HWBOT Rev7.1 point recalculation in progress

  • article
  • 64

The hwbot point calculation or revision 7 was wonky, and messed up the rankings under load. Now the bug are ironed out rev7.1, a full recalculation of approx 870.000 rankings is in progress. As the site is under quite some load from the day to day submissions and visitors, this takes a long time. In the meanwhile, please be patient. The full recalc was started 48 hours ago. At a rate of one ranking every 2 seconds, approx. 90000 have been recalculated up till now.


Belgium richba5tard says:

Currently 171200 / 870000 rankings recalculated.

Russian Federation temrus says:

please. answer me on my pm

Belgium richba5tard says:

check pm

Belgium richba5tard says:

199600 / 870000 done

United States Mr.Scott says:

This is gonna finish on April 1st.

Oh the irony. 

 

United States yosarianilives says:

Hi, I have 2 subs that are getting the wrongful pts. http://hwbot.org/submission/3724805_yosarianilives_wprime___1024m_sempron_150_18min_13sec_124ms and http://hwbot.org/submission/3622335_yosarianilives_gpupi_for_cpu___1b_xeon_e5_1607_14min_40sec_748ms I only mention because the wprime score just got it's pts today. Which means that something is still broken with the rankings, as that would suggest it got it's pts AFTER getting recalculated by the giant recalc you forced.

 

Also could you take a look at 1c x265? There seems to be a bug with people subbing things for "no pts no rank" then becoming part of the 1c rankings. If you look at 1c rankings here, http://hwbot.org/benchmark/hwbot_x265_benchmark_-_4k/rankings?cores=1#start=0#interval=20 it only shows 1c cpus. However on the main page for the bench this sub is shown as the first place for 1c http://hwbot.org/submission/3615254_samsarulz_hwbot_x265_benchmark___4k_core_i5_7600k_8.97_fps and because of that no actual 1c cpu gets global pts for x265 4k.

Czech Republic Sheppix says:

Global points are broken ...

United States yosarianilives says:

7 minutes ago, Sheppix said:

Global points are broken ...

There is currently a recalc going that's supposed to fix them, but it seems that it may not be completely successful as my sempron wprime score suggests.

Belgium leeghoofd says:

Richbastard should give a heads up when the recalc is completed, plz start reporting afterwards. The more people hit that recalc button, the longer it will take

United States yosarianilives says:

12 minutes ago, Leeghoofd said:

Richbastard should give a heads up when the recalc is completed, plz start reporting afterwards. The more people hit that recalc button, the longer it will take

Exactly why I haven't even been subbing anything until the smoke clears. However I brought this up because the wprime sub got it's pts in the MIDDLE of the whole site recalc, it didn't have pts a week ago. That means that it got it's pts AFTER being touched by the recalc. This needs to be looked into as it suggests that rankings are still fundamentally broken and that a whole site recalc may not fix them.


As for the issue with 1c rankings, I think that's a separate issue to what's been going on with rankings. It seems that basically if you sub something for x265 4k and you select no pts or rank it still gets ranked, as 1 core cpu. No matter the true core count. 

Austria basco says:

there should be more info on this whole recalc and submission problem. plz or is the bot leaving?

Belgium richba5tard says:

On 3/27/2018 at 2:25 PM, yosarianilives said:

Hi, I have 2 subs that are getting the wrongful pts. http://hwbot.org/submission/3724805_yosarianilives_wprime___1024m_sempron_150_18min_13sec_124ms and http://hwbot.org/submission/3622335_yosarianilives_gpupi_for_cpu___1b_xeon_e5_1607_14min_40sec_748ms I only mention because the wprime score just got it's pts today. Which means that something is still broken with the rankings, as that would suggest it got it's pts AFTER getting recalculated by the giant recalc you forced.

 

Also could you take a look at 1c x265? There seems to be a bug with people subbing things for "no pts no rank" then becoming part of the 1c rankings. If you look at 1c rankings here, http://hwbot.org/benchmark/hwbot_x265_benchmark_-_4k/rankings?cores=1#start=0#interval=20 it only shows 1c cpus. However on the main page for the bench this sub is shown as the first place for 1c http://hwbot.org/submission/3615254_samsarulz_hwbot_x265_benchmark___4k_core_i5_7600k_8.97_fps and because of that no actual 1c cpu gets global pts for x265 4k.

I see, it is indeed because of #1 to 5 is marked "no pts no rank". I'm implementing a fix.

Greece OGS says:

BTW what was the reason behind some 3x 3DMark Benchmarks getting 160+ points for a while (like last week) and now back to 125?
Asking cause it made/makes much more sense compared to 3x/4x 3d05/3d03 (that have gotten hw capped for years gpu-wise) etc points.

Belgium richba5tard says:

9 hours ago, FireKillerGR said:

BTW what was the reason behind some 3x 3DMark Benchmarks getting 160+ points for a while (like last week) and now back to 125?
Asking cause it made/makes much more sense compared to 3x/4x 3d05/3d03 (that have gotten hw capped for years gpu-wise) etc points.

I think so. The logic to find the top score against which a submission is compared was not always correct. As points are relative to top score, some where incorrect.

Belgium richba5tard says:

11 hours ago, richba5tard said:

I see, it is indeed because of #1 to 5 is marked "no pts no rank". I'm implementing a fix.

Fixed!

http://hwbot.org/benchmark/hwbot_x265_benchmark_-_4k/rankings?cores=1

If you spot other rankings where number #1 is someone who marked his submissions as "no pts" let me know.

United States yosarianilives says:

1 hour ago, richba5tard said:

Fixed!

http://hwbot.org/benchmark/hwbot_x265_benchmark_-_4k/rankings?cores=1

If you spot other rankings where number #1 is someone who marked his submissions as "no pts" let me know.

Will do! Also thanks rankings look much better now for that, also I noticed that the wrongful pts I had a few days ago for gpupi and wprime have gone away so the recalc appears to be working.

Greece TASOS says:

On 3/24/2018 at 10:46 PM, richba5tard said:

199600 / 870000 done

What's the status today ?

 

 

India ksateaaa23 says:

10 hours ago, richba5tard said:

Fixed!

http://hwbot.org/benchmark/hwbot_x265_benchmark_-_4k/rankings?cores=1

If you spot other rankings where number #1 is someone who marked his submissions as "no pts" let me know.

only no.1 rank fixes. x265 4k rank 2,3,4,5,6 is 0.6 and all should get the same no of points. but it's not.

similarly g470 xtu scores also only no.1 rank fixed. rank 2,3,4 is 80 & 5,6,7,8,9 is 79 and the same scores different points.

 

g470.PNG

xtu.PNG

Belgium richba5tard says:

1 hour ago, ksateaaa23 said:

only no.1 rank fixes. x265 4k rank 2,3,4,5,6 is 0.6 and all should get the same no of points. but it's not.

similarly g470 xtu scores also only no.1 rank fixed. rank 2,3,4 is 80 & 5,6,7,8,9 is 79 and the same scores different points.

 

g470.PNG

xtu.PNG

Seems correct to me. Rank 1 to 5 get bonus points.

United States yosarianilives says:

15 minutes ago, richba5tard said:

Seems correct to me. Rank 1 to 5 get bonus points.

Although I have nothing to worry as I'm not tied with anyone, for future reference how is 1-5 determined in the case of a tie? Sub date?

United States Splave says:

no. tie for first both gets first place points and new 2nd place guy gets 3rd place points I believe like xtu locked subs 

United States yosarianilives says:

22 minutes ago, Splave said:

no. tie for first both gets first place points and new 2nd place guy gets 3rd place points I believe like xtu locked subs 

That's not the case in the screenshot above, 2-5th get pts according to rank even though they have the same score. I just took the easy solution of beating everyone's score to avoid the chance of not getting full pts :D

United States Splave says:

then they changed it or its another bug :)

India ksateaaa23 says:

screenshots from the uat server x265 4k & xtu single core globals. gives full globals for ranks 2,3,4,5.......

anyway x265 v 2.2 can be the tie-breaker. will re-run and try to do 0.601+........ 

xtu will remain as a problem. not only for one core but for many non-k chips.....

 

x265 uat.JPG

xtu uat.JPG

Greece George_oc says:

17 hours ago, yosarianilives said:

Although I have nothing to worry as I'm not tied with anyone, for future reference how is 1-5 determined in the case of a tie? Sub date?

Damn dude, I haven't been following hwbot for the past 2-3 months due to work, haven't benched anything! I was amazed though that you managed to get 81 points in XTU and that f*cking 0.61 in x265 :D I must have had like 30 runs at 80 marks for XTU, I tried everything! In the end I told my self that it was not possible to get a better score with 109MHz bclk, you would need even more. You've found the ultimate G470, congrats, nice runs :) My G470 unfortunately only works with single channel now, so I imagine the IMC got hit hard by all this pushing I did :P 

United Kingdom GeorgeStorm says:

10 hours ago, Splave said:

then they changed it or its another bug :)

 

8 hours ago, ksateaaa23 said:

screenshots from the uat server x265 4k & xtu single core globals. gives full globals for ranks 2,3,4,5.......

anyway x265 v 2.2 can be the tie-breaker. will re-run and try to do 0.601+........ 

xtu will remain as a problem. not only for one core but for many non-k chips.....

 

x265 uat.JPG

xtu uat.JPG

Whilst you do get equal points, there's a bonus for places 1-5 for global scores, so your total points will be great for 2nd even if scored the same as 5th for example.

Ranks are done by submission date I believe.

United States yosarianilives says:

30 minutes ago, George_o/c said:

Damn dude, I haven't been following hwbot for the past 2-3 months due to work, haven't benched anything! I was amazed though that you managed to get 81 points in XTU and that f*cking 0.61 in x265 :D I must have had like 30 runs at 80 marks for XTU, I tried everything! In the end I told my self that it was not possible to get a better score with 109MHz bclk, you would need even more. You've found the ultimate G470, congrats, nice runs :) My G470 unfortunately only works with single channel now, so I imagine the IMC got hit hard by all this pushing I did :P 

Lol, was the first run for me on that chip, but was the 4th chip. The secret for me was just out clock everyone :P Was running 109.9 and some tight timings on my psc. Didn't even use the optimal OS for xtu. I actually binned on x265, which was awful to bin on with 40 minute runs. I just ran xtu as an after thought while I ran everything else. 

France bolc says:

4 hours ago, GeorgeStorm said:

 

Whilst you do get equal points, there's a bonus for places 1-5 for global scores, so your total points will be great for 2nd even if scored the same as 5th for example.

Ranks are done by submission date I believe.

date will sort out yes, but since when there is a bonus for first 5 places...? :cry:

global looks ok on the ua for x265 4k, but for xtu, there is obviously a pb with scores, why would #8 score less than 9 or 7 ?

United Kingdom GeorgeStorm says:

There's been a bonus on the top3/5 for a while, just if people ranked the same they'd all be given the bonus type thing as far as I'm aware, whereas now things have changed to try and avoid issues with hundreds of people getting '1st' and reaping the high points associated with it. Gaming the system so to speak.

 

Not an official view but that's my thinking :)

France bolc says:

Luumi and I submitted on 1/25 both 0.6 fps, I believe my screenshot (3am) is earlier than  his (8.30 morning or evening) but I am #6 and he is #5 :D
we don t see time stamp though for the submission, I guess I submitted right away :D
http://hwbot.org/submission/3768609_luumi_hwbot_x265_benchmark___4k_celeron_g470_0.6_fps
http://hwbot.org/submission/3769054_bolc_hwbot_x265_benchmark___4k_celeron_g470_0.6_fps

United Kingdom GeorgeStorm says:

Maybe @richba5tard can comment on that :P

France bolc says:

http://hwbot.org/benchmark/hwbot_x265_benchmark_-_4k/rankings?start=0&cores=1#start=0#interval=20
 

as of #7 and 0.59 fps, all 0.59 get 28.5 pts. glad I worked so hard for 0.6... given I had 0.6 at 12.19 AM :cry:

United States yosarianilives says:

12 minutes ago, bolc said:

http://hwbot.org/benchmark/hwbot_x265_benchmark_-_4k/rankings?start=0&cores=1#start=0#interval=20
 

as of #7 and 0.59 fps, all 0.59 get 28.5 pts. glad I worked so hard for 0.6... given I had 0.6 at 12.19 AM :cry:

.61 wen? :P

France bolc says:

I am left to do 0.6051, but given you did 0.606ish with 109.9 bclk...? I am screwed :D

or hope 2.2 can do miracles :D

United States yosarianilives says:

27 minutes ago, bolc said:

I am left to do 0.6051, but given you did 0.606ish with 109.9 bclk...? I am screwed :D

or hope 2.2 can do miracles :D

I didn't use it at the time, but in later testing I found that the sandy no avx preset may improve performance. Also I didn't use hypers cause I didn't think they'd work, but now I've seen with some better testing that the preset is just crap for hypers and if you all manual they might work some wonders.

France bolc says:

16 hours ago, yosarianilives said:

I didn't use it at the time, but in later testing I found that the sandy no avx preset may improve performance. Also I didn't use hypers cause I didn't think they'd work, but now I've seen with some better testing that the preset is just crap for hypers and if you all manual they might work some wonders.

good to know, I might give a try at 2280+ cas8, but 8 10 8 24 1t with tight subs with psc, against 8 8 8 24 but perhpas with losser subtimings. will see :D

United States yosarianilives says:

1 hour ago, bolc said:

good to know, I might give a try at 2280+ cas8, but 8 10 8 24 1t with tight subs with psc, against 8 8 8 24 but perhpas with losser subtimings. will see :D

Well hyper will be hard to get to higher mhz like 2280, I'm only considering it cause once you get to 109.9 you have to use the 1866 strap which gets me 2050 mhz, where hypers could shine.

France bolc says:

5 hours ago, yosarianilives said:

Well hyper will be hard to get to higher mhz like 2280, I'm only considering it cause once you get to 109.9 you have to use the 1866 strap which gets me 2050 mhz, where hypers could shine.

my kingston hyperx with hyper can do
image.png.b49b55aa6446aef1d2e1fbd8df5c1376.png
(voltage for each stick) so no issue to get 2300 888 24 1t if the imc wants ;)

 

i am more afraid that subtimings will be harder to tighten on hyper than on psc, and I have no bbse capable of similar results, hum,  let me see

I have these crazy bbse that do
image.png.771c617cd679c8aaebc26e2c6aff3ceb.png

didn t know they could go to 2666 cas9 :D might do 2280 cas8.... to check one day

Belgium richba5tard says:

All rankings have been recalculated by yesterday morning. I did some spot checks and did not find any which look odd. Does anyone else find any?

United Kingdom GeorgeStorm says:

I thought things may have been finished as suddenly my rank etc seemed more in line to what it was before the update.

 

Glad to see things are back up and running :D

Noxinite says:

My 32M globals were still broke earlier, but seem fine now.

I also seem to be getting WR pts on subs that get 0 globals??

https://hwbot.org/submission/2915203

Belgium richba5tard says:

31 minutes ago, GeorgeStorm said:

I thought things may have been finished as suddenly my rank etc seemed more in line to what it was before the update.

 

Glad to see things are back up and running :D

Hooray, a lot of effort for status quo! Everybody happy! O.o

Italy hardmad says:

2 hours ago, richba5tard said:

Hooray, a lot of effort for status quo! Everybody happy! O.o

Are Road to Pro Points / 2 included in this ? I've been stuck at 0 since the change began
http://hwbot.org/user/gtijason/

United States niobium615 says:

Overall the rankings are looking a lot better, thanks rich!

The 2C XTU global rankings still seem to be a bit messed up though(lots of subs getting 38.8 global points) : http://hwbot.org/benchmark/xtu/rankings?start=100&cores=2#start=100#interval=20

Finland Tame says:

9 hours ago, richba5tard said:

All rankings have been recalculated by yesterday morning. I did some spot checks and did not find any which look odd. Does anyone else find any?

Take for example Fire Strike global points, if you scroll down enough the rankings (in rank 100+ range) you suddenly start encountering 38.6 gl points. Looks like there's still some calc to do.

Australia jordan.hyde99 says:

http://hwbot.org/submission/3485313_jordan_hyde99_3dmark03_geforce_gtx_580_107823_marks This doesn't seem right?

Belgium richba5tard says:

7 hours ago, niobium615 said:

Overall the rankings are looking a lot better, thanks rich!

The 2C XTU global rankings still seem to be a bit messed up though(lots of subs getting 38.8 global points) : http://hwbot.org/benchmark/xtu/rankings?start=100&cores=2#start=100#interval=20

Ah yes, submissions ranked 100 to 200 have not explicitly been refreshed. Doing this now...

stivut says:

Submissions with fake global points:

http://hwbot.org/submission/2334234_anderson6060_3dmark2001_se_geforce_gtx_580_117894_marks

http://hwbot.org/submission/3762089_adventurepo_cinebench___r15_core_i7_8700k_1699_cb

http://hwbot.org/submission/2353756_stu_a_msd_3dmark11___performance_2x_radeon_hd_7970_18248_marks

http://hwbot.org/submission/2355817_stu_a_msd_3dmark___fire_strike_radeon_hd_7970_8731_marks

http://hwbot.org/submission/2355820_stu_a_msd_3dmark___fire_strike_extreme_radeon_hd_7970_4254_marks

http://hwbot.org/submission/2412349_stu_a_msd_xtu_core_i7_3770k_942_marks

http://hwbot.org/submission/3471106_deocer_gpupi_for_cpu___1b_core_i7_7700k_5min_16sec_268ms

http://hwbot.org/submission/3471346_deocer_wprime___1024m_core_i7_7700k_2min_6sec_887ms

http://hwbot.org/submission/3471053_deocer_pifast_core_i7_7700k_12sec_440ms

http://hwbot.org/submission/3801276_avalanche_3dmark___fire_strike_geforce_gtx_970_11558_marks

http://hwbot.org/submission/3714371_avalanche_memory_frequency_ddr4_sdram_2201.2_mhz

http://hwbot.org/submission/3486311_iji_superpi___32m_core_i7_7700k_5min_58sec_936ms

http://hwbot.org/submission/3493678_iji_wprime___1024m_core_i7_7700k_2min_4sec_86ms

http://hwbot.org/submission/2474679_atisoc0936_3dmark03_radeon_hd_6950_89892_marks

http://hwbot.org/submission/2474681_atisoc0936_3dmark05_radeon_hd_6950_37984_marks

http://hwbot.org/submission/2844405_nata_58_wprime___1024m_core_i7_4790k_2min_24sec_952ms

http://hwbot.org/submission/2825166_nata_58_3dmark2001_se_geforce_gtx_580_129876_marks

...and many many others that put their owners in higher position in league rankings.

The rankings are still messed up.

Belgium richba5tard says:

Read my post above, if hardware rank is >100 points are not yet recalculated.

United Kingdom tom.randle.10 says:

im gunna say something here, no offence to anyone but this is getting out of hand, do you think these guys are not trying to fix this problem on BOT>>>>?????.  why the hell dont you all just shut the fuck up, wait and let them do their job theyre tryin to do to get it right, ????, richbastard. and all the others in BOT  that give their time . effort, for nothing, you ungrateful pricks, ) this is a free show here) and have been left with a heap of shit to clean up, yes its going to take time, so stop all the bullshit with and why isnt my scores or points been calculated and done yet, the more you fools do this , the more time it takes for them to look at all your whinging and deal with that, then get on with doing what theyre trying to do for every one,  and fot bot, try to get things in order, deal with CPUZ' s  crap. and all the the rest of the other benches that arent up to scratch, no fault of BOT, they just gotta deal with it for who ??? YOU that want to bench who and when what you want to, so get off your high horses and get down to ground level again, become real and give the guys a go at it, all this bullshit youre going on with is just making things take longer for your benefit, so support it , let them do it right and get it done, is it gunna kill ya not to see "your" points there ??????????? or are you that VAIN, that you need to see that in  BOT to make you feel worthwhile, if so you got big problems already , and nothing to do with BOT ??????!!!!!! 

and for you blokes that dont know what i and it means

The best predictor of future behaviour is past behaviour...unless you make change !

 

Canada P_6_Canuck says:

Your points are well taken Tom, we the greater community are mostly either ignorant or incidentally take for granted the huge amount of work that goes on behind the scoreboards to keep the Bot up, running and growing. It's no doubt a lot of real work, that requires knowledgeable expertise, determination and patience that I suspect isn't adequately, if at all appropriately enumerated. Perhaps we can all take a moment and thank all those past and present who have and continue to labor making this 24/7, 365 days a year, global venue possible, and for any who have been bitten by the OC bug a hell of a lot of fun... SOOO a big sincere thanks to you all. That said I would like to suggest for future reference; if there is going to be a big and prolonged overhaul of the engine, take a little more time and make a concerted effort to adequately explain it to the community. The more detailed the better, and considerably more than has currently been the case might nip a range of problems in the bud. State a schedule, even if it has to be revised, underscore that members ranking will potentially be fluctuating wildly as the changes progress throw the system. My own have in the course of this update nearly quadrupled at a point (that was thoroughly entertaining) before finally settling down about where I would say is a factually accurate. Anyway my 2 cent is now spent, Cheers Bots, thanks for your hard work.

Belgium richba5tard says:

Thanks grantgibson for the support!

I'm on holiday at the moment so I was not able to check the recalc for submissions with a rank > 100 properly, and it seems to have quit shortly after I pushed the button a few days ago. Darn! I've rescheduled it again and it seems to work fine currently. So to summarize, all submissions with a hardware rank <= 100 are properly recalculated for rev7.1, but if the hw rank is > 100 it is still being calculated.

Russian Federation Antinomy says:

Strangely enough, UCbench results give points though the whole benchmark was removed from HWBot.

http://hwbot.org/benchmark/ucbench_2011/rankings?cores=2#start=0#interval=20

Greece TASOS says:

On 4/5/2018 at 11:00 PM, richba5tard said:

.... So to summarize, all submissions with a hardware rank <= 100 are properly recalculated for rev7.1, but if the hw rank is > 100 it is still being calculated.

 

On 4/6/2018 at 12:54 AM, Antinomy said:

Strangely enough, UCbench results give points though the whole benchmark was removed from HWBot.

http://hwbot.org/benchmark/ucbench_2011/rankings?cores=2#start=0#interval=20

 

Frederik

Any updates on these ?

Belgium leeghoofd says:

He's on holiday till today... he will get back to stuff

Belgium richba5tard says:

9 hours ago, TASOS said:

 

 

Frederik

Any updates on these ?

Yes, I'm testing the code on the uat server.

France bolc says:

Hi :)

Has the dust settled done, are all HW and GL now properly calculated?

 

I have a question on the side. If someone has mis-submitted a score placing him, say 5th. When he deletes/edits his sub, will previous #6  7 etc becoming #5 #6 etc, get more points than what they had ?

Cheers

++

Aleslammer says:

Have multiple places where this Cat is not working right. 

1-3503 WP32.jpg

Strunkenbold says:

Is there a recalculation planned for results which dont appear in a ranking? There were quite some results which got not properly synced to a ranking.

Belgium richba5tard says:

3 hours ago, Strunkenbold said:

Is there a recalculation planned for results which dont appear in a ranking? There were quite some results which got not properly synced to a ranking.

Can you give an example?

Strunkenbold says:

3 hours ago, richba5tard said:

Can you give an example?

http://hwbot.org/submission/3807142_k.j._superpi___1m_core_i7_8700k_6sec_219ms

http://hwbot.org/submission/3757303_alexwu_superpi___1m_core_i7_8700k_5sec_679ms

http://hwbot.org/submission/3783335_pyon400_superpi___1m_core_i7_8700k_6sec_272ms

 

Could the engine trigger a recalculation for results with better score than the actual best? 

Belgium richba5tard says:

If i trigger recalculate now than the better submission is ranked properly. I think there was a period of time in March where the best submission logic did not work wel. Does this still occur with submissions made in April?

Strunkenbold says:

In my experience, the rank synchronization is now working better than it ever did in the past years. 

But in the last 3 months it was pretty non functional for some benchmarks. Like 3D mark firestrike was working quite good but superpi or 3dmark 03 needed manual recalculation. 

Seems this is a problem since the beginning. I see 10 year old submissions floating around which never appeared in a ranking. 

Strunkenbold says:

Have to take this back.

Still seeing results which get not synced. Really wondering whats the problem. SuperPi 1M seems to be problematic.

http://hwbot.org/submission/3836738

http://hwbot.org/submission/3837231 <-- I tried to recalculate this

Please log in or register to comment.