Xtreme Grand Prix discussion

[saaya]

EDIT:
If you have a Question about the competition please first read the competition rules to see if they answer your question(s) HERE
If you want to register to join the competition please please visit the registration booth HERE



Troman had the idea to make a new 3d contest not using only one benchmark but several ones to make it more demanding. i had thougt about a similar thing before and we both worked on the idea.
this is what we came up with so far:

we thougt we would use 3dmark 2001, 2003 and 2005 as well as aquamark, superpi and pifast. the results of all scores will be used to calculate a final score.

the current scoring system we thougt off looks like this:
3Dmark05 result divided by 15
3Dmark03 result divided by 30
3Dmark01 result divided by 40
Aquamark result divided by 100
Superi (1000-result)X10
Pifast (1000-result)X10

all scores will be added up and form the final Ultimate 3d score wich will be meassured in Xtreme Marks

we are still tweaking the rating system to make each benchmark as important as the other ones and to adopt it to each benchmark, as some scale more than others.

the contest will go as following:
to take part you have to register first, post the specs of the systems you will be competing with and details about the cooling, and if possible a pic of the system if any part in your system dies you can only finish the competition by replacing it with an identical piece of hardware.
the contest will run like a grand prix, but instead of having several race tracks we will have several benchmarks
we will run one benchmark for one week, and then switch to the next benchmark in line.
you will have a 1 week time window to tweak your system for a benchmark and submit your best shot, after that week is over all scores for that benchmark get locked, and your highest result in that benchmark for that week will be used to calculate the final score.

categories:
we will naturally split the field of racers into several categories from a "formula1" to a "demolition derby"
we havent decided about the categories yet though, and we came to the conclusion that it would be best to watch the systems of the participants of this competition and then create categories to evenly split the participants to. (if we make a cacade categopry and only 2 people use a cascade this would be pretty pointless for example...)

prizes:
honor, glory, loads of women and a pic of you on the front page

start and end date:
not set yet

please let us know what you think!
any feedback is welcome

we would appreciate it if some of you could post results of your system at stock speeds and at oced speeds in as many of the benchmarks we are going to use as possible, wich will be very helpfull to see what bench scales how, this will help us to tweak the formula to calculate the Xtreme Marks

and if you think we should drop a benchmark from the line or add another one, please let us know!

- Saaya and Troman

[2fink]

count me in.

not the best system, but i think it will really make fun to tweak the system for every benchmark.

question: if i have a very good old run with the same hardware, can i use it or must i re-run the bench in the "one" week?

[Ackbar]

prizes:
honor, glory, loads of women and a pic of you on the front page

I'd recommend including prizes proportional to the category you enter. So if you enter the lightweight X300/6200 category and win, you get a lesser prize than if you entered the X850 XTPE/6800 Ultra category.

Actually, along these lines, I'd rather see an endurance test competition! Not just one run, but you'd have to be able to loop the test at least some arbitrarily large number of times so people who aren't stable will have to work to get their systems stable and not just able to make one good run!

Let the games begin!

[wittekakker]

Nice idee of u Troman. I realy going to watch this, and maybe bench if I find some time. This should be on frontpage to saaya

[Highland3r]

Will the benchmarks be weighted? ie. Those of us with say a 9800pro cant really compete in 01/03/05/aquamark against those running say an X800 or SLI/crossfire...

[Wiggy McShades]

could you make a seperate sli/amr division its kinda stupid running 1 card against 2. how exactly are you going to place people into divisions ? by gpu and cpu speed or by cooling?

[Wiggy McShades]

Look at my first post....
there you see every classe that wil exist(for now,Saaya and I will work out pretty soon)

ohhh i thought that was a sugestion, sorry. can you enter two different rigs into two different classes ?

[alexio]

Maybe you can find a sponsor for this competition and include some nice rewards.

I think I will participate in this contest, since I have the time for it this summer.

[2fink]

Good question....
yet i think (after speaking with saaya)that this question can be easy answered....rebench that week....

Its like the olympics or a world championship..that's what this is actually all about...see who's the best of all at that moment...

good to know. cant wait to start benching.

[faruquehabib]

sounds fun. downloading programs now...

[Entity_Razer]

I'm in,just let me take my big room fan

AMD 2500+
A7n8X-X
1GB DDR 3200+ twinnmos (non UTT,just Mtec's)
HDD's don't matter I suppose?
9800pro
Creative Soundblaster Live! 5.1

cooling:
SP 97 + 92 mm EnermAX fan

[LilGator]

saaya, it seems the Pi tests are a little over weighted. Most people should be able to score 40secs or lower ... so 1000-40 = 960x10 = 9600 ...

Lets say we have a system that does: 34,056 - 15,517 - 7,135 - 89,223 ...[eva2000's sig ]

01 => 851.4
03 => 517.23
05 => 475.67
AM => 892.23

Then you throw in a 42sec PiFast, and a 32sec Super_Pi...

[(1000-42)*10] = 9580
[(1000-32)*10] = 9680

The 2 Pi results add up to 10660. The 4 3D tests add up to 2736.53.

I think you should drop the x10 in the Pi equation, leaving a total of 958 for the 42sec PiFast and 968 for the SuperPi run, both numbers much more comparable to the 3D runs.

Total score would be: 3802.53 XtremeMarks [13396.53 if you leave the x10 for the Pi]

Now for scaling, lets say eva2000 bumps up his clocks and cranks out these new scores: 36,342 - 16,721 - 7,862 - 93,275 - 39sec - 29sec

Total score would be: 4854.8 XtremeMarks [22242.8 if you leave the x10 for the Pi]

It seems to me like the jump from ~3800 -> ~4800 is much more reasonable, than the 13K -> 22K. So I suggest removing the 10x in the Pi formula at a minimum.

^ THiNKiNG on the FLY ^
----------
[I like the proposition below better]
----------

OK, I really like how the 3D tests are scaled. The dividers are all relative to the world records in each test. IE: eva2000s, 7K '05 score is approx 47.6% of the WR. I think the 3D01 divider should be bumped to Score/45 ... since the WR is approx. 45K. AM3 should be bumped to Score/120. Then we should be set.

----------

OPTiON B - (I like this better), for PiTests ...

Super_PI
[120-(score)]x10

Ok, for consistency sake, we want the scores to be relative to their WRs. A score of 45,000 in 3D01 would result in a sub-score of 1000 XtremeMarks; or a perfect score (tying the WR supposedly). A score of 36,000 would result in a sub-score of 800; or equivalent of 80% of the WR.

Now SuperPi. The WR is currently somewhere around 19s I think. So a score of 19-20s should land you around a 1000, just like in the 3D tests, right? K, cool, let's do it.

[120-(20s)]x10 = 1000.

Lets say we score 32s, an 'average' score, we would get [120-(32)]x10 = 880 ... or about 88% of the WR (this is much rougher estimate is lower scores are better here, but you get the idea).

Same will work for PiFast, I think the WR is somewhere 10s higher (~29-30s?), so a formula of [130-(score)]x10 should work the same here.



After all are tallied, the WR score should be somewhere close to 6000. Or 1000 (a 'perfect' score) in all 6 tests. To make things more interesting, and to not have to rely on decimal places for close scores, I think 60,000 should be the scale. So my proposed final calculations are as follows

3DMark05 result divided by 1.5
3DMark03 result divided by 3
3DMark01 result divided by 4.5
Aquamark3 result divided by 12
Super_Pi (120-result)X100
PiFast (130-result)X100

Using eva2000 as a lovely example again, here is our final outcome:

34,056 '01 - 15,517 '03 - 7,135 '05 - 89,223 AM3 - 42s PF - 32s SP

01 => 7568.00
03 => 5172.33
05 => 4756.67
AM => 7435.25
PiFast => 8800.00
Super_Pi => 8800.00

TOTAL: 42,532 XtremeMarks for approx. an aircooled 2.7GHz S754 A64 and 270MHz 2-2-2-5. w/ x800XT.

[LilGator]

Yes, the problem there is SLi. 15K is reachable with SLi/Crossfire making his score (single card, in comparison) much lower. But seeing as there will be separate categories for Dual GPUs / Single GPUs, it shouldn't be a problem.

[QuikSilver]

This is going to be awesome! I'm glad I have a high clocking chip and a 6600GT that can also get insane clocks with a few mods.

[i found nemo]

count me in

but to keep it fair, i think each speed should be a class, eliminating who has what cooling


example for cpu

2000mhz-2200mhz:
2201mhz-2400mhz:
2401mhz-2600mhz:
2601mhz-2800mhz:
2801mhz-3000mhz:


so on so forth, sure it's more classes, but on the down side, there would have to be more prizes, as there is more classes....also, classes by core and mem clock speed, and by pipes enabled on the card....??

[LilGator]

AiR | H2O | PHASE | EVERYONE

Seems like fair separations. I agree, a category for Cascades would be silly if there were 2 people. Instead, there should just be a category with everyone, LN2/DI/Cascades can all compete for the top there. No rules, anything goes in this category Also, I'm sure there are high-end aircooled systems that compete with some water. Same with chilled water and low end phase. An everyone category would be interesting to see, obviously it will be won by someone with LN2/DI/or a cascade, but this way we don't need a separate category ... just a thought ...

After those basic separations, some sub-categories will be necessary. None of these benches take advantage of multi-threads, so no need to separate for multi-CPUs. Multi-GPUs on the other hand will be needing a separation. Side note, I think it would be silly to differentiate between ATi and nVidia as well.

AiR SiNGLE GPU | AiR MULTi GPU
H2O SiNGLE GPU | H2O MULTi GPU
PHASE SiNGLE GPU | PHASE MULTi GPU
ALL SiNGLE GPU | ALL MULTi GPU

Separate categories for iNTEL and AMD I think is also over the top, way too many categories, and neither one has an advantage over the other like multi-GPUs have over single, or cascades have over air.

Personally, I'd stop it right here, 8 categories, should work well.

If we have hundreds of people in this contest, we might be able to have more sub-categories. But this would be silly for a low number of contestants, since some categories might only have one person in them (like PHASE MULTi 4 PiPE GPU). But here is the next breakdown anyhow...:

4 PiPES | 8 PiPES | 12 PiPES | 16 PiPES

Anything less than 4 pipes competes in the 4 PiPE category, anything with more than 16 pipes ( ) competes in the 16 PiPE category.

So here we have it:

AiR SiNGLE 4 PiPE GPU | AiR SiNGLE 8 PiPE GPU | AiR SiNGLE 12 PiPE GPU | AiR SiNGLE 16 PiPE GPU

AiR MULTi 4 PiPE GPU | AiR MULTi 8 PiPE GPU | AiR MULTi 12 PiPE GPU | AiR MULTi 16 PiPE GPU

H2O SiNGLE 4 PiPE GPU | H2O SiNGLE 8 PiPE GPU | H2O SiNGLE 12 PiPE GPU | H2O SiNGLE 16 PiPE GPU

H2O MULTi 4 PiPE GPU | H2O MULTi 8 PiPE GPU | H2O MULTi 12 PiPE GPU | H2O MULTi 16 PiPE GPU

PHASE SiNGLE 4 PiPE GPU | PHASE SiNGLE 8 PiPE GPU | PHASE SiNGLE 12 PiPE GPU | PHASE SiNGLE 16 PiPE GPU

PHASE MULTi 4 PiPE GPU | PHASE MULTi 8 PiPE GPU | PHASE MULTi 12 PiPE GPU | PHASE MULTi 16 PiPE GPU

ALL SiNGLE 4 PiPE GPU | ALL SiNGLE 8 PiPE GPU | ALL SiNGLE 12 PiPE GPU | ALL SiNGLE 16 PiPE GPU

ALL MULTi 4 PiPE GPU | ALL MULTi 8 PiPE GPU | ALL MULTi 12 PiPE GPU | ALL MULTi 16 PiPE GPU

Like I said, I wouldn't go this far, maybe a breakdown like this would be OK, but actual "prize-winning" categories would be silly, being there are 32 of them.

[LilGator]

For the question that may come up sooner or later; with an aircooled A64 and a watercooled 6800GT, which category am I in ?

I would propose that the moment you slap a waterblock on the GPU of your all aircooled system, you bump up to the water category, regardless if you have your CPU watercooled or not.

Same goes for phase on the CPU and air on the GPU. You are bumped up to the phase change category.

Make sense ?

[=[PULSAR]=]

I absolutely agree with you gator w/e the highest level of cooling on the system is, thats the category it should be in. Just my opinion but the categories should just be split up into cooling and SLI/non-SLI. No point in having an overall category cause obviously phase/DI/waterchilled is going to win.

[Wiggy McShades]

For the question that may come up sooner or later; with an aircooled A64 and a watercooled 6800GT, which category am I in ?

I would propose that the moment you slap a waterblock on the GPU of your all aircooled system, you bump up to the water category, regardless if you have your CPU watercooled or not.

Same goes for phase on the CPU and air on the GPU. You are bumped up to the phase change category.

Make sense ?

i would think if you have water on one part of ur system its considered a water cooled rig, it would be unfair to put you in air cooling because no one with an only air cooled system could compete with a watercooled system even if its only the gpu.

[LilGator]

I absolutely agree with you gator w/e the highest level of cooling on the system is, thats the category it should be in. Just my opinion but the categories should just be split up into cooling and SLI/non-SLI. No point in having an overall category cause obviously phase/DI/waterchilled is going to win.

Right, should we have a EVERTHiNG-COLDER-THAN-PHASE category ? LOL

Or individual categories for Cascade / Dry Ice / LN2 ?

I think an anything goes category would be a better way for the big boys to compete as it doesn't matter what the heck you cool with, just get the top score.

i would think if you have water on one part of ur system its considered a water cooled rig, it would be unfair to put you in air cooling because no one with an only air cooled system could compete with a watercooled system even if its only the gpu.

Exactly.

[Wiggy McShades]

why exactly do you need the seperate intel and amd sections ?

[LilGator]

So you're saying chilled water (-5c) ... will compete with LN2/DI/Cascades ?

Seems like quite a range there.

Categories are created for fairness sake. AMD vs. Intel is perfectly fair, especially when separated into cooling categories Same with ATi vs. nVidia ...

I see where you are going though, but I don't think it's necessary to break down the individual tests, seeing as that's the whole purpose of these XtremeMarks and the entire contest

So here's what I would say:

If you don't complete all the tests, your score is still tallied and totaled up the same way. That's the idea in this Grand Prix, at the minimum, complete all the tests, or "finish the race". If you don't, then you don't; no need for a breakdown of the individual tests, then you might as well just show the real scores, and we already have threads upon threads for that sort of thing

[LilGator]

Thats what i said

amd vs intel-->3dmark01?fair?

Absolutely, ever heard of Dothan ?

[Fission]

I am definitely in. With all the great concerns being brought up, I will throw in my 2 cents.

I have a VIA chipset mainboard, so I am functionally impotent. Could there not be a class called "Welfare Class" for those of us who bought a VIA? If not, I will still be proud to come in last.

2ndly, AMD N-Bench-3 is not that popular on this forum it seems, and is essentially AMD only, but it is a pretty intensive utility, and it definitely showcases the quality of overclock on AMD machines.

Both of these concerns can be answered by a simple "yes" or "lol" but I thnk that it would help me feel like a big man if VIA users could stand some chance at glory, even though we bought the worst chipset out there (we'll assume no one on this forum has an SiS).



EDIT - Where do we register our specs/pics?

[[XC]thewildblue]

Im in as well. Out of interest how do we stand with duallys and such like with super Pi etc ?