• Please review our updated Terms and Rules here

Video: 386DX 40 MHz vs 486SX 25 MHz

Are you sure the 486 has cache?
Those cache sockets are emptified.

Unless you soldered in some sockets w/ RAM at a later point after the photo was taken.
 
EverythingIBM said:
Are you sure the 486 has cache?
Those cache sockets are emptified.

Unless you soldered in some sockets w/ RAM at a later point after the photo was taken.
Click to expand...

Thanks for pointing this out!

I added a comment in the video :)

Yes I took the photo a while ago. The board wasn't working and today I fixed it:

- Got the jumper settings online
- Jumpered it correctly
- Fitted 9 cache chips 15ns for 256KB Cache
- De-soldered the RTC chip
- Soldered in a socket
- Mounted a DALLAS RTC
- Set the clock, loaded BIOS defaults, auto detect CF card

After this the board was working like a charm :)

The board certainly isn't a race horse. I have a fast Biostar PCI board and an OEM Acer with integrated Mach32, PCI and VLB and they will likely score a bit higher. But the idea was to compare a fast 386DX 40 MHz as an alternative to a basic 486SX 25 MHz system.

Or put another way, with better hardware the performance gap would only increase.
 
Last edited:
The 40 MHz BUS of the 386 had its advantages, but the main bottlenech those days was CPU performance and the 486SX-25, even without L2 cache is faster than 386DX-40. In fact your 386 system seems quite advanced scoring 16.6 in 3dbench. My old 386DX-40 only scored 12.8 there (that was with 60ns 8MB RAM, 128K cache at 20ns and a VIA chipset, No FPU). I could only increase the performance with a more modern chipset and faster cache memory, with parts I got from ebay.
Another thing about the 486 is the impact of the L2 cache, where in low-end systems it was not worth the money. L2 cache was expensive. You'd probably save your money for an SX-33 rather than spend it on L2 cache those days. The 386DX on the other hand was in another level without the on-board cache.
 
Yes the L1 Cache in the 486 shielded it against poor or missing L2. I might follow up with using a fast PCI based 486 motherboard that eliminates all bottlenecks.
 
Try putting a 486DLC-40 CPU in the 386 board.

Even better would be to have a 386 board with VLB slots, but those are rare.
 
vwestlife said:
Try putting a 486DLC-40 CPU in the 386 board.

Even better would be to have a 386 board with VLB slots, but those are rare.
Click to expand...

The reason I have 386 stuff is because they are SLOW :) If I need more speed that's what the 486 is for.

Doom will never be fluid on a 386 no matter how much time and money you poor into it. I go straigt to a Pentium MMX for Doom.

But for Wing Commander I use the FX-3000 board, swap out the crystal oscillator with a 50 MHz one (= 25MHz clock speed) and off I go.
 
Doom doesn't make use of MMX. Get a DRx2 for that 386 :D

Similarly to you, I have a 386 for it being slower which is the sole reason I leave it set up mostly as it is although it needs a few upgrades.

Also, you copying? Or is it another coincidence. (I'm just joking here)
 
Haven't seen the video yet, but I'm curious if you also tested Doom in the half-res mode (I think it's F5 or F6 to toggle). This reduces by half the amount of memory transfers to the video card, and I think it would be interesting to see if that affects the outcome of the comparison.
 
High_Treason said:
Also, you copying? Or is it another coincidence. (I'm just joking here)
Click to expand...

Subscribed :)

Well there are only so many benchmarks you can run. And the popular ones do pop up all the time :)

Trixter said:
Haven't seen the video yet, but I'm curious if you also tested Doom in the half-res mode (I think it's F5 or F6 to toggle). This reduces by half the amount of memory transfers to the video card, and I think it would be interesting to see if that affects the outcome of the comparison.
Click to expand...

Nope, just the default quality of Shareware version 1.9. This is the setting you see pretty much all timedemo results based on. Makes it easier to compare.

PS: I've removed the 486SX 25 MHz, but can definitely do a low resolution video on the 386DX.
 
Last edited:
High_Treason said:
Get a DRx2 for that 386 :D
Click to expand...

The ultimate 386-based system was the IBM "Blue Lightning" 486BLX3-100, a.k.a. 486DLC3. It was the first x86 CPU to reach 100 MHz, beating Intel's 486DX4 by a few months.

My PS/2 Model 56 has been upgraded with the lesser 16-bit-external-bus variant, the 486SLC3, and with its hefty 16 kB L1 cache, it is also surprisingly speedy, especially with the Model 56's onboard SCSI hard drive controller and the XGA2 graphics card I have installed. The only thing letting it down from being a good DOS gaming machine is the lack of any affordable Sound Blaster-compatible MCA-bus sound cards!
 
orion24 said:
The 40 MHz BUS of the 386 had its advantages, but the main bottlenech those days was CPU performance and the 486SX-25, even without L2 cache is faster than 386DX-40. In fact your 386 system seems quite advanced scoring 16.6 in 3dbench. My old 386DX-40 only scored 12.8 there (that was with 60ns 8MB RAM, 128K cache at 20ns and a VIA chipset, No FPU). I could only increase the performance with a more modern chipset and faster cache memory, with parts I got from ebay.
Another thing about the 486 is the impact of the L2 cache, where in low-end systems it was not worth the money. L2 cache was expensive. You'd probably save your money for an SX-33 rather than spend it on L2 cache those days. The 386DX on the other hand was in another level without the on-board cache.
Click to expand...

Actually, the IBM 486SLC @ 50Mhz (without cache) is faster than the 486SX-25 with cache in this instance.
So a 386 can beat a 486.

vwestlife said:
The ultimate 386-based system was the IBM "Blue Lightning" 486BLX3-100, a.k.a. 486DLC3. It was the first x86 CPU to reach 100 MHz, beating Intel's 486DX4 by a few months.

My PS/2 Model 56 has been upgraded with the lesser 16-bit-external-bus variant, the 486SLC3, and with its hefty 16 kB L1 cache, it is also surprisingly speedy, especially with the Model 56's onboard SCSI hard drive controller and the XGA2 graphics card I have installed. The only thing letting it down from being a good DOS gaming machine is the lack of any affordable Sound Blaster-compatible MCA-bus sound cards!
Click to expand...

The microchannel soundblasters have problems with a lot of the faster CPUs and are very glitchy (just a shoddy port from ISA to MCA by creative). Also I discovered that the soundpiper isn't compatible with all DOS games, like DOOM-- so you won't get any adlib in them [!]

--> the most compatible commercial MCA soundcard is the chipchat thus far.
 
Last edited:
Mau1wurf1977 said:
Subscribed :)
Click to expand...
Same. Always good to see someone else running hardware properly and experimenting.

vwestlife said:
The ultimate 386-based system was the IBM "Blue Lightning" 486BLX3-100, a.k.a. 486DLC3. It was the first x86 CPU to reach 100 MHz, beating Intel's 486DX4 by a few months.
Click to expand...
Ah, but to me it loses by default for being in an OEM box, also I don't have one. A very interesting upgrade though, it's interesting to observe the many different methods people had to invent to keep things up-to-date back then... Inversely around a decade later you had IDT effectively installing a 486 onto a Pentium platform.
 
The IBM SLC and DLC chips are quite fast. I've personally used the SLC2 50,66, and DLC3 75/100. Back in the early to mid 90s, I had a 486DX-33 based on Symphony chipset with 64kb cache. My friend had an Alaris/IBM 486SLC2-50 with 0kb L2 cache (although 16kb L1 cache). Except for memory performance, the two systems were more or less equal. In some instances his was faster. Very impressive for 16-bit motherboard with no cache.
 
Not true, my last dx-40 386, an amptron, smeared the floor with anything lesser than my dx-50 486, throw a coproc with that 386 and a difference will be made, also your giving the 486 an unfair advantage with the vlb controller. As for running doom, personally i got it fluid on a 386-33 whilst my 486 SLC-50 i couldn't get it very playable. god help you if you wanted level 15 of doom 2 to work at more that 2fps on a slc-50

Mau1wurf1977 said:
The reason I have 386 stuff is because they are SLOW :) If I need more speed that's what the 486 is for.

Doom will never be fluid on a 386 no matter how much time and money you poor into it. I go straigt to a Pentium MMX for Doom.

But for Wing Commander I use the FX-3000 board, swap out the crystal oscillator with a 50 MHz one (= 25MHz clock speed) and off I go.
Click to expand...
 
k2x4b524[ said:
Not true, my last dx-40 386, an amptron, smeared the floor with anything lesser than my dx-50 486, throw a coproc with that 386 and a difference will be made, also your giving the 486 an unfair advantage with the vlb controller. As for running doom, personally i got it fluid on a 386-33 whilst my 486 SLC-50 i couldn't get it very playable. god help you if you wanted level 15 of doom 2 to work at more that 2fps on a slc-50
Click to expand...

Maths CPU doesn't do anything for the tests in the video. Controller card also has no impact, I can put the ISA controller in a Pentium and still get 50+ fps. Beating a 486DX 50? Doom fluid on a 386DX - 33? Maybe in low detail and smallest image size.

I'd like to see some footage of your claims though :)
 
Mau1wurf1977 said:
Doom fluid on a 386DX - 33? Maybe in low detail and smallest image size.
Click to expand...

Low detail and largest image size, and adding a GUS helped offload some processing as well. Doom ran at ~15fps normal and ~30fps in lowres mode on my 386dx-40, both of which appear to exceed the 486sx-25 in your video. I had 128k cache installed, and a Cirrus Logic VGA, both of which definitely made a difference, however.
 
Trixter said:
Low detail and largest image size, and adding a GUS helped offload some processing as well. Doom ran at ~15fps normal and ~30fps in lowres mode on my 386dx-40, both of which appear to exceed the 486sx-25 in your video. I had 128k cache installed, and a Cirrus Logic VGA, both of which definitely made a difference, however.
Click to expand...

Doom -timedemo demo3 is run without a soundcard to help consistency.

I'd like to see a run of your 15fps 386. Double the speed of my FX-3000 with 256KB Cache? I don't think so :) Maybe your memory is tricking you a little bit or maybe it was a 486DX 40?
 
A while ago I noticed that Star Wars Dark Forces claims to run on a 386DX-33, at least according to the box. That one has some polygon graphics too, though I doubt they use the FPU. Some time ago I set myself a mission to run it on my 386 system with a DX33 in... I'm quite confident that coupled with a TSeng ET4000 and fast 16MB RAM that the DLC-40 I use can do it, but the 386DX I am unsure of. Interestingly, that game has a system speed selector, it is automatic for the most part but can be changed manually.

There are a lot of factors that can sway it according to my research, for example, adding 256K cache (opposed to the 128K I have) actually takes away 10% performance on average and installing the Trident VGA removes about 45% in games. I don't yet have the 16MB RAM I want, so I can't yet say how much that would change things - currently some very latent 2MB SIMMS (8MB Total) are installed.

I also think chipsets were a big factor, I have an OPTi-based Shuttle board.

Edit: Doom runs pretty well on the Ti486DLC-40 anyway.
 
Last edited:
You must log in or register to reply here.
Back
Top