I.nfraR.ed

Hate that "feature" of gigabyte boards. I always try and save important profiles on external drive (usb stick), otherwise there's a high chance the dual-bios reflashes main chip from the backup one, wiping all saved profiles.

I am with Alex here. Trying to look at it as neutral as possible. To me, your mistake is first (as others mentioned) not using e.g. paypal's "pay for goods or services", however I can really understand the reasoning behind this. I've bought several times from respectable hwbot members and most of the times (if not always) I've used the "friends" option. The next mistake is to directly attempt to test it with an "unofficially supported" method or even attempting any test when you saw the scratch (or crack). Based on other comments in the thread, I could definitely believe the possibility this CPU was damaged prior shipping with e.g. excessive force from a pot mounting, however, realistically speaking, noone could prove that. As for the shipping claims and other ways of recovering, I can't comment, since I don't have any experience with that. Good luck recovering your money though!

Great one, mem clock does really pay off! Can't be done on AMD, since pre-Thuban CPUs are limited to around 1000MHz memory clock.

That mem clock! Awesome score.

I only have one s.775 board (P5E3 Premium) and never had REX myself, but my board worked 1000 6-7-5-20 first boot 2x2GB Hypers. Excuse my lame question, but I've only tested it with E6400 on water for 4GHz 32M low clock. I see you have modded it, so which voltmods are "mandatory" for this board if I want to bench e.g. E8500 on LN2? I see vPLL and vCPU on the picture.

I've never owned Dom GT 2000C7, but I've seen there are better ones, especially if you have MNH-based kit still running. I consider my GTX2 better than the ones listed, but I haven't tested on the same platforms - might be better, might be worse. If I manage to get these (sent PM), I will compare directly to mine. I've bought some ST 2200C8 from websmile and other sources in the past and they all need significantly higher voltage for same tight timings on AM3 platform, so it would be interesting for me to compare Sam's set against mine. I only have 3xGTX2 sticks - 2 original ones I bought new when released and one spare from Splave (perhaps a reject, I don't know).

I will get the GTX2 if available, please.

Depends. Usually you can bin for FSB on air/water and in my limited experience they scale pretty linear on LN2, so I can predict what would be the max frequency possible. There are other factors though - voltage scaling, NB coldbug (yes there are such CPU's), IMC coldbug.

Yeah, 2 out of my 4 4850's have the 03 bug, unfortunately my best one (MSI HD4850) has it too. Same applies to 4870 and 4890, but don't remember which has it and which not. Good luck with sale!

There's not much difference between "ambient" or current situation with temp set at 10°C on chiller. It still can boot those clocks, but the lower temperature helps stabilize 3.6GHz uncore. What would help though is if I drop the "real" temperature to e.g. 2-3°C, which might actually help up the CPU-NB one notch. I can control it down to 3°C (with a step of 1°C) which is the minimum of the chiller. From that point going down is "full-on" mode with bypassing the temp sensor of the chiller. So, by design, I still would fit in the competition restrictions, unless I override it manually. That's why I asked if it is ok to use HWInfo in case it shows something else (like on Bones' screenshot). He's using a FX, though. I completely understand the sensors are not reliable, plus there is difference in motherboards implementations as well.

Btw, what is the temperature sensor we're not allowed to drop under 0? Because it's physically impossible that my CPU is at 1-2°C when the chiller is set to e.g. 10°C. Real CPU temperature is like 3-4°C higher, which in my case is temp sensor #2 and #3: Is it ok if I use HWInfo which e.g. doesn't show that "Core #1 - #2" temperature, but something else (have to test first what it shows)?

Thanks. Probably won't run it again, although I can match that CPU-NB with the spare 1090T. Will see if I have time next weekend, but certainly no cold on the memory.

Just for reference, my golden Sempron does 4.7GHz 1.35V wprime32 IIRC. But max HTRef is the more important thing.

I use AMD Overdrive and usually up 5 MHz at a time.

Old, but gold :)

Sorry if anyone interpreted my post the wrong way, it wasn't anything personal. It was totally unrelated to any score posted or any other post in the thread. @Bullant Your efficiency is pretty good. I have improved my score at same settings on new OS, still have to drop a second, but my good 1090T died. Will try the other one later.

On my 970A-UD3 I can't POST with Deneb with Hypers > 850 or so and have to work my way up in windows. No matter the settings, it always goes to fail-safe if I set memory clock at anything higher than 850. It seems Sam encountered the same. It's not an easy board to bench. Recently I've bought 990FXA-UD3 rev 1.2 and it is very similar in efficiency, but setting mem frequency works as expected and can boot bench settings without the need of adjusting refclock within windows, which in fact doesn't always work on 970A - sometimes it hard-locks before reaching target frequency. I've noticed 970A worked with 1090T and Hypers at 947C6 straight from bios, however the CPU died. Don't know the reason, but I suspect CPU-NB voltage. I was running 3550-3600 1.55V for extended period of time on chiller and positive temps. Perhaps the board overshoot it, or it just died due to the stress conducted from running at that voltage on other boards.

Probably a bad board, especially if no high voltages used. Can't be 100% sure, but my Hypers are still ok after benching on AM3 numerous times. I'm the one who "resurrected" one of my GTX2 sticks 2 or 3 times, but as far as I understood his problem is a little different. As for the SPD flashing, I've found a second board that boots with one good and one dead stick - Crosshair V Formula. The good stick doesn't really need to be Hyper-based, a random one works too, as long as you have the needed SPD file saved on the drive.

This whole monitoring doesn't do much, except showing min/max temperature and min/max core clock at the time of taking the screenshot, basically displaying values in a finite period of time, which doesn't necessary match the time frame when the bench was running. So in other words, it doesn't really prevent cheating, unless it's obvious from the screenshot. I guess everyone is playing fair, but that just came up in my mind.

If you have enabled the Internal PLL Overvoltage, than it's game over. For easier testing leave just 2 cores active, so you have better chance to rule out voltage/temp. I don't remember if there were CPUs which didn't need PLL Overvoltage enable untill x53, best I found myself were couple of x55 CPUs.

Nice, eliminated the competition, haha. Try it, it was long time ago and I want to try it myself again.

XP was slow for me, because of no ACPI.

Have you slipstreamed the needed controller drivers? For my external pci-e controllers I have to integrate the drivers in order to get the SSD detected.

Yes, my CPUs (1090T and 960T) CPU-NB tops at about 3600 (or a little higher) and chiller doesn't help much compared to normal water. So I'm already on the limit there. The other option would be frozen RAM. Think I've done 2000+ 6-6-6-18 with just cold air before, might try. Or maybe see how high could go with 6-7-6.

BTW, there was a similar competition long time ago: https://hwbot.org/competition/hwb_ddr3_hcc/stage/142_thuban_-_superpi_32m Already beat the top score and I think 14.2x is possible.