Adjustment for Global Points - Work in progress

[Strong Island]

agreed, legacy with a 5850 I gained nothing from 5.8ghz skylake to 6.1ghz finally.

 

also 3 of 4 5850 DCU cards died when I took the stock cooler off

 

When you buy 4 cards and none of them have the right ram IC you are left with unsalable garbage and you have to eat it.

 

 

ya when I was benching the 8800 gts 512mb, especially with 3dmark 05, 06, and vantage, the cpu didnt have a big impact on the score at all. I had to go thru 5 cards, most had quimonda mem, some died from learning how to epower.

 

But just because the cpu is at a high speed doesnt mean the gpu wasnt pushed really hard.

 

I just felt like making that point, because I think there is a great place for this kind of stuff here. People seem really excited with the new cheap gpu comp.

 

It's really fun to me and I never thought to myself 1 time, oh man a sponsor might not care about this.

 

I'm not saying I dont want this hobby to grow, I mean who wouldnt want to make a few dollars off something they love to do.

 

But it's not what pushes me, just for some reason it's fun. I sometimes dont even understand why.

 

But just because you dont like a benchmark doesnt mean it should go. I mean maybe someone really loves benching it.

 

There are plenty of other avenues to get points, we need something for everyone. If I dont like a bench I dont care if someone else has points in it.

 

I will just choose what I love and roll with it. A lot of choice isnt bad.

[Strong Island]

I'm not totaly agree but anyway, base hwbot ranking mostly on hwboint made on old vga and bunnyextractionty loocked cpu on xtu, is a totaly waste of time and effort.

Don't care much about hwbot ranking right now, will care less with new revision.

 

I get what your saying, I dont think it should be based mostly on hwboints, but a good balance is good right?

 

What is so wrong with benching old gpu's? I mean how many times can you bench a 980 ti before you want to try something else.

 

It's also a nice place to start for people learning about volt mods because things are bigger and less complex.

 

Not everybody was around when the cards first came out so it's good to go back and bench these things and gain that same experience that veterans got.

 

I think it should be balanced so everyone has a chance. Maybe 25 hardware slots is a good start.

 

And I agree on the locked xtu. But it's funny because now most people in the top rankings have it so when 743 comes out we will all drop evenly, haha.

[Alan_Alberino]

I said in one of my comments that some old 3D benches are pure CPU benchs, I said this looking it from the point of view of taking out global points, 3DM2001 SE and Aquamark WR are with a stock GPU, 3DMark05 air cooled GPU, cause the thing to make the WR in this benchs is a high CPU.

 

I'm not discrediting this WR, cause they involve lot of tweaking knowledge and effort looking for a good CPU, but they are supposed to be GPU benches and they are actually CPU benches.

 

Hardware Points obviously must stay, cause like lot of people said you need to search for a good GPU, and then mod it without killing it, with isn't a easy job at all. I killed two cards at my first attempts to make a zombie GPU with G-Powerboard, and I finally managed to get working my 3rd card with a lot of effort, and my mod is very stupid beside the professional mods that a lot of people make here to get a 1st hardware place, so I imagine the effort behind that and think that of course it should be rewarded with Hardware points.

[RULE]

I get what your saying, I dont think it should be based mostly on hwboints, but a good balance is good right?

 

Defenily

 

What is so wrong with benching old gpu's? I mean how many times can you bench a 980 ti before you want to try something else.

 

Nothing against old card, i've also my toys but 01-aqua with your cpu will be easy win. So when you put one score with you cpu at 6.6 cut the fun to all the others.

 

Totaly different story for cheap new hardware, is a competition and hwbot can chose the right benchmark to let the cpu out of score. 50boints for this? hell yes

[Splave]

its hard to make it all level point wise. a 6.6cpu wont help a 8600GT like it will a 980ti so its hard to give them both 50HW points for 01 and aq3.

 

if we make the rules too complex it makes it impossible to follow, say if we tried to make a 8600gt aq3 50hw points and a 980ti aq3 25hw points.

 

I though maybe making like a pre 2010 graphics card list have 50 points legacy and newer no legacy or something would be cool since CPU is king but it is hard to follow imo.

[Strong Island]

Defenily

 

 

 

Nothing against old card, i've also my toys but 01-aqua with your cpu will be easy win. So when you put one score with you cpu at 6.6 cut the fun to all the others.

 

Totaly different story for cheap new hardware, is a competition and hwbot can chose the right benchmark to let the cpu out of score. 50boints for this? hell yes

 

I see what your saying but 1008mhz gpu clock is the highest I can find right now for those benches at least in the top 20. The score I beat had 36mhz lower core clock.

 

So just because the cpu is clocked high doesnt take away from the gpu acheivement. I could have taken those scores with lower cpu clock but I was pushing everything to the max.

 

And 05 and 06 I did downclock the cpu a bit and it didnt make 1 difference in the score. So that's why I dont understand why these things are just looked over "like oh he has no skill, it's all the cpu"

 

Even though it took me 4 days to mod that card and 6 different cards to find a nice one with Sammy mems, some came dead. Others didnt make it.

 

Everything was pushed to the absolute limit. isnt that what this place is supposed to be about. Making our hardware cry.

[Strong Island]

I said in one of my comments that some old 3D benches are pure CPU benchs, I said this looking it from the point of view of taking out global points, 3DM2001 SE and Aquamark WR are with a stock GPU, 3DMark05 air cooled GPU, cause the thing to make the WR in this benchs is a high CPU.

 

I'm not discrediting this WR, cause they involve lot of tweaking knowledge and effort looking for a good CPU, but they are supposed to be GPU benches and they are actually CPU benches.

 

Hardware Points obviously must stay, cause like lot of people said you need to search for a good GPU, and then mod it without killing it, with isn't a easy job at all. I killed two cards at my first attempts to make a zombie GPU with G-Powerboard, and I finally managed to get working my 3rd card with a lot of effort, and my mod is very stupid beside the professional mods that a lot of people make here to get a 1st hardware place, so I imagine the effort behind that and think that of course it should be rewarded with Hardware points.

 

 

I agree with some gpu benches being won with stock cards. I hate even saying it because I know the people benching them still had to do a lot of work and it took a lot of tweaking, but it is silly when they are supposed to be gpu benches.

 

I would love so much to see firestrike ultra and cat 4k gain points, and then maybe remove cat 720p and regular firestrike. Could help clean up some duplication while releasing more pure gpu benching

 

I think we have a nice system now that a lot of people are used to. We should tweak it more instead of completely revamping it.

 

And to the guys who have a totally different opinion, I completely respect it and see where your coming from. I think it's cool we can have an open discussion with differing opinions that doesnt turn ugly.

Edited July 22, 2016 by Strong Island

[Alan_Alberino]

I would love so much to see firestrike ultra and cat 4k gain points, and then maybe remove cat 720p and regular firestrike. Could help clean up some duplication while releasing more pure gpu benching

 

Yeah, this would be great, I gave a very similar idea for a transition some pages ago:

 

"maybe adding globals to Fire Strike Ultra and/or Catzilla 4K wont be bad, and once they got a decent number of users (nobody bench them now cause they dont give points), take out points of Fire Strike and Catzilla 720P."

 

New cards are designed for 4K so it would be nice to test them with 4K and get globals for that, and not globals for 720p where they have a ridiculous performance.

Edited July 22, 2016 by Alan_Alberino

[Strong Island]

Yeah, this would be great, I gave a very similar idea for a transition some pages ago:

 

"maybe adding globals to Fire Strike Ultra and/or Catzilla 4K wont be bad, and once they got a decent number of users (nobody bench them now cause they dont give points), take out points of Fire Strike and Catzilla 720P."

 

New cards are designed for 4K so it would be nice to test them with 4K and get globals for that, and not globals for 720p where they have a ridiculous performance.

 

ya it's a good idea and also takes the 6950x out of the equation for those benches, which I like. It's totally cool if someone wants to use 6950x but it will be nice to not have to if you cant.

 

With 980 ti and cat 720p you could get a really competitive score with a 4 core cpu.

 

So cat 4k should be even more like this obviously. And fsu will be similar. Could add a little life to gpu benching

[der8auer]

The whole 2D vs 3D discussion is absoulte nonsense. No matter how you tune the points, you can't shift the user activity towards 3D when the amount of money involved is several times higher.

 

Even if you push the focus towards 4K benches you still need a 800-1200 USD (or even more) VGA to be competetive and most likely you also have to mod it. I mean seriously how can you belive that anybody is willing to do this when you can get a whole 2D rig with 6700K for ~ 600 USD and dont have to mod it...

 

HWBOT revisions have a big impact on the OC community and activity and you should think how you can get more people started and I can guarantee you that this is the wrong path.

 

The new Titan X will be out in few weeks which will be priced above 1000 €. If you want to be competitive in 3DM11, 3DMVantage, 3DMFS you need X99 with 6950X + Titan X + quad memory kit. That's at least 4000 € just to participate. How many people do you think will be able to participate there? 5? 10?

 

It's cool if you like 3D but if you force people to buy this kind of hardware for the rankings, hwbot will be totally dead within months.

[websmile]

We do not agree that often, Roman, but you write exactly what I thought when I followed the discussion. Also the pecking on legacy benchmarks doesnt make that much sense, this all leads hwbot into a direction even more dominated by money. We already stripped the most benched 3ds and old 2ds of wr points to level things, we have 6/12 wr and global for 2d and 9/13 wr and global points for 3d already. If we focus even more on new benchmarks for ranking that need near 3k rig without vga and any binning effort, we will kill hwbot faster than any other way...

[der8auer]

Yes I also thought exactly the same about legacies. Also the amount of benchmarks we have on the bot also does not matter. Even if we have 100 benchmarks, you still only need 15 globals for your profile.

 

If you don't like 3DM2001, bench something else. It's much better to have a big selection of benchmarks and everyone can bench what he wants. Removing legacies won't fix anything

[sofos1990]

Also, legacies should stay because you need less money for a WR which makes the benchmark more "attractive".

For me the main reason though is the skills vs pure clocks. Show me a newer benchmark that somebody took the first place with 100-150Mhz on CPU/GPU less. I'm not saying I don't want pure clocks in the game(otherwise I would be against XTU) but with the legacies there you balance the situation.

[Strong Island]

Yes I also thought exactly the same about legacies. Also the amount of benchmarks we have on the bot also does not matter. Even if we have 100 benchmarks, you still only need 15 globals for your profile.

 

If you don't like 3DM2001, bench something else. It's much better to have a big selection of benchmarks and everyone can bench what he wants. Removing legacies won't fix anything

 

I totally agree, that's why I love the legacy benches and older hardware because it gives people a chance with less money to better there profile.

 

I also agree that the amount of benches doesnt matter. it gives everyone a chance to bench something they like, if you dont like it, let someone else have fun benching it and move on to something else.

 

But ya it would be terrible to move hwbot to a place where you need 6950x+titan x to be able to move up the rankings. I agree.

 

My only thought was maybe releasing points for 4k benches will help a little bit with the cpu costs.

 

But you still would need a $1000+ gpu and 4 core cpus arent that cheap, so you still need a lot of money.

 

I honestly love things the way they are.

 

I was just trying to throw some ideas out.

 

Also, legacies should stay because you need less money for a WR which makes the benchmark more "attractive".

For me the main reason though is the skills vs pure clocks. Show me a newer benchmark that somebody took the first place with 100-150Mhz on CPU/GPU less. I'm not saying I don't want pure clocks in the game(otherwise I would be against XTU) but with the legacies there you balance the situation.

 

ya it's true, the legacies take a lot more tweaking skills. I still have a lot of trouble in efficiency with them even with a good cpu.

 

With the legacies you do have to squeeze every last bit of efficiency out to make a top score.

 

I love them, it gives me something to look forward to, to learn how to become efficient. And gives more things to tweak.

Edited July 22, 2016 by Strong Island

[Rauf]

Excellent post from der8auer, the one about 2d vs 3d. I think that from some kind of fairness point of view, 3D should receive the higher points. But points won't fix the 3D decline, or the decline of all XOC. We have previously discussed this and had some ideas on how to make 3D cheaper and thereby more accessible and popular. It is a discussion not for this topic really, but maybe the most important to have. Sadly only a few people seem to care... All money driven sports have already come to the conclusion that too much money hurts the sport. Be it player salaries, money backing, equipment or engineering/development costs. When they get too high and start to influence too much it has to be limited.

 

But I dont agree with the "keep all benchmarks" thinking. We have a point system which is based on the activity of each benchmark. We have to keep the ones we have busy for the points system to be viable. Also there is the aspect of competitiveness. Even now we don't see people with the best cpus/gpus bench all relevant benchmarks because there are too many so they don't need to. Seeking out the categories with the lowest level of competition is a very viable strategy here on hwbot, but the phenomenon is not good for the sport. A high level of competitiveness should be a nobrainer for all sports that wants to thrive.

[Strong Island]

Excellent post from der8auer, the one about 2d vs 3d. I think that from some kind of fairness point of view, 3D should receive the higher points. But points won't fix the 3D decline, or the decline of all XOC. We have previously discussed this and had some ideas on how to make 3D cheaper and thereby more accessible and popular. It is a discussion not for this topic really, but maybe the most important to have. Sadly only a few people seem to care... All money driven sports have already come to the conclusion that too much money hurts the sport. Be it player salaries, money backing, equipment or engineering/development costs. When they get too high and start to influence too much it has to be limited.

 

But I dont agree with the "keep all benchmarks" thinking. We have a point system which is based on the activity of each benchmark. We have to keep the ones we have busy for the points system to be viable. Also there is the aspect of competitiveness. Even now we don't see people with the best cpus/gpus bench all relevant benchmarks because there are too many so they don't need to. Seeking out the categories with the lowest level of competition is a very viable strategy here on hwbot, but the phenomenon is not good for the sport. A high level of competitiveness should be a nobrainer for all sports that wants to thrive.

 

there is a small incentive to bench more popular categories because you get more points in those benches for top scores.

 

But i get what you mean, if traffic was more condensed to a smaller amount of benches, the competition would be higher and top scores would be harder to get. I just dont see it as a huge problem because there are good submissions in all the global point benches

[der8auer]

But I dont agree with the "keep all benchmarks" thinking. We have a point system which is based on the activity of each benchmark. We have to keep the ones we have busy for the points system to be viable. Also there is the aspect of competitiveness. Even now we don't see people with the best cpus/gpus bench all relevant benchmarks because there are too many so they don't need to. Seeking out the categories with the lowest level of competition is a very viable strategy here on hwbot, but the phenomenon is not good for the sport. A high level of competitiveness should be a nobrainer for all sports that wants to thrive.

 

Maybe one solution would be to adjust the activity aspect. Maybe increase the influnce of activity. So this way the unactive benchmarks would "disable" themselfes.

[Mr.Scott]

Maybe one solution would be to adjust the activity aspect. Maybe increase the influnce of activity. So this way the unactive benchmarks would "disable" themselfes.

That would bring a whole new topic of discussion to the table, How much activity before a bench would be considered inactive and disabled.

[Alan_Alberino]

That would bring a whole new topic of discussion to the table, How much activity before a bench would be considered inactive and disabled.

 

Mmmmmmmmmmmmmm, have an idea for relating points and activity, dont know if its good but can be a base:

 

See which bench had more users in the last year, take that as 100% users, and then divide benchs in 3 categories:

 

- Benchs with 75% or more of that 100%: Keep Normal Puntuation

 

- Benchs between 25% and 75%: {[(Percentaje-25)*2)/100]}*normal puntuation.

 

- Benchs less than 25% = no points.

 

Why this formula in the second case? 75%-25% = 50%, 50%*2=100% of normal puntuation, but having less than 75%, for example 65%, you get 65-25 = 40, 40*2 = 80, 80/100= 0.8 and then you get 80% of the normal puntuation of the bench, sucessfully reducing the puntuation of a bench with not that much activity.

 

Of course that 75% and 25% can change if the benchs above 75% are only 2 or 3, its just an idea.

[der8auer]

@Massman what is the current influence of activity on the points? Is there even an activity influence? Talking about activity over time, not plain amount of results.

[Mr.Scott]

@Massman what is the current influence of activity on the points? Is there even an activity influence? Talking about activity over time, not plain amount of results.

Excellent question. I am interested also.

[aerotracks]

I took a look at my point distribution on live and uat. Right now I have about twice as many 3d points as I have 2d points. With the new revision, I will be losing a few hundred 3d points while my 2d points will almost double. The way this is going, I don't see myself focusing on 3d when it's taking another stab, but that's not what this new revision should be about?

 

*Edit: I should add, my 3d points don't come from $1000 GPUs but rather cards that people can actually afford without becoming an organ donor.

 

Live:

Uat:

Edited July 26, 2016 by aerotracks

[Aleslammer]

Just looked on UAT and a little confused, as it exists without R7 my HW points are primarily 3D as shown on UAT this has shifted to 2D although I will say scores I have a little pride in. The most confusing my best global is a crap HWB Prime score with a 5820K that I sub’d just to put a 5820K in my hardware library with no more effort than switching on a preset in BIOS for 4.0, so is 20 points in R7 like 2.0 in R6?

[Alan_Alberino]

Any updates on this?

[Massman]

Found a bit of time to read through the feedback in this thread. You guys wrote massive walls of text, awesome! I won't go in to specific replies for now, but I will make some general statements below that should address most of the feedback.

 

For feedback on specific submission point comparisons please always link to the scores you are comparing. Otherwise it's difficult to comment

 

1) Global versus Hardware Points

 

This discussion permeates the HWBoints discussion since the very first revision. There are two points of view, generally. The first point of view is that the balance should be leaning towards the Global Points as overclocking is driven by new hardware releases and the fastest benchmark scores. The second point of view is that the balance should be leaning towards Hardware Points since it tends to require more knowledge on a wider variety of platforms. I think most people are in the middle of both points of views and want a 'fair' balance between Global and Hardware Points. Ultimately this is a decision for the Community Leader @Christian Ney). I don't think there is a 'right' decision here, since there will always be people who want more of the one or the other.

 

On the topic of the industry, I want to get the following point across: the decision of how much Global or Hardware Points are allocated to the Overclockers League is entirely up to the community to decide. The fact that the industry prefers the latest generation of hardware to be used/promoted has no relevance to the Overclockers League. Even if we would have a League that is based on solely the cheapest graphics card and the oldest hardware, the industry would still be able to use the benchmark records for product promotion. The Overclockers League should reflect solely the point of view of the overclocking community.

 

2) Why so many changes at once?

 

Some people argued that we introduce too many changes at once and it would be better to do small changes.

 

As @george\.kokovinis pointed out in an earlier post, for any economy it's important to have a stable and predictable environment for people to invest time and resources into it. It is in everyone's best interest to not change the algorithm every couple of months with small tweaks and changes. One big change that holds long-term allows everyone to get used to the environment and find out how to get the best results.

 

Based on the past ten years at HWBOT, I think this is indeed the best approach.

 

3) 2D and 3D required for the League

 

Svictorcc proposed to use a League algorithm that requires a certain amount of 3D benchmarks for the League points. From a design point of view, this is no problem. It is up to the community to decide if this is a good idea. I think some of the replies to that suggestion bring up a very important argument against it: the cost of competitive 3D overclocking. The cost of a competitive 3D system is now USD $1700 (CPU) + USD $1200 (GPU). With the unlocked Skylake dual cores, it's possible to be competitive in the 2D benchmarks with a USD $100 Core i3 6100. Even the Core i7 6700K is priced 'low' compared to the 3D systems.

 

Requiring people to bench 3D for a high rank in the League will drastically increase the minimum cost of overclocking. I think a better approach to promote 3D overclocking is by hosting regular Cheapaz Chips competitions with low-end hardware. The winner gets 50pts for that and the entry cost is minimal.

 

4) Adding/removing points from benchmarks

 

Several people have suggested to add/remove points from legacy/similar benchmarks because they are outdated, low in activity, or just too similar. Some argued that currently it the competitive scene is too spread out, causing very little direct/focused competition. Others argued that it's better to have a wide selection because then you can choose what you prefer. Just like the Global vs Hardware points, this is a discussion that does not have a right answer and will therefore always be a point of discussion.

 

In the new algorithm we have the capability of adjusting arbitrary weight to the point slopes in favor of specific benchmarks. This allows us to increase the points for the 'most important' benchmarks once or twice a year, encouraging direct and focused competition in specific rankings that the community finds important. For example, on January 1st and July 1st of every year we pick 3 CPU benchmarks and 2 GPU benchmarks that get 125% of the points. The others remain at 100% or below. Below an example:

• 125%: Geekbench Multi Core, Cinebench R15, Y-Cruncher 1B, Time Spy, Catzilla 4K
  
• 100%: others
  
• 80%: XTU, 3DMark01, 3DMark05, SuperPI 1M, CPU Frequency
  

The choice can be based on a community poll or as determined by the community leader @Christian Ney

 

@Massman what is the current influence of activity on the points? Is there even an activity influence? Talking about activity over time, not plain amount of results.

 

If I remember correctly, the activity is measured over a time-span of 1 year ("amount of people participating in the last 365 days"). This is certainly the case for R6 and I think this did not change for R7. But let me check the code before confirming this.