Yes, ok, but getting it to run dos don't require drivers, i like to know if quake and blood etc loves it.
Would it post at all? that is the most important factor, proper 486able windows drivers are 2nd.

hello guyz, i am also on my long haul of putting together a 486 rig, and i would like to know:
assume that i have -60 dram and -10 cache, with their rated standard timing, would 50*3 or 40*4 be faster in 3d game tests(not synthetic benchmarks)?

-10 cache would allow fastest settings on 40MHz FSB ?

Good 15ns cache should allow you to run at 40MHz at the fastest timings. In theory the 10ns cache could provide advantages if you run 50MHz or faster, but I am not totally convinced the 10ns cache is legit. I have a feeling it could be fake.

Anonymous Coward wrote:

Good 15ns cache should allow you to run at 40MHz at the fastest timings. In theory the 10ns cache could provide advantages if you run 50MHz or faster, but I am not totally convinced the 10ns cache is legit. I have a feeling it could be fake.

well, actually most 486 boards i can find here would simply come without cache chips, so i would have to search for sram cache chips by myself anyways.

Anonymous Coward wrote:

Good 15ns cache should allow you to run at 40MHz at the fastest timings. In theory the 10ns cache could provide advantages if you run 50MHz or faster, but I am not totally convinced the 10ns cache is legit. I have a feeling it could be fake.

If Quake would sometimes hang the computer, would it come more likely from cache, ram, or vcore ?
Knowing that 3dbench2, pcpbench and doom do always complete ?

JaNoZ,

There aren't any GeForce cards which will work on UMC 8881-based motherboards that I have tested, that is, they did not even POST.

I have tried several Cyrix 5x86-100 chips, none of which worked well at 133 MHz. I have had better luck with the IBM 5x86c-100 chips.

Check the signal noise and swing of your CLK signal using an osciliscope before you determine if you need to condition the signal. Any in-line ckt you add will likely add some undesirable delay. I have found 2x60 to be pretty reliable on the MB-8433UUD, but you need to run the L2 at 3-2-2-2 and at least have 1WS on DRAM read. Try using only a single stick of RAM. Be sure to drop the PCI bus to 2/3 or 1/2.

The AmDX2-66 BGC I have ran at 166 MHz (2x83.3), so it is not likely a derated X5-133. A derated X5-133 with a 2x multiplier would imply that there is a way to change chips from 3/4x to 2/3x post fabrication, which I doubt. The 83.3 MHz signal is not clean though and there were fragments on the screen. I do not recommend using this FSB speed.

noshutdown, it is more likely that 40x4 is faster than 50x3. With 40 MHz, you can likely run the L2 at 2-1-1-1 and you may be able to keep 40 MHz FSB. At 50 MHz, you need to add extra L2 wait states and either drop the PCI bus down to 25 MHz, or 33 MHz if you have the MB8433.

sunaiac, I have not found that cache faster than 15 ns allows for faster L2 timings.

Anonymous, I am 50/50 on the 10 ns cache. I do not have solid proof either way. Do you think the branding is fake, the speed rating is fake, or both? I used the 10 ns cache in a 430TX board as the TAG RAM for 83 MHz operation and it held up well for all of the 686 benchmarks - nearly a year of testing.

feipoa wrote:

sunaiac, I have not found that cache faster than 15 ns allows for faster L2 timings.

Ok.
My DX4 is fine at 120 with 2-1-1-1 timings and 1 write WS, but quake hangs like 2 out of 5 times with 0 write WS.
I hoped faster cache would help. Maybe I should run more tests with 1 WS to see if it also hangs...
I'd really like to get stable 0WS, because of how faaast it is 😁
(Shuttle HOT 419 R2)

A lot of things can cause system hangs. Feipoa always tells me to get a cache tester to make sure my SRAMs are good. I haven't yet gotten around to actually doing so, but I highly recommend it. SRAMs are very sensitive to static electricity and are easily damaged by improper handling. I am actually quite certain that a large percentage of my 10n SRAMs are defective, but I need a tester to weed them out.

feipoa wrote:

Mau1wurf1977, which desolder station do you have? JaNoZ, which hot air rework station do you recommend? […]
Show full quote

Mau1wurf1977, which desolder station do you have?
JaNoZ, which hot air rework station do you recommend?

JanoZ, why do you want the AM5x86 to run at 2x? I do not see much value in running such a chip at even 2x66 when 4x40 has better performance. At 160 MHz using a 27 MHz PCI bus, the 160 scores 17.3 fps in Quake 1, while the 2x66 receives 15.1 fps with a 30 MHz PCI bus. The Am5x86-16KB-WB does exist. This is the one from my collection,

AMD_5x86-2x.jpg

Look at that mid-2002 datecode. I think the B1's were the last editions.

I think the most optimal Am5x86 setup is to run the chip at 60x3. Of about a dozen chips I have, I could not obtain a stable chip at this speed at 4V or less. I'm sure some do exist though. I know some people who have been able to run the Am5x86 at 3x50 with the L2 on the fastest 2-1-1-1 timings without problems provided they restricted the L2 cache to 256 KB of double-banked. You should use a board with a 2/3 PCI multiplier in this case, otherwise it seems silly to run a 50 MHz bus on a PCI-based board. 4x40 is probably the best compromise for the Am5x86. I have at least one board which will run 1024K L2 WB cache on the fastest L2 settings, 128 MB RAM, and a 40 MHz PCI bus.

To determine if a board will do 60 MHz "properly" depends on many factors. First is the PCI bus, if it can be set to 1/2 and 2/3 of your FSB. At 60 MHz, the slowest L2 cache timings should be fine, however it will yield slower L2 throughput compared to the same chip at 40 MHz and the fastest timings. On the up-side (and if I remember correctly), I beleive the RAM speed is slightly faster. Usually a single stick of 64 MB RAM and double-banked 256 KB cache should be best for testing purposes.

how does your enhanced dx2-66(16k wb) overclock? does it run at 50*3 or higher? i guess it should have the same core as the 5x86-133.

I haven't thoroughly tested the DX2-66 yet. It is in the pipeline of things to do. I have been trying to determine what changes were made with the B1 revisions of Am486 chips.