Right, that figures. I need to get 100 mbps cards to stress the bus. I'm still going to install the EISA card because this is the only machine I have with that bus and I'm curious to learn how it works.

majestyk wrote on 2024-04-21, 17:56:

PC Hoarder Patrol wrote on 2024-04-21, 14:23:

Is it still possible to find correct VRM modules for these boards - sure I have an old Siemens board somewhere, just no processors (easyish?) or VRMs (hard?!)

At the moment you have the chance to get some really cheap, it´s article number 135001060040 on Ebay. These VXI VRMs are top quality, but they have quite tall caps. This might cause problems when the VRM socket on the mainboard is positioned in the wrong orientation - with the VRM´s component side towards the cpu cooler.

Horun wrote on 2024-04-21, 19:14:

Those are good ! The caps appear to be 20mm (TI UCC3882 datasheet) so you need nearly an inch from vrm socket slot to heatsink if orientation is toward the cpu.....

Thank, those VRMs do look like a good deal. If my memory serves (been a few years now since I've even seen the board...recall it was stripped from an old SNI Primergy server) the VRM sockets are off to the side so shouldn't foul the processor sockets. Really must find some time to hunt it down, assuming I even still have it!!

I still have a box of Compaq VRMs if shipping from germany is desirable?
I Confess beeing multi cpu addicted such as luckybob, so there was no way bypassing the PPro.
I was lucky purchasing my systems in the early 0-years when they were availiable cheap even on ebay.
Meanwhile I own 20+ dual and quad PPro systems, but just 2 single PPro ones.

As for the OP:
looking for an socket 8 board is neither worth the efford nor the money.
Simirar to socket 4 the boards are overpriced due to their rareness and so are coolers and VRMs.
Socket 8 sytems are bottlenecked as they do not feature AGP graphics and most of them use PS/2 memory.

eisapc wrote on 2024-04-23, 07:54:

Simirar to socket 4 the boards are overpriced due to their rareness and so are coolers and VRMs.

But the (new) VXI VRMs I recommended for € 6.50 each aren´t exactly overpriced, I would say.

Errius wrote on 2024-04-22, 04:46:

Right, that figures. I need to get 100 mbps cards to stress the bus. I'm still going to install the EISA card because this is the only machine I have with that bus and I'm curious to learn how it works.

1GBS Ethernet still is within PCI specs

I've been looking for an IBM 365 (dual Pentium Pro) for years without success, it's nice to have in the collection to install NT4.

As far as the Mac is concerned, I also fully vote for a Powermac that can boot natively with both OS9 and OSX.

I would also add that even if the Powermac G5 and the Intel MacPro were once very fascinating, I find that after having all become vintage, the case of the Powermac G3/G4 is much more beautiful and original and that it has aged much better than the G5 and Intel Macpro

rmay635703 wrote on 2024-04-23, 18:22:

Errius wrote on 2024-04-22, 04:46:

Right, that figures. I need to get 100 mbps cards to stress the bus. I'm still going to install the EISA card because this is the only machine I have with that bus and I'm curious to learn how it works.

1GBS Ethernet still is within PCI specs

Apparently VLB could run as fast as 1.7 Gbps. (200 MiB/s)

Errius wrote on 2024-04-25, 03:16:

rmay635703 wrote on 2024-04-23, 18:22:

Errius wrote on 2024-04-22, 04:46:

Right, that figures. I need to get 100 mbps cards to stress the bus. I'm still going to install the EISA card because this is the only machine I have with that bus and I'm curious to learn how it works.

1GBS Ethernet still is within PCI specs

Apparently VLB could run as fast as 1.7 Gbps. (200 MiB/s)

VLB 2.0 spec supported 66mhz (too bad nobody used it that I’m aware of)

Would be amusing to include some type of embedded all in one device with high speed WiFi gig Ethernet and usb 3 bridged on a vlb card.

Sadly drivers likely would need to be made from scratch to actually ping the vlb bus.

Technically it is possible for a VLB device to appear as a PCI one, by listening PCI conf mechanism accesses and responding to them as if it was a PCI device.

Tiido wrote on 2024-04-25, 04:55:

Technically it is possible for a VLB device to appear as a PCI one, by listening PCI conf mechanism accesses and responding to them as if it was a PCI device.

This idea is similar to how the 3c509 can be set into EISA mode and configured using the ECU by listening to the EISA configuration mechanism, even though the card only has ISA contacts.

Does anyone else think this thread has gotten COMPLETELY derailed?

dionb wrote on 2024-04-19, 15:19:

What mistake? The PPro is older than the MMX instruction set; Intel's engineers were at the top of their game back in 1995, but even they couldn't time-travel and include something developed a year later.

Moreover, it was useless and the fact all Intel CPUs introduced from 1996 onward had it doesn't in any way change that. All humans are born with an appendix but it doesn't do anything and we can live without it. MMX was touted as 'multimedia extensions' but in fact that type of SIMD operations were rare

The “mistakes “ Intel made with the ppro were strategic ones like orphaning socket 8 and not offering at least one speed grade beyond 200mhz during its relevant lifetime (overdrive was too late),

electrically there was minimal difference between signals found on socket 8 vrs 370 which was another kick in the pants to socket 8 owners when celeron released (similar to socket 4)

Another mistake was that AGP, sdram support and Intels Starfighter 740 3d video card should have been pushed out as a prosumer thing for ppro as its main advantages beyond being a fast pci slot were mostly relevant circa 1995, all were in a halfhearted place long before eventually releasing too late in the lifecycle to be relevant. -1998 way too late to matter.

Something that I long wondered having seen thousands of systems brought back sometimes for tear down/scrap or even return under an nda during my formative years hanging out at one of the major pc support and distribution companies 1997-2007 was why were Pro150/166 extremely uncommon compared to pro180/pro200. I only saw one pro133 returned under NDA.

Also interesting was that the ppro was not as egregiously expensive in single cpu form as many note (late 95 yes but prices eroded rapidly on pro180’s)
, late 1996 the local tech bought around 250x ppro 180’s with Micronics motherboards and they weren’t significantly more expensive than a top of the line standard pentium as I had an original invoice to compare that came in the pallet I bought nearly a decade later.

rmay635703 wrote on 2024-04-26, 20:30:

[...]

The “mistakes “ Intel made [...]

The discussion you are quoting me from was specifically about the lack of MMX support in PPro being called a mistake, not any of the other things you are referring to.

luckybob wrote on 2024-04-26, 17:59:

Does anyone else think this thread has gotten COMPLETELY derailed?

+1 🤣