Deliberately disabling GPU Cores for 3D Benchmarks

[Splave]

Would you say 3090 benching 3000usd cards is more or less competitive than people benching 500 cpus submissions wise? ?

[FireKillerGR]

I was analyzing the current system, how it favors 2D and how more unbalanced it will end up being with 3D legacies being killed.

Now, regarding your question. 18 people (not 500) benched a 12900KS (8p category on r20) on ln2 and between your score and #3 there is a gap of almost 100 mhz.

With that in mind, I would say that buying a random 6900XT-XH (less than 1.5k usd) or a 3090/Ti (less than 2.5k usd) would give you more chances to enter the top-3 in 3D than buying 10-15x KS (10~k usd) to enter the top-3 there.

If people don't want to put the effort to bench 3D due to time/difficulty or because it isnt as rewarding as it should have been then that's a different issue/question.

PS. Went a bit off topic. ?

Edited July 4, 2022 by FireKillerGR

[Splave]

5 hours ago, FireKillerGR said:

I was analyzing the current system, how it favors 2D and how more unbalanced it will end up being with 3D legacies being killed.

Now, regarding your question. 18 people (not 500) benched a 12900KS (8p category on r20) on ln2 and between your score and #3 there is a gap of almost 100 mhz.

With that in mind, I would say that buying a random 6900XT-XH (less than 1.5k usd) or a 3090/Ti (less than 2.5k usd) would give you more chances to enter the top-3 in 3D than buying 10-15x KS (10~k usd) to enter the top-3 there.

If people don't want to put the effort to bench 3D due to time/difficulty or because it isnt as rewarding as it should have been then that's a different issue/question.

PS. Went a bit off topic. ?

just look at our profiles, I have pretty much every 2d gold and you have some 3d golds and you are ahead. How is that undervalued for 3d?

37 people benched R20 on 8c alder lake ln2 and gold is worth 161pts...3 people benched FS on ln2 and gold is 169pts and as you said its cheaper to just go buy a card right? so its worth more and cheaper with less competition. Maybe I should focus more on 3d haha. I would say that is quite rewarding in every way is it not?

Same story in 11, 2 ln2 results in the top 5 still worth 169 points. 

32M WR is 161 points....top 80 subs are ln2...why would anyone pay to bin cpus when 3d is worth this much. 

 

[Guest hammertone]

15 hours ago, Splave said:

No one wants to lose points or to make benches die because they are no longer valuable. But what about people that spend time and money to compete on the top end. When it's too easy to cheat it needs to get removed. I had 01 dual card gold worth 200+ points all it required was a 7ghz cpu and still voted to remove points after spending $750 on 590s. 

I think people like to trash ln2 benchers a bit what we call "elite" hate behind closed doors...but I don't see companies sponsoring hwbot comps for ddr1 ddr2 platforms or Intel sponsoring 775 competitions. So ln2 users should have some say I think ? 

Yes and no

To be quite honest when I did water cooling, I thought you ln2 guys were a pack of wankers ?

That's the yes part.

The other part is extreme and elite still make up less that 5% of HWbot demographic.

Most of overclocking is done ambient. That's all around the world.

So the biggest market for sponsors is them. Mum and dad can buy kids pc's

______________________________________________________________________________________

Myself. I never had any hate for top overclockers in the world. Neither an ounce of jealousy.

Just not in my realm of scores before. So I did not look what a Splave or OGS had done.

Sorry to burst your bubble ?

That's the no part.

Now the good part.

Cold liquid gold. Nitrogen is so easy. So now I can see why guys like you Splave are so fast.

... and I do encourage you to keep doing the good scores 01 and 03 for global.

That's all anyone has to do. They want more points? Then try harder + go faster. 

 

edit is fix spelling ounce

Edited July 4, 2022 by hammertone

[Guest hammertone]

15 hours ago, Splave said:

No one wants to lose points or to make benches die because they are no longer valuable. But what about people that spend time and money to compete on the top end. When it's too easy to cheat it needs to get removed. I had 01 dual card gold worth 200+ points all it required was a 7ghz cpu and still voted to remove points after spending $750 on 590s. 

Sorry to quote you again.  The easy cheating bit.

Are most of us here grown ups? Know right from wrong? 

Now my sceptical side. That usually gets me in trouble. Is removing these points going to affect

someone else we know worse than you. ie done more 3Dstuff. 

What's 750 on some old gpu's. That's nothing. Do not see a real sacrifice there sorry.

Not convinced. ?

[FireKillerGR]

6 hours ago, Splave said:

just look at our profiles, I have pretty much every 2d gold and you have some 3d golds and you are ahead. How is that undervalued for 3d?

37 people benched R20 on 8c alder lake ln2 and gold is worth 161pts...3 people benched FS on ln2 and gold is 169pts and as you said its cheaper to just go buy a card right? so its worth more and cheaper with less competition. Maybe I should focus more on 3d haha. I would say that is quite rewarding in every way is it not?

Same story in 11, 2 ln2 results in the top 5 still worth 169 points. 

32M WR is 161 points....top 80 subs are ln2...why would anyone pay to bin cpus when 3d is worth this much. 

 

You are partially right. In a sense a big part of the active ocers is going for the 2D because of the convenience/time its required to set everything up. Also, some do occasional xoc based on the one cpu/mobo they get per gen from vendors. That doesn't mean that the rankings themselves are competitive. 

So quantity of users doesn't answer the competitive-ness question/issue.

If 2D gets 3x the points of 3D (due to the quantity of benchmarks with 80+ global points for the top spot) then nobody will ever need to bench 3D.

To sum it up, we need a balance between 3D and 2D (obviously 2D will be getting more points due to the quantity of the benchmarks + core quantity) till a better formula is integrated. Maybe I am wrong, but no ocer should be able to end up in top10-20 by doing exclusively 2D or 3D.

Edited July 4, 2022 by FireKillerGR

[Leeghoofd]

4 hours ago, FireKillerGR said:

You are partially right. In a sense a big part of the active ocers is going for the 2D because of the convenience/time its required to set everything up. Also, some do occasional xoc based on the one cpu/mobo they get per gen from vendors. That doesn't mean that the rankings themselves are competitiv

To sum it up, we need a balance between 3D and 2D (obviously 2D will be getting more points due to the quantity of the benchmarks + core quantity) till a better formula is integrated. Maybe I am wrong, but no ocer should be able to end up in top10-20 by doing exclusively 2D or 3D.

Been there, tried that and it failed. Complex algorithms, formulas are a thing of the past. Minor tweaking to the point system can happen (extra thresholds eg to be added) However both of you need to bench some 3D or 2D to keep the edge.

Rankings are close, plus, more importantly more rewarding for our biggest group of benchers. The points setup is easy to understand, even for a rookie, no more engineering degree required. There's more to this than the fight at the top the ranking.

Maybe to stop the current muppet show we can make it best 15 global 2D submissions and best 15 global 3D submissions next year for career and hardware ranking

[Obijuan83]

3 hours ago, Leeghoofd said:

Maybe to stop the current muppet show we can make it best 15 global 2D submissions and best 15 global 3D submissions next year for career and hardware ranking

This is the better solution from my point of view. 

We will never be happy with rules/point system etc...there must be a balance to cover the majority of members.

But dont forget that Hwbot is a "hardware overclocking" database. When first submitted an score back in late 2006 was excited for the existence of such a competitive database. We all trying to push to the limit cpus and Gpus with relatively same hardware. But its sad(almost shame) that too many Legendary overclocking 3D efforts and result are gone because now you simply buy an i7 12900 put it on LN2/Cascade/even AIO WITH  Gpu card @ stock and you are in top10... 

This is not competitive overclocking for me

[saltycroissant]

Thats why adding vantage/FS gpu score only is a step in the right direction to get more ppl into 3D.

[Leeghoofd]

26 minutes ago, Obijuan83 said:

This is the better solution from my point of view. 

We will never be happy with rules/point system etc...there must be a balance to cover the majority of members.

But dont forget that Hwbot is a "hardware overclocking" database. When first submitted an score back in late 2006 was excited for the existence of such a competitive database. We all trying to push to the limit cpus and Gpus with relatively same hardware. But its sad(almost shame) that too many Legendary overclocking 3D efforts and result are gone because now you simply buy an i7 12900 put it on LN2/Cascade/even AIO WITH  Gpu card @ stock and you are in top10... 

This is not competitive overclocking for me

Previous owner called it "evolution", I proposed back than a clock limit on modern CPUs on the legacy 3D... Guess a decade later that wasn't maybe such a bad idea ? Too little too late

[der8auer]

1 hour ago, saltycroissant said:

Thats why adding vantage/FS gpu score only is a step in the right direction to get more ppl into 3D.

if the hurdle is too high with obscure tweaks it will not help I think

 

1 hour ago, Leeghoofd said:

Previous owner called it "evolution", I proposed back than a clock limit on modern CPUs on the legacy 3D... Guess a decade later that wasn't maybe such a bad idea ? Too little too late

would've been interesting to limit GPUs with the available CPUs at the time

[saltycroissant]

10 minutes ago, der8auer said:

if the hurdle is too high with obscure tweaks it will not help I think

 

 

Well i think that vantage or firestrike arent the super tweakable benchmark if you consider that you only have tess/LOD, unlike 01se where you need a 2 page guide just to cover the basic.

[Aleslammer]

How the first post phrased the statement was rather interesting as it has nothing to do with what is done in 2D, If the intent was to run 2 cards and only one used and posted as two sounds pretty black & white to me. Then again why are global points available for benches that don't scale well with multiple cards, just makes the gray thinkers minds start to wonder.

Caught a statement about 2D making more points, actually does. 54.24% of HW points earned by the top 1000 from the HW masters list was based on 2D. 2D is cheaper to run, haven't looked at global points. Data set used about a month old now.

Bench tweaking, have read some pretty long winded SuperPi guides over the years, in-fact more involved than most of the 3Dmark01 guides I've stashed for reference.

[Guest hammertone]

On 7/5/2022 at 4:24 AM, Obijuan83 said:

We all trying to push to the limit cpus and Gpus with relatively same hardware. But its sad(almost shame) that too many Legendary overclocking 3D efforts and result are gone because now you simply buy an i7 12900 put it on LN2/Cascade/even AIO WITH  Gpu card @ stock and you are in top10... 

This is not competitive overclocking for me

Correct.

The legend overclockers efforts are diminished. Better cpu IPC. Even before Sandybridge. ie 2500K

Members new to oc want to try that, older sockets and going for gold. That is more challenge ?

[speed.fastest]

Just some thoughts, maybe if futuremark/UL not supporting older 3DMark, some wrapper like benchmate can be good idea? Plus this will ease for newcomers to set some tweaks.

[TerraRaptor]

1 hour ago, speed.fastest said:

some wrapper like benchmate can be good idea?

Please no. These wrappers are always a mess to make them work on older platforms. I.e. benchmate, after months and months of development, still has numerous issues with older gens (775 etc). Aquamark wrapper is also a good pile of bugs (though having some bells&whistles to ease benchmarking).