DenizOezmen wrote on 2020-07-05, 09:17:

dionb wrote on 2020-07-05, 09:03:

Of course Asus did make one i8xx board that has it all: the Asus P3C-E, with native 133MHz FSB, performance on par with i440BX and an ISA slot too. But rare as hen's teeth as RDRAM, which these days is no more expensive than SDRAM, but 20 years back cost far, far more. I have one, but it's dead :'(

But is still only goes up to 1 GB of RAM because of the i820, doesn't it?

Unless you can get it stable on 3 RIMMs (which, let's face it, isn't possible), no.

Edit: Maybe to clarify this a bit: When I was talking about a "sweet spot", this was purely from a personal standpoint. Obviously, there is no such thing as the perfect mainboard, and everyone has to settle for their own compromise. Having worked with virtual machines early on made chipsets with a high RAM limit attractive. RDRAM was prohibitively expensive, and pairing an i820 with an MTH to use affordable SDRAM meant basically losing the performance gain over earlier Intel chipsets or those of competitors. (And in any case, the RAM limitation stayed in place.)

Not competely sure what the MTH does in terms of limits. I have a working i820 board (P3C-D) and an MTH-on-a-stick which is rumoured to take 512MB DIMMs. The mind boggles. Maybe I'll give it a try one day. Still, no doubt that the MTH was a monstrosity, combining the high latencies of RDRAM with the relatively low bandwidth of SDRAM, and adding a bit of instability for good measure.

In the end, the i8xx chipsets never made much sense me (regarding my intended usage).

What on earth were you doing back then that needed >1GB of RAM? Even with virtualization, that's pretty heavy.

PARKE wrote on 2020-07-04, 16:34:

Therefore the later revisions are likely more reliable than the rev. 1.03 - not less.

Another aspect is that ASUS also catered for an overclocking audience and some of the later improvements may well have focused on those users. Think for example running a 1Ghz/133 cpu with 4x256Mb of PC133 SDRAM.

P3B-F rev. 1.03 is capable running Tualatin 1.4S /133 + 4x256 RAM w/o any issues. I got this MB brand new, about 20 years ago. Can confirm it had no modifications except 'unofficial' BIOS.

DenizOezmen wrote on 2020-07-05, 09:17:

But is still only goes up to 1 GB of RAM because of the i820, doesn't it? […]
Show full quote

dionb wrote on 2020-07-05, 09:03:

Of course Asus did make one i8xx board that has it all: the Asus P3C-E, with native 133MHz FSB, performance on par with i440BX and an ISA slot too. But rare as hen's teeth as RDRAM, which these days is no more expensive than SDRAM, but 20 years back cost far, far more. I have one, but it's dead :'(

But is still only goes up to 1 GB of RAM because of the i820, doesn't it?

Edit: Maybe to clarify this a bit: When I was talking about a "sweet spot", this was purely from a personal standpoint. Obviously, there is no such thing as the perfect mainboard, and everyone has to settle for their own compromise. Having worked with virtual machines early on made chipsets with a high RAM limit attractive. RDRAM was prohibitively expensive, and pairing an i820 with an MTH to use affordable SDRAM meant basically losing the performance gain over earlier Intel chipsets or those of competitors. (And in any case, the RAM limitation stayed in place.)

In the end, the i8xx chipsets never made much sense me (regarding my intended usage).

Which exact i8xx chipsets are you referring to?
The i820 also had a lot of trouble running 3 RIMMs stable. I tried using an ES VC820 and could never get it to run an OS stable. But when I switched it with a retail VC820 the stability issues were gone.
I never had a i820 chipsetted board with the MTH but I heard pretty bad things about it.

I suppose Intel deliberately limited the maximum amount of RAM the i815 could support to limit future expandability and Intel's marketing push to 'force' people to buy RDRAM was one of their really bad stunts. It really made a mess of things back then.
I remember RDRAM being something like 5 times as expensive compared to SDRAM and DDR? Nuts! xD

dionb wrote on 2020-07-05, 15:07:

What on earth were you doing back then that needed >1GB of RAM? Even with virtualization, that's pretty heavy.

It is, but it wasn't as much room for future expansion compared to many of the older chipsets that went before it. When the 440BX was new, I got 128MB SDRAM with it but it would be expanded to 6x that amount (my 440BX board had only 3 RAM slots). And people who got a run-of-the-mill Pentium 1 board would often pair it with something like 8MB or 16MB, though the boards could cache at least 4x that amount (and support much more though outside of the cacheable area).
But when i815 was new, 256MB was already decently common and close to the maximum one could expand the RAM to. Even my Phenom 2 with which I started with 4GB ended up having 16GB (so 4x the starting amount).

Of course back then 1GB was a lot!

It's a shame really, i815 would have been so awesome if it had not been engineered with this artificial RAM limitation 😒

Tetrium wrote:

I suppose Intel deliberately limited the maximum amount of RAM the i815 could support to limit future expandability and Intel's marketing push to 'force' people to buy RDRAM was one of their really bad stunts. It really made a mess of things back then.
I remember RDRAM being something like 5 times as expensive compared to SDRAM and DDR? Nuts! xD

The last iteration of the Rdram chipsets, the i850 (and the i860) never supported 1GB Rdram sticks, so the limit was 2GB. For early 2003 a low limit. I had the bad luck (Intel would rather say "honour") of investing in a 232-pins 32bit PC4200 rdram solution, which got dumped by Intel just month later when they introduced the Dual Channel DDR-SDram chipsets...
I still have a working P4T533-R system, which earned its place in my collection by merit of that strange ram. Maxed out at 2*256 MB...

Just a quick note. 1.04 Version was a PCB layout change to fix the "Photoshop Bug". It may have also incorporated the additional changes mention in the previous posts (VRM change etc).

Hello, I have a P3B version without an ISA slot. I wonder, if you can solder ISA slot to it, and flash BIOS from version with ISA slot, and get working ISA slot?

I believe soldering ISA slot should work just "plug and play" as it's the same BIOS for all boards. Before doing that I would check for missing components as manufacturer might done some cost reduction work when left that slot NA. Soldering that ISA connector is another story as every hole needs to be opened... I would personally just find another board with ISA.

I was thinking about soldering a missing SBLINK header to my P3B-F 1.03 with ISA slot.

Stability issues on the P3B-F v 1.03... These only occur when I try to run my 933 coppermine at 150MHz fsb. 140 is totally stable with 1,85v and massive socket A cooler on a slotket 😀
So this one is probably drawing a little more current than intended and gives me no problems at all. Not even with the also demanding Ti4200.

Just for the info 😀

dionb wrote:

And that southbridge with its issues vs various PCI devices (SBLive...).

P3V4X has VT82C596B south bridge, unlike most other boards on Apollo Pro 133A, which were paired with VT82C686A. It's more stable, but has very lackluster ATA66 performance.

Has anyone tried the P3 1.0B Ghz (133 fsb) on the P3B-f rev 1.03?
From what I have gathered from multiple forums some had luck some did not with 133fsb cpus on the Rev 1.03.
Could it be possible later production of the Rev 1.03 with one ISA work ok with 133FSB cpus, than possible older Rev 1.03 with 2 ISA slots?

I don't see why P3 1GHz 133FSB wouldn't work.

I'm running P3 1,4GHz @150FSB with 2x256MB Apacer sticks. Memtest gives error if memory settings are tightest so needed to set precharge to 3T.
Mobo is P3B-F v.1.03 with 2 ISA slots.

3DMark01SE with Geforce 5900XT: 11108

With 4x256MB sticks I can run @140FSB.