TerraRaptor

@Leeghoofd it was 2500+. Usual s462 issue.

Great one, another 8GHz club member!

Next time Albrecht should kill hwbot in a more sophisticated way?)

Epowered. I replaced crap infineon af-14 with h5rs5223cfr and did timing mod too to make it work.

Fastest celD on s478, great job!

Strong!

Oh bro, very impressive!

Ace;)

Great to see you again on front page Aristidis!

Has noticed myself that 3dmark window shows software T&L while other scores have hardware T&L in 3dmark settings. This is not an alteration of 3dmark settings by me, just a setting in driver (it allows selection of vertex processing as a set of 3 choices - application controlled/default/software processing). Having enough cpu power I used software processing. Obviously not a cheat as it is driver level tweak that doesn't affect render.

Strong

Maybe you bench yourself to realise that primaries (the only part that is possible to fill in during submitting) are nowhere essential if the rest is set wrong? Who cares of Cas latency of 6 if PL is set to 8? Who cares of trcd of 7, if secondaries are 12-9-9-10-14-10-14? Prove me I'm wrong, lol. No one needs these cl-trcd-trp-tras really. And if you need it, open screenshot - me personally think of this information like misleading and useless.

what video ram chips you have, hynix?

Like:)

https://youtu.be/UlAJiXIW6PE - as I got several personal messages, here is an explanation how I was changing timings with gt218. Important is to open a bios of previous generation as nibitor fails to parse newer tesla 2.0 bios correctly and denies to activate "test timings" box

I would say 104.0 is always doable with air. Higher is a matter of colder temps very often.

Good one. I was hoping for someone to break 1600 points, you proved it is possible.

Biggest performance impact is (I suppose) tREF, the last timing like 01CF etc - setting it FFF or FFFF gives the most.

Well done!

Make 3vsb mod feeding clockgen Define optimum nbgtl Several mchbar tweaks

Thanks bro!

Well done bro!

Like mine standard 1.8v. Main problem for me was skews that time. I should retest it soon with all the new ideas I have.

Think problem is cpu side. Highest I did with c2q and REX is 534mhz.

Just to compare. I also think it's not a good idea to skip attempts to teach REX to OC quads:)