knopflerbruce

It could be that there is no limit, it just doesn't test 40 threads (next step after 32 is 64?). If this is indeed true, UCBench needs a patch:)

THis is OT, but i'll ask anyway - why not buy the FX4100 for example+ I bet it's the cheapest chip - but most guys I see benching BD use the 8 core ones. Any particular reason why?

To beat a TWKR it must do above 9 GHz:p

And then Pentium 4 mobile, and then... maybe a few more models where there is actually some sort of competition

The ranking would be somewhat interesting, something unusual to aim for - even if there are no boints (apart from HW) to collect.

No idea, I have a friend in Hannover, so I'm visiting him - just happened to be so lucky that his holidays matched the CeBit dates:) I can ask Stoolie if you like.

Just ordered plane tickets to Hannover I guess I'll be there a day or two. At least 10th if I can get a cheap student ticket for that day, said it was available at 19 euro on the site.

OK, didn't see that difference.

Run wprime with 8 threads and then with 3. You'll see if there is a huge difference or not.

Then it should be added as Geforce 210 GDDR3:)

I don't think revision is important. DDR type and bus width must match the category, but not the amount of video memory.

If you've got thoughts about stages (benchmarks and HW limitations for example), feel free to post them here:) I guess the input here will be important when Massman makes the final decisions. Some stuff i posted in a thread not accessible to everyone:p

56x is way above average, but not golden. I think about 1/10 of those can do 5600.

Nice find! Thanks for sharing!

Am I the only one who doesn't listen to music while benching?

Retail availability = in stock in at least one shop. If it's available in ONE shop, even in very limited quantities, it's fine. Just to define that term. The NDA rule works pretty well, if it's replaced it must be with something that improves the current situation. And more fair doesn't necessarily mean better in this case, you can have a very fair rule that's very difficult to moderate. NDA is a known date, so very easy for us to tell if it's lifted or not. Retail availability is more difficult.

I wouldn't recommend doing experiments with new benchmarks if this is a somewhat serious competition. We don't want weird bugs in those.

I agree that it shouldn't be top priority atm, but it can't be THAT much work. We don't have to recalculate exisiting subs, as they're all below 220mb/s, so those scores will remain unchanged. What you need is a box where people enter the XP startup score, then some stuff like if x > 220 [correction formula], else nada - and that's pretty much it I think. It's probably a bit more complicated "IRL", but compared to other things related to coding i can't imagine this is that heavy:p

THe problem is that you run an SSD without much tweaks, and you can get a very nice score removed because "oh, Xp startup is a tad too fast", which itself is silly. It might have been OK in ancient times, but today it's a joke. There's no OC related skill in crippling performance.

The same thign could be said when the 220 cap was invented - using this method is not worse in any way than having a cap present without haveing a correction formula. Plus, it's one of MANY subtests, it's not like 3Dmark05 where you have just 3...

At least this thread shows who is photoshopping and who is not:D

I got one:p

Exactly. No-one is EXPECTED to share, but alot of guys still do. Still, people must be allowed to have some secrets - after all, figuring out the mods is a skill, too. If EVERYTHING is shared, all tweaks and all mods, then in the end it's about who has the fastest chip, and that would be kinda boring if you ask me. If people actually share two mods and keep one secret, let's be happy about the two we got for free, instead of bitching about that one that's hidden. And try to figure out the last one on our own. That's the way we should really think. As long as the white papers are available anyone can figure these things out, only knowledge (and unwillingness to risk your own card for mod testing, perhaps) is the limit of what you can do.

I agree that BIOSes should be shared - unless the bencher did the coding himself, which is highly unlikely.

How does systeminfo help? Someone needs to try 7970:D