Go Back   HWBOT forum > HWBOT Information > Benchmark Rules > Rules Updates

Rules Updates Discussion about ongoing/future rules updates

Reply
 
Thread Tools Display Modes
  #51  
Old 04-19-2017
Antinomy Antinomy is offline
HWBOT Staff
 
Join Date: May 2007
Location: Vladivostok, Russia
Posts: 1,424
Thanks: 28
Thanked 33 Times in 19 Posts
Antinomy
Send a message via Skype™ to Antinomy
Default Re: How graphics cards categories are determined or what makes for a new category

Quote:
Originally Posted by K404 View Post
would RX570 and RX580 get lumped in with 470 and 480?
Yep, exactly my point. Merging flavors of one model seems O.K. for me but mixing different models doesn't seem too good.

As for memory size - I've looked through NV and it seems there's only 10-20 categories all before Geforce 9. I'll comply a list so others can see how fine is it to merge each one of them.
I'd like to remind that HWBot is about a hardware performance database (including overclocking) and competition and hwboints being second. If we mix different hardware models, it kinda kills the initial idea.
__________________
Quote:
Originally Posted by Turrican View Post
Yes, Alexander is right, as always.

Last edited by Antinomy; 04-19-2017 at 00:28.
Reply With Quote
The Following 3 Users Say Thank You to Antinomy For This Useful Post:
macsbeach98 (04-26-2017), Mr.Scott (04-19-2017), websmile (04-19-2017)
  #52  
Old 04-19-2017
Strunkenbold Strunkenbold is offline
robot overlord
 
Join Date: Apr 2009
Location: GERMANY
Posts: 1,355
Thanks: 155
Thanked 316 Times in 253 Posts
Strunkenbold is on a distinguished road
Default Re: How graphics cards categories are determined or what makes for a new category

A rebrand is not different hardware for me. But it would be okay if we would have a chance to see something different with GPU-Z. There might be small differences like the new RX 580 look like to achieve higher clocks but shouldnt that do a 80nm 7600GT vs a 90nm 7600GT also? Yet I see them being all at the same performance level. So shouldnt the last RX 480 that came out of the fab not also being built with the new "matured" manufacturing process? This all looks like a process where the first cards clocked bad and the last produced did more and more better.
I didnt see GPU-Z screenshots with the new official driver. If AMD did like with the 7970 / 8970 / 280X, we will have to merge. If AMD choose to give the cards a new unique deviceIDs, they stay separated.

Id like to argue that we are not a hardware database. If we would be, we had made different decisions in the past. Like the Mars had its own category. Separation of die shrinks and so on. Yet we did not. We have the problem that we have to look how things get recognized by software. And we have somehow to deal with that. We have to draw the line somewhere.

I also like to mention that no one can take any performance data from hwbot. We allowed tessellation tweaks, LOD and so on. So for me, hwbot and especially GPUs are only about Boints. Because if we would make the db as exact as possible we would have a fine db but "without any users", just to quote someone from the Mars thread.

Edit: RX 480 and RX 580 has again the same DeviceID but different SubsystemID. Still GPU-Z makes a difference. Wonder if thats a change in GPU-Z or the driver makes differences here.
__________________

Last edited by Strunkenbold; 04-19-2017 at 12:39.
Reply With Quote
  #53  
Old 04-19-2017
K404's Avatar
K404 K404 is offline
nVidia FTW!
 
Join Date: Jan 1970
Location: Glasgowvic, Scotistan. Average Kingdom of Un-United Britain..
Posts: 3,855
Thanks: 16
Thanked 146 Times in 85 Posts
K404
Default Re: How graphics cards categories are determined or what makes for a new category

As PJ has stated a few times, HWB only makes rules that it can enforce. The issue technically wasn't about allowing LOD... but about being powerless to spot/stop it.

I think i'm quoting PJ again here.... he once said that the smart cheat only cheats a little bit (apologies if he didn't, or if i've misquoted)

Subtle tweaking falls in-line with good efficiency....and Windows has too many variables.

HWBot is all about making a big deal out of a 1% difference in performance...when in the real-world...gaming, coding etc... no-one would notice. Yesterday, I was mocking a certain brands video card release material because it mentioned a 1.5% overclock......

I think this post is pointless... but the question still is... what does HWB want to achieve?
__________________
Quote:
Originally Posted by Hondacity View Post
i believe in K404! :P
https://twitter.com/K404ExtremeOC

I do not believe in 1st place by any means, at any cost.

Quote:
Originally Posted by der8auer View Post
Knowing workarounds to get a benchmark to run is not a skill. It's a combination of knowledge and wasting time.
Reply With Quote
  #54  
Old 04-19-2017
buildzoid's Avatar
buildzoid buildzoid is offline
robo cop
 
Join Date: Jan 1970
Location: United Kingdom
Posts: 371
Thanks: 143
Thanked 62 Times in 39 Posts
buildzoid
Default Re: How graphics cards categories are determined or what makes for a new category

Uh I think RX 480s and 580s might be able to submit in both categories: https://www.techpowerup.com/232498/r...?cp=2#comments

Since you can just flash them to look the same as far as I can tell at least.

Also what about flashing a Fury X to look like a Fury? I don't think that would show any difference in GPU-z either.

Honestly if things like a 280X and 7970 got lumped together I would welcome it as it would give a boost to the value of benching HD 7970s or 280Xs. But I don't really have a very strong opinion on this stuff since I mostly bench stuff based on my own interests anyway.
__________________
Keep calm and raise Vcore


My Blog
Reply With Quote
The Following User Says Thank You to buildzoid For This Useful Post:
Strunkenbold (04-19-2017)
  #55  
Old 04-19-2017
Strunkenbold Strunkenbold is offline
robot overlord
 
Join Date: Apr 2009
Location: GERMANY
Posts: 1,355
Thanks: 155
Thanked 316 Times in 253 Posts
Strunkenbold is on a distinguished road
Default Re: How graphics cards categories are determined or what makes for a new category

Thanks for the link. RX 580... just another nightmare category, congrats AMD! Of course now every RX480 owner will flash RX 580 Bios for better oc... And no one can distinguish, again...
__________________
Reply With Quote
  #56  
Old 04-20-2017
buildzoid's Avatar
buildzoid buildzoid is offline
robo cop
 
Join Date: Jan 1970
Location: United Kingdom
Posts: 371
Thanks: 143
Thanked 62 Times in 39 Posts
buildzoid
Default Re: How graphics cards categories are determined or what makes for a new category

Quote:
Originally Posted by Strunkenbold View Post
Thanks for the link. RX 580... just another nightmare category, congrats AMD! Of course now every RX480 owner will flash RX 580 Bios for better oc... And no one can distinguish, again...
Uh I've been running custom 480/Fury BIOSs for a while now. It gives pretty much the same results as a 580 BIOS. Well actually better once you copy the 1750 or 1625Mhz timings into the 2000MHz strap.
__________________
Keep calm and raise Vcore


My Blog
Reply With Quote
The Following User Says Thank You to buildzoid For This Useful Post:
Antinomy (04-20-2017)
  #57  
Old 04-20-2017
Antinomy Antinomy is offline
HWBOT Staff
 
Join Date: May 2007
Location: Vladivostok, Russia
Posts: 1,424
Thanks: 28
Thanked 33 Times in 19 Posts
Antinomy
Send a message via Skype™ to Antinomy
Default Re: How graphics cards categories are determined or what makes for a new category

LOL, even the GPU revision is changed, I was hoping it'll help us. Too bad for GPU-Z.
Quote:
Originally Posted by Strunkenbold View Post
A rebrand is not different hardware for me. But it would be okay if we would have a chance to see something different with GPU-Z. There might be small differences like the new RX 580 look like to achieve higher clocks but shouldnt that do a 80nm 7600GT vs a 90nm 7600GT also?
We need to get an agreement what is a rebrand and what is a different model and what is neither. For example - the 7600GT B1 was sold as usual 7600GT, no vendor sold it as a separate model. Same goes for us, no need to separate (same here - Deneb C2 and C3, sold within one model). Then we have cases like 4870 vs 4890, the latter being another respin of the GPU but sold as a different model. That's where I'd like to draw a line. It's how manufacturer sell it. By manufacturer I mean GPU vendor (ATI/AMD, NVIDIA). Not all those VGA vendors and not the Chinese mutants. Renamed cards fall somewhere between.
Quote:
Originally Posted by Strunkenbold View Post
We have the problem that we have to look how things get recognized by software. And we have somehow to deal with that. We have to draw the line somewhere.
Recognition by software is a problem, I agree. But we shouldn't put it on top. Software can be updated, combined with other and so on. GPU-Z doesn't detect 3-rd party IGPs correctly but that doesn't mean we should make a junk yard from those submissions putting them all together. If we merge everything we can't deal with, there's no going back.

What I like to do is split big problems into smaller one.

So far the VGA problem falls in four:

1) Different memory size. On one hand memory size itself shouldn't be a big issue if it's just a variation of a model. But there are two buts:
1.a) There are cases, where memory size matters. But it shouldn't be 1 or 2 benchmarks (points-awarded). I think above 33% of benchmarks with benefit over 7-8%. There are not too many of such categories so we could discuss all separately.
1.b) Sometimes different memory size is a different model. For example, Geforce 7800GTX and Geforce 7800GTX 512MB. The latter was released later, has better o/c and higher clocks.
1.c) Different memory size might mean different memory chips and clocks. Most notable - Geforce Ti 4200 64MB (222MHz mem) and Ti 4200 128MB (250MHz mem) and hence different memory chips. Merging them will make 128MB non-competive. I want to point out that this is a reference design. I'm not talking about all those Chinese or noname vendors who made a zoo with 512 and 1Gig cards with very slow memory chips, it's their problem and I think those should fall with general memory size. I'm talking about a reference model case.

Yet again, all of cases from 1) are countable and can be named and dealt with one by one.

2) Different clocks (same model name):
2.a) Overclocked and only overclocked card by vendors (Goes like hell, TOP, OC, you name it). Go with reference. I think we should do the same with silent o/c without name changed like the current GTX 1060 with faster memory.
2.b) Are there any fishy cases where one model has the same name but ruins the balance only by clocks?

3) "Renamed" cards (one of the hard ones):
3.a) Different models (within one family) with different clocks. Like 7600GS DDR3 and 7600GT, 6800GT/Ultra, 8800GTX/Ultra. I think we should keep these separate because the are different cards pretty much like different CPUs are separated only by multiplier (say, Celeron D 3.066 and 2.53).
3.b) Different models with differences in dies. Like 4870 vs 4890. Should be left separate. Problem comes if software has problems detecting this (RX480/RX580)
3.c) Rebranded OEM cards. Open to examples and suggestions here.
3.d) Rebranded models and families (like Geforce 1xx, X1050/X300, 1050/9600, 7800/7900/R2xx/R3xx series). Open to examples, counter-examples and suggestions. Software detection problem too but don't focus on it solely.

4) Unique cards. Suggest we keep it for the latter.


Please, keep things as separate as possible. Don't bump everything together. I understand the "merge everything where GPU-Z can't tell the difference" but are there other options? That's what I want to discuss.
One thing I'd like to point out - no matter where the line between categories is drawn there always will be a case that kicks in and messes everything up. There's no perfect solution but we can find a good enough.
__________________
Quote:
Originally Posted by Turrican View Post
Yes, Alexander is right, as always.
Reply With Quote
  #58  
Old 04-20-2017
buildzoid's Avatar
buildzoid buildzoid is offline
robo cop
 
Join Date: Jan 1970
Location: United Kingdom
Posts: 371
Thanks: 143
Thanked 62 Times in 39 Posts
buildzoid
Default Re: How graphics cards categories are determined or what makes for a new category

Specifically in the case of the 580 and 480. On one hand we can't prove which card is which. On another hand I personally wouldn't really care either. I had a reference 480 that did 1480Mhz core on water(low voltage like 1.285V). I had a 480 that did 1580 on water(1.45V). So far the 580s seems to clock exactly the same as the 480s so people posting 580s in the 480s and 480s in the 580s honestly wouldn't give them much of an advantage. If they really want to do well on those cards they will still need a BIOS with modded memory timings, more voltage and possibly even higher power limits than what the 580s ship(580s are IIRC 180W I was running 300+W BIOSs on my 480s).

So with 580s and 480s I honestly say just let people flash them and submit them as whatever the hell they want. It's less work than trying to catch people cheating. It's not particularly unfair as RX 580s are just voltage bumped 480s(stock voltage is now up at 1.2V with 1366MHz instead of 1.15V 1266MHz) and both cards end up being more viable for points. At the same time the database can stay consistent to the rule of dividing up all cards based on manufacturer names instead ending up with ugly categories like "Polaris 10 and 20" where you would find 580s and 480s anyway.
__________________
Keep calm and raise Vcore


My Blog
Reply With Quote
  #59  
Old 05-29-2017
K404's Avatar
K404 K404 is offline
nVidia FTW!
 
Join Date: Jan 1970
Location: Glasgowvic, Scotistan. Average Kingdom of Un-United Britain..
Posts: 3,855
Thanks: 16
Thanked 146 Times in 85 Posts
K404
Default Re: How graphics cards categories are determined or what makes for a new category

Have any solid and final decisions been made about this topic?

I know... it's Computex week, I am asking at the worst possible time
__________________
Quote:
Originally Posted by Hondacity View Post
i believe in K404! :P
https://twitter.com/K404ExtremeOC

I do not believe in 1st place by any means, at any cost.

Quote:
Originally Posted by der8auer View Post
Knowing workarounds to get a benchmark to run is not a skill. It's a combination of knowledge and wasting time.
Reply With Quote
  #60  
Old 05-30-2017
Strunkenbold Strunkenbold is offline
robot overlord
 
Join Date: Apr 2009
Location: GERMANY
Posts: 1,355
Thanks: 155
Thanked 316 Times in 253 Posts
Strunkenbold is on a distinguished road
Default Re: How graphics cards categories are determined or what makes for a new category

We are discussing things internally, Im waiting for a response from Alex. But I agree, we really have to decide on every card merge individually. Which makes things in general very slow.
__________________
Reply With Quote
Reply


Thread Tools
Display Modes

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off

Forum Jump

Similar Threads
Thread Thread Starter Forum Replies Last Post
[AT] SD Association Announces UHS-III (up to 624 MB/s), A2 Class, LV Signaling hwbotrss Tech News RSS 0 03-11-2017 21:10
[AT] XFX Starts to Sell Single-Slot Mini-ITX Radeon RX 460 Slim Graphics Cards hwbotrss Tech News RSS 0 02-25-2017 19:10
[AT] Sony Announces SF-G UHS-II SD Cards: Up to Nearly 300 MB/s Read/Write Performance hwbotrss Tech News RSS 0 02-24-2017 20:10
[AT] MSI Adds Low-Profile AMD Radeon RX 460 Graphics Cards to Lineup hwbotrss Tech News RSS 0 02-18-2017 19:10
[AT] GIGABYTE Quietly Launches Low Profile GeForce GTX 1050, 1050 Ti Graphics Cards hwbotrss Tech News RSS 0 01-10-2017 23:10


All times are GMT +1. The time now is 03:31.


Powered by vBulletin® Version 3.8.4
Copyright ©2000 - 2017, Jelsoft Enterprises Ltd.
Copyright HWBOT 2004 - 2015