HWBot Prime 2016 Tweak

[speed.fastest]

Hello, I want to share HWBot Prime Tweak to the next level performance. So i think will be alot of people re-benching

 

Check at my score efficiency with i3 6100 @ 4707MHz : speed.fastest`s HWBOT Prime score: 4893.54 pps with a Core i3 6100

 

The magic is JRE 9 64bit combined with Windows 10 x64

I hope this will easier to who that find HWBot Prime Tweak.

 

Edit: Build 108 seems like not working with HWBot Prime. Try build 107 or 106.

 

Edit 2 : Link added : JRE 9 EA Build 106

https://drive.google.com/file/d/0B2mrL4IrCzQEb3dCTlRmYTgyWlU/view?usp=sharing

 

Edit 3 : Link added : JRE 9 EA Build 107

https://drive.google.com/file/d/0B2mrL4IrCzQEcjZSUTFVTmRmMEU/view?usp=sharing

 

Added guide how to fix broken saving score function for JRE9 118 or above from @Spiedie

tl:dr https://dl.dropboxusercontent.com/u/45025086/hwbot/HWBotPrimeFor9.zip, extract and double click run.cmd

 

get activation.jar out of JAF 1.1 from Java Archive Downloads - Java Platform Technologies

get jaxb .jars from lib folder, downloaded from https://jaxb.java.net/2.2.11/

place jars in same folder (need hwbotprime-0.8.3.jar, activation.jar, jabx-api.jar, jaxb-core.jar, jaxb-impl.jar)

create file run.cmd or run.bat with contents:

start javaw -cp .\* org.hwbot.bench.BootStrap

double click run

 

if you still can't save datafiles, replace javaw with java to see the errors

Edited August 16, 2016 by speed.fastest

[zeropluszero]

It works on W7 as well, and is Skylake only as far as I'm aware.

[speed.fastest]

Yes it worked in win 7 x64 too. But in windows 10 scored more constant.

[Leeghoofd]

Yep testid it with Hsw-e and Ivy... no boost there

[Strunkenbold]

For me those scores looks buggy. The old behavior was, first run slow, second run faster by 20-40 points. Now with java9 the first run is at least 100 points faster than the second one and you simply cant reproduce the first one.

[websmile]

Why buggy? I mean it is different Java, and I tested and results were reproducable within margins after every reboot, frequency scaling was logical and I simply think as much as 9 changed results it might as well have changed the characteristics of rerun orgys. Sure it needs observation, but when software base changes, also characteristics of result optimization can change. Let´s simply watch it, for the start I think all works normal

[RULE]

Thanks for sharing man.

[speed.fastest]

You're welcome man. Enjoy it

[subaruwrc]

congrats on spoiling this one that fast.. lol

[speed.fastest]

Lol

 

Update : Only build 107 or 106 works. Let me upload it for you if you cant find it.

[raules009]

test g3258

version : 8.0.110.12

os : win7 64

4,7ghz 4228.16

 

test g3258

version : 9.0.0.53

os : win7 64

4,7ghz 4337

 

test g3258

version : 9.0.x.x

os : win7 64

4,7ghz 4429

[Guest Digg_de]

Tested on Win7 64Bit with my 24/7 Setting

 

My old Version

 

New Version (JRE 9)

 

[Spiedie]

This is what you get if you throw every java you can get at every system you can get your hands on. 5 runs per java version, 80+ different java versions. Ill let everyone else test skylake

 

Min 3583.77, max 3591.22, diff 7.45, ratio 0.21%, best jre8u31 i5-2400S

Min 3602.00, max 3608.58, diff 6.58, ratio 0.18%, best jre7u51 i7-2630QM

Min 7458.67, max 7499.40, diff 40.73, ratio 0.55%, best jre7u15 2xE5-2670

Min 4589.02, max 5390.52, diff 801.50, ratio 17.47%, best jre9u83 i5-3570k 4Ghz

Min 4919.96, max 4958.63, diff 38.67, ratio 0.79%, best jre7u76 i7-4712MQ

Min 1555.62, max 1592.32, diff 36.70, ratio 2.36%, best jre6u3 D925

Min 3989.60, max 4384.42, diff 394.82, ratio 9.90%, best jre6u20 G3258 4.7Ghz

Min 1639.32, max 1643.54, diff 4.22, ratio 0.26%, best jre8u51 G470

Min 1666.20, max 1698.01, diff 31.81, ratio 1.91%, best jre7u71 Atom Z3740D

[Leeghoofd]

test g3258

version : 8.0.110.12

os : win7 64

4,7ghz 4228.16

 

test g3258

version : 9.0.0.53

os : win7 64

4,7ghz 4337

 

test g3258

version : 9.0.x.x

os : win7 64

4,7ghz 4429

 

How many runs is this ? I tried 10 consecutieve runs on a 2600K, 3770K, 4770K & 5960x and didn't get any significant boost with the new 9 106-107 Java version versus my preferred ones ( version depending on the CPUs )... usually the new 9 version was even slower lol...

 

We are looking for a whopping boost of +400/500points as with Skylake...

Edited March 6, 2016 by Leeghoofd

[Guest Digg_de]

Now i've compared two Version of Java with Win10 64Bit. Hitting the "Quick Benchmark" Button 3 times.

 

JRE8u73/64

 

JRE9ea+106/64

[TX_OC]

Lol

 

Update : Only build 107 or 106 works. Let me upload it for you if you cant find it.

 

So that was the reason hwbot prime wasn't launching for me...

[speed.fastest]

@Digg_de Nice test! Identical to me, Windows 10 x64 is more constant & better score.

[FireKillerGR]

Tried by myself and its slower on X99 by 50 to 90 points

[speed.fastest]

So only work with skylake? Thanks for the test. because at the moment i dont have any other platfotm.

[MetalRacer]

I can confirm its working on Ivy-E with Win7 64bit.

Edited March 7, 2016 by MetalRacer

[GtiJason]

So Spiedie's best efficiency is with jre6u20 (Win7?)

Min 3989.60, max 4384.42, diff 394.82, ratio 9.90%, best jre6u20 G3258 4.7Ghz

 

And Raules had better luck with version : 9.0.x.x on W7 64

test g3258

version : 9.0.x.x

os : win7 64

4,7ghz 4429

 

I guess it's time to do some testing and see what I get

[Leeghoofd]

@MetalRacer: Interesting development your old java was the fastest one or did you use an 8 version ? What was the points increase from that version to the 9 one? Edited March 7, 2016 by Leeghoofd

[MetalRacer]

I was using 8 and then switched to 9. I compared the best of five scores between the two and version 9 was 300 points higher.

[speed.fastest]

Nice to see other platform get 300 points boost. I hope i can try to other platform.

[Guest Digg_de]

I'll try on s775 soon.

 

edit:

No boost with s775 here:

 

other JRE:

 

JRE9:

Edited March 8, 2016 by Digg_de