Massman

Normal vs tweaked, if possible

efficiency vs memory frequency (categorised per Cas latency) efficiency vs memory frequency (categorised in efficiency groups) -Sub60k = under 60k efficiency -Tweak/bug = results with the known tweak/bug = 60,2k - 62k -Tweaked = results with good efficiency, but no tweak/bug = 62k - 64k -Minimal tweak = normal scores without/little tweaking = 64k - 67k -Slow = efficiency sucks Both graphs show no significant difference in efficiency between DDR2 and DDR3 memory. Now, I did find one other score that had an efficiency below 60k: http://hwbot.org/result.do?resultId=656720 19.515 - 3068 - 575 CL4 - 59872 (Masell) 16.203 - 3693 - 821 CL7 - 59837 (Stealth - http://hwbot.org/result.do?resultId=804884) So, like I said before, around 60k is not impossible, although one needs the tweak/bug and additional tweaking. 58,9k is almost 1k lower than that ... which still leaves me in the unknown .

Only the basic tweaks (performance mode and system cache - ok)

623 CL5 is quite fast as well, don't underestimate it . 1M is different from 32M, by the way, it's a lot less memory oc dependant. However, you do make a valid point as it may be possible that the memory has more influence than I can currently give it credit for. After all, the Celeron isn't that powerful. I'll check more score and see what I find out .

Would it be possible to run a few 1M's using a normal, clean XP? Just for comparison reason No it's not: http://hwbot.org/result.do?resultId=805789 I've been checking out other DDR3 runs as well and the difference between efficiencies was around 0.5k-0.6k, not way over 1k like in this case. Like I said, most of the scores are not really out of line (outliers, yes, but not by far), there's just one that's so much faster than anything I've seen before that I want to know why exactly it's so much faster . 18.219 - 3322 - 623 CL5 - X48 (60523) 17.500 - 3368 - 842 CL7 - X48 (58940) An efficiency of 60200 is doable with a (very specific tweak/bug). If the efficiency would've been the same, his second run would've been .47s slower. Can one shave off half a second only by switching to DDR3, or is there something else playing a role (1M boost bios, another oddity or so)?

I never said you should stop benching . I got the information I wanted regarding the first 'bug', but that does not explain why your score is even more efficient (you're 0.2-0.25s too fast) Are the scores in post #25 very tweaked or just a clean XP boot and run?

Not for every session you do you have to make pictures (although it's always nice to see people benching), but in situations where no detailed information is available having pictures is mostly the best to avoid confusion . If you think that there might be a problem, you can always post pictures (with or without paper). Only a matter of avoiding confusion

Thanks for the update guys, I really appreciate this! I've unblocked the score

Yes, but as far as I know, the 5000+ BE = 5000+ Brisbane (65nm), whereas the 'normal' 5000+ = 5000+ Windsor (90nm). (same goes for 5400+)

I'm actually trying to help you out here, Stealth, so I would appreciate it if you co-operate with me

If he had the same efficiency as you have, which for me is still a to be proven valid object. Do you have an explanation why you have been able to hit that low efficiency? (I'm not asking for an explanation, I ask if you have one )

I'm excluding the possibility of the Rampage Extreme just being faster: http://hwbot.org/result.do?resultId=805789, there's a huge difference between both runs. Don't know what bios he was on, though

Hm, and now?

http://www.hwbot.org/result.do?resultId=795443 It's blocked

Intel .

I'm asking you to help me find out if the efficiency anomality has something to do with the bios, just like with the 45nm wolfdales, which are given a 0.1-0.125s boost by using older biosses. If it's indeed the bios which is giving you the anomality, there's no problem putting back your score. The bios updates are just for the 1M comparison: after you've got the results with the 0121 and 1003 bios, you can use 0301 again ;-). There's no other argument than "pls unblock my score" to extend my (subjective) scale. You are the only person who I've seen hitting an efficiency lower than 60k. It may be nothing, I'm fully aware of that, but I'd rather find out what exactly is causing you to have a greater efficiency . At the moment, the bios bug is my only hypothesis.

The 60.3-60.5k actually defines the scale: these are the outer values that I use to determine what data fits and what data doesn't fit the profile of plausible Celeron 4x0 efficiency. Note that within the group of plausible efficiencies, these are indeed the outliers. You could do that OR you could help me find out what's causing the anomality . Would it be possible to test the same frequency using an even older (0121) and the newest bios (1003)?

Because the 940 Extreme has the unfair advantage of having an unlocked mulitplier. It would be unfair to those who have a retail 940 when both locked an unlocked are placed in the same category

Not according to what I've seen in the past. Even with DDR3, the hardware bug leads to an efficiency of 60.5k, not 58.9k. And before this turns into a believer vs non-believer battle: I use the efficiency calculation as a method, not as proof.

No one calls you a cheater, no one is flawless. By the rule: "Seems off scale". If a score doesn't follow the general performance scaling, we have a reason to block it, regardless of it's validity. When we block it and turns out we're wrong, we unblock it . 18.937 - 3304 - 500 CL4 - 62567 18.562 - 3316 - 500 CL4 - 61551 18.219 - 3322 - 620 CL5 - 60523 17.500 - 3368 - 842 CL7 - 58940 Your first score is right on spot when comparing to other results; the efficiency is 62.5k, perfect on line with the other scores. Now, due to a hardware bug it's indeed possible to shave off time to an efficiency around 60.5k, which indeed fits your third best score. The last one is just way beyond the best efficiency I've ever seen on a Cel420/430 (fastest is around 60.3k), so I had my doubts, hence why I blocked it.

This is not a spam forum

MFT disabled: 55MB/s MFT enabled: 1350MB/s

SSD is the future, there's no doubt about that. We're discussing the usage of MFT here: an application that partially uses the system memory to increase HDD performance.

We're still having the discussion in the crew forums, nothing has been decided yet.

No a bug in the bot