64 MB RAM and 1 MB L2 cache is beneficial for Windows only, in DOS you can hardly see the difference vs 32 MB and 512 kb.

The Cyrix 5x86 is simply better suited for later gen. PCI boards - but not even those support it to its full capability. Take a look at feipoa's benchmark results, with the major M1 features switched on by software tweaks, the Cx 5x86 is much faster per clock than the AMD.

In my experience (testing over 20 VLB cards on overclocked high-end 486 systems) compatibility weighs a lot more than performance. What cannot be run smoothly on a 160 Mhz 5x86 with a VESA card usually needs a fast Pentium (or better!).

Jolaes76 wrote:

64 MB RAM and 1 MB L2 cache is beneficial for Windows only, in DOS you can hardly see the difference vs 32 MB and 512 kb.

The Cyrix 5x86 is simply better suited for later gen. PCI boards - but not even those support it to its full capability. Take a look at feipoa's benchmark results, with the major M1 features switched on by software tweaks, the Cx 5x86 is much faster per clock than the AMD.

In my experience (testing over 20 VLB cards on overclocked high-end 486 systems) compatibility weighs a lot more than performance. What cannot be run smoothly on a 160 Mhz 5x86 with a VESA card usually needs a fast Pentium (or better!).

thanks for the info. when you say 1mb cache helps in Windows does that include 3.1?

I'm not to concerned with compatibility, as odd as that sounds. I have a 66mhz 486 and a Pentium 133 machine so I think I have compatibility covered. this machine is mostly just a fun side project and a machine aimed at Win 3.1

I guess a Pentium overdrive would also be something to take into consideration. I did read Feipoa's post last night. its good and detailed but i'm not a super technical guy so it was a little long and complicated to read. Maybe theres a "high end 486's for dummies" somewhere. The charts were pretty though. I was tired though so I think i'll take anouther look.

If you want to know what a high-end 486 looks like without PCI, you should have a look at my VL/EISA 486.

If you want to build a "high end" system around a VL-ISA board you should probably avoid PC Chips. Go with something based on the SiS471 chipset with good quality PCB. These boards are not that hard to find, and they shouldn't be expensive.

I strongly disagree that the Cyrix 5x86 is less suitable as an upgrade chips for older 5V only motherboards. I have actually encountered more compatibility issues using AMD 5x86 in older systems.

when you say 1mb cache helps in Windows does that include 3.1?

I meant Win9x. I have no 3.1x installed on such a PC.

Less suitable - I meant speedwise. Greatest overclock results are achieved on the Biostar 8433-UUD, a very late PCI motherboard...

ah, well...maybe i'll just throw my amd 133mhz in the vlb board and pick up a cyrix 5x86 down the road for my M919 board. I just hate the whole L2 cache situation with that board. I even have the damn cache stick and I still get no cache.

I have a VLB board with the same UM8498F chipset. This one. But mine has the "-BS" code behind de week/year code, whereas the picture of the M912 shows a "-HYS".
It is surprising that it claims 1MB L2 support, and I wonder if that depends on a later stepping of the chipset.

Do you know if write-back processors are actually working as such on this board? If so, I would be very grateful if you could trace the 3 cache control lines from CPU to chipset.

hasn't actually arrived yet. should in the next few days. I'll check it out and let you know when I get it.

soviet conscript wrote:

hasn't actually arrived yet. should in the next few days. I'll check it out and let you know when I get it.

Okay, great. Did you perhaps buy the one Robin4 linked too? http://www.ebay.com/itm/271493239236

gerwin wrote:

soviet conscript wrote:

hasn't actually arrived yet. should in the next few days. I'll check it out and let you know when I get it.

Okay, great. Did you perhaps buy the one Robin4 linked too? http://www.ebay.com/itm/271493239236

bought it off a guy over at the vintage computer forums but could be the same guy. boards is exactly the same, even same CPU.

Correct me if I'm wrong but if I want write back cache enabled I can't also be using a SCSI controller card? I remember a post that said something about it conflicting.

Yeah, VLB and Writeback cache is generally a no no. Unless the motherboard chipset specifically supports it, it will crash the system. I think there may be a few very late VLB boards out there that can handle it.

Anonymous Coward wrote:

Yeah, VLB and Writeback cache is generally a no no. Unless the motherboard chipset specifically supports it, it will crash the system. I think there may be a few very late VLB boards out there that can handle it.

will a VLB video card and IDE controller be okay?

In most cases, yes. Rare incompatibility issues might arise but in most cases you can settle the the cards in master / slave position, adjust wait states to 1 if necessary and you are done.
Some of the the SiS 471 VLB boards can handle the later Adaptec SCSI cards and stay in write-back mode but I cannot tell which mb, which SCSI controller. But I definitely met such combo in the past.

soviet conscript wrote:

Anonymous Coward wrote:

Yeah, VLB and Writeback cache is generally a no no. Unless the motherboard chipset specifically supports it, it will crash the system. I think there may be a few very late VLB boards out there that can handle it.

will a VLB video card and IDE controller be okay?

Just don't buy combined VLB IDE controller + VGA, they tend to be slower than similar separate VLB VGA and ISA controller.

RacoonRider wrote:

soviet conscript wrote:

Anonymous Coward wrote:

Yeah, VLB and Writeback cache is generally a no no. Unless the motherboard chipset specifically supports it, it will crash the system. I think there may be a few very late VLB boards out there that can handle it.

will a VLB video card and IDE controller be okay?

Just don't buy combined VLB IDE controller + VGA, they tend to be slower than similar separate VLB VGA and ISA controller.

I didn't even know they made combined VLB VGA and multi I/O controllers.

board came in today. was sent in a bubble pack and not a box so the AT keyboard connector was bent back but luckily only the front two tabs were broke and not the actual solder points to the MB. I haven't got to play with it much. another thing that sucks is it only came with a whopping 32kb of L2 cache.

The BIOS looks exactly like the BIOS in my M919.

now this may sound stupid but how does one go about checking WB? cachechk? I have it set to WB in BIOS and it seems to be posting just fine. haven't had a chance to put it in a proper case and load an OS yet.

another question. It works just fine if I have an ISA video card installed but it refuses to display when I try one of my VLB cards it wont display anything. I've tried 2 confirmed working cards in all 3 slots. is there something I need to set? in my other setup it just worked like any other video card.

1) if you change only this option in BIOS to write-thru and benchmark results change (degrade) in several tests (CTCM, Speedsys etc.) then you can be sure that you are able to switch between WB and WT caching

2) check the hardware first. If ISA VGA cards work in a VLB slot then at least the 16bit part of the slot is OK - but the VESA connector part may be damaged. Inspect the pins of the connector. If these look OK, then you have a configuration issue and it can likely be resolved. Try setting everything to standard values in BIOS first and adjust all compatibility options - wait states, clocks etc and do NOT install any other card while trying to have it POST.

You might want to reset the CMOS first (discharge by unplugging PSU and taking out battery) AND the first time you boot, press DEL or INS LIKE HELL (or whatever keypress combo lets you enter the BIOS.) I know it might sound strange but on a few pickish motherboard I always had to make this ritual - entering the BIOS to get a VESA VGA card work. After that (and making sure the correct values were set) the motherboards booted up fine.

Keep in mind that if the same jumpers exist on a VESA card and on the motherboard, then the motherboard settings are overriden by what is set on the card.

(Edit: wrote this before Jolaes76's reply appeared)

Anonymous Coward wrote:

Yeah, VLB and Writeback cache is generally a no no. Unless the motherboard chipset specifically supports it, it will crash the system. I think there may be a few very late VLB boards out there that can handle it.

Sorry to hear that. WB-cache is the last thing I still wanted to figure out for my VLB board, but I now understand it is probably not ever gonna work...

soviet conscript wrote:

now this may sound stupid but how does one go about checking WB? cachechk? I have it set to WB in BIOS and it seems to be posting just fine. haven't had a chance to put it in a proper case and load an OS yet.

The main thing to enable WB on a supporting CPU is to set the WB/WT pin of the CPU high. This is done with a jumper. See the black dots in This chart. In case of an AMD chip SpeedSys will detect this setting and write 'Am486DX4WB' on the top left. Now I have never managed to get it to work properly on my UMC board, but I think that in SpeedSys the green 'moving' line in the memory throughput graph should appear raised, in the ranges covered by L1 and L2 cache.
(I can only see a raised L1 thoughput myself, wheras L2 actually performs worse. Also the system will start to hang when parsing config.sys/autoexec.bat at boot)

Here is some general info on 486 WB cache: J. Steunebrink: AMD 5x86.

soviet conscript wrote:

I didn't even know they made combined VLB VGA and multi I/O controllers.

Here is an example. The graphics benchmarks are actually on par with other VLB systems, no problem there. I haven't measured the disk thoughput, but it is acceptable at least.

Alright. I probably won't get around to testing until Tuesday since I have 2 12 hour shifts this weekend and the holiday but I'll definatly give it a full testing Tuesday and report. 16 bit cards definatly work in the vlb slots. The link to the and 5x86s was interesting. I have a "raw" chip as they put it in my machine but my spare is a Kingston turbochip. I just assumed its was the same chip just rebranded but apparently the turbochip only supports write-through. Never knew that.