Fascinating stuff - the college I attended at the time had a single DX50 machine hooked up to a scanner as their multimedia machine, and it had an air of exclusivity about it. The concept is emotionally satisfying - it's running at a "real" 50mhz - but as you point out I don't remember it feeling any faster than the DX2/66 machines in the rest of the college.
If you (or anyone) run into the 512MB, or 8.4GB, or 127GB limit for HD size you can use disk overlay software like Seagate Ontrack to make full use of the disk. (It is operating system independent) This software had allowed me to boot to CF cards as well on systems with limited drive selections in the BIOS such as a T3200/5200, or those that dont offer a custom drive types or auto detection. Of course you are limited to 2GB partition size when using Fat16 but there are ways around that like using MSDOS 7 and fat32. As for your issue, the DX2 66 had a modified base bus speed of 33MHz and uses a 2x clock multiplier to derive the CPU clock speed; because Intel discovered that there was significant stability issues at bus speeds above 40MHz. So it seems the bottleneck is in the design . If you get a DX2-80 or a DX4-120 (AMD) you can have the max 40MHz Bus speed and then whatever CPU clock you want. I enjoy your videos, you are knowledgeable and articulate! Keep at it 👍 experimentation is good fun.
Great video! The DX50 was mainly a paper tiger but could have maybe done more damage if the systems were cacheless(which few were at the time). The cache and the fact that most work is done on the cpu decreases the "wunderkind" potential severly. I had the DX50 running at 50(VLB and PCI), 60 and 66(both PCI) fsb and even at 66 it is not turning the DX2-66 to dust. The picture changes completely though if you up the cpu speed. A DX2-100 is a very capable combo and a DX3-150(X5-AMD only) even moreso. Here the fsb makes a world of difference and the systems become, relative to 486 margin, very Fast. A cacheless intel DX2-133 for example comes close to a DX3-150 AMD with cache set at the best timings in some applications. It is no coincidence that intel did not push for faster bus speeds on the 486 since this would have severly sabotaged the early Pentiums - especially in non FPU intense applications. Btw, I also ran the PCI bus at 50/60 and 66mhz - a feat that only the Matrox Mystique/Millenium withstood... the card was screaming in more way than one. The 60/66 mhz option was run on an SiS49x chipset which is known to have lax timings. 50mhz PCI clock on an UMC888x is a true screamer. Doom also hinted that a good ratio of cpu to fsb speed may lie between 2:1 and 3:1. No wonder the engineers clock-multiplied their chips since the buses were, in many cases, fast enough to deliver to/from the 486-architecture at "slow" speeds without hassle.
I wouldn't have expected the overclocked bus to help much in that situation. It increases available bandwidth but the 133MB/sec you get at 33MHz is already overkill for a 486-class system. 60FPS VGA only takes around 4MB/sec and an EIDE interface maxes out around 16MB/sec. Also the video card only needs to push frames as fast as the processor can supply them, so when you're CPU-bottlenecked you don't see much improvement. Those VLB cards are essentially just sitting around on their overclocked bus waiting on the spinning rust and CPU.
I'm not really surprised that the VLB I/O controller didn't help at all. It really shouldn't hinder gameplay, for the very reason you say in the video. Seeing how DX-50 lags behind the DX2-66 - is informative. Of course it's faster than the DX2-50 - but apparently not by too much.
Funny some one would suggest using tseng 4000 over the cirrus logic, in terms, as long as both are vlbus cards performance wise they are actually very close definitely not slow enough to call it a bottle neck.
No jumper setting on any of the video cards? I seem to remember mine having a jumper to set it to "Equal or lower than 33MHz" and a "Higher than 33MHz" for 40 and 50 MHz bus speeds. I also believe that the whole "50MHz bus, whooo!!!" malarky is based largely on myths from back in the day. I mean, a supposedly superfast CPU, that was also known to be very temperamental and which hardly anyone owned. It's not hard to see that it became something "more" in people's minds than it actually was. 16MHz extra actual CPU clock speed, is a pretty high % boost at these low overall clock speeds, so it would be REALLY hard to beat that with just boosting the FSB.
I don't recall which cards had that jumper but I was aware of it and made sure it was set correctly if the card had the particular jumper option. Guess I just never mentioned it in the video.
An interesting test could be to take a PCI 486 mobo and to crank the bus up to 66 MHz and set a 1/2 divider on the PCI bus to compare it with the DX2 66 ^^ (or perhaps a PCI 2.1 video card would work ?) Another cool test could be to take a DX4 100 and run it with 2x 50MHz to see how much of an improvement the faster FSB does on the performances
@@AncientElectronics Yeah... It is a machine that is hard to build. It is on a DX-33, DX2-66, DX2-80 and DX4-120 that I have better results with the two S3-805 cards that I have. They are branded "Spea Mirage7" or something like that. I have no idea if that is a brand or model name.
Fascinating stuff - the college I attended at the time had a single DX50 machine hooked up to a scanner as their multimedia machine, and it had an air of exclusivity about it. The concept is emotionally satisfying - it's running at a "real" 50mhz - but as you point out I don't remember it feeling any faster than the DX2/66 machines in the rest of the college.
If you (or anyone) run into the 512MB, or 8.4GB, or 127GB limit for HD size you can use disk overlay software like Seagate Ontrack to make full use of the disk. (It is operating system independent) This software had allowed me to boot to CF cards as well on systems with limited drive selections in the BIOS such as a T3200/5200, or those that dont offer a custom drive types or auto detection. Of course you are limited to 2GB partition size when using Fat16 but there are ways around that like using MSDOS 7 and fat32.
As for your issue, the DX2 66 had a modified base bus speed of 33MHz and uses a 2x clock multiplier to derive the CPU clock speed; because Intel discovered that there was significant stability issues at bus speeds above 40MHz.
So it seems the bottleneck is in the design . If you get a DX2-80 or a DX4-120 (AMD) you can have the max 40MHz Bus speed and then whatever CPU clock you want.
I enjoy your videos, you are knowledgeable and articulate! Keep at it 👍 experimentation is good fun.
Great video!
The DX50 was mainly a paper tiger but could have maybe done more damage if the systems were cacheless(which few were at the time). The cache and the fact that most work is done on the cpu decreases the "wunderkind" potential severly.
I had the DX50 running at 50(VLB and PCI), 60 and 66(both PCI) fsb and even at 66 it is not turning the DX2-66 to dust. The picture changes completely though if you up the cpu speed. A DX2-100 is a very capable combo and a DX3-150(X5-AMD only) even moreso. Here the fsb makes a world of difference and the systems become, relative to 486 margin, very Fast. A cacheless intel DX2-133 for example comes close to a DX3-150 AMD with cache set at the best timings in some applications. It is no coincidence that intel did not push for faster bus speeds on the 486 since this would have severly sabotaged the early Pentiums - especially in non FPU intense applications.
Btw, I also ran the PCI bus at 50/60 and 66mhz - a feat that only the Matrox Mystique/Millenium withstood... the card was screaming in more way than one. The 60/66 mhz option was run on an SiS49x chipset which is known to have lax timings. 50mhz PCI clock on an UMC888x is a true screamer.
Doom also hinted that a good ratio of cpu to fsb speed may lie between 2:1 and 3:1. No wonder the engineers clock-multiplied their chips since the buses were, in many cases, fast enough to deliver to/from the 486-architecture at "slow" speeds without hassle.
I wouldn't have expected the overclocked bus to help much in that situation. It increases available bandwidth but the 133MB/sec you get at 33MHz is already overkill for a 486-class system. 60FPS VGA only takes around 4MB/sec and an EIDE interface maxes out around 16MB/sec. Also the video card only needs to push frames as fast as the processor can supply them, so when you're CPU-bottlenecked you don't see much improvement.
Those VLB cards are essentially just sitting around on their overclocked bus waiting on the spinning rust and CPU.
Maybe it starts to matter with faster chips. The DX4 120 seems to be faster than the AMD 5x86 for example.
if you look at 9:05 you have a blown chip on the lower left of the trio64
I'm not really surprised that the VLB I/O controller didn't help at all. It really shouldn't hinder gameplay, for the very reason you say in the video.
Seeing how DX-50 lags behind the DX2-66 - is informative. Of course it's faster than the DX2-50 - but apparently not by too much.
Funny some one would suggest using tseng 4000 over the cirrus logic, in terms, as long as both are vlbus cards performance wise they are actually very close definitely not slow enough to call it a bottle neck.
No jumper setting on any of the video cards? I seem to remember mine having a jumper to set it to "Equal or lower than 33MHz" and a "Higher than 33MHz" for 40 and 50 MHz bus speeds.
I also believe that the whole "50MHz bus, whooo!!!" malarky is based largely on myths from back in the day. I mean, a supposedly superfast CPU, that was also known to be very temperamental and which hardly anyone owned. It's not hard to see that it became something "more" in people's minds than it actually was. 16MHz extra actual CPU clock speed, is a pretty high % boost at these low overall clock speeds, so it would be REALLY hard to beat that with just boosting the FSB.
I don't recall which cards had that jumper but I was aware of it and made sure it was set correctly if the card had the particular jumper option. Guess I just never mentioned it in the video.
An interesting test could be to take a PCI 486 mobo and to crank the bus up to 66 MHz and set a 1/2 divider on the PCI bus to compare it with the DX2 66 ^^ (or perhaps a PCI 2.1 video card would work ?)
Another cool test could be to take a DX4 100 and run it with 2x 50MHz to see how much of an improvement the faster FSB does on the performances
Those are some good ideas. I'll keep them in mind for some later projects. thanks.
@@AncientElectronics you could even go a step forward with a DX4 120 overclocked to 2x66 and compare it with the amd 5x86 ^^
I have that very same Samsung desktop sata drive just under 300gb
My Samsung desktop sata drive is under 300gb
I have way better results, using S3-805 compared to CL-5428.
unfortunately, I couldn't get my 805 to play nice with the 50MHz FSB
@@AncientElectronics Yeah... It is a machine that is hard to build. It is on a DX-33, DX2-66, DX2-80 and DX4-120 that I have better results with the two S3-805 cards that I have. They are branded "Spea Mirage7" or something like that. I have no idea if that is a brand or model name.