Updating my feeling

First all, sorry about my poor English.

People active in Hwbot for a long time, knows that I was very pro active here few time ago (maybe 2 years), then after Pro League was removed, I have participate very few in comparison to before.

You see me no more complaining... or even suggest...

Some of my suggestions, I think, never will be applied (for example, to allow ES CPUs, or even to worth points with 1, 2, 3, or even 4 videocards)... but also it`s not change my life, or even will not change the league.

The Hwbot are doing a great job. related to a new way, OC e-Sports, but I really cannot understand why there are nothing related to 3D and 2D.

The way we are now, is clear, 2D worth much more than 3D, and people that knows deep about OC, knows that`s not the truth.

Maybe is time to separate 2D and 3D League...

I don`t think I need to play XTU... or even Pifast... regardless I think is nice...

My focus is 3D and always will be.

 

Just my 2 cents...

Personally I'm a fan both 2d and 3d and I think it's good benches are good at both (but each to their own) and I know a lot of people see 3d as something that is easy and requires no skill and a lot of money although from my experience while the benchs aren't very tweakable it takes a certain skill to clock the cards as high as possible.

 

On term of points I personally think 2d benchs like XTU receive way too many points although that seems to be because of the extremely large amount of subs for it hwbot get

Have you ever thought that your feelings are very important to....you and no one else? Srsly if every user on hwbot would express their feelings in suggestion area it would be a huge spam here...

 

A complete bencher should be top qualified for 2d and 3d.I agree that 3d benchmarks should get an adjustment since it involves tweaking the whole platform not only cpu and ram .

 

Also people move away from 3d benchmarks due to insanely costs,playing with a multi-card setup nowdays is like buying a decent second-hand car which is quite stupid tbh.

 

To complete my absolutely uselless post , a little adjustment to 3d benches and also lowering the challenger series in value of competition points would be benefical to hwbot.

Edited September 16, 2015 by Alex@ro

Thanks for sharing, Ronaldo. I also saw some discussions spark at Facebook. To check the information, I went to the database to look up the following information: a list of results with the highest GL+WR points. This is what I found:

 

[table=head]GL+WR pts | benchmark | xCPU or xGPU | OCer | link

184.9 | Unigine Heaven - Xtreme Preset | 2 | Dancop | link

182.0 | 3DMark03 | 2 | der8auer | link

166.4 | XTU | 2 | 0.0 | link

166.4 | XTU | 4 | Splave.ROM | link

163.7 | HWBOT Prime | 4 | Splave.ROM | link

163.7 | XTU | 6 | Xtreme Addict | link

153.9 | Catzilla - 720p | 3 | der8auer | link

152.7 | 3DMark - Fire Strike | 4 | Dancop | link

150.8 | 3DMark Vantage - Performance | 4 | elmor | link

150.5 | CPU Frequency | 8 | The Stilt | link

149.6 | 3DMark11 - Performance | 4 | 8 Pack | link

149.6 | 3DMark - Fire Strike Extreme | 4 | k|ngp|n | link

148.1 | 3DMark03 | 1 | Dancop | link

144.5 | SuperPi - 1M | 4 | Chi-Kui Lam | link

139.7 | 3DMark - Fire Strike | 1 | k|ngp|n | link

139.5 | Catzilla - 1440p | 4 | 8 Pack | link

136.9 | SuperPi - 32M | 4 | Dancop | link

133.8 | GPUPI - 1B | 4 | 8 Pack | link

129.7 | HWBOT Prime | 2 | Strat | link

128.7 | Cinebench - R15 | 4 | Splave.ROM | link

127.4 | 3DMark11 - Performance | 1 | steponz | link

127.1 | Catzilla - 720p | 1 | steponz | link

126.7 | PiFast | 4 | dRweEz | link

123.3 | Aquamark | 1 | Dancop | link

122.1 | Unigine Heaven - Xtreme Preset | 3 | 8 Pack | link

121.8 | Cinebench - R11.5 | 4 | Splave.ROM | link

121.2 | Unigine Heaven - Xtreme Preset | 1 | Splave.ROM | link

120.8 | 3DMark - Fire Strike Extreme | 1 | k|ngp|n | link

119.7 | 3DMark Vantage - Performance | 1 | steponz | link

119.1 | GPUPI - 1B | 1 | sofos1990 | link

118.0 | Catzilla - 720p | 4 | 8 Pack | link

117.1 | 3DMark03 | 3 | TeamAU | link

116.8 | 3DMark06 | 1 | der8auer | link

115.9 | 3DMark05 | 1 | Dancop | link

115.7 | Catzilla - 720p | 2 | Rbuass | link

114.2 | XTU | 8 | Dinos22 | link

112.1 | Geekbench3 - Multi Core | 36 | Toolius | link

111.0 | GPUPI for CPU - 1B | 32 | dhenzjhen | link

110.6 | Cinebench - R11.5 | 2 | der8auer | link

110.2 | 3DMark2001 SE | 1 | elmor | link

109.9 | Cinebench - R11.5 | 60 | PreacherMan | link

109.9 | wPrime - 1024m | 60 | PreacherMan | link

109.9 | HWBOT Prime | 60 | dhenzjhen | link

109.2 | Cinebench - R15 | 2 | michel90 | link

108.7 | Cinebench - R15 | 60 | dhenzjhen | link

108.3 | wPrime - 32m | 2 | der8auer | link

108.1 | Geekbench3 - Multi Core | 4 | dRweEz | link

106.9 | 3DMark - Fire Strike | 2 | Dancop | link

103.5 | HWBOT Prime | 6 | Xtreme Addict | link

102.2 | 3DMark Vantage - Performance | 2 | Dancop | link

101.5 | GPUPI for CPU - 1B | 4 | Dancop | link

100.8 | Cinebench - R15 | 6 | Xtreme Addict | link

100.7 | HWBOT Prime | 8 | der8auer | link

99.8 | wPrime - 1024m | 2 | der8auer | link

96.7 | Cinebench - R15 | 8 | der8auer | link

96.1 | Geekbench3 - Multi Core | 8 | der8auer | link

95.4 | Cinebench - R11.5 | 6 | Xtreme Addict | link

94.6 | 3DMark - Fire Strike Extreme | 2 | Dancop | link

93.1 | 3DMark11 - Performance | 2 | Dancop | link

90.8 | Geekbench3 - Multi Core | 2 | ObscureParadox | link

90.0 | Cinebench - R11.5 | 8 | der8auer | link

88.2 | HWBOT Prime | 36 | rsnubje | link

87.7 | HWBOT Prime | 1 | ObscureParadox | link

85.6 | wPrime - 32m | 6 | BenchBros | link

85.1 | Aquamark | 2 | sofos1990 | link

84.9 | Cinebench - R11.5 | 40 | dhenzjhen | link

84.0 | Catzilla - 1440p | 1 | steponz | link

81.8 | 3DMark05 | 2 | sofos1990 | link

81.8 | wPrime - 32m | 8 | der8auer | link

81.4 | 3DMark06 | 2 | TeamAU | link

79.9 | GPUPI for CPU - 1B | 8 | der8auer | link

79.6 | wPrime - 1024m | 6 | nacho_arroyo | link

78.1 | wPrime - 1024m | 8 | der8auer | link

77.4 | GPUPI for CPU - 1B | 2 | der8auer | link

77.2 | wPrime - 32m | 1 | I.nfraR.ed | link

76.8 | GPUPI - 1B | 2 | sofos1990 | link

72.4 | Cinebench - R11.5 | 1 | I.nfraR.ed | link

71.7 | Geekbench3 - Multi Core | 6 | Xtreme Addict | link

71.0 | 3DMark11 - Performance | 3 | Wizerty | link

70.2 | Geekbench3 - Multi Core | 18 | sofos1990 | link

70.2 | Cinebench - R15 | 36 | rsnubje | link

69.2 | wPrime - 1024m | 48 | knopflerbruce | link

69.2 | Cinebench - R11.5 | 36 | rsnubje | link

68.2 | HWBOT Prime | 48 | tiborrr | link

67.9 | GPUPI for CPU - 1B | 6 | Xtreme Addict | link

67.1 | GPUPI for CPU - 1B | 48 | tiborrr | link

66.4 | 3DMark2001 SE | 2 | sofos1990 | link

65.6 | wPrime - 1024m | 1 | I.nfraR.ed | link

64.2 | Catzilla - 1440p | 2 | Rbuass | link

59.6 | Cinebench - R15 | 1 | I.nfraR.ed | link

58.9 | 3DMark - Fire Strike | 3 | Wizerty | link

55.2 | GPUPI for CPU - 1B | 18 | sofos1990 | link

54.2 | 3DMark - Fire Strike Extreme | 3 | Wizerty | link

53.9 | 3DMark Vantage - Performance | 3 | 8 Pack | link

53.2 | Cinebench - R11.5 | 48 | Patriot | link

52.7 | GPUPI for CPU - 1B | 1 | ObscureParadox | link

52.0 | Geekbench3 - Multi Core | 1 | Ananerbe | link

43.5 | GPUPI - 1B | 3 | sofos1990 | link

42.6 | 3DMark05 | 3 | mtech | link

42.5 | HWBOT Prime | 16 | FlashG | link

41.6 | 3DMark06 | 3 | 8 Pack | link

41.6 | Unigine Heaven - Xtreme Preset | 4 | 8 Pack | link

40.6 | Aquamark | 4 | sofos1990 | link

40.6 | 3DMark06 | 4 | 8 Pack | link

40.0 | Aquamark | 3 | mtech | link

40.0 | Catzilla - 1440p | 3 | Slinky PC | link

38.7 | Cinebench - R15 | 80 | clmns | link

37.1 | Cinebench - R11.5 | 64 | OVIZ Hardware Lab | link

36.8 | 3DMark05 | 4 | Smoke | link

36.0 | wPrime - 1024m | 32 | clmns | link

35.6 | Cinebench - R15 | 3 | subaruwrc | link

35.6 | HWBOT Prime | 3 | Moose83 | link

35.0 | 3DMark03 | 4 | Dinos22 | link

34.4 | wPrime - 32m | 3 | mtech | link

34.4 | Cinebench - R11.5 | 3 | subaruwrc | link

34.2 | wPrime - 1024m | 36 | 3dwallop | link

32.6 | 3DMark2001 SE | 3 | delly | link

32.6 | Cinebench - R15 | 12 | Ace123 | link

32.1 | Cinebench - R15 | 28 | Blackbadger | link

31.9 | HWBOT Prime | 5 | topdog | link

31.3 | XTU | 16 | DJForza | link

29.3 | Cinebench - R11.5 | 12 | loul | link

28.6 | 3DMark2001 SE | 4 | delly | link

28.6 | wPrime - 32m | 12 | k|ngp|n | link

27.9 | wPrime - 1024m | 3 | mtech | link

26.4 | HWBOT Prime | 12 | loul | link

26.4 | Geekbench3 - Multi Core | 3 | TheMadDutchDude | link

24.1 | wPrime - 1024m | 16 | CtrlFix | link

23.7 | Cinebench - R15 | 64 | OVIZ Hardware Lab | link

23.3 | Cinebench - R15 | 16 | FlashG | link

23.3 | wPrime - 32m | 16 | CtrlFix | link

22.5 | Cinebench - R11.5 | 16 | CtrlFix | link

22.5 | GPUPI for CPU - 1B | 3 | NoMS | link

20.8 | XTU | 1 | True Monkey | link

19.9 | Geekbench3 - Multi Core | 12 | spider220075 | link

19.0 | XTU | 12 | dhenzjhen | link

18.1 | Geekbench3 - Multi Core | 16 | dhenzjhen | link

16.2 | Cinebench - R15 | 24 | arobase | link

16.2 | XTU | 18 | S14 | link

16.2 | Geekbench3 - Multi Core | 5 | topdog | link

15.2 | HWBOT Prime | 10 | dhenzjhen | link

14.2 | Cinebench - R15 | 20 | dhenzjhen | link

14.2 | XTU | 32 | dhenzjhen | link

14.2 | wPrime - 32m | 24 | dhenzjhen | link

14.2 | Cinebench - R15 | 48 | rvborgh | link

14.2 | Cinebench - R11.5 | 5 | topdog | link

14.2 | HWBOT Prime | 20 | dhenzjhen | link

14.2 | GPUPI for CPU - 1B | 16 | dhenzjhen | link

14.2 | wPrime - 1024m | 24 | dhenzjhen | link

14.2 | HWBOT Prime | 32 | dhenzjhen | link

14.2 | wPrime - 32m | 5 | topdog | link

13.2 | wPrime - 32m | 36 | rsnubje | link

13.2 | Cinebench - R11.5 | 20 | dhenzjhen | link

13.2 | Cinebench - R11.5 | 24 | dhenzjhen | link

13.2 | wPrime - 1024m | 5 | topdog | link

13.2 | Cinebench - R15 | 18 | Sn0wMe | link

13.2 | GPUPI for CPU - 1B | 5 | topdog | link

13.2 | wPrime - 32m | 48 | OVIZ Hardware Lab | link

12.1 | Cinebench - R11.5 | 10 | dhenzjhen | link

12.1 | HWBOT Prime | 28 | dhenzjhen | link

12.1 | Geekbench3 - Multi Core | 48 | WesM63 | link

12.1 | HWBOT Prime | 24 | dhenzjhen | link

12.1 | Cinebench - R15 | 32 | dhenzjhen | link

12.1 | wPrime - 32m | 32 | dhenzjhen | link

12.1 | HWBOT Prime | 18 | NAMEGT | link

12.1 | wPrime - 1024m | 18 | Sn0wMe | link

12.1 | Cinebench - R11.5 | 18 | Sn0wMe | link

12.1 | Geekbench3 - Multi Core | 32 | dhenzjhen | link

12.1 | Cinebench - R15 | 10 | borandi | link

12.1 | wPrime - 32m | 18 | Sn0wMe | link

12.1 | GPUPI for CPU - 1B | 24 | dhenzjhen | link

12.1 | Cinebench - R15 | 5 | topdog | link

11.0 | wPrime - 32m | 28 | dhenzjhen | link

11.0 | wPrime - 1024m | 20 | utilizatori | link

11.0 | wPrime - 1024m | 10 | SteveRo | link

11.0 | wPrime - 1024m | 14 | borandi | link

11.0 | Cinebench - R11.5 | 28 | dhenzjhen | link

11.0 | XTU | 10 | legato3 | link

11.0 | wPrime - 32m | 10 | dhenzjhen | link

11.0 | wPrime - 32m | 20 | utilizatori | link

11.0 | Cinebench - R11.5 | 32 | dhenzjhen | link

11.0 | HWBOT Prime | 64 | dhenzjhen | link

11.0 | wPrime - 32m | 14 | borandi | link

11.0 | wPrime - 1024m | 28 | dhenzjhen | link

11.0 | Geekbench3 - Multi Core | 24 | dhenzjhen | link

9.9 | Cinebench - R15 | 14 | dhenzjhen | link

9.9 | Geekbench3 - Multi Core | 15 | dhenzjhen | link

9.9 | Geekbench3 - Multi Core | 40 | dhenzjhen | link

9.9 | Cinebench - R15 | 45 | dhenzjhen | link

9.9 | wPrime - 32m | 45 | dhenzjhen | link

9.9 | Cinebench - R11.5 | 14 | dhenzjhen | link

9.9 | Cinebench - R15 | 40 | dhenzjhen | link

9.9 | wPrime - 32m | 15 | dhenzjhen | link

9.9 | GPUPI for CPU - 1B | 14 | dhenzjhen | link

9.9 | Cinebench - R11.5 | 45 | dhenzjhen | link

9.9 | HWBOT Prime | 30 | dhenzjhen | link

9.9 | Geekbench3 - Multi Core | 14 | dhenzjhen | link

9.9 | wPrime - 1024m | 15 | dhenzjhen | link

9.9 | XTU | 64 | edgarallenphro123 | link

9.9 | GPUPI for CPU - 1B | 45 | dhenzjhen | link

9.9 | Geekbench3 - Multi Core | 10 | dhenzjhen | link

9.9 | GPUPI for CPU - 1B | 15 | dhenzjhen | link

9.9 | Geekbench3 - Multi Core | 60 | dhenzjhen | link

9.9 | Cinebench - R15 | 30 | dhenzjhen | link

9.9 | wPrime - 32m | 60 | dhenzjhen | link

9.9 | wPrime - 32m | 40 | dhenzjhen | link

9.9 | HWBOT Prime | 45 | dhenzjhen | link

9.9 | Cinebench - R11.5 | 15 | dhenzjhen | link

9.9 | HWBOT Prime | 15 | dhenzjhen | link

9.9 | wPrime - 1024m | 45 | dhenzjhen | link

9.9 | XTU | 30 | dhenzjhen | link

9.9 | GPUPI for CPU - 1B | 28 | dhenzjhen | link

9.9 | wPrime - 1024m | 40 | dhenzjhen | link

9.9 | GPUPI for CPU - 1B | 60 | dhenzjhen | link

9.9 | GPUPI for CPU - 1B | 40 | dhenzjhen | link

9.9 | Geekbench3 - Multi Core | 30 | dhenzjhen | link

9.9 | HWBOT Prime | 14 | borandi | link

9.9 | wPrime - 1024m | 64 | knopflerbruce | link

9.9 | Cinebench - R15 | 15 | dhenzjhen | link

9.9 | wPrime - 32m | 30 | dhenzjhen | link

9.9 | GPUPI for CPU - 1B | 20 | dhenzjhen | link

9.9 | XTU | 28 | dhenzjhen | link

9.9 | Geekbench3 - Multi Core | 20 | dhenzjhen | link

9.9 | HWBOT Prime | 40 | dhenzjhen | link

9.9 | wPrime - 1024m | 30 | dhenzjhen | link

9.9 | GPUPI for CPU - 1B | 30 | dhenzjhen | link

9.9 | Geekbench3 - Multi Core | 28 | dhenzjhen | link

9.9 | Geekbench3 - Multi Core | 45 | dhenzjhen | link

8.7 | wPrime - 32m | 80 | dhenzjhen | link

8.7 | wPrime - 1024m | 80 | dhenzjhen | link[/table]

The current situation seems to be the result of a natural progress to me. The majority seems to like 2D more than 3D. To compete seriously in the 3D Rankings you need a good 980Ti + X99 with 5960X. So the Setup requires more than 2000€ which is obviously way above the average gaming PC. However, you need only about 600€ to compete in the 4 Core 2D ranking with a 6700K and Z170.

The result is a low competition and activity in the X99+5960X+980Ti categories which is understandable.

 

edit: Nice info PJ. So there is not even an issue atm. 3D still gives a lot of points.

That's the most realistic approach. Also for some people you have to take into consideration the price of LN2,benching a 2d with a 6700K will consume far less than a 980TI+5960X which adds to the cost.

So there is not even an issue atm. 3D still gives a lot of points.

 

I wouldn't go as far as to say there is no issue. Because intuitively I think (personal opinion!) that single GPU benching is more difficult than single CPU benching. So, intuitively, I do think the single GPU categories are undervalued compared to the single CPU stuff. Especially considering how most CPU benchmarks nowadays come down to bin-bin-bin-to-win strategy.

 

A solution for this is not as simple as "just add more points", though.

I wouldn't go as far as to say there is no issue. Because intuitively I think (personal opinion!) that single GPU benching is more difficult than single CPU benching. So, intuitively, I do think the single GPU categories are undervalued compared to the single CPU stuff.

 

Thanks

 

My opinion is just an opinion...

 

Go back to old revision where just the most valuable submission each benchmark 3D (Single or 2 Way, 3 Way or even 4 way)...

Give more for new benches (Ok 3Dmark 01 is nice, but is EOL, since Windows XP have no support even from Futuremark)... So can add for example FS Ultra, 3DM11 Extreme and even, if is possible, think about gaming benchmarks for the future... if the reason is to grow up the community...

And for the last, can make more valuable subs not only for the Monster GTX 980Ti, but even for cards like GTX 750 (example)...

Well... I think this way, regardless some people think people bench 2D because is less expensive, "WHAT IS NOT TRUE", since CPU binning happens a lot and is no cheap... (ohoho,,, and even buy binning ES chip, that should be forbidden).

Again... for my way of view... respect who thinks different, but many more people join to 2D because is easier... and it's the first step for the overclocker...

You can consider Occker 3D the guy that buy a videocard and look for the factory limits... but in the truth, you need to go beyond factory limits.

The reason of the topic is... to have separate leagues...and if you think to be good overclocker need to run 2D and 3D... go ahead and be top ranked in both categories.

 

And pls... nothing about to tell "good overclocker is 2D and 3D" ,... nobody told nothing about this... or looks for someone think that will break 3DMark11 WR with the CPU and memory "stock".

 

Best wishes for all.

Edited September 16, 2015 by rbuass

As someone who benches mostly 2D but has started doing a bit more 3D recently (mostly for Team Cup), it's hard as hell... 3D is way more difficult than 2D. 2D is tweaking and putting in the time... 3D just has so much to learn and it requires absolutely incredible focus and practice to get pours right, etc.

 

Just look at this post by Vince describing benching the 980 TI Kingpin Edition, arguably one of the most well-rounded and complete cards of all time:

 

Sounds like solid testing procedure! Think maybe your loading volts are still on the low side, should be around 1.7v+ to hit the really high clocks

Next session try this..

-Leave both droop jumpers at default, and leave enabled on evbot. (this is 2nd from lowest level of droop and will give slightly more than your getting now)

-Pull card down to -65c and put about 1.4v set on the core with evbot, also about 1.82vmem.

-Time the ln2 pouring so it will run empty around -65c, so youll need to get it to -72c or so then let it go empty

- Use the torch to glaze and put a nice layer of frost over entire copper base of gpu pot, now splash some ln2 in there and pull it down again to -80c and have it go empty, now bump evbot to 1.5v.

- Use the torch again for glazing the base of gpu pot when Ln2 runs out, then splash more ln2 in the pot and pull the temps down again. Pot should be ready for any loading now.

-set 1.7v +400/+698 on the card and start the benchmark about -90c and maintain that temp for all of gt1, when gt1 is almost done, exit out back to desktop. This should give around 1850/2100

-set 1.75v +500/+698 on the card and start benchmark again about -95c and keep it between -95c and -100c and run whole benchmark. This should give around 1950/2100 in clocks btw

-Set 1.775v +550/+698 on the card and start benchmark -105c to -10c. This is about 2ghz/2100 and should run easy.

-Set 1.8v +573/+698 on the card and start benchmark about -110c. This will give 2030/2100

- Repeat raising the gpu at this point in around 26mhz increments at a time MAX..so the steps are smaller after 2g-2026-2055-2080-2106-etc.

With this droop setting, you card may take 1.85v set (1.72-1.73v loading) if u can keep it running that whole time without much crashes. Some cards can max out on this voltage/droopsetting around 2100. I think you will find 1.775-1.8v set with above droop settngs to be a nice sweet spot for max clocks.

Once you can leapfrog it quicker to 2000+ more, it will clock much nicer. The thermal grease freezes when you crash or sit idle for long time so the key point to the best results is to keep the card loaded and running always. Once you lose the grease, its obvious because not even 1900 will run. You can get it back tho by warming the pot to -50c or so then pulling it back down and trying again

 

All cards are different with regards to loading voltages and droop. Some like more droop, some less. This is why we have a wide range adjustment on kp980ti.

 

One other important note: Always bring the voltages up and down gradually on the card and the vrm will stay healthy and benching over and over again without issue. Some of my lab cards have literally hundreds of hours of benching on them at ridiculous voltages and speeds and no failures!. The card design is extremely robust, like a tank, but you still need to exercise care when applying such high voltages/currents.

Incrementally is always best.

 

 

Imagine trying to figure that out by yourself! And once you've figured that out... imagine actually trying to do it!

 

3D, especially multi-card 3D, has a steep barrier to entry if you're competing for globals. However, it's a part of this hobby and it's factually more difficult than 2D.

 

 

 

My only suggestion is lowering the submissions to points ratio for 3D... So say a 2D bench is worth 50 points after 500 subs... perhaps make a 3D bench worth 50 points after 100 subs. That way the points are still there, even if there are less subs.

 

 

 

Also, can we cap the points a single bench gets somehow? Some of these XTU subs are worth more than my entire profile which is absurd.

Edited September 16, 2015 by xxbassplayerxx

I agree with rbuass, Man benching the 980 ti kpe this weekend along with the 5960x really made me have a new found respect for the top results. it's not easy at all to push to the max. I had to keep my cpu at a safe speed just to try and concentrate on the gpu, the amount of ln2 I used and money spent was insane. I have a new found respect for the guys hitting 2100 with 980 ti's and titan x's.

 

And most of the top guys who bench 3d are basically benching 2d at the same time because they are maxing there cpu's out also and running physics tests. I mean steponz was hitting 5.9ghz during his 3d runs which is 2d clocking also, what is the difference between 3dmark physics and xtu, not much.

 

Also just to throw it out there I think the points given out for the oc esports stuff has really been excessive, the rankings feel weird now.

 

I think the points for 3d should be worth a lot more than 2d, but then I look at the chart Massman showed and it looks pretty equal, maybe there is a way to add a modifier for the amount of pots used for the submission, like steponz suggested.

Edited September 16, 2015 by Strong Island

DISCLAIMER: EXPERIMENTAL IDEA - THIS IS NOT PLANNED FOR PRODUCTION - JUST TESTING NEW IDEAS

 

I've been thinking about how to make the point algorithm more intuitively. From a technical perspective, the problem is that there are two parameters for calculating the points. These are: amount of competitors ("popularity") and position ("rank"). So the higher you ranking in a competitive category, the more points you will get. Neither of those parameters account for the "difficulty" of a certain benchmark.

 

I tried to make an algorithm which takes into account difficulty. I split up each global ranking (benchmark x core category) and counted how many LN2 pots were used for the top-50 results. Then I normalized this value across all rankings, so the "difficulty multiplier" is minimum 1 and maximum 2.

 

* IF GPU bench: sum(cpu ln2) + sum(gpu ln2)xGPUs + sum(mem ln2)

* IF CPU bench: sum(cpu ln2) + 0 + sum(mem ln2)

* Normalization: difficultyMP = totSumLN2 / max(totSumLN2)

* Points: currentPoints * difficultyMP

 

In the table below you can see what it would look like. Imho, from the looks of it I think it's a bad idea to measure the difficulty of a benchmark by the amount of people use LN2. We need a different method of defining difficulty.

 

* SuperPI 32M: in top-50, 47x CPU-LN2 and 23x MEM-LN2

* Fire Strike 1xGPU: in top-50, 24x CPU-LN2 and 22x GPU-LN2

* Aquamark 1xGPU: in top-50, 49x CPU-LN2 and 18x GPU-LN2 and 10x MEM-LN2

 

* Points for 4xCPU: currently 1371.4, would become 2149.9

* Points for 1xGPU: currently 1543.6, would become 2571.7

 

(click on the table headers to filter by column)

 

[table=head]Benchmark | Category | Difficulty | PointsNew | PointsCurrent

SuperPi - 32M|4|1.83|251.2|137

HWBOT Prime|4|1.52|249.3|163.6

SuperPi - 1M|4|1.65|239.3|144.6

XTU|6|1.45|237.8|163.7

Aquamark|1|1.92|236.5|123.4

CPU Frequency|8|1.57|236.3|150.4

3DMark03|1|2|236.2|118.1

XTU|2|1.38|229.8|166.4

3DMark05|1|1.96|227.7|115.9

XTU|4|1.31|217.9|166.4

3DMark - Fire Strike|1|1.55|216|139.6

3DMark2001 SE|1|1.95|215.2|110.2

3DMark Vantage - Performance|1|1.79|213.8|119.7

Cinebench - R15|4|1.65|213|128.7

PiFast|4|1.67|211.2|126.7

3DMark06|1|1.79|208.6|116.8

3DMark11 - Performance|1|1.63|207.9|127.5

HWBOT Prime|2|1.56|202.3|129.7

Cinebench - R11.5|4|1.6|194.3|121.8

Catzilla - 720p|1|1.5|190.8|127.2

Unigine Heaven - Xtreme Preset|1|1.56|189|121.2

3DMark - Fire Strike Extreme|1|1.49|179.8|120.8

Cinebench - R15|2|1.62|176.8|109.2

wPrime - 32m|2|1.6|172.8|108.3

Cinebench - R11.5|2|1.56|172.5|110.6

XTU|8|1.49|170.1|114.3

3DMark - Fire Strike|2|1.56|166.7|106.9

Geekbench3 - Multi Core|4|1.46|158.3|108.1

HWBOT Prime|6|1.52|157.7|103.5

Cinebench - R15|6|1.56|157.4|100.9

HWBOT Prime|8|1.54|154.6|100.7

3DMark Vantage - Performance|2|1.67|153.7|92.2

wPrime - 1024m|2|1.54|153.3|99.8

3DMark03|2|1.85|151.3|82

Cinebench - R15|8|1.55|149.5|96.6

Cinebench - R11.5|6|1.56|148.8|95.4

wPrime - 32m|4|1.63|145.8|89.4

3DMark - Fire Strike Extreme|2|1.52|144.2|94.6

Aquamark|2|1.68|142.8|85.1

3DMark05|2|1.73|141.2|81.8

3DMark11 - Performance|2|1.49|138.5|93.1

Unigine Heaven - Xtreme Preset|2|1.63|138.5|84.9

Cinebench - R11.5|8|1.54|138.1|89.9

3DMark06|2|1.69|137.6|81.4

GPUPI - 1B|1|1.14|136.2|119.2

wPrime - 32m|6|1.58|135.5|85.6

GPUPI for CPU - 1B|4|1.33|135.3|101.5

wPrime - 1024m|4|1.61|134.4|83.6

wPrime - 1024m|6|1.57|125.4|79.8

wPrime - 32m|8|1.51|123.4|81.6

wPrime - 32m|1|1.6|123.2|77.2

Geekbench3 - Multi Core|2|1.35|122.1|90.8

Catzilla - 720p|2|1.35|122|90.7

Catzilla - 1440p|1|1.36|114|84

wPrime - 1024m|8|1.45|113.3|78

3DMark2001 SE|2|1.61|106.7|66.4

HWBOT Prime|1|1.2|105.4|87.7

wPrime - 1024m|1|1.56|102.3|65.6

3DMark Vantage - Performance|4|1.87|94.9|50.8

GPUPI for CPU - 1B|2|1.21|94.2|77.6

GPUPI for CPU - 1B|8|1.35|94|69.9

Cinebench - R11.5|1|1.3|93.9|72.4

Geekbench3 - Multi Core|6|1.3|93|71.7

3DMark11 - Performance|4|1.87|92.7|49.6

3DMark - Fire Strike Extreme|4|1.77|88|49.6

Geekbench3 - Multi Core|8|1.26|83.8|66.4

GPUPI - 1B|2|1.04|79.8|77

3DMark - Fire Strike|4|1.51|79.2|52.4

GPUPI for CPU - 1B|6|1.15|78.4|67.9

3DMark Vantage - Performance|3|1.43|77|53.9

3DMark11 - Performance|3|1.36|76|56

Unigine Heaven - Xtreme Preset|3|1.61|75.7|47.1

3DMark - Fire Strike|3|1.27|75|58.9

Catzilla - 1440p|2|1.17|74.9|64.2

Cinebench - R15|1|1.2|71.7|59.6

3DMark - Fire Strike Extreme|3|1.27|69|54.2

Unigine Heaven - Xtreme Preset|4|1.57|65.4|41.6

Catzilla - 720p|3|1.18|63.5|53.9

3DMark03|3|1.45|61.1|42.1

3DMark06|4|1.5|60.9|40.6

3DMark05|3|1.42|60.4|42.6

3DMark06|3|1.36|56.5|41.6

Geekbench3 - Multi Core|1|1.05|54.5|52

GPUPI for CPU - 1B|1|1.01|53.3|52.7

3DMark05|4|1.44|53|36.8

Catzilla - 1440p|4|1.3|51.9|40

Aquamark|4|1.26|51.2|40.6

Aquamark|3|1.27|51|40

3DMark03|4|1.45|50.8|35

wPrime - 32m|3|1.48|50.8|34.4

Catzilla - 720p|4|1.18|50.7|43

Catzilla - 1440p|3|1.14|45.7|40

GPUPI - 1B|3|1.05|45.6|43.5

3DMark2001 SE|3|1.25|40.8|32.6

Cinebench - R15|3|1.14|40.7|35.6

HWBOT Prime|3|1.14|40.7|35.6

Cinebench - R11.5|3|1.14|39.3|34.4

wPrime - 1024m|3|1.38|38.5|27.9

GPUPI - 1B|4|1.07|36.2|33.8

3DMark2001 SE|4|1.23|35.1|28.6

HWBOT Prime|5|1.07|34.2|31.9

wPrime - 32m|12|1.14|32.7|28.6

XTU|16|1|31.9|31.9

Cinebench - R15|12|1.01|31|30.6

Cinebench - R11.5|12|1.01|29.6|29.3

Geekbench3 - Multi Core|3|1.02|27|26.4

HWBOT Prime|12|1.01|26.7|26.4

wPrime - 1024m|16|1|24.1|24.1

Cinebench - R15|16|1|23.3|23.3

wPrime - 32m|16|1|23.3|23.3

wPrime - 1024m|12|1.07|22.9|21.4

Cinebench - R11.5|16|1|22.5|22.5

GPUPI for CPU - 1B|3|1|22.5|22.5

HWBOT Prime|16|1|22.5|22.5

XTU|1|1|20.8|20.8

Geekbench3 - Multi Core|12|1|19.9|19.9

Geekbench3 - Multi Core|16|1|19|19

XTU|12|1|19|19

wPrime - 32m|5|1.19|16.9|14.2

Cinebench - R15|24|1|16.2|16.2

Geekbench3 - Multi Core|5|1|16.2|16.2

XTU|18|1|16.2|16.2

Cinebench - R11.5|5|1.11|15.7|14.2

wPrime - 1024m|5|1.17|15.4|13.2

Cinebench - R15|36|1|15.2|15.2

Geekbench3 - Multi Core|18|1|15.2|15.2

GPUPI for CPU - 1B|18|1|15.2|15.2

HWBOT Prime|10|1|15.2|15.2

HWBOT Prime|32|1.01|14.4|14.2

wPrime - 1024m|24|1.01|14.4|14.2

wPrime - 32m|24|1.01|14.4|14.2

Cinebench - R11.5|36|1|14.2|14.2

Cinebench - R15|20|1|14.2|14.2

Cinebench - R15|48|1|14.2|14.2

GPUPI for CPU - 1B|16|1|14.2|14.2

HWBOT Prime|20|1|14.2|14.2

wPrime - 1024m|36|1|14.2|14.2

wPrime - 1024m|48|1|14.2|14.2

XTU|32|1|14.2|14.2

GPUPI for CPU - 1B|12|1|13.6|13.6

Cinebench - R11.5|20|1|13.2|13.2

Cinebench - R11.5|24|1|13.2|13.2

Cinebench - R11.5|48|1|13.2|13.2

Cinebench - R15|18|1|13.2|13.2

GPUPI for CPU - 1B|5|1|13.2|13.2

HWBOT Prime|36|1|13.2|13.2

HWBOT Prime|48|1|13.2|13.2

wPrime - 32m|36|1|13.2|13.2

wPrime - 32m|48|1|13.2|13.2

Cinebench - R15|5|1.07|13|12.1

Geekbench3 - Multi Core|32|1.01|12.2|12.1

Cinebench - R11.5|10|1|12.1|12.1

Cinebench - R11.5|18|1|12.1|12.1

Cinebench - R11.5|64|1|12.1|12.1

Cinebench - R15|10|1|12.1|12.1

Cinebench - R15|28|1|12.1|12.1

Cinebench - R15|32|1|12.1|12.1

Geekbench3 - Multi Core|36|1|12.1|12.1

Geekbench3 - Multi Core|48|1|12.1|12.1

GPUPI for CPU - 1B|24|1|12.1|12.1

GPUPI for CPU - 1B|48|1|12.1|12.1

HWBOT Prime|18|1|12.1|12.1

HWBOT Prime|24|1|12.1|12.1

HWBOT Prime|28|1|12.1|12.1

wPrime - 1024m|18|1|12.1|12.1

wPrime - 32m|18|1|12.1|12.1

wPrime - 32m|32|1|12.1|12.1

Cinebench - R11.5|28|1.01|11.1|11

Cinebench - R11.5|32|1|11|11

Geekbench3 - Multi Core|24|1|11|11

GPUPI for CPU - 1B|32|1|11|11

HWBOT Prime|64|1|11|11

wPrime - 1024m|10|1|11|11

wPrime - 1024m|14|1|11|11

wPrime - 1024m|20|1|11|11

wPrime - 1024m|28|1|11|11

wPrime - 1024m|32|1|11|11

wPrime - 32m|10|1|11|11

wPrime - 32m|14|1|11|11

wPrime - 32m|20|1|11|11

wPrime - 32m|28|1|11|11

XTU|10|1|11|11

Cinebench - R11.5|14|1|9.9|9.9

Cinebench - R11.5|15|1|9.9|9.9

Cinebench - R11.5|40|1|9.9|9.9

Cinebench - R11.5|45|1|9.9|9.9

Cinebench - R11.5|60|1|9.9|9.9

Cinebench - R15|14|1|9.9|9.9

Cinebench - R15|15|1|9.9|9.9

Cinebench - R15|30|1|9.9|9.9

Cinebench - R15|40|1|9.9|9.9

Cinebench - R15|45|1|9.9|9.9

Geekbench3 - Multi Core|10|1|9.9|9.9

Geekbench3 - Multi Core|14|1|9.9|9.9

Geekbench3 - Multi Core|15|1|9.9|9.9

Geekbench3 - Multi Core|20|1|9.9|9.9

Geekbench3 - Multi Core|28|1|9.9|9.9

Geekbench3 - Multi Core|30|1|9.9|9.9

Geekbench3 - Multi Core|40|1|9.9|9.9

Geekbench3 - Multi Core|45|1|9.9|9.9

Geekbench3 - Multi Core|60|1|9.9|9.9

GPUPI for CPU - 1B|14|1|9.9|9.9

GPUPI for CPU - 1B|15|1|9.9|9.9

GPUPI for CPU - 1B|20|1|9.9|9.9

GPUPI for CPU - 1B|28|1|9.9|9.9

GPUPI for CPU - 1B|30|1|9.9|9.9

GPUPI for CPU - 1B|40|1|9.9|9.9

GPUPI for CPU - 1B|45|1|9.9|9.9

GPUPI for CPU - 1B|60|1|9.9|9.9

HWBOT Prime|14|1|9.9|9.9

HWBOT Prime|15|1|9.9|9.9

HWBOT Prime|30|1|9.9|9.9

HWBOT Prime|40|1|9.9|9.9

HWBOT Prime|45|1|9.9|9.9

HWBOT Prime|60|1|9.9|9.9

wPrime - 1024m|15|1|9.9|9.9

wPrime - 1024m|30|1|9.9|9.9

wPrime - 1024m|40|1|9.9|9.9

wPrime - 1024m|45|1|9.9|9.9

wPrime - 1024m|60|1|9.9|9.9

wPrime - 1024m|64|1|9.9|9.9

wPrime - 32m|15|1|9.9|9.9

wPrime - 32m|30|1|9.9|9.9

wPrime - 32m|40|1|9.9|9.9

wPrime - 32m|45|1|9.9|9.9

wPrime - 32m|60|1|9.9|9.9

XTU|28|1|9.9|9.9

XTU|30|1|9.9|9.9

XTU|64|1|9.9|9.9

Cinebench - R15|60|1|8.7|8.7

Cinebench - R15|64|1|8.7|8.7

Cinebench - R15|80|1|8.7|8.7

wPrime - 1024m|80|1|8.7|8.7

wPrime - 32m|80|1|8.7|8.7

GPUPI for CPU - 1B|36|1|8.3|8.3

wPrime - 32m|64|1|7.4|7.4[/table]

Edited September 17, 2015 by Massman

The result of the table kinda looks like before - just with a higher amount of total points. Not really a solution.

Difficulty is only relative to the user. 32m is easy for me. Single card fse more difficult.

More people benching 2d then award more points to 2d because there is more competition.

 

VGA expenditures is causing a rift in 2d vs 3d.

 

Also if you have deep pockets you can still bench air 4 way and get wr points no problem, how is that fair compared to someone working hard to push 2d on ln2.

 

I think it's fine the way it is now.

I usually don't bench purely for points, but looking from some opinions provided here I can understand the frustration of the 3D benchers (which I totally respect, I mainly does 2D as my current job is to test motherboards and RAM, but I'm aiming to be better 3D bencher as well).

 

A 'difficulty multiplier' as suggested could work, but as Pieter said:

 

Imho, from the looks of it I think it's a bad idea to measure the difficulty of a benchmark by the amount of people use LN2. We need a different method of defining difficulty.

 

As an idea, maybe the difficulty multiplier need to take into account:

- How the benchmark scales to non-hardware tweaks (OS choice/OS Tweaks, Driver Tweaks/adjustment, anything non-hardware related), in this case Pi 32M and 3DMark2001SE should be getting on the high side just because the sheer number of tweaks needed to be done to get it ran right.

- How the benchmark scales to some hardware parameters. Is the benchmark Really CPU-bound and doesn't really scale much to other components?(e.g like GPUPI & Hwbot Prime shows little scaling to memory compared to Geekbench/XTU). Or how Fire Strike Ultra is really GPU-oriented(when you do 1GPU) and does not scale that much according to cpu, compared to 3d11 1GPU when CPU and RAM also mattered a lot.

- How many hardwares need to be configured/monitored properly during an all-out bench session. In this case, anything 3-way/4-way is much more difficult compared to doing a single card session.

 

 

The hard part is, some benchmark's "difficulty' will also depend on the hardware. E.g:

Benching 3DMark Fire Strike on GT 730 or lower don't need that much attention to CPU(you can leave CPU on air), compared to you benching the same benchmark on GTX 980 Ti where the CPU start to give score scaling because the big GPU Score on GTX 980 Ti need much CPU Score to give an optimal score.

 

That's what i thought(sorry for the bad English),

One important point is also that Nvidia is locking the cards. There are not many users who like to epower a 1000$ card or solder on these. Much easier to play with a K-CPU and a Board with a nice BIOS.

As an idea, maybe the difficulty multiplier need to take into account:

- How the benchmark scales to non-hardware tweaks (OS choice/OS Tweaks, Driver Tweaks/adjustment, anything non-hardware related), in this case Pi 32M and 3DMark2001SE should be getting on the high side just because the sheer number of tweaks needed to be done to get it ran right.

- How the benchmark scales to some hardware parameters. Is the benchmark Really CPU-bound and doesn't really scale much to other components?(e.g like GPUPI & Hwbot Prime shows little scaling to memory compared to Geekbench/XTU). Or how Fire Strike Ultra is really GPU-oriented(when you do 1GPU) and does not scale that much according to cpu, compared to 3d11 1GPU when CPU and RAM also mattered a lot.

- How many hardwares need to be configured/monitored properly during an all-out bench session. In this case, anything 3-way/4-way is much more difficult compared to doing a single card session.

 

The hard part is, some benchmark's "difficulty' will also depend on the hardware. E.g:

Benching 3DMark Fire Strike on GT 730 or lower don't need that much attention to CPU(you can leave CPU on air), compared to you benching the same benchmark on GTX 980 Ti where the CPU start to give score scaling because the big GPU Score on GTX 980 Ti need much CPU Score to give an optimal score.

 

Thanks for your notes Alva, always thoughtful . Conceptualizing "difficulty" is already a big challenge, but the hardest part for us is to quantify it and make it measurable. We can only adjust our hwboints algorithm based on numbers.

 

I think we also need to make a good distinction between "difficult" and "dedication". A certain result may be difficult to achieve for one person, because it takes weeks of dedication to get to it, while for another person it just takes re-running the benchmark at higher clock frequencies. The difficulty of a benchmark should not be measured by "how much time it takes" to get a certain result, because that varies by silicon quality.

 

Take for example: a 30th place in a benchmark ranking requires a 5.2GHz Skylake CPU. For someone with LN2, this will be really easy and not difficult at all. But for someone with only water cooling, it will be very difficult. As for the LN2 guy, it's a matter of just pouring the LN2. But for the water guy, he'll need to figure out how to get the water as cold as possible, etc.

VGA expenditures is causing a rift in 2d vs 3d.

 

 

One important point is also that Nvidia is locking the cards. There are not many users who like to epower a 1000$ card or solder on these. Much easier to play with a K-CPU and a Board with a nice BIOS.

 

I would have to agree with both allen and roman. No competition in these rankings is mainly because of the price. Dropping $1k on a 980ti or modding a $1k titan is a lot of money. Even when they were $700-800 it was a lot but it keeps going up.

 

I see the lack of "reward" in terms of points is a direct result of the costs involved. If one were to look at the top of the leaderboards for single vga 3d (consisting of 980ti, titanx, etc), the only competition around is that which comes from vendors or people working closely with the vendors. The average enthusiast just doesn't have the means to compete.

 

On the other hand, 2d is more affordable and thus more submissions. You get a boost in points for those same guys working alongside the vendors or those who are working for the vendors.

Down the rabbit hole we go....

3D is hardest, and the most 2D Technical benchs, are under Windows XP.

Even I like, it's end of life.

I agree rbuass, and easier should be separate the 2D and 3D League.

If someone prefer 2D or 3D, go ahead.

Just checking the front page

 

Maybe implementing the 3D and 2D leagues as separate leagues like hardware masters, still conserving actual league as the main? Something like 3D Masters and 2D Masters... Same point calculation as now but only with 2D or 3D scores depending on ranking. Just an idea.

Would be interesting to are 2d master and 3d master.