The Cyrix story is a tragedy. Their M3 was an absolute masterpiece of engineering that supported both 3Dnow, and MMX, and had an integrated memory controller. Intel's compiler shenanigans were directly aimed at Cyrix who was finding some significant commercial success. My first job was selling computers for Office Depo starting back in 1995. Cyrix was selling like hotcakes.
Finally. Somebody who was there in the 90s when home PC building was picking up and enthusiasts were really humble. Not your current horde of overconfident and misleading RU-vidrs.
I loved this video. I would like to see these compared to an IBM Blue Lightning CPU. I used to have a BL75 chip and board back in the day. That was the board that taught me the four black power connections go together.
He seems to be like Techmoan - every time you think there is nothing left of odd tapes or formats, Techmoan pulls another strange format or device out of the drawer and presents it. CPUGalaxy must have an huge stock of CPUs in all differnet versions and ages so I think we will get to see some nice and odd CPUs in the next years. ;-)
Thank you for all that you do. These videos are great for looking back at all the hardware we saw but never got to own back in the day. I have seen these UMC CPUs in a magazine called Asian Computer Sources from 1992, which was a magazine full of ads from taiwanese hardware makers. Never thought the UMC was a pretty good CPU. It's also notable that in 1995 or thereabouts UMC spun off their chip design center to become Mediatek.
Few videos back I commented your graphs, especially one kind of graphs. Today, I do have to reiterate that I love your work and I also truly appreciate that you've put the percentage difference on the graphs as well. Thank you!
I had a UMC U5SX 40MHz CPU back in 1994ish and it always felt like a flying machine compared to SX33 and even DX40 systems I used at the time. This explains why!
Not surprised that Intel sued UMC. I'm guessing there must have been some infringing stuff there if UMC settled out of court and basically gave up on the US market. Also I the MOD music in Epic Pinball. That stuff takes me back! I really miss music made that way!
" I'm guessing there must have been some infringing stuff there if UMC settled out of court and basically gave up on the US market."... that's not how it works, if they were infringing Intel would have stopped them from selling UMC chips EVERYWHERE. Id say UMC would have actually won but intel being the much bigger company it is, has more money to pay lawyers drag things out in the courts, so as a compromise gave up on the US market rather than fight it out in the courts, draining what money they had
That's an interesting 486 motherboard you have there, especially that blue socket! I found a motherboard quite similar recently, it performs really well. Great video!
Wiedermal ein schönes Video, die besten Grüße für dich und bleib gesund! Ich warte immernoch sehnsüchtig auf die CPU :o kannes kaum erwarten. Habe extra einen kleinen Sockel für meinen Schreibtisch gebaut auf dem dieser seinen Platz finden soll. Lg,
The 486DX-33 averaged 3.88 watts. The U5S-33 averaged 2.15 watts. Let's assume 20 cents per khw (reasonable in the US). 8760 Hours in a year. The 486DX-33 would consume (8760 / (1000/ 3.88))=33.98kwh, or about $6.80/yr. The U5S-33 would consume (8760 / (1000/ 2.15))=18.83khwh, or about $3.77/yr. The best you could possibly hope for is saving $3.03 in power if it's on 24/7. In the Pacific Northwest, we're at $0.105 per kilowatt hour, that'd be around $1.50/yr. This is how absurd labeling the CPU "green" really is. Green hard drives are almost as bad. And by comparison, the rest of the motherboard - chipset, RAM, add-in cards, monitor, fans, etc... DWARF the CPU utilization.
Love waking up watching one of your new videos! 😊 my first pc was a Syntrex (intel) 80286 8mhz/16mhz with 640kb RAM and 384kb extended memory for a entire 1024kb of RAM!! Then my aunt gave me quantity 2 intel 80486 dx66 entire systems! I destroyed one of the 486 chips by bending a pin while examining it. I was 10 yrs old I was so clumsy. I learned that day to BE CAREFUL with the cpu pins. I love your channel! Thanks so much for sharing. Much respect from Florida.
Ah, so you have discovered the magic of the wonderful U5. It is a bittersweet thing, amazing because it flattens everything else, the 40MHz version will keep pace with an alarming number of Intel DX2 systems and no single-clocked SX will catch it. At the same time it's sad, because you have to wonder what UMC might have achieved if they stuck around. You can sure bet Intel were soiling their pants, it would be fun to read the notes from that panicked meeting in the boardroom, upon Intel realizing that not only had UMC made a fast CPU, but they could sell it at a lower price also. I see an ECS UM486V AIO, just like my DX-50's board, but with onboard I/O. You can keep your fancy onboard IDE, because the danger factor of VLB SCSI at 50MHz is just too appealing. The U5SD is actually another SX chip, though some have theorized they were DX models that had their FPU's connections cut by laser to disable them. They use the DX pinout (two pins are swapped) though. Hopefully this naming conundrum wasn't indirectly my fault, as you seemed to be quoting from an article I wrote, which was cool to hear, though it seems someone has since added a note about the U5SD to the info panel and I'm not sure what they mean by it. The U5D would be the DX version, but they're impossible to find, even records of them, the only one I knew of in a collection was dubious. Probably as much chance of finding one of those WinBond W89K chips that were meant to fit in modified 486 motherboards, which is to say, almost none.
The Intel 486 SX 33 was my first CPU in 1994 and was built in a Packard Bell PC with 4MB RAM. Damn was it slow! I was so happy to be able to upgrade it in 1995 to a Intel 486 DX4 100 with then 12MB of RAM ;-)
Ok as soon as he dropped the CPU all I could see in my head was his wife smacking the back of his head. And with her Russia saying butter fingers. Love the channel, also I think you and your wife look awesome together 👍🤘🤘
Remember these bad boys. Most of them had no problem running at 50Mhz on suitable board. Way cheaper then iDx2-66, but way faster at doom,hexen, heretic and duke. I was selling pcs in that day and wondered why they suddenly disappeared. Likewise the cyrix 5x86-100. ALL of them had no issues running at 120Mhz. Wondered if you could find a wierd 5x86 cpu. Long after they'd stop being made, I came across a weird cyrix 5x86 that had an additional multiplyer available. I can't remember the precise details, but I seem to recall it overclocked to a weird speed like 133Mhz. Honestly not making this up. I loved that 586 and collected them for a while. There was a weird 586-120 you could overclock to 133Mhz. Unfortunately it didn't seem any faster, but it was still super cool having a 133Mhz cyrix 586. Happy Days :) Thanx for this video, keep up the good work
Just curious on one thing, let's say some US citizen is quite interested with this vintage UMC CPU, and wants to collect, does the restriction printed on the CPU still apply today?
Please do a video on a fully decked out 486SX2/50 with it also running Doom please.. I'd love to know how a sx2/50 with 8mb ram and a decent VLB video card stacked up against an average DX2/66.. Also I do know that the ET4000 is one of the best video cards of that era, but maybe also bench Doom with a speedy S3 805 VLB accelerated video card too as I had one of them and it perfomed somewhat better in DOS games like Doom than my ET4000 Isa card did (although my ET4000 Isa card was still by far the quickest of the 16bit video cards)
Just wondering would the under clocking of the Cyrix chip play a role at all? I know it’s most likely it’s 2mb of cache, but I’d love to see it running at 40MHz up against other 40MHz chips, preferably in 40MHz designs of the other two chips if available! Love your videos!
Makes me wonder if intel or AMD were to make a special edition 486 with modern features...... 10X internal multiplier, 512KB L1 cache... We can dream :)
I can't believe how bad Doom runs on them.. I thought I read from iD software at the time that a 33Mhz 486 can run Doom at the same frame rate a movie plays which from memory is like 24fps.. It looks like it runs only slightly faster than my 386SX-25 back in the day! Which was playable, .. barley.. I remember running Doom on a 486DX/33 and it running rather smooth.. Maybe my memory is bad who knows.. I do remember it ran flawlessly on my DX2/66 though.
@@mstcrow5429 Interesting and good to know.. Sometimes I think maybe I've become to accustomed to 60+ fps gaming and so see things differently now.. But in the video you can see the shotgun skips frames and doesn't do the pump Action occasionally.. well that's exactly what Doom did when running like garbage i.e. in my 386sx 25 if I didn't reduce the quality to low and window size the shotgun occasionally skipped frames when reloading... Also the original Doom version 1.1 ran or appeared to run faster on older hardware to me than 1.2 onwards especially version 1.9 and I heard somewhere that depending on your video card version 1.1 was faster
@@mstcrow5429 Yeah it would have run perfectly on your system.. when Doom came out I have a 486dx2/66 with 8mb ram and an S3 805 VLB video card and Doom ran flawlessly on that.. I also had a 386sx/25 with a ET4000 video card and Doom was playable on that in low mode with sound blaster turned off. But Ultimate Doom v1.9 ran worse on my sx 25 than original Doom V1.1
Fantastic video. I am also surprised about the result and would appreciate a hint on why the UMC is faster. Was it different in the chip design? Some cache extensions?
It should be nice to test this CPU on an Intel chipset motherboard for checking if these good results are linked to have the same chipset and CPU mark.
Do you have a CPU Museum which can be visited when this 2020/2021 human malware thing is sort of over? Can you compare and benchmark the Cyrix 486DLC-33 with an ISA 1MB Tseng ET4000 vs Intel 486SX-25 with ISA 512K Trident VGA Card? This would solve an 30 year old dispute with a friend which system would be more powerful back in these days. ;-)
I have a general question/suggestion. Why were the early CPU's in a ceramic package, and how and from what is the ceramic material made? And is there is difference between the white and the purple variants?
Speedsys will rate stuff out of wack when it cant detect an L1... And it absolutely shoud show one for the cyrix. Cache Check shoud show it as well. Not sure if a tsr is needed on some motherboards.
I used to have a 486SX25 overclocked to 40, 8 MB EDO RAM, a Cirrus Logic 5428 VLB with 1 mb, SB16 MCD. It runned very well, I even played Doom and Diablo 1 to the end. I tweaked the ram a lot in autoexec.bat and config.sys menu. It whould be great if you benchmarked a setup like that.
What would you recommend as the best half-height VLB videocard? I have Trident and S3. Despite being fast, they produce visual artifacts. Not the best cards. 50.0 points through in 3Dbench on Trident on an AMD DX5 133MHz overclocked to 160MHz.
"I have Trident and S3. Despite being fast, they produce visual artifacts." That doesn't sound right. You might coincidentally have two bad cards or a bad motherboard for them to be artifacting.
@@anomaly95 There are vertical stripes of difference in luminance that repeat maybe every 30 pixels or so. Sometimes it's visible, sometimes it's not. Depends on the contrast of the image. I may have to check what controls the horizontal refresh and see if maybe there is a depleted capacitor on there or something. It's mostly on the Trident. S3 is generally alright. There bugs are of different nature. Nothing unexpected. Both were new old stock. Unused with 2MB memory each, but I don't remember exact models. To me it looks like the stripes represent clusters out of which the videocard assembles the picture. They shouldn't be visible, and they haven't increased over the years. The card's always been like that.
I had this CPU, it had only two big issues. First, no FPU, so you couldn't play at time with 3D Studio, and other rendering software (well FPU emulator was super slow), and second was Windows 95, instant freeze during installation.
Den UMC Super33 habe ich auch seit dem Jahr 2000 in meiner Sammlung und war und bin überrascht wie stark der gegenüber Intel und Co ist. Leider keine FPU und zu wenig Takt.
Never heard of the UMC chip, very interesting! Do you have an AMD K-6 III in your CPU treasure box? It was very interesting because of it's cache design compared to the K-6 II. Thank you very much!
I think Speedsys has some bug in the cpu speed routine, be careful. I was experimenting with a few different DX4 variants (AMD, Intel and Intel Overdrive) and sometimes it gave double CPU score, which can not be the result of me tinkering with BIOS settings (clock freq was the same). I could not find the buggy setting combination because the socket on my motherboard started to get unreliable and did not want to risk damaging it with the swapping.
Hi. Is there some significant difference between 486 DX4-75 and DX2-80? What's the real difference between the SX, DX, DX2 and DX4? Thank and regards from Brazil!
The difference is in the front side bus and the clock multiplier in the chips. Dx4-75 was 25mhz x 3, dx2-80 was 40mhz x2, only difference with the dx and so was the presence of a floating point unit in the chip. The sx cpus lacked it and so were much cheaper as a result.
Is anybody able to proof the SD version with an FPU really exists? I have found a page about this CPU line on Wikipedia and it says SD has the DX pinout to ensure older boards compatibility, but has no actual FPU inside. (?)
They tried to stop amd quite a number of times. But they were hindered by the agreements they had to make with IBM in the contract for production of the chip for the xt, IBM didn't want a single source so insisted that intel get other companies to make compatible chips. Intel tried to stop the sale of the amd 386 and 486, but was unsuccessful and it lead in part to the pentium brand and forced amd and others to make their own chips, which amd did with the k5 chips and the k6 (not actually developed by amd and no relation to the k5). Of course we know what the eventual outcomes were as intel licenced x86-64 from amd in a licence exchange after the abysmal itanium.
wow UMC really kicked intel's but, its a shame they got killed off. love that ziff socket what an easy elegant design, not like this pain in the ass TR4, Threadripper socket.
It wasn't nonsense. Back then the price difference between a 486 with a FP and one without was appreciable, and also back then, almost no applications used the floating point unit, so you'd save a decent chunk of change.