No luck. Will try again tomorrow.

Now that the toasted EPROM has been replaced with the Winbond EEPROM, flashed to the 2012 BIOS and the system speed on bootup is set properly in the BIOS... (cough)... 😵

Let the benchmarking resume! This time I'm circling back to Build #1 of 4, the BIOSTAR MB-8433UUD with the X5-133ADZ.

BIOSTAR MB-8433UUD-A
AMD AM5x86-P75 / AMD X-5-133ADZ
32MB EDO RAM
256KB L2 Cache
STB S3 Trio64V+ (2MB)
4GB CF Card (CF->IDE Adapter)

AMD X5-133ADZ @ 133Mhz - Un-Tuned
L1 Cache is 16KB - 136.9MB/s
L2 Cache is 256KB - 55.8 MB/s
Main Memory Speed - 28.9 MB/s
RAM Access Time (Read) - 145ns
RAM Access Time (Write) - 92ns

0

AMD X5-133ADZ @ 133Mhz - Tuned
L1 Cache is 16KB - 137.4MB/s
L2 Cache is 256KB - 62.1 MB/s
Main Memory Speed - 39.7 MB/s
RAM Access Time (Read) - 105ns
RAM Access Time (Write) - 60ns
Main Memory Speed (Write) - 69.3MB/s
SpeedSys Overall Score - 50.27

Let's try our luck running the processor @ 160Mhz

Attempt 1: FAIL - L2 Cache not Recognized

Attempt 2:
L1 Cache is 16KB - 164.3MB/s
L2 Cache is 256KB - 67.0 MB/s
Main Memory Speed - 34.8 MB/s
RAM Access Time (Read) - 120ns
RAM Access Time (Write) - 76ns

Attempt 3: Added 1 : 2/3 divider
L1 Cache is 16KB - 164.3MB/s
L2 Cache is 256KB - 67.0 MB/s
Main Memory Speed - 44.3 MB/s
RAM Access Time (Read) - 145ns
RAM Access Time (Write) - 92ns
Main Memory Speed (Write) - 54.8MB/s
SpeedSys Overall Score - 60.33

Looks like you at least got the 133 MHz speed optimised.

The 160 MHz speed should be able to run with the L2 cache wait state set at 2-1-1-1 and HOST/PCI clock at 1:1 provided that you are using double-banked 256 KB cache. What speed is your cache brand and speed?

Leave the EDO DRAM Read Speed set to 4-2-2-2. You want to Enable Slow Refresh (it is more optimal). Try setting L2 cache scheme to write-through and ALT Bit in TAG to 8+0.

Can you graphics card not handle 40 MHz operation? I have not experienced the L2 cache not being recognised when the FSB is run at 40 MHz. Usually the system gives off memory errors. Try using only one stick of RAM of FPM memory for now. In my experience, the less number of sticks the more successful 40 MHz operation is on any PCI-based 486.

DRAM Read/write should be at 0 WS.

Your Cachechk results should be,
L1/L2/Memory: 165/75/48
Ram access time (write): 51 ns

Be sure to use 50 or 60 ns memory .

Your Quake score is the same as I get on a 133 Mhz system.

Back in business thanks to you feipoa. This is why I love VOGONS!

Switched to a single 32MB EDO 72-pin SIMM. The BIOS screenshot below shows the L2 cache set to Wr-Through with the 8+0 bit, but I changed it back to Wr-Back and it's working just fine.

AMD X5-133ADZ @ 160Mhz (P90 Speed)

Cache: UM61256FK-15

Tuned BIOS

Attempt 4: Switched to a single 32MB EDO SIMM and followed, tuned BIOS
L1 Cache is 16KB - 164.9MB/s
L2 Cache is 256KB - 74.5 MB/s
Main Memory Speed - 47.7 MB/s
RAM Access Time (Read) - 88ns
RAM Access Time (Write) - 50ns
SpeedSys Overall Score - 60.34

Try running MemTest 4.0, at least 2 passes of all tests. I know when I was testing EDO vs. FPM and all slots filled, FPM was more reliable. I'm not sure about 1 stick of EDO vs. 1 stick of FPM though.

So are we back to the Saturn II board now? !

Yep... I think we are back to the Saturn board now thanks to all of your help. Unless...... 😀

I have two Asus PVI-486SP3 boards and I believe the Cyrix 5x86 120 "M1sc" is supported. I really wanna get this processor working in one of these setups. If you recall, I got pretty abysmal performance when running it in one of my Asus VL/I-486SV2GX4 boards.

So right now, the systems I've completed (and I'm happy with) are:

Build 1: 1 x BIOSTAR MB-8433UUD-A Rev. 3.1 with AMD-X5-133ADZ (Am5x86-P75) in Write-Back Mode
BIOS tuned to run at 133Mhz and 160Mhz stable, with good cache/memory benchmarks.

Build 2: 1 x Asus VL-i486SV2GX4 Rev. 2.0 with AMD X5-133ADZ Write-Back Mode
BIOS tuned to run at 133Mhz, with good cache/memory benchmarks.
May upgrade to 1024MB cache

Build 3: 1 x Asus VL-i486SV2GX4 Rev. 2.1 with Cyrix 5x86 120 in Write-Back Mode
Slow memory/cache performance
May upgrade to 1024MB cache

Build 4: 1 x Asus PCI I-486SP3G Rev. 1.8 with AMD X5-133ADZ in Write-Back Mode
Processor isn't "officially" supported on this board
Issues with L2 cache - some articles say it must be set to write-through
Slow memory/cache performance in numerous configurations
No "Official" 40/50Mhz option in BIOS, need to try jumpers

feipoa wrote:

Do you recall what FSB speeds corresponded to the other two undocumented jumper settings?

It has been a long time since I've tried this board, but I recall that the bus speed jumpers did not work like on other boards - there were two jumpers, and they both had to be set to the same position - so the two jumpers yielded only two settings - 25 and 33 MHz - and not four, as you would normally expect. I also remember that there was another hardwired jumper pad nearby, but switching that one to the unused position apparently did nothing.

Will you test the Asus VL-i486SV2GX4 @ 160mhz?

Could you post the gamebenches for the Biostar board @ 160mhz?

vetz wrote:

Will you test the Asus VL-i486SV2GX4 @ 160mhz?

Could you post the gamebenches for the Biostar board @ 160mhz?

Certainly can do both!

5u3 wrote:

feipoa wrote:

Do you recall what FSB speeds corresponded to the other two undocumented jumper settings?

It has been a long time since I've tried this board, but I recall that the bus speed jumpers did not work like on other boards - there were two jumpers, and they both had to be set to the same position - so the two jumpers yielded only two settings - 25 and 33 MHz - and not four, as you would normally expect. I also remember that there was another hardwired jumper pad nearby, but switching that one to the unused position apparently did nothing.

So.. I tried messing with the jumpers. No dice - still stuck at 25 or 33Mhz. I did find this article below:
http://www.experts-exchange.com/Hardware/Misc … Q_10015632.html

I havent followed this thread closely but why aren't you running a cyrix 5x86 2x66 in the BIOSTAR MB-8433UUD2? Seems like the most logical use of resources.

Artex, do you have a high resolution detailed photo of this board? I'd like to determine which chip they are using to generate their FSB.

dirkmirk wrote:

I havent followed this thread closely but why aren't you running a cyrix 5x86 2x66 in the BIOSTAR MB-8433UUD2? Seems like the most logical use of resources.

AMD 5x86 can clock higher (apparently 4x40 isn't uncommon)
And cyrix 5x86 are kind of fake 486 because they are a cut down pentium like core

Running an IBM/Cyrix at 2x66 will requiring modifying the motherboard's voltage regulator for a variable voltage source. While it is a straight-forward modification, I suspect Artex is not up to this. There is also a chance his Cyrix 5x86-120 can't quite cut it at 133 MHz.

You do not need to modify the voltage regulator on the board to run a 66 MHz bus, but you do need to modify the motherboard's CPU voltage regulator to run the Cyrix 5x86-120 or IBM 5x86-100 at 133 MHz. The Biostar board only has options for 3.45 V, 4.0 V, or 5.0 V. To run an IBM 5x86-100HF at 133 MHz, I found that somewhere in the 3.65 - 3.85 V range was optimal. You can also use an interposer board with a VRM and modify the VRM on the interposer. Modifying the voltage regulator on the Biostar board is a simple matter of removing a resistor and putting in a trimmer.

On the other hand, if you wish to run an Am486 DX2-66V16BGC at 2x66 and 3.45 V, you need not modify the voltage regulator. Photo attached. Note the September 2002 datecode on this chip.

This chip, when used in the Biostar board, also POSTED at 2x83 MHz, but there were artifacts on the screen. Lost the photo.

The SiS 496/497 boards I tested did have undocumented jumper options to run at 66 MHz, but I could not get the board to function reliably. This is also true of some UMC 8881/8886 boards. The only UMC 8881/8886 board I have been able to function well with a 66 MHz FSB is the Biostar MB8433-UUD. I suspect the issue is not so much the chipset but the motherboard's PCB design. I do not think any board designer was considering operation at 66 MHz when plotting the traces, sourcing amplifiers/drivers/inverters, impedance matching ckts, etc. It just so happens the Biostar board is functional. Perhaps the designers put a lot of weight on stable 50 MHz operation, or just got lucky.

You can also mod the Biostar board to run with 1024K cache. Photo.

For some reason, I had to increase the L2 cache timing from 2-1-1-1 to 3-1-1-1 to get stability to my liking with 1024K cache and 33 MHz FSB. I planned to resolve this issue using buffers but ran out of time. The issue may be related to weak interconnects, which could be resolved with soldering. I should have tried it with 512K cache double-banked to determine if it was simply a size limitation.

For those wanting a Cyrix 5x86-133, you can mod. some QFP packaged Thinkpad IBM 5x86 chips to work in a socket 3. Photo.

This IBM 5x86c worked so well at 150 MHz that I considered epoxying a huge Socket 370 cooler to the top. Unfortunately, one of my wires desoldered with all the playing around and shorted out an adjacent pin. This IBM is now toast, but I have 5 more with later date codes to do the mod to.

And some work has been done to get an extra 2x multiplier onto the Biostar board, which would allow you to run a 33 MHz FSB, an IBM 5x86 at 2x, and an external multiplier at 2x for a total of 133 MHz. This would save the motherboard from running an overclocked FSB. Photo.

While the circuit worked, it only POSTed at 25x2x2 for 100 MHz operation, not 33x2x2, for 133 MHz operation, which is what I was hoping for. I'm not sure exactly why, but it may have something to do with too much phase delay. The external PLL is supposed to take care of that by aligning the rising edges of the clock. Photo.

The blue waveform is the 33 MHz FSB generated from the motherboard, just before the signal enters the CPU. The yellow is the external PLL chip intercepting the clock line. The output from the PLL chip then enters the CPU. The CLK path to the CPU is cut and rewired. Notice how the vallies don't quite line up. Approx. 2 - 4 ns delay. Anyone have any ideas what I'm doing wrong? I used a 33 Ohm series termination resistor. I also tried varying the series termination resistor using a trimmer to find the sweet spot, but no dice.

Lots of desires, lack of time.

Here is the IBM 5x86-100 at 150 MHz. I could not get it to run Quake. This is probably the best Speedsys score you'll find on a 486-class motherboard. All run on a Biostar MB8433-UUD.

Holy shizzz.... you ARE hardcore. 😀 I don't think I'll take this testing quite as far, but I certainly commend you for all of your work there. Wow.

Would my Cyrix 5x86 120 (at stock speed) compare to the AMD X5-133 (at 160Mhz) on the MB-8433UUD? Still wondering if I should move that Cyrix chip to this board.. but only if it's gonna be faster than the AMD processor.

feipoa wrote:

Here is the IBM 5x86-100 at 150 MHz. I could not get it to run Quake. This is probably the best Speedsys score you'll find on a 486-class motherboard. All run on a Biostar MB8433-UUD.

Impressive results 😀 Do you think a POD running at 50mhz FSB could beat that? It has 73 points in Speedsys at 100mhz (40mhz FSB).

Vetz, no, I do not believe so. I seem to recall FSB not considerably affecting the Speedsys score for the same internal CPU frequency. Or do you mean a POD at 125 MHz? If a POD at 125 MHz, then it will surely have a higher Speedsys score. Are you able to POST at 125 MHz?

Artex, test results indicate that the X5-160 is, on average, about 15% faster than the Cx5x86-120 in ALU-based operations, whereas the Cx5x86-120 is about 15% faster in FPU-based operations. If you decide to weigh FPU and ALU operations equally, the two chips have equal performance.

feipoa wrote:

Vetz, no, I do not believe so. I seem to recall FSB not considerably affecting the Speedsys score for the same internal CPU frequency. Or do you mean a POD at 125 MHz? If a POD at 125 MHz, then it will surely have a higher Speedsys score. Are you able to POST at 125 MHz?

I meant POD at 125mhz. I haven't tried it. Did you give it a try under your Ultimate 486 Benchmark?