The official UCBench 2011 (beta) thread.

[Massman]

So, you guys like the benchmark?

[Christian Ney]

On my side yes, like WPrime core count and frequency no other things

[Chiller]

is not bad this benchmark.

 

think that the results are good in order to what cpu and OC you use.

 

Not much bugs in it or none at all.

[Christian Ney]

and there is full online details and there isn't the full CPU identification but with a good SS

[gnidaol]

I like it, I just wanted that the rules let us choose to run only the test that gives the best score.

example: only selecting to use 8 threads and sse3 with a 2600k, or 3 threads and sse2 with a amd x3... The score is based only on the best run anyway so it could be really good to let the OCer to choose the test he wants to get validated instead of making us run lots of tests that doesn't take any part on the final score.

 

The validation already works that way but the rules seems to tell us to not choose anything else but the stock options...

 

 

 

EDIT:

Other thing I like about this benchmark is that it is the "only cpu" benchmark that makes my system get the highest power consumption among lots of others "only cpu" benchmarks.

Edited July 5, 2011 by gnidaol

[Chiller]

stock options is good.

 

For everybody the same test is the best that can be.

[gnidaol]

stock options is good.

 

For everybody the same test is the best that can be.

 

I agree with you, but the test changes itself depending on you cpu anyway. An intel quad core will peform different tests in comparison with an amd quad core cpu even in stock options.

[Chiller]

for everybody the same test,

 

everybody has another cpu, just like Wprime, Superpi, every cpu will give another score, just like UCbench.

 

Wprime, Superpi pifast are all also the same test for everybody, so will be with UCbench.

[gnidaol]

for everybody the same test,

 

everybody has another cpu, just like Wprime, Superpi, every cpu will give another score, just like UCbench.

 

Wprime, Superpi pifast are all also the same test for everybody, so will be with UCbench.

 

 

 

I think I didn't make myself clear. What I wanted to say is that this test doesn't uses the same calculation method to every cpu...

SuperPI, Wprime, Pifast uses always the same exact iterations to find the result. UCbench doesn't do that. It uses different methods for different cpus by default.

[Chiller]

but the tests are still the same for every cpu.

 

The test is for everybody the same.

 

Only a amd will perform better in another sub-test then a sandy bridge for example, but the test is still the same for everybody.

 

That's the main thing

[gnidaol]

but the tests are still the same for every cpu.

 

The test is for everybody the same.

 

Only a amd will perform better in another sub-test then a sandy bridge for example, but the test is still the same for everybody.

 

That's the main thing

 

But don't you think that getting the result of a different subtest as final score for each cpu is kinda unfair?

 

Anyway we won't get to any point here, I like the bench and that is what matters for real =)

[Massman]

Closing the competition end of this month. The competition was to beta-test the application and that application now is generating points

[Chiller]

GL or HW points ?

[Massman]

HW.

[gnidaol]

Hello guys, after a couple months I come back again to rise up one thing that really bothers me about this bench:

 

In 3dmark03 we just need to run the game tests, in 3Dmark vantage we can disable feature tests, in wprime 1.55 I need to select the number of threads (otherwise I could end up running it as a single threaded benchmark), now please, make my hearth calm down by explaining to me why do I need to run single thread, dual thread, xxx thread until the correct thread count and even more, sse, sse2 sse3 and sse4 if ONLY THE BEST subtest is taken to the final score???

 

To me it's the same as not allowing wprime thread selection or running 3dmark subtests that doesn't matter to the final score. As far as I can see is really pointless to stress my cpu calculating stuff that won't make any difference at the final score.

 

Please staff, do the test: run it at default options and then select just the set of instructions of the highest score and the correct number of threads of your CPU and run it again. You will receive almost the same score and you be able to successfully validate it at the official website.

 

I love this benchmark, but I see that we are obligated to do things here that in a lot of other benchs we can avoid without breaking the rules.

 

thanks if you prove that I'm wrong and thanks if you accept my suggestion!

[IanCutress]

Simply put, it's how the program works. It's a GUI based on a command line program. The GUI doesn't detect what the processor can do, only the command line can. And sometimes (not always), it's not the most up to date instruction set that gets the best score. So the program runs through them all to find that score.

 

It's true it can be adapted, but this benchmark wasn't made specifically for the overclocking community. What I assume is that with everyone that runs it, it provides spine points on graphs for all the different instruction sets, architectures, and overclocks.

 

Benchmarks made for overclocking are invariably synthetic (let's use all the processor has to offer), whereas this is an example of something real world, which may not use everything, and thus some architectures are better suited to it than others.

 

However, the GUI could probably be updated for manual selection of scores and stuff, but it would require reinterpretation of the official website to accept those scores (which may or may not be what the creator wants).

[knopflerbruce]

So the highest score is always the final score? Tbh I have no idea - it doesn't say anything about that part anywhere.

[gnidaol]

Simply put, it's how the program works. It's a GUI based on a command line program. The GUI doesn't detect what the processor can do, only the command line can. And sometimes (not always), it's not the most up to date instruction set that gets the best score. So the program runs through them all to find that score.

 

It's true it can be adapted, but this benchmark wasn't made specifically for the overclocking community. What I assume is that with everyone that runs it, it provides spine points on graphs for all the different instruction sets, architectures, and overclocks.

 

Benchmarks made for overclocking are invariably synthetic (let's use all the processor has to offer), whereas this is an example of something real world, which may not use everything, and thus some architectures are better suited to it than others.

 

However, the GUI could probably be updated for manual selection of scores and stuff, but it would require reinterpretation of the official website to accept those scores (which may or may not be what the creator wants).

 

You already can do that at current GUI at "adv..." button. The score is validated as well with no issues. What I'm asking is for hwbot to accept these specific and validated runs to be considered valid and not only the ones with lots of "feature tests".

[gnidaol]

LOL I'm alone here trying to make this bench have the same rules that the others does... =(

 

Come on, staff say something? Did you try it only choosing the right number of threads and the best optimization for each cpu on adv button? Did it affect final score at the same configuration at all??

[PsySc0rpi0n]

If i could only submit my score, tolds me every time this:

 

competition limitation: your submission does not comply A verification link is required.

 

Verification url must match pattern: http://anrieff.net/ucbench2011/viewitem.{0,4}?\?id='>http://anrieff.net/ucbench2011/viewitem.{0,4}?\?id=[0-9?]*

 

The ucbench site is generating links like http://anrieff.net/ucbench2011/viewitem.{0,4}?\?id=[0-9?]* or http://anrieff.net/ucbench/viewitem.{0,4}?\?id=[0-9?]*

 

Just make sure to add "2011" after "http://anrieff.net/ucbench" like this "http://anrieff.net/ucbench2011..." and it should work...