Gigabyte GA586VX Socket 7 - Bad caps

[Topcat]

Yeah, you read this right, SOCKET SEVEN!!! This board was sent to me for repair, and the owner was insistant on it being fixed. This board is vintage 1996 with bad caps branded 'D.S.', which I've seen bad in a bunch of Chaintech and other knock-off Nvidia graphics cards. The cap values were 1000uF 6.3v and 470uF 10v.

This was a big time stroll down memory lane!! I didn't even have an AT power supply to test it, I had to go and hunt a used one. This is also the first time in about 5 years I've even touched a SIMM module...

Here's a couple pics:







Ok, here's one of it with new caps... It works, but DAMN it's SLOW!

[Ark42]

Mmm... is that a COAST slot?

[kc8adu]

Mmm... is that a COAST slot?

harder to fake than srams.
remember the fake cache fiasco?
btw i still get those old socket 7 boards in for service.
just did one for a mangotech truss saw.
had a funky bios.

[Topcat]

AHH YES!! A coast module.... For you young whippersnaps who's first system was a P3 or P4, 'coast' is short for 'cache on a stick'.... HEHEHE I remember that from YEARS ago when I got my A+ cert...

Another one that stuck was the old 'twain' acronym, which is a universal scanner driver... It stands for 'technology with an insignificant name'... Who thinks of this crap!! HEH

[krazykev64]

Re: Gigabyte GA586VX Socket 7 - Bad caps

and i was taught that TWAIN stood for THING WITHOUT AN INTERESTING NAME

[krazykev64]

Re: Gigabyte GA586VX Socket 7 - Bad caps

Anyway why wold some crazed fool want a 13 year old board fixed. must be very desperate. bring on the 286's or earlier

[Wizard]

Re: Gigabyte GA586VX Socket 7 - Bad caps

Yow, bring back 5 YEAR old thread!

Good thing I remembered and checked the data and tell you why Topcat found it extremely slow. If the memory is larger than 64MB on that 430VX chipest, outside of 64MB is uncached, this is because this horrible VX chipset is hardwired, same thing with TX as well. If the board is 430HX (Best I ever had!) with 512K with two tag cache chips installed, can cache full 512MB memory and speed is blisteringly fast.

Cheers, Wizard

[jpdoe]

Re: Gigabyte GA586VX Socket 7 - Bad caps

Hi, Wizard
You mean the RAM above 64MB is not cached by the onboard cache? I would guess it's cached by the cache on the CPU anyway, right?

I'm interested in this because I have several k6-2 systems at work that some genius purchased equipped with 128MB of RAM, when the L2 cache only caches half of it. I did some tests and found out that the RAM above the 64MB mark is accessed four to five times slower than the ram below that mark.
I had a Cyrix 486SLC with a broken L1 cache. With this cache disabled, the machine was slow as hell.
What I find quite impressive is the fact that some evil genius installed WinXP on one of those ill configured k6-2, and the thing has been working that way for more than a year. The employee who uses it is currently under psicological treatment
Adding insult to injury, the motherboard is a PCChips M598, known and loved for his SIS southbridge with a broken UDMA implementation that results in the hard disks always working in PIO mode.

[pfrcom]

Re: Gigabyte GA586VX Socket 7 - Bad caps

The only Socket7 processors which don't rely on motherboard chipset to cache RAM are the AMD K6-3 series and the AMD K6-2+ (non-Plus K6-2 excluded)

With one of those processors, the motherboard's on-board cache acts as 3rd level cache (similar concept to Intel's P4 Extended Edition processors)

I'm still using an Asus P/I-P55T2P4 running a K6-3/450 underclocked to 400, with 6 multiplier and 66mhz FSB

That's a 2.2 volt processor on a board whose manual says minimum is 2.5v - my thanks to the German enthusiasts who published hidden jumper settings

That board uses the HX chipset, which is Intel's only Socket7 chipset able to cache more than 64mb

My board is fitted with the optional TAG SRAM required to cache 512mb instead of default 64mb

IMO the HX, not the BX, was Intel's first "legendary" chipset

[gdement]

Re: Gigabyte GA586VX Socket 7 - Bad caps

That's a 2.2 volt processor on a board whose manual says minimum is 2.5v - my thanks to the German enthusiasts who published hidden jumper settings

A good way to find undocumented settings like that is if you can find the datasheet for the VRM chip. Same also for clock generators. Typically they have a chart of the settings, which aren't hard to match up with jumpers on the board.


I once set up a 430TX with > 64MB RAM, which has the same cache limit. It was worth it, 64MB RAM wasn't enough, cached or not.
I've seen a couple 430HX boards but none which are properly equipped to use more memory.

[jpdoe]

Re: Gigabyte GA586VX Socket 7 - Bad caps

@pfrcom
Thanks for the info. I wasn't aware of the k6-2+ existance.

@gdement
I guess you're right about 64MB not being enough. I think on these machines I'll give NT4 a shot. I wanna find out if the SiS IDE NT4 drivers use MWDMA2. That alone would make a hell of a difference. But this experiment will have to wait.

[Per Hansson]

Re: Gigabyte GA586VX Socket 7 - Bad caps

Yes, DMA should work in NT4
And yea, those old systems really where crap
It took up to Super Socket 7 with AMD K6-2+ 550Mhz to make me somewhat happy, but in all honesty I still felt it was kinda slow

When the original Athlon got released though, oh my, stuff took off back then

[Scenic]

Re: Gigabyte GA586VX Socket 7 - Bad caps

i recently recapped my Tekram Super Socket 7 board (out of my retro games PC).
had those D.S. Caps everywhere... some of them looked like they were slightly bulging..
works fine again allthough i recapped it with GP Rubycons (Rubycon YK 16V 1000uF).. oh well.. anything would be better than 10+ years old crap brand caps i guess..

[linuxguru]

Re: Gigabyte GA586VX Socket 7 - Bad caps

Most of the AMD K6-II/II+/III/III+ CPUs will not run in garden-variety HX boards. Apart from the BIOS support, they would also need 2.2v Vcore, 100 MHz FSB, appropriate high multipliers, etc. The best bet for a HX-board is a Pentium-233/MMX, and you need SDRAM, not EDO, for it to run halfway decently.

Apart from the speed issue, they're fine for infrastructure stuff, including Linux Routers, etc.

[gdement]

Re: Gigabyte GA586VX Socket 7 - Bad caps

The newer K6 cores have a 6x multiplier remapped under 2x, so you can get 400MHz on a 66FSB board.
But voltage can also be a problem.

I had a really good experience with my 430VX board. That was an M-Tech R533 which I had with a Cyrix 6x86. Everything worked correctly, and it was fast at the DOS games I was playing. The board before it (PCChips 486) and after it (Tyan with VIA MVP3) both had annoying problems.

[hardwareguy]

Re: Gigabyte GA586VX Socket 7 - Bad caps

The HX was amazing....went well even with a K6.

Pop in a couple 32M EDO 50ns SIMMS and a 512KB COAST module and you had a seriously fast system.

If they only made an HX SDRAM support with the TX southbridge.....now that would have been amazing. (or just let the TX cache more than 64MB)

The ALI Aladdin IV was the only real upgrade path....the TX was a joke with the 64MB cache limit.

[pfrcom]

Re: Gigabyte GA586VX Socket 7 - Bad caps

The best bet for a HX-board is a Pentium-233/MMX, and you need SDRAM, not EDO, for it to run halfway decently

I don't think HX supported SDRAM, only EDO or Fast Page DRAM

I've got 2 of 64mb and 2 of 32mb EDO in mine - seriously expensive at the time

Still does the job though, running Win2000 on a couple of wide SCSI disks

[pfrcom]

Re: Gigabyte GA586VX Socket 7 - Bad caps

Most of the AMD K6-II/II+/III/III+ CPUs will not run in garden-variety HX boards. Apart from the BIOS support

Patched BIOSes here http://web.inter.nl.net/hcc/J.Steunebrink/k6plus.htm


I'm using Jan Steunebrink's patched BIOS for the Asus P255T2P4