Massman

Ok, figured out what's going wrong. Instead of 820-ish, I now get 950-ish with same timings/clocks!

BH-5, yes. 3.7-3.8V

MRL doesn't fix it ... I tried that :-/ Boot 234 is possible, one OC Knob higher = freeze

At first sight, it looks like mem-write and latency are not what they should be, what makes me think some IMC issue. Any idea on how to fix it manually?

Now testing a new bios - the wall is even more apparent than before (although I can now boot at 235 1:4 (1880). 234Mhz HTT stable boot to bench, 235MHz boot to windows load.

.5 cpu multiplier = upper MEM divider 10.5 CPU = /11 MEM 9.5 CPU = /10 MEM

Tested this CPU under phase-change already ... had forgotten how cold affects HTT overclockability. AMD coldbug SUCKS! :-D

Are you sure you're running that frequency? I just tested 10,5x280 with 200 divider and it reports 263MHz memory ... MaxxMem score also indicates much lower performance than 280Mhz memory

I've taken the liberty of giving BenchZowner the weekend off to take a rest. It's one thing to complain about "the (HWBOT)" system", but a whole other topic to mess up threads by calling cheat on everything that moves.

It scales upto 3.7V, but after that it just gets more unstable.

This board isn't really good for Maxxmem. There's a wall AND the performance CPC isn't what it should be

I did it in the past because I overclock

You've repeated this multiple times over the last couple of days. Maybe it's time for you to go on a break and come back when you don't need cursing to make a point. It's just a suggestion ...

Just submit a few scores ... too bad I apparently killed ALL of my DDR2 boards.

... and I have no idea who to root for. No Belgian Devils present

Splitting points up based on cooling is a very nice idea in theory, but in practice just not possible. Software temperature read-outs are oh so cheatable and not at all a good verification tool.

Ouh, this looks bad. So basically, you're changing the NB clock in windows (which is impossible since change NBclk needs hard reboot) and all application read that it's set like this, so you get an unreal score. Good catch!

Good idea. I'll leave the URL untouched, though.

Just send a mail to the coder of the application. Hope he knows what's up. This looks to be a problem of unsupported architecture. We had the same issue when the Gulftown appeared.

That latency looks bugged. Can you verify with everest?

I'm starting to think this score might be bugged. The latency is very low in comparison to Chew*'s scores (which is a phenomenon occuring with bugged runs). Can you try V1.82? It may just be that V1.74 doesn't support Regor well enough.

With the GF2 GTS I get 1/2 of your 3GHz score with my 3.8GHz.

Quite frankly ... HWBOT doesn't get ANY hardware from manufacturers. Everything you see from me is either through Madshrimps or directly to me (because I did the work for Madshrimps).

Tested the platform in 3DMark01. It's worthless, don't waste your time.

The mission statement can be found on the 'about hwbot'-page: First of all, HWBOT is designed to fit the needs of the community and not the manufacturers. That's the reason why I agreed on spending my day working for this site ... to make it better for the community. For those who don't realize: I'm not paid to sit here all weekend and discuss in topics like the ES. I do because I want to. Maybe the best way to explain that we are build for the community: more than once we got the proposal of working with paid memberships on HWBOT and every single time we declined. This means that no one has to pay a single dime to be part of HWBOT. Given our tight financial situation, I think that says enough. Now, concerning the sponsorship issue. I think the split up will be on voluntary basis for most of you at the beginning (maybe some criterium like #boints to keep out newcomers), because it's difficult to draw lines. It feels like some of you pretend that PR benching is the very easy task, whereas it still is hard work. It's not just click 'n' bench ... Also, please let's not believe in the fairy tale where the amateur league will be completely free from all free hardware. In the end, I can lend a few nice memory sticks to my friend and let him bench with it to achieve nice scores ... and there's nothing wrong with that. Also (in my situation), I have the support from a performence-minded PC shop in Belgium, which allows me to bench hardware I don't have to pay (note: I don't GET the hardware, but I can BENCH it once in a while), which also gives me the upper hand. Sponsorships come in all forms and is simply uncontrolable. The biggest problem is quantity of hardware ... getting one untested chip, even for free, is the same hard work as buying one untested chip. If you get a free chip from you girlfriend as present, it doesn't mean you're sponsored. From what I've seen, many refer to the easy RMA process for the sponsored guys (read: just get a new card). Well, I think that people who pay for their hardware have a much easier RMA process than any of the sponsored guys! Really, manufacturers HATE it when you kill hardware, even if you just broken a world record with it ... you need to come up with a very good explanation if you want to get a new card to continue benching. For the regular guys, it's just a matter of cleaning up and sending it back to the shop. Very true. At the moment, they try to get publicity through the normal rankings. The aim is to still give them a way to reach the community with their overclocking results (after all: it's still nice to see all the top results gathered in one place), but not have them affect the points and ranking of the normal folks (which actually include 99% of the community). An upper league could solve this issue.