The coolest socket 7 motherboard that you’ve never heard of

if this one hasn't been mentioned yet.

in the aiming for the stars thread I think I recall it getting around 5000 record which is among the highest results next to the record holder asus and gigabyte models who got to mid 5000s

Tetrium wrote on 2021-11-28, 10:01:

Well, putting a V800 engine into a lawnmower doesn't automagically make it teh fastast lawnmower in all of whooman history 😜

You're right.
You'll need a jet engine!

stef80 wrote on 2022-04-25, 07:48:

Does anybody have more info on EPoX EP-51MVP3E-M. Good, bad or in-between? It has 1MB of L2 and supports K6-3+ / 127GB disks with latest bios.
It's either that or DFI K6XV3+/66 [512KB L2].

The EPoX only has UDMA33, while the DFI has UDMA66.
As far as I know the DFI K6XV3+/66 is the only SS7 board with UDMA66 and 3 ISA slots (more soundcards!). All other boards I've seen so far either have only one/ two ISA slots or lack UDMA66. I don't see the point in having 6 PCI slots.

The K6XV3+/66 also has variants with 1MB or 2MB cache! But the 2MB variant is very rare. Have only seen one so far and it was described as being defect. Seller stated that it would often freeze during post. Hirsch did bid on it, hoping a recapping might solve the issues, but got outbid. He didn't want to risk too much.

The 512KB and 1MB variants are way more common. We've seen around 10 of them pop up during the last year.
I have three of them. 1x 1MB and 2x 512KB. But one of the 512KB ones isn't working, while the other one needs a recap.
Hirsch also has three, with one being broken. Not sure about the cache. I think his are all 1MB.
It's true that the caps aren't the best. But that's not a reason for/against any board imho. In the log run all old boards will need a recap.

As stated before, it can't go beyond 100MHz FSB.
But it can do something which most SS7 boards can't: Go below 2V.
So you can use one these low voltage 400MHz K6-III at 1.6V. Or downclock+downvolt one of these 570MHz K6-2+. 6x66MHz, since you can change the multiplier via SetMul but not the FSB. At 1.6V you don't even need a heatsink. 😁 But I would still use one as passive cooler.
Excellent choice for a DOS/Win combo machine.

It's true, DFI has gazillion combinations for jumper + DIP switch for voltage regulation. It starts from 1.3V or something about.
Why is DFI the only board to support UDMA66, isn't chipset the same on all boards (VIA MVP3) ? Or is it SMSC chipset on DFI that integrates IDE controllers?
BTW, EP-51 also has 3 ISA slots.

stef80 wrote on 2022-04-26, 09:33:

It's true, DFI has gazillion combinations for jumper + DIP switch for voltage regulation. It starts from 1.3V or something about.

Yes, 1 jumper and 5 switches.
2^6 = 64 combinations
1.3V - 3.5V with 0.05V increments

stef80 wrote on 2022-04-26, 09:33:

Why is DFI the only board to support UDMA66

It's not the only board which supports UDMA66.
It's the only with UDMA66 AND three ISA slots, afaik.

stef80 wrote on 2022-04-26, 09:33:

isn't chipset the same (VIA MVP3) ?

It's the same chipset, but there is a separate controller chip for IDE I guess.

stef80 wrote on 2022-04-26, 09:33:

BTW, EP-51 also has 3 ISA slots.

Yes it has. But only UDMA33.
There are plenty of boards with 3-4 ISA slots and UDMA33.

Does someone actually reached 33 MB/s speed with these old boards and cpus to make UDMA66 more than paper trait?
From my experience all disc operation heavily depends on cpu and they are slow.. I tried 1 Gbit network cards with SS7.. and with cpu limits i never reached speed of 100 Mbit..

ruthan wrote on 2022-04-26, 12:36:

Does someone actually reached 33 MB/s speed with these old boards and cpus to make UDMA66 more than paper trait?
From my experience all disc operation heavily depends on cpu and they are slow.. I tried 1 Gbit network cards with SS7.. and with cpu limits i never reached speed of 100 Mbit..

There are IDE controller cards with their own CPUs (ie. Intel's i960RM like on this SuperTrax SX6000) to offload main CPU for most of I/O related work so 66MB/s is achievable. Earlier 486-era cards often used regular 286 CPUs for the same purpose. As for the integrated components, well, that would certainly be harder.

HanJammer wrote on 2022-04-26, 12:47:

ruthan wrote on 2022-04-26, 12:36:

Does someone actually reached 33 MB/s speed with these old boards and cpus to make UDMA66 more than paper trait?
From my experience all disc operation heavily depends on cpu and they are slow.. I tried 1 Gbit network cards with SS7.. and with cpu limits i never reached speed of 100 Mbit..

There are IDE controller cards with their own CPUs (ie. Intel's i960RM like on this SuperTrax SX6000) to offload main CPU for most of I/O related work so 66MB/s is achievable. Earlier 486-era cards often used regular 286 CPUs for the same purpose. As for the integrated components, well, that would certainly be harder.

That's interesting, but on retropc with msdos and windows98 would it be useful for anything?

Nemo1985 wrote on 2022-04-26, 13:20:

That's interesting, but on retropc with msdos and windows98 would it be useful for anything?

"retropc"? You meant vintage PC right? Under Win9x it's just a matter of drivers - manual for SuperTrax 6000SX states there is no support for Win9x because there are no drivers, although I found some drivers so maybe they release it later. Under DOS it will use int 13h BIOS extension. I/O processing is AFAIK done by the card independently of the drivers. But honestly I don't think that under DOS HDD speed matters alot as the games are usually small anyway (I wonder what is the biggest HDD installable game for DOS? I'm not counting multi-CDROM releases with alot of digitized video content, because those while being large, didn't really installed much on the HDD anyway).

HanJammer wrote on 2022-04-26, 12:47:

There are IDE controller cards with their own CPUs (ie. Intel's i960RM like on this SuperTrax SX6000) to offload main CPU for most of I/O related work so 66MB/s is achievable. Earlier 486-era cards often used regular 286 CPUs for the same purpose. As for the integrated components, well, that would certainly be harder.

But its just work around.. some additional card, which bypass chipset / cpu limits, you dont need any MB UDMA 66 support to make it running, just PCI bus. Same as some SCSI PCI card, this one is probably only quicker and more rare and expensive... Maybe someone can make some test, if even some cheap PCI IDE controller helps to decrease CPU load and increase real disk performance.. Because some fancy IDE 100 MB/s / 133 MB/s tag and theoreticall speed support means anything.

I have SSD inside some one these slimmer pcs, but it seems not help too much.. or these cheap IDE to Sata adapter are bad for performance.. I never so i mood to make more testing in this area, people usally make zillion tests for cpu and graphics cards, but very few for disk subsystem and network card transfer speeds etc..

Any opinion on GA-5AX rev 4.1?

PROS I noticed:
Memory performance is excellent compares to DFI k6xv3+/66
CONS:
ATA 33, but with ultra66 PCI it is not a problem anymore.

What about AGP performance? I forgot to do the tests on that MVP3 chipset.
I have Voodoo 3 2000

RaiderOfLostVoodoo wrote on 2022-04-26, 09:49:

stef80 wrote on 2022-04-26, 09:33:

isn't chipset the same (VIA MVP3) ?

It's the same chipset, but there is a separate controller chip for IDE I guess.

It´s a matter of the southbridge, there´s no seperate controller chip.
The one capable of ATA66 is "VT82C596B" (BGA package).
Earlier versions that were also "MVP3" had the "VT82C586B" (QFP package) and were limited to ATA33.

(Epox MVP3G5)

(FIC VA503+)

So if you see a VIA MVP3 mainboard with BGA southbridge it´s ATA66.

candle_86 wrote on 2021-11-25, 02:00:

Jose_Luis wrote on 2021-11-24, 21:07:

candle_86 wrote on 2021-11-23, 15:41:

considering how CPU bottlenecked it is, you ideally want a card that uses a sub 10 series driver, so I'd stick to TNT2/GF2/GF3, or a Voodoo because they have really good 3dnow support, avoid ATI as their 3dnow support was always trash

I'm not sure about the graphics performance wouldn't be improve because the CPU is bottlenecked.. The graphics functions are made for the GPU of the graphics card. I'd like to test what you told me.

I tested a beginning of this year a Tyan S1834D , 2 Pentium III 1GHz, 2GB SDR, Windows 7 Professional 32 bits. The motherboard and the CPUs are from year 2000. First, I checked that with an Ati Radeon HD3850 AGP 512MB GDDR3 (released in 2008), the Windows permonce index in graphics 3D was 6.8 finally, I checked that with a Geforce 7600GS 512MB DDR2 (released in 2006), the Windows performance index was 3.7. I repeated the test and the values were identical.

https://www.youtube.com/watch?v=HRhm4aGNI3o

He covers it very well

That´s right but that comparative has been made with Windows 98, you wouldn´t be able to test a Geforce 7600GS 512MB DDR2 on Windows98. It´s the point you can use Windows 7 Professional 32 bits acceptable in a Super Socket 7 motherboard nowadays with a graphics supported on Windows 7.