Kingston Technology 2x 512MB DDR1 suggested timings

Guys, these lame Kingston Technology two 512MB sticks have bad even the SPD values: 2.5-3-3-8 (at least at 2.5V), but I wonder, if someone can suggest how to push them beyond 229MHz.

Because they did not make the 229MHz even stable. They seems to work:

...but then fail bad

I did not exactly want to reduce the O/C, so I was tweaking as much, as I can and get them at these timings:

To run semi-stable. Good for benches, bad for usage. What shocked me, that for example even 2.90V to them did not improve stability a bit (in fact, things get worser...!!!). So auto (therefore around the 2.5V) seems to work best.

Then I tried relaxing the TRRD, TRWT from 2 to 3, but again things get - worser! WTF!

Then I tried the Async Latency from 7 to 8nS and it get... worser. Unsurprisingly things get worser at 6nS too... But the 8nS was surprising to me. Someone can enlighten me there...?

As the mobo, it is a Asus A8R32-MVP Deluxe (Radeon Xpress 3200) and the Opty can take over 2800MHz at 1.425Vcore, tested, so no problemo there. But the rams do suxx, so...

There are the important bios settigs:

(on the sidenote, when I saw the SPD, then I realized that the TRAS values for 200MHz was 8 and I used 7, as recommended for AMD64 memory controler, so might there I should start...? SPD values for 166MHz are 2-3-3-7, but at 229MHz I'm far from 166MHz...)

pic of chips? I think these are just crap ones (tccc or smth like that) on jedec pcb (which also plays role).

Regarding voltage as far as i remember - only hynix d5/d43 and winbonds (and probably mosel) liked high voltage. Microns liked something like 3v, SECs liked something like 2.7-2.8.

Nope, did not have pic of chips... There are heatspreader:

Funny fact: relaxing the TRAS to 8 did not help as well. Oh, well...

So as next I trying relaxing the TRC from 7 to 9 and TRFC from 11 to 12, but it still looks bad as WOT suddently quit on me, so...

Ideas need.

And once again this weird behaviour - relaxed timings TRC from 7 to 9 and TRFC from 11 to 12 - give error just sooner...!

WTF if going on there

Don't look at time to first error. Use memtest and look at total number of errors after 100% test. Again, some chips are just not designed to work at high frequency.

Please check if your sticks are single sided? If YES then you most probably have infineon chips (ce ones?) - they don't like high voltage, don't go over 215-230 Mhz even with relaxed timings but can do nice CL2 3-2-7 @ app.215Mhz. then you need to lower your RAM clocks but with better timings.

Nope, both sticks are double sided. IMHO the time to first error is important, since it seems to be more stable with tighter timings (it can take the Prime 95 memory test longer and fail only when getting into the bigger FFT...) ...

And given the settings, relaxing the memory frequency did not seems to be a option, as 229MHz fit very nicely with the 250/183 divider and using 166MHz divider will kill the memory speed way too much.

The question is - what timings to relax, when these ALMOST work stable. Not what frequency to change...

Tried TRFC relaxing to 13 from 11 and it seems to work more stable:

But it still failed. So I tried TRFC 14 and TRC 8 (as compared to well working TRFC 11 and TRC 7) and it crashed even in YouTube video... so, very bad.

More relaxed timings = worser stability. What it this?!