two over the years.

[@Flashy7]

Did you test the other Cyrix-core CPUs also to see if they are really same to the last transistor?

[@CPUGalaxy]

yes, and i got the same values.

[@Windows3x]

Very nice video ☺️

[@ruthlessadmin]

My second CPU was a Cyrix 486DX2 80mhz... This kind of makes me sad to know the only reason it seemed so fast, was because it was clocked a bit faster than the 66mhz chips I was otherwise used to.

[@ewetzlma]

Where can I get these led and power buttons? Thanks 🙏

[@dabombinablemi6188]

I'll probably never have a board for it - but at least I know how my Ti 486DX4-100 performs. Unfortunately, my 386DX-40's board only officially supports up to 486DX2-66.

[@zosxavius]

They are a lot closer matched than I realized. I always knew the AMD was slightly inferior, but the difference is pretty minor isn't it? I didn't realize Cyrix was that competitive by then. Their first few 486 chips were much worse.

[@charonunderground8596]

Very good video ! thx ! Maybe try PCI graphic card and compare to VLB

[@CPUGalaxy]

I did already and used a PCI ET6000. The performance of the VLB ET4000 was much better. Although the ET6000 is a much better card. On a Pentium mainboard I can get much better results with the ET6000 and it seems that is an implementation issue of the pci bus on the 486 mobo.

[@bstar777777]

@@CPUGalaxy I was looking for a comment on this... this shocks me. My ET6000 PCI card destroys my VLB cards (using a dx4 100 in both cases), but I do have to use different motherboards to test (not so lucky to have this Asus board). I wonder if Asus did some voodoo magic on this board to get VLB performing so well. Back in the day I bought a very high end VLB Number9 card and it couldn't compete with budget PCI video cards.

[@dcikaruga]

You didn't talk about any cache RAM on the motherboard, I could see 256K on boot-up but didn't some support 512K?

[@big0bad0brad]

I could be a bit off on this but I seem to remember some cache restrictions where when using a larger cache configuration, it would split the cache into two banks that handled the first and second chunk of the maximum RAM size (could have possibly been an issue being short a bit in the cache tag RAM). So you had to really load up on RAM to have any impact, and even then you'd almost have to specialize RAM allocations to get any kind of consistency and get a boost beyond "at least you have cache for the upper RAM now". It may have been specific only to some chipsets. But with this board only having 2 SIMM slots maybe it wouldn't even apply.

[@chrochtislavchrochtov1169]

what about adding oc AMD 100dx4 oc to 120mhz or 140mhz ? and pentium overdrive for some comparsion :) what is faster ... im running 486dx133 wb on ~160MHz stable now im not sure what is faster, if something ...

[@arthurmann578]

Though Cyrix CPU's seemed to run slower and sometimes WAY hotter than its competitors, I wish that they were still in business to give Intel and AMD a "run for the money!" I still have a few of them lying around somewhere in my home. 🤔

[@laurdy]

I wonder how much bus speed effects performance: try 486's at 4x 25MHz, 3x33MHz and 2x50MHz. As far as i'm aware AMD used the 5x86-133 core in some of there late model 486DX2 chips with the multiplier set to 2/3

[@ccanaves]

Do the same test with the DX2's

[@UnrealVideoDuke]

Who is going to win the race? @0:27 My bet is the Intel CPU. Others used work-arounds with the limited 486 reference documentation. Unfortunately we do not see a NEC 486 CPU or a few others that used to make CPU's. Nice to see the Texas Instruments and ST CPU's as they were not very commonly used. IBM used the Cyrix brand for a while and I am not too clear on the history but VIA technologies still make the CPU's to this day. AMD soaked up ATI and are producing both CPU's and GPU's. Intel soaked up S3 and had a very decent GPU back in the late 90's but they were literally behind in GPU processing and went to AMD for the Vega GPU reference... Too much history to talk about but leave that up to other to tell

[@UnrealVideoDuke]

Interesting that both TI & ST used the exact Cyrix reference and I'm not surprised that they were behind in most of the benchmarks as they are not really used in high-end applications but used in basic desktop applications like data processing. Kinda surprised at the math-co-processor of the Cyrix to be a little better than I thought. Intel was always known as High-End for Desktop applications, processing & games. AMD was always known to be a close competitor to Intel but for those who are on a budget.