UCBench Globals?

[I.M.O.G.]

Interested in finding out how global points are decided to be added.

I have personal motivations because I have a good global UCBench score, but that said, once global points were added my score would likely be whipped by many others as the competition level increases.

[Ananerbe]

I agree!

[Massman]

When the community asks for it, we can vote

[GENiEBEN]

Originally Posted by I.M.O.G.

Interested in finding out how global points are decided to be added.

I have personal motivations because I have a good global UCBench score, but that said, once global points were added my score would likely be whipped by many others as the competition level increases.

So you still want globals? Been sittin on this since Feb, still not fixed (author knows).

[I.M.O.G.]

Originally Posted by GENiEBEN

So you still want globals? Been sittin on this since Feb, still not fixed (author knows).

Yes.

With the screenshot and online submission requirement, it seems relatively easy to police. In order for that submission to be mistaken for legit, one of two things would need to happen: hwbot staff would need to go full specialbunny, or you'd need to do really good work on doctoring the screenshot and make a submission that is properly scaled for the frequency you are running at.

UCBench scores scale very directly with cpu freq, so scores pretty much have to be in tune with the frequency, or hocus pocus is going on.

[Christian Ney]

Originally Posted by I.M.O.G.

hwbot staff would need to go full specialbunny

The sad part is that yes we are

[GENiEBEN]

Originally Posted by I.M.O.G.

Yes.

With the screenshot and online submission requirement, it seems relatively easy to police. In order for that submission to be mistaken for legit, one of two things would need to happen: hwbot staff would need to go full specialbunny, or you'd need to do really good work on doctoring the screenshot and make a submission that is properly scaled for the frequency you are running at.

UCBench scores scale very directly with cpu freq, so scores pretty much have to be in tune with the frequency, or hocus pocus is going on.

While you make a point, I will run this bench at the same frequency and still outscore you, no cheats involved, so the frequency scaling does not prove much Taking #1 with .1 lead should be easy. As for the cmd window, no need to photoshop that either
I do want globals for this but it's just not ready.

[I.M.O.G.]

Less talky, more clocky. If you can beat my score then do it... I'd congratulate you, while I go post something better. That's the fun part.

When these scores are worth globals, it will be worth fine tuning the score... LN2 benching isn't like air benching - I don't sit around burning off LN2 to tweak for no benefit in points. I know the same legal UCBench efficiency tweaks you do... When they are worth applying, I'll apply them.

The frequency scaling point does prove something - at a given frequency, you can look at valid submissions and tweaks and determine what is in a valid range and what is out to lunch. Same with sp, wprime, and pifast... This isn't like pcm05.

[knopflerbruce]

I dunno... there is just one number that can be used to detect cheats/bugs. I wish there were some more details available. maybe some sort of online verification will do the trick?

I wouldn't die if this benchmark got globals already, but it won't be easy to moderate properly.

[I.M.O.G.]

Could we get a poll on this to see what people think?