TaPaKaH

Well, I did. Didn't mean it as an offense to DJ, it's common that big multithread scores are done with ES stuff so report was "automatic".

240 is the Auto setting on Z97X-SOC, I guess other subs are pretty horrid as well

You can run 1.9Vgpu on the card with a chiller, it should load around 30c

It seems that scores submitted for competitions like Pro cup do not appear in hardware rankings. Example score: http://hwbot.org/submission/2569232 Example ranking: http://hwbot.org/benchmark/superpi_-_32m/rankings?hardwareTypeId=processor_3641

twcl8, I presume? How high can you suicide?

you sure?

let's see your scores first

wow, better than my sandbag http://valid.canardpc.com/uzhxal

Selling a pair of E2140 CPUs that are not good enough by my standards but are still better than most chips out there. #1: Q738A 460x6 boots, 470x6 does not. 400x8 boots at 1.225V, 450x8 boots at 1.375V. ...SOLD on 20Jun #2: Q744A 450x6 boots, 460x6 does not. 400x8 boots at 1.300V, 450x8 boots at 1.475V. Moved to new thread.

there's a black market for passports, you know?

I thought you're Swiss?

Is this your old chip?

You're completely missing the point of the argument. What I am trying to say is that noone on hwbot can come even close to your efficiency, this is why I find it suspicious and ask for a proof. Please stop dragging the argument towards my Q9550 result - unlike your scores on E8400, it is perfectly in line with what is known to be possible.

I see that you run slightly different RAM settings on your air run. What happens when you run single channel in 1:1 with PL15 like you did on LN2? Speaking of my 37139 pp on Q9550, as promised, here is a run with 37073-37200 (depending on CPU speed wprime was actually ran at), here is another run with 37003, here is 37091 and here is 37196.

http://item.ebay.com/261476853344 Someone (not me) was lucky enough to score thirteen 1250s at once not too long ago.

Is that one also unlocked?

An amazing result. Congrats on the overall percentage overclock title!

Why should I be kidding? Don't you think it's natural to ask questions whenever someone claims a 1.3% (71852 vs 72808) improvement in efficiency over best cumulative effort of thousands of people over six years, especially in a very popular 2D benchmark? I am not asking you to reveal what this "secret" actually is, I just want to see confirmation that these runs are legit. So far there is none. I think you will agree that LN2 does not make things faster on its own. So whenever you decrease the frequency from ~6GHz to ~4GHz while maintaining the same relation betwen CPU/FSB/RAM clocks and RAM timings, efficiency will remain the same. That is, if you were to run 445x9 with exact same RAM ratio, timings and performance level, you should get 12.078*661/445=17.941s (or a close time) if things are to make sense. Also, let's keep personal offense out of this discussion.

I don't see anything on the video. You make wPrime runs with unknown CPU at unknown frequency. Again, you don't need to run LN2 - 17.963s or faster at 4GHz on air is sufficient.

Your comparison with my Pi runs is not really valid. I'm quite sure that for any "example" you can come up with, I can find at least two or three other guys with similar efficiency at lower clocks. And this is exactly my point - your efficiency is far ahead of everyone else's. I'm in no way trying to say that this is impossible, but it's very unexpected to see from someone whose previous success came primarily from raw clockspeeds. You don't need to put things on LN2 to demonstrate efficiency - in wPrime it stays pretty much the same as the benchmark doesn't react to lower RAM dividers and/or higher timings that change from air to LN2. If you can show that a time of 17.963s or lower is possible at 4000MHz (17.963 * 4000 = 71852 performance product) then you're going to get a lot less complaints from me and the other guys

Could you please explain how you got 12.078s in 32M at 5949MHz? That's a performance product of 71852 and I haven't seen anyone else on hwbot do less than 72808 on a Wolfdale 6M.

All I see so far is that all teams uniformly lose points hour after hour. Shouldn't the recalculation assign "greater or equal" number of points to every score than it was previously?

not a bad efficiency, more cpu freq and you're there

sold

pending