TerraRaptor

That is the same pcb with some bells&whistles missing - agp instead agp-pro, no cnr, no integrated vga. Never tried it though. Should be identical to CUSL2 in terms of overclocking.

That could be a dram vrm bug of a partucular board to kill it - not a platform issue. I had s478 board recently with pll mod that was accidentally killing cpus although pll voltage was in line with no-mod boards.

Dead sure this is the same card as here - https://hwbot.org/submission/4083883_ Most likely, cheat involves nvstrap driver. Same bios, same score, same clocks

I had several mobos on i815 with integrated video and they were no go for overclocking. However I was quite noobie that days so don't know what to expect with proper approach. Video is last thing one should consider with this socket anyway.

External clock may have less jitter/phase noise (i.e. high end audiophile gear sometimes use external clocks for their DAC of a femtosecond grade). That is my guess. At least, the different quality of pll ic may be the reason why TUSL/CUSL boards have certain difference in max clock between samples - the reason may be not chipset but pll. GraduS should know more - he was really deep into digging JP forums on s370.

Those were turboPLL/JordanPLL to allow overclocking on the boards with no-overclock stock pll ic. I believe you cannot run pci/ide in asynchronous mode if chipset doesn't support it which is the case for i815

Easiest way to check what limits you is to use different "straps" and low multi coppermine d0. First check fsb limit of cpu with 133/100 and write down clocks for pci (if waza locks system at given fsb - then it is ide limit). Then go 133/133 to find memory limit. Then check 66/100 to see if you can break the pci clocks of 133/100 while staying with dram below limit of 133/133. PS. IDE is definitely a problem. Japanese were using IBM fireballs if i'm not mistaken in early 00's which were known to stand 50-55 mhz ide clocks.

Why have temperature restrictions if we only have 4.5GHz limit? More to that, I guess monitoring software like HWINFO should (must) have a possibility to offset the values so it will be easy to trick it. Another trick is - run the test with eg. SS, run the test, open tabs, switch SS off and let the temperature rise. I think temperature limit should be removed and the frequency limit should be adjusted to meet "average air overclock" for the given DIV#X platforms. Let's say, if the stage assumes the use of Pentium III Coppermine, let the LCC frequency be 1200MHz or 3600MHz for Core 65nm.

I have also bought one yesterday. It is 620 fsb stable 1:1 but won't go higher than 1600 dram no matter what. Will sell it asap.

Yes, boots fine with flash drives prepared in Winsetup2USB to install w2k/wxp/2k3 if boot device set to USB-HDD.

@sammurai will post pictures as soon as I get to the board So far, I had at least 2.15v measured. Probably just my board. Anyway, ovp-mod should not be hard to implement

I remember promo of Giga lga1366 ud9 (i guess) - with demonstration vrm is capable of delivering 1.5kW jvhW-5BzoE0 So not the first time the board has such a monster vrm

Is this board really so weak compared to Gigabyte and Biostar in terms of fsb overclocking or is it just shadowed by REX being almost as expensive?

I guess LOD tweak was available for me in 3DMark Vantage and 3D11 several years ago (HD7970 era). That was a combination of utility to change LOD + certain level of catalyst AI - buggy one but working sometimes. Do not know what is the situation now with modern hardware and Crimson drivers.

@DeDaL @gradus