What brand upgrade kit did that IBM 5x86c come on?

Does the Windbond cache printing look a little weird to anyone else?

Is there anything printed on the bottom of these cache chips? Did you finally get them from some Chinese-based IC reseller?

Is the second number on the ISSI chip a serial number? If so I'm starting to wonder why they have the same serial number and date code as mine.

AC, does the 10 ns ISSI cache you just ordered also have the same serial and date code?

I'm not sure what kind of upgrade kit it came on. I didn't get a box/manual etc. Just the chip, heat sink and voltage regulator. The heat sink is generic with nothing printed on it.

Do you think it was a gain bery kit?

I'll get another cyrix 586 100mhz in a couple of days but it looks very different to this one and has a blue heat sink stuck to it which is labeled cyrix. Its also very banged up with a number of bent pins.

The second number on the ISS chips. Yeah all of mine ...30 in total have the same number however they each have different numbers on the underside.

The winbond print is very weak isn't it?

Yeah they came from one of the million chinese IC sellers out there named Lisa, Linda, Sam, or Amy.

Word of advice to you who want to use seekic. Dont use any of your main email accounts. Like 200 emails a day is a tad annoying.

feipoa what heat sinks and fans do you use on your Cyrix 586 chips?

How can I get a decent fan/heatsink combo firmly on these chips?

It could have been Gainberry. Actually it doesn't really matter. There isn't anything special about Gainberry kits. They are just using off the shelf 5x86Cs and VRMs.

Nice cache. But the winbond chips do look weird compared to the older ones. Maybe it's just because of the late production.

oi, was that Yamaha card on eBay? I was bidding on one like that because it was full size, so you don't need to make a crappy cable to a DB. Then the seller cancelled my bid and pulled the auction - that's happened a few times lately. It was in Bris.

sorry mate yeah it was on ebay

well i still cant run my cyrix 120-gp at 3.6 volts...🤣....what a pain in the ass

I can run the 120 chip at 3.45 volts and it seems to work fine with all cyrix enhancements turned on except BTB_EN - divide overflow in speedsys.

However the Cyrix 100-GF chip runs perfect OCed to 120 with all enhancements on and at 3.3 volts.

rgart wrote:

well i still cant run my cyrix 120-gp at 3.6 volts...lol....what a pain in the ass

I can run the 120 chip at 3.45 volts and it seems to work fine with all cyrix enhancements turned on except BTB_EN - divide overflow in speedsys.

However the Cyrix 100-GF chip runs perfect OCed to 120 with all enhancements on and at 3.3 volts.

You will probably want to read Cyrix Register Enhancements Revealed. I suspect that you enabled BTB while leaving LOOP and RSTK enabled. For branch prediction (BTB) to work, LOOP and RSTK need to be disabled. Depending on the particular CPU, BWRT and DTE may also need to be disabled. I think for the Step1. Rev. 3 chips, BWRT, LOOP, and RSTK need to be disabled. For the Step0. Rev. 5 chips, it may have just been BWRT and LOOP which need to be disabled.

When overclocking these Cyrix/IBM 5x86c chips, be aware that some of the special features overclock well, and not others. So if you are having some stability issues, you'll want to cycle on/off features to determine which feature needs to be disabled. For my Cyrix 5x86-133/4x chip, DTE was the culprit, but only at 133 MHz.

Can you define what you mean by "work fine"? I like to have all hardware in a proper case, install WinNT 4.0, W2K, all the updates, let some benchmarks go for a few hours, then let an mp3 play overnight before I feel confident with a setup. EDIT: A few passes of MemTest 86+ v4.00 is also useful.

I don't understand how you are trying to run the Cyrix 5x86-120GP at 3.6 V. I've never seen a Cyrix 5x86-120GP run at 3.45 V and not 3.6 V. What measuring device are you using to ensure the chip is receiving 3.6 V? You can hook up a multi-meter to Vcc and Vss.

I have never seen a Cyrix 5x86-100 with a proper blue Cyrix heatsink. Even the official Cyrix 5x86-120 heatsinks were green.

What heatsink do I use? I just use the blue IBM heatsink that is stuck onto the chip and bought some higher flow fans from the local electronics store. I am working on something more elaborate now, but I just need to get some 168-pin 486 PGA's, which has been tricky.

Ill do some reading and testing and get back to you.

Why do you use Windows 98 to test your cpu is stable and not Windows 95 C?

rgart wrote:

Ill do some reading and testing and get back to you.

Why do you use Windows 98 to test your cpu is stable and not Windows 95 C?

Everything runs a little slower on Win98 compared to Win95. The slower the OS, the higher the CPU strain. I find that Win98 is forgiving on some stability issues, so NT4.0 and W2K are better choices to check into once you feel Win98 has passed this stage in qualificaiton.

feipoa wrote:

I suspect that you enabled BTB while leaving LOOP and RSTK enabled. For branch prediction (BTB) to work, LOOP and RSTK need to be disabled.

Yeah I was doing that...
Windows 98 on a 486...ouch 😜

the 100GF is a Step 0 Rev 5 chip.

feipoa wrote:

Can you define what you mean by "work fine"? I like to have all hardware in a proper case, install WinNT 4.0, W2K, all the updates, let some benchmarks go for a few hours, then let an mp3 play overnight before I feel confident with a setup. EDIT: A few passes of MemTest 86+ v4.00 is also useful.

I guess when I say it "works fine" I mean it does not crash when I put the system under load doing the things I like to do 😜

Doom1 and Doom2, Monkey Island, Lands of Lore, Darksun, Ultima
general dos usage.
playing mp3's and messing with sound cards

I will do some benches and tests and post the results when I think the system is stable.

the ECS UM8810 board has been nightmare and I have had all sorts of crazy stability issues Ive never seen before....but Im confident I can get it stable.
the target OS is win95 or dos 6.22
the quality control in some of these old companies is laughable.

a ctrl-alt-del consistantly shows a screen full of coloured ascii characters before rebooting. its only a quick glimpse but little things like that make me think this board has not been refined at all by the makers. Im not the only one with that complaint about the ascii i notice.

Im currently using:

IBM 5x86-3V3100GF
ECS UMM8810 PAIO mobo rev 2.1(pheonix bios 4.04)
128k Cache 15ns (cringe I know...)
128 MB Fast Page Ram (should i drop back to 64 or 32?)
S3 virge pci on board n1e3bd
YAMAHA YMF718-S ISA
Adaptec SCSI 2940U2W bios 2.00
SCSI Quantum Atlas 18gb 68pin (awful noisy.ruins gaming experience. Replacing soon)
Generic IDE DVD Drive (ide cd and scsi hd. Is that ok?)
Windows 98 SE

multiplier is 3x
fsb is 33x

( id rather not pull my only MG200 Matrox Millenium out of my K6-2 system so unless I can get another cheap..........)

(Ive got to get my hands on 512K cache as 64kx8 or 128kx8 like the board supports)

feipoa wrote:

I don't understand how you are trying to run the Cyrix 5x86-120GP at 3.6 V. I've never seen a Cyrix 5x86-120GP run at 3.45 V and not 3.6 V. What measuring device are you using to ensure the chip is receiving 3.6 V?

All I meant about the 3.6 volts is that I STILL dont have the voltage module for the UM8810 board. I cant find one. So Im unable to test my Cyrix 5x86 120-GP chip. I can run the 120-GP chip at 3.45 volts and it boots but I have no idea about the long term stability.

I was messing with an oc of 120mhz on this 100-GF and 3.45 volts and i had issues so Im going back to its bare roots. Im back at 100mhz on 3.3v and Im going to start by getting that stable.
I am using ibmm9.exe cyrix utility
Feipoa: when I use your default settings of...

RSTK_EN 1
BTB_EN 0
LOOP_EN 1
LSSER 0
WT1 1
BWRT 1
LINBRST 1
FP_FAST 1
MEM BYP 1
DTE_EW 1
IORT 000
USE_WBAK 1

I dont even make it out of the program. When I hit F3 to save some funky text comes up and then I get a super hard lockup.

I have no idea if the motherboard supports LINBRST - Theres very little text available and experimentation on the ECS UM8810 rev 2.1 board.

Im currently using:

RSTK_EN 1
BTB_EN 0
LOOP_EN 1
LSSR 0
WT1 1
BWRT 1
LINBRST 0
FP FAST 1
MEM BYP 1
DTE_EW 0
IORT 000
USE_WBAK 1

so far so good. fast and stable. windows 98 has not had any BOD's.

I have tried running DOOM2 but I noticed the keyboard feels wierd. As if theres larger than usual keyboard latency and Im sliding around the floors......I can only feel it ever so slightly. Its difficult to explain but its not "zippy" and responsive like my VLB AM5x86 running at 3x50mhz.

my other system is :

AM5x86 133ADZ chip OC'ed to 150.
1MB 10ns Cache
64MB Fast Page ram
Generic Motherboard - Asaki/hmc chipsets
Generic IO card
ET4000W32i VLB 1MB
CT1600 Sound Blaster Pro 2
CT1740 Sound Blaster 16 (MT32/SC55 interface)
Maxtor 1.01 GB Hard disk

multiplier x3
fsb 50mhz

I know the cyrix 100 pci based system is faster than the amd 150 vlb system but Ill keep testing and trying to get a full stable system.

Where can I download the windows 98 update file your using feipoa?

The unofficial Win88SE service pack is everywhere online. It is even on those annoying download sights which have 10 "download" buttons, whereby you need to figure out which download button is the correct one.

I think an AMD X5-150 is faster than a Cyrix 5x86 at 100 MHz. Are you using a VLB/ISA-only board for the AMD X5-150 setup? Can the VL bus tolerate this kind of speed long-term? What memory speeds do you get for cachechk -d -t6 and cachechk -d -w -t6? Are you able to pass at least one round of MemTest86+ v4.00? Are you using just 1 stick of 64 MB for the memory? I've been able to get more stable system with just one stick and fast BIOS timings.

Yes, the older UMC 8881/8886 chipsets and/or BIOS implimentations do not support linear burst modes on the Cyrix 5x86. The lock-up you are describing is certainly due to the board not supporting this feature. Not to worry though, the performance impact is rather minimal. Are you able to run the system with LINBRST at 0 and BTB at 1?

Perhaps some IBM 5x86-120 chips might be stable at 3.45 V. I know Gainbery had their voltage regulator at 3.45 V and sold them with overclocked IBM 120 MHz chips.

For testing stability, I've noticed the 1024K cache on at least one motherboard does not tolerate fast cache timings at 40 MHz very well. I recommend first setting the L2 cache to write-through mode. Put the cache and memory timings on auto, which will be slow, then run Memtest through Test 6 (or all the way through test 9 if you wish). If it does well, put your timings on the fastest possible, run Memtest again. if it fails, try to find the middle ground with the timings and passing of MemTest. It is also a good idea to check your L1/L2/RAM read/ Ram write with each timing change.

feipoa wrote:

The unofficial Win88SE service pack is everywhere online. It is even on those annoying download sights which have 10 "download" buttons, whereby you need to figure out which download button is the correct one.

Those pages drive me crazy.

Its even funnier when you have to signup and THEN get the run around and the files not even there.

feipoa wrote:

I think an AMD X5-150 is faster than a Cyrix 5x86 at 100 MHz. Are you using a VLB/ISA-only board for the AMD X5-150 setup? Can the VL bus tolerate this kind of speed long-term? What memory speeds do you get for cachechk -d -t6 and cachechk -d -w -t6? Are you able to pass at least one round of MemTest86+ v4.00? Are you using just 1 stick of 64 MB for the memory? I've been able to get more stable system with just one stick and fast BIOS timings.

speedsys indicates the cyrix 100 at 3x33 with enhancements turned on is faster than the AMD 150mhz at 3x50.

Yes it is the VLB/ISA board we have talked about before. I cant fault it when the bus is running at 50mhz. Long Term use? I have no idea. It seemed like the ET4000 VLB was the only card willing to take the 50mhz bus. The S3 805i VLB did not agree with a 50mhz bus.

I really like the board but Im willing to abuse it some. I have a really nice Forex 486 board to replace it if need be but I hope it lasts a long time. I have left it running doom, hexen, duke3d, heretic demos for around 48 hours without any issues. the PC itself has never crashed under any circumstances.

I definitely have not put it through your rigorous testing process feipoa.

My 1024k Chinese cache does some strange things to this board. Many system reporting programs report 0 cache. cachechk reports 1024k cache and everything is above board except the RAM speed.

cachechk reports RAM speed is 0.0mb/second and I get that message "RAM at megabyte 3 is very very slow" It annoyed me in the beginning so I swapped the cache chips with other Chinese reprints and swapped the RAM but nothing I did changed it so I gave up on that one. Im sure if I put the original 256k back the ram speed goes back to normal.

However the PC runs at optimal speed and I have had no stability issues so I let it be.

I will post some pics, testing and results on my days off.

I noticed the ET4000 was getting quite hot after 48 hours of doom so I swapped the CPU back with an Intel 100 DX4 but I still noticed the ET4000 chip gets very hot so I went back to the AMD 150mhz.

With regards to the other PC - the cyrix - I will test LINBRST 0 and BTB 1 and get back to you. Im sure BTB works fine but not with other enhancements.

I will also get back to the Cyrix 120-GP chip and attempt to over clock it at some point in the future. I will try 3.45volts extensively when I have time.

Unfortunately I think I pm'ed you about it - the bios on the 3x50fsb system is AMI and after trying 10 or so AMI programs I couldnt find anything that allowed me to change cache timings and its certainly not available in the simple bios. again I dont have a manual or can find very little details on the board.

I will run memtest and let you know how it goes.

I have only ever run memtest from a bootable CD and most 486's cant boot from a CD. Can I download it and run it from a dos prompt? Ill take a look.

I'd give anything for a manual. Im not methodical enough to write a manual on the jumper settings by trial and error. Im sure the board supports cyrix 5x86 somewhat.

ok Ill take a few pics now until I can put some serious effort into these two pc's.

A well configured AMD X5-150 should be faster than a Cyrix 5x86-100. If the L2 and RAM timings need to be dropped too low, then I suppose the Cyrix 5x86-100 could be faster. The Speedsys score is not necessarily indicitive of real world performance.

Are you using 'cachechk -d -t6' to check your cache and memory speeds? Did you put the L2 to Write-through mode in the BIOS? Can we see your speedsys chart for the AMD X5-150? Expect 1 in 10 of the Chinese cache to be bad. You may need to cycle through them to find the bad one. Use HIMEM and MEMTEST for htis. Do you get any errors in MemTest86+ 4.00?

MemTest86+ 4.00 comes with a downloadable bootable floppy disk. Do not use v4.20 on a 486. 4.00 seems to be the latest version to work on 486's.
http://www.memtest.org/download/4.00/memtest86+-4.00.iso.zip

I will mem test tonight

I cycled through all the cache chips. Still has issues with regards to RAM speed.

Ok:

these pics are NOT about the cyrix machine

this is the AMD 133 running at 3x50mhz

[imghttp://imagizer.imageshack.us/v2/1024x768q90/211/dsc00944o.jpg[/img]

I really dont have many options with that board.

Im unable to change cache timings.

I cant use WB mode.

If I change a few jumpers I can get the CPU registering as a AM5x86 when the PC boots but external cache turns off automatically and I am unable to turn it back on even when I set WB on the motherboard

Then I get a speedsys result of this:

Here are the jumper settings I used to get 150mhz. Dont ask me what half of them do it was trial and error and took some time.

the board...great little board.

this is an old pic. cache is different now.

I am currently playing with a SiS board. I have witnessed this funny RAM speed slow down if the L2 cache is set to write-back mode and if the TAG is set to 8-bit. Did you put your L2 cache onto write-through mode? If that doesn't fix it, try putting 256 KB cache back and jumper the board for 256 KB cache.

[Something odd I just witnessed now for this SiS board is that the Cyrix 5x86 chip is so fast that L2 and RAM timings need to be slowed down for the system to be stable, whereby an AMD or Intel chip at the same frequency could run with the fastest timings. I hope not all SiS 496/497 boards are like this. The makers of the DTK PKM-0033S really didn't put much time into testing this thing. There's another odd problem whereby using the POD at 83 MHz will lock-up the keyboard after a few minutes of uptime. I can't image why this board had no BIOS updates.]

I'm not sure why there is a RAM slow-down with the AMD X5-150. The turbo button jumper is sometimes to blame. Try the L2 cache on WT mode. Also, try slowing down your RAM timings in the BIOS. I've seen some crazy things happen with high FSB frequencies whereby the memory read speed would cap out at about the middle speed setting in the BIOS.

I'm pretty sure setting the L2 to write-through mode will solve things for you. I just experience this same event and my speedsys graphs looked just like yours until I put the L2 into WT mode.

EDIT: You randomly trying jumper configurations? Perhaps make a list of all the jumper settings silk screened on the PCB and try to make sense of it. Keep looking for the manual. Don't forget to try jumpering the turbo button.

Ill try jumpering the turbo pins.

I used the silk screens on the motherboard as much as I could but there are many jumpers that are not labeled.

I will reread your post in the morning and see what I can do to change the RAM being so slow.

Back to the Cyrix System.

everytime I turn on the PC WT1, USE WBAK and LOCK NW are already turned on (1)

Is that usual?

Do i have to use autoexec.bat or command line to turn enhancements on. Theres nothing more permanent?