RealBench Challenge 2016 - Challenge II Discussion

[websmile]

Yes

[havli]

Well, I understand that... although funny thing is "Server HW" is forbidden in general to keep things affordable despite many Xeons hold no performance advantage over regular i7 and in fact are significantly cheaper. Like E5-1650 being exactly the same as i7 3930k, only having different writing on the IHS.

 

Also for example old Maxwell Titan X is not allowed while more expensive and faster GTX 1080 is ok.

Edited October 5, 2016 by havli

[jpmboy]

No, above 4200 is not allowed, it would not make sense to use lower frequency and then bump to 4200. Any particular reason you ask?

 

I think the issue Menthol refers to is jacking up the core after RB reads the SIO for frequency, then backing down to 4200 after the run...

 

Question: cache frequency must be =< 4200 also, correct?

[websmile]

Yes, cache=core. The max core frequency has to be 4200 or lower all the time

[jpmboy]

Yes, cache=core. The max core frequency has to be 4200 or lower all the time

 

Thanks for the reply.

[Wizerty]

I don't understand why some people already have +1 pt . I was top10 on oct4 ( 1st sumbition 1oct with 145k). even now (8oct) it still top8.

Can you plaese fix it

[websmile]

You have +1 point, the problem is the ranking is messed up - you are fourth which is also shown but ranked fifth and only get 21 points, same for methol for example.

Sigh...

[jpmboy]

6 core 2011 but no 6 core 2011-3. Rationale? Arbitrary ?

[websmile]

I already explained that we have four stages to cover all possible newer systems including amd. I see no rational reason to prefer sponsored 1700 dollar cpu ocers by allowing 2011-3 on a multi threaded low clock stage because there is no core limitation on this stage. The program is well balanced

[Calathea]

Invalid data file: Unable to parse the datafile

 

I guess I'm screwed. I hate benchmarks with validations files...

 

it shall be known that I tried, and reached 147k

 

Imgur: The most awesome images on the Internet

Edited October 10, 2016 by Calathea

[Rauf]

@Calathea Bench with internet connection and log in before saving file. Don't know if there are other ways to do it but that worked for me.

[Calathea]

@Calathea Bench with internet connection and log in before saving file. Don't know if there are other ways to do it but that worked for me.

 

Thanks Rauf, I'll remember that next time. Made the result this morning and then went to work so it's too late. I'm always benching last minute

[websmile]

There is no way to submit atm than bench with internet connection and directly submit. We are working at a fix to make save file and submit work, it is a problem in the program itself, the competition is really helpful to find stuff like this (and others) and improve realbench for the user and future points enabled - Sorry for inconvenience, your feedback is appreciated

 

@meankeys and @Slinky PC Your test 1 results are much too high for the given frequency and system used. I would strongly recommend to re-bench, because these will not stay at the competition for me, I will task these to head of moderation for second check but for me these are invalid - sorry, but plausibility check is part of the game not only for xtu

Edited October 10, 2016 by websmile

[Calathea]

There is no way to submit atm than bench with internet connection and directly submit. We are working at a fix to make save file and submit work, it is a problem in the program itself, the competition is really helpful to find stuff like this (and others) and improve realbench for the user and future points enabled - Sorry for inconvenience, your feedback is appreciated

Websmile, I wish this Internet connection thingie should of been mentioned in the competition rules. I'm not mad, but I did spend several hours on this comp only to be greeted by the "computer says NO" this morning. And that's a bad start on any morning.

 

That said, Realbench has grown into a pretty neat bench. I like it!

[websmile]

If the issue still remains for challenge 3, we will make a heads up at rules section or stage rules

[Wizerty]

@Calathea : sorry to heard this .

Yes, it should be add to rules, I was trap, my teammate was trap and pretty sur that many other also

If it's to difficult to fix it (I can understand that maybe it's not easy), then just ask to remove the option. No "save file" option -> no one will be trap anymore.

 

@Asus - Hwbot : if next round is on another rig can you please give us rules before the begginig of the contest. Ocer always have many rig , but not necessary ALL rig available on the market and not all EOL rig neither . I could be nice to have one week to find it This stage was unfortunate because X99 was "legit" and then it was remove . Tks, wiz

[websmile]

Those french guys, always complaining - you need a better hwstock, I offer nice x38 775 board at mp . We will see if announcement can be made quite early, next stage starts at 28th of october

[Wizerty]

french = complaining ? Don't know what you are talking about

(I'm not complaining, at least this time :wistle: only give you my feedback and hope to improve this contest . If I was not happy I won't be there). This contest is nice, even if it give me nightmare...

wiz

[websmile]

OK, I checked top20 preliminarily. For the results I removed with reason "bugged test 1", feel free to contact Christian Ney and offer video proof or appeal the removal, I took in a 2-3% margin for real good tweaking over other results and lucky runs, but when there are 10% over other top and average results with same systems (and these are the only results with this characteristics) I have no choice then to see plausibility check failed at given settings

[Menthol]

Not being able to load a saved file was discussed on the ASUS ROG forums at the start of the competition, sorry didn't think to bring it up here, there is a forum section on settings and tweaks on the same forum for those that want more information ( Wizerty. your giving me a nightmare on this) Congrats

Edited October 10, 2016 by Menthol

[Wizerty]

@Menthol : I spent ALL the week-end on this stage. Sunday I stop benching only 3hours for my running race (10km), then fu**** tired I had to start benching again, trying to catch meankeys. Your score is really good, nice subtest4 . wondering, do you use AMD SDK for openCL ? it give me 10k boost on subtest3 but average score is lower on subtest 2and4, wondering if you see it also ? And anyone know how to fix memory drop on opencl test ? with my 980Ti frequency drop from 2200 to 16xx. With my 1080 frequency drop from 1450 to 1250. this happen only with opnecl, on 2d or 3d benchmark gpu-z show full speed. but when opencl start gpu-z report big drop.

[Calathea]

And anyone know how to fix memory drop on opencl test ? with my 980Ti frequency drop from 2200 to 16xx. With my 1080 frequency drop from 1450 to 1250. this happen only with opnecl, on 2d or 3d benchmark gpu-z show full speed. but when opencl start gpu-z report big drop.

 

FYI I noticed this on my 970 too, dropped to ~ 1500 MHz. Although my experience from gpgpu (bitcoin ) is that memory clock does not affect performance.

[Wizerty]

Well memory help on RB, at least on 1080 seems to help. On 980ti I don't know because all OCing give 16xx MHz so I can't tell if it will be better or not. I have an XOC bios with 1750MHz but rig shutdown as soon as test start. I also tried Rx480 but score was under 100k and my strix fury was not available. OFF : 2-3y ago I also used many VGA for mining (LTC, dodge...). It was the heating system for my house (true story) during all the winter. I used 2-3 VGA per room , office, kitchen, lobby, living room... it was funny. mostly 290x and 280x. I don't think I earned money with this but at least it refund electricity -> free heating system

[Menthol]

using intel sdk runtime for cpu only, my subset 2 is always lower for some reason, I don't remember the memory drop on older drivers but also have it now

Edited October 10, 2016 by Menthol