I.nfraR.ed

Ok, I have VID mod, FB mod and 2 for mem - VDDQ and VDDC, but I have no time right now to understand what I've really done to that card. Seems I haven't done OCP. OVP is eliminated with VID mod. http://imgur.com/a/HYL4s/all I can add some guidelines/explanation later, if you want. Overall the mod(s) are similar for all Asus cards (DK/TOP) based on this controller/PCB. PS: BTW, you need to also cut 2 traces on the PCB.

I have 9800GTX+ DK and can check when back home. Have it full-modded, if I remember correctly. For GPU I think it was VID mod + FB mod for fine tune. For MEM it should be just FB mod. And maybe OCP...but it was long time ago, so I don't remember http://hwbot.org/submission/2634311_i.nfrar.ed_3dmark2001_se_geforce_9800_gtx%28%29_55893_marks

About -80 for me, but it might be the board. Haven't tested another CPU to tell though... LN2 helps, but voltage does not. Colder it gets, lower voltage tolerance it has. At least my CPU.

Turrican has 800: http://forum.hwbot.org/showthread.php?t=6520, but this sub is slightly higher (~0.9MHz ) I could not get higher FSB on my board before, so I was stuck at 780, but now 800+ should be possible and I plan to try next week. Last time I was able to POST at 840MHz and boot at 810.

Yes, that should be correct, but it would be good if we have official confirmation. Technically you can submit in all stages and only your best sub will be counted in your country total. Unless the engine is broken, if anyone else from your country submits with another card different from your best, then your score with that one should not block his score. Example: 1. you submit: - 560Ti: 44pts - GTX280: 39pts 2. A fellow submits: - GTX280: 28pts Even when your GTX280 sub is better than his, your current best is with 560Ti, so your GTX280 gets invalidated for the comp and his subs count to the total. That's how the engine is supposed to work, if I'm not wrong. If points update due to other users inflating points or pushing your sub to lower place, then it should adjust itself. But...maybe we need to carefully pick manually. Honestly I don't know for sure, so an explanation would be needed. I don't think there had been such case since now and maybe it is not implemented at all...

Vantage does not matter. Stage is all about 3DMark 06. Best submission per person for a country should win as always. If you have other scores (lower HW points) with different VGA in the stage, then they should not be taken into account.

Maybe it is just wrongly named. Try to rename the "DOS" bios to AM2N3VT3.30. PS: Hmm, no..it's not. Phenom II are supported in 3.20 too, so why not flash this one and then upgrade to 3.30 after you get the board working.

You have the names here: http://forum.hwbot.org/showthread.php?t=93537 and some description from Antinomy in the next posts.

For manual screenshots I usually downclock, open cpuz, gpuz etc, make screenshot, then up again to real clocks and make another screenshot (if it doesn't crash).

I had my Gigabyte prepped, but had to remove the eraser, because I needed it for the X4 Phenom. But will try to push it to the max soon. Still have the OS installed and tweaked. Can't find a cheap K6-III+ though. I have one with stripped gold from the pins So many of these CPUs went to "gold-diggers"...

Yes, it is FSB + PCI. With that "divider" it reaches roughly the same RAM frequency, but on lower FSB. I have some Qimondas laying around and can try if it could go higher, but I think chipset/pci/agp is not able to go higher. First thing I tried was to use the modded PSU with higher 3.3V line, but it did not help. With the FSB + PCI option I could only boot at about 122MHz FSB and it stops at about 133 within windows. Which is the same SDRAM freq I could reach with 1:1, ~133*1.33 Of course performance at 1:1 is better on these platforms. At 133MHz FSB PCI/AGP should reset back to normal, but the "extra 33MHz" options is only available up to 122 in bios, then disappears. Also bios has frequency addition (up to +28). Maximum you can set is 155 + 28 = 183MHz.

I/O source is from 5V line (stock board allows up to 3.9V from bios), so I did a simple FB mod. My 3.3V-modded PSU is Fortron 400 PNF and can go higher than 4V, but for this sub I used Seasonic 1000W Platinum.

I tried chiller on NB - did not help at all. Volts helped a little. This is with 4.4V I/O. Maybe it's the mem controller, but there's no lower divider - only 1:1 and FSB + 33MHz. Second one not available at high FSB though. Theoretical maximum for this board is 183MHz. This was the first time I play with KT133(A).

Yes, chipset limit. No difference between 3-3-3 and 2-2-2. Or possibly some of the buses/controllers, since nothing is locked. You can see I've benched maxmem with tight timings at almost the same freq.

This was one of the things that got me out of the cup.

5GHz CPU-NB is about the maximum for 32M with "reasonable" volts and IIRC it does not scale much past that.

I managed to save it on the third attempt. First one froze while opening cpuz, second one restarted while opening cpuz. I have some old LN2 profiles saved, so had to only adjust clocks. Could run 940 6-6-6-18, but there was no suitable combination with cpu and nb clocks, so I had to use yours . And yes, Hypers are tricky with Thuban on cold, while on Deneb you can bench tight without a problem (on most chips). Disabling cores on Thuban could gain you about 200-300MHz on the NB with same volts, but Deneb is faster clock per clock.

Seems that I still remember how to overclock AMD

PSC are better on Thuban. Much more stable. Disabling 4 cores also helps with CPU-NB clocks.

Why do you need LN2 on the chipset?

I use one Bulgarian tool, called WinSetupFromUSB: http://www.winsetupfromusb.com/files/ You can try to force it as HDD with the included tools. I've only used it with flash drives though.

It's no guarantee I can finish it. One of my boards (the good one) is buggy on cold, the other one did not do 140MHz FSB last time I tried. Will mod it and see if it can go higher, but might have to settle at 780MHz again, like the previous time. Although that 780MHz was pretty stable in the long 32M run, so I guess if the board allows me, I can probably complete benchmarks at 810, why not at the maximum of 840... Btw, what is the theoretical maximum of the boards with VIA chipset?

In my testing (at least on the Aladdin board), there was no gain in enabling external cache for pifast. Higher Mem/FSB frequency helped more. I was actually booting at 810MHz, but the board got wet. I will try to improve my score soon. It posted at the maximum of 840, too

Yes, it shows "random" time every time I open a stage (more specifically the time of the previously opened stage). Didn't even manage to bench pifast and aquamark on the k6. I don't have the scores, it was not a sandbag here too.

Great one, congrats!