Lod Or No Lod

[ncc-1709]

do you use lod tweeks while benching or not?

and which benches?

[Massman]

All benches; I usually spend over ten hours of testing for the best lod settings for each card, especially in 3dm01

[RyderOCZ]

No LOD ever.....to me, its.... well, we don't go there because I don't want to start an argument

 

The OCZ team does not run LOD, we will never be at the top

[ocztony]

Reducing the level of detail a card will render a frame below that what is available thru approved driver release is cheating.

 

I said what Ryder would not, OCZbenching team does not use LOD and I have spoken to Futuremark expressing they should make LOD tweaks illegal and detectable...whether they agree with my POV is upto them though.

[Massman]

No LOD ever.....to me, its.... well, we don't go there because I don't want to start an argument

 

The OCZ team does not run LOD, we will never be at the top

 

Very good to hear you guys don't use lods ... it's true, you'll never be at the top without lods. I see as an essential part of the 01 overclocking as a good lod can give you a 50fps boost, even with lower end cards.

 

All the arguments have already been mentioned in countless thread about lods and a lot of them make sense. Though, mild lod changes (the ones you can set in Rivatuner) don't changes the bench that drastic. Okay, less detail, but you can still see what the test is all about.

[harleybro]

Yup I do use it. Not that I am happy about it but to have competitive scores you need to. I personally see it as another tweak. Running a dual or quad core cpu and setting the affinity for the fastest core in a single threaded bench is that a cheat? The cpu was made to run with multiple cores. I do run benches without LOD but the ones I submit are adjusted.

 

Though, mild lod changes (the ones you can set in Rivatuner) don't changes the bench that drastic.

 

Does that include tweaking rivatuner to allow settings of 9 and higher?

[RyderOCZ]

getting high scores is part hardware, part software, tweaking the OS and vidcard driver is part of the game.

 

what's next, banning Windows 2000 because you get higher scores in 3D01 with it?

The code for the bench is written.. the driver for the Vid Card is written.. Using LOD, you are altering the code and how it is rendered, to me its like re-writing the bench to suit the output you want, name another "Tweak" that does that...for any bench.

Copy Wazaa, maxmem, etc for Pi, all they do is set certain parameters, clear out memory, etc to make things more efficient and make the bench faster, that is a tweak.

 

Yup I do use it. Not that I am happy about it but to have competitive scores you need to. I personally see it as another tweak. Running a dual or quad core cpu and setting the affinity for the fastest core in a single threaded bench is that a cheat? The cpu was made to run with multiple cores. I do run benches without LOD but the ones I submit are adjusted.

 

Does that include tweaking rivatuner to allow settings of 9 and higher?

Setting the Affinity...doesn't change how the CPU runs does it? It just tells the bench specifically where to run, thus fitting in the tweak category.

[Massman]

The code for the bench is written.. the driver for the Vid Card is written.. Using LOD, you are altering the code and how it is rendered, to me its like re-writing the bench to suit the output you want, name another "Tweak" that does that...for any bench.

Copy Wazaa, maxmem, etc for Pi, all they do is set certain parameters, clear out memory, etc to make things more efficient and make the bench faster, that is a tweak.

 

So disabling AA and AF in rivatuner would't count either?

[RyderOCZ]

If the bench was designed to run with them on...then I don't agree to disabling them.

 

I mean do we accept scores not run at default resolution? Why can't I just lower the resolution so I get a better score? Same as reducing the level of detail, isn't it?

 

I don't have a problem with tweaking the timings, the speeds, enabling or disabling things in the OS, running a different OS (like win2000 mentioned above)

 

I mean there is no other bench where we actual change how the bench is run, just the 3D benches.

SuperPi...nobody changes the program, they change the environment the program runs in.

 

IMO LOD, and the related, modify the execution of the code which means you are no longer running the same program.

[RyderOCZ]

I am not trying to say you guys need to police it. Although if you could...that would be great

 

I am just saying...that IMO, its a cheat, not a tweak. My opinion is very much in the minority, but discussions rather than arguments are always good

[ocztony]

So you all think reducing LOD that much is renders butterfly's as large + is ok...come on now thats just plain stupid.

 

Its a cheat, if you were not able to do it you would not miss it, trust me.

 

3D benches should test for default LOD and not run if over ridden IMO and I expressed the same to the futuremark guys. Everyone on the same page and same playing field...end users and manufacturers alike.

[ncc-1709]

The reason i started this thread was because ive been doing some testing.

 

i use NVhardpage for the testing. it allows a lod of 48. but does it make a diffrence on high end hardware, no. nothing at all any more. it may bive you 1k in 01, but nowhere else. infact, AQM3 gains from oposite LOD. making it sharper improves the score over poor quality.

 

i still use it, as it allows slightly higher clocks, due to being less demanding on the cpu, but otherwise, your not gaining anything. its just like reducing the quality settings in the drivers another notch.

 

it never use to be possible to turn AA and AF off on ATI cards, yet i never had it on. ATItool used to have a nice setting that turned it completally off. now was that a cheat? another driver to be tweeked.

 

 

if you check the nvidia drivers, LOD is allready in there, same for ATI, they just set it slightly diffrently, better performance or better IQ. either way, its in there and the manufactureres use it to sell there cards.

 

ok, i went off at a bit of a tangent.

[Massman]

IMO LOD, and the related, modify the execution of the code which means you are no longer running the same program.

 

Lod doesn't really alter the code as it's defined by driver settings. It's kinda like the AA and AF settings: it changes the way the videocard has to render the given course, it doesn't change it.

 

Theoretically, changing the resolution should be considered as a tweak IF it's all about the highest 3DMark01 score. But we can't compare those results to the ones we currently have. In order to be able to compare, you always need a set of ground rules, which are easy to verify. FM doesn't offer the details of the lod used in a bench, so no one would really know if the score was lod tweaked or not.

 

If FM decides to prevent lod tweaks by inserting a lod prevention tool in their next 3D bench, I would be 100% ready to bench without the lod tweaks ... but untill then, they just help us to push the hardware a little further.

 

So you all think reducing LOD that much is renders butterfly's as large + is ok...come on now thats just plain stupid.

 

Its a cheat, if you were not able to do it you would not miss it, trust me.

 

(I guess, 3dmark03?) You can still see it's a butterfly, you can still see the water coming down from the hill, you can still see the turtle, you can still see the rock where the butterfly is landing on ... you still know what it's about. It's not like you're running the wireframe and can't see what the creators intended to show.

 

Last sentence is absolutley correct, but why shouldn't we use the tools nVidia itself gave us?

 

3D benches should test for default LOD and not run if over ridden IMO and I expressed the same to the futuremark guys. Everyone on the same page and same playing field...end users and manufacturers alike.

 

That's indeed the point when talking about real ingame performance and benchmark performance. When you want to compare two videocards and want to see the performance ingame, you use a real game and all standard settings. If you want to test and compare just how fast the videocard can be compared to another one, you can use synthetic benchmarks and try to let it as fast as possible. Lods are given to us as they're inside the nVidia drivers, so we can use them.

 

i use NVhardpage for the testing. it allows a lod of 48. but does it make a diffrence on high end hardware, no. nothing at all any more. it may bive you 1k in 01, but nowhere else. infact, AQM3 gains from oposite LOD. making it sharper improves the score over poor quality. i still use it, as it allows slightly higher clocks, due to being less demanding on the cpu, but otherwise, your not gaining anything. its just like reducing the quality settings in the drivers another notch.

 

Lod48 in NVHardpage = lod3 in riva, if I'm not mistaken. You can set lods to 15 in rivatuner with a small tweak in the power users' tab.

 

It's true, though, that AM3 doesn't like lods at all. I remember lod 1.9(!) gave me the best result when I was benching the Ti200 a long time ago. Prolly faster because it took away a little of the cpu load, not because the gpu hade to render fewer details

 

Higher clocks is definitly true, especially for the 3dmark01. If you tweak each test (7 in total), you normally have 7 different lod settings and 7 different frequencies. In nature, for example, I'm mostly able to set the clocks 20mhz higher than I would when benching lod0.

[ocztony]

I remember when NVIDIA released a special driver to reviewers that set LOD for 3d01 real low...it made their cards faster than ATI's at the time. Only issue was the industry gave them a bashing for it and many reviewers had to redo their reviews. This is the whole reason why FM asked you use apporoved drivers only...to stop the cheating.

 

LOD is a way of speeding up a bench just like *** does, to me its no different to use either, one speeds up internal clock and the other lower the resolution of rendered frames...same crap different day.

 

We are going to be at each others throat here as OCZbenchteam has taken a 0 cheat approach, this includes LOD and we are firmly against it.

 

last post from me on this subject

[ncc-1709]

Lod48 in NVHardpage = lod3 in riva, if I'm not mistaken. You can set lods to 15 in rivatuner with a small tweak in the power users' tab.

 

 

there is a lot of diffrence, i have both installed. lod 3 is the same on both programs.

[Massman]

I remember when NVIDIA released a special driver to reviewers that set LOD for 3d01 real low...it made their cards faster than ATI's at the time. Only issue was the industry gave them a bashing for it and many reviewers had to redo their reviews. This is the whole reason why FM asked you use apporoved drivers only...to stop the cheating.

 

44.03 drivers? I remember those drivers giving a huge boost in nature due to some modifications. They were not approved by the community, but in the end, many benchers used them to get their top position.

 

Does FM regard lods as cheating?

[RyderOCZ]

yup, if the 3D application has ways to DETECT it, trying to circumvent LOD settings can be considered a cheat. Until this happens, altering LOD through regedit/rivatuner/*insertwhateverTWEAKapp* it's not a cheat!

In your opinion

 

so I guess you run NVIDIA/ATI driver settings with HIGH QUALITY too? don't you dare set it to "low quality/high performance" as it will render the scenes with less detail, and that's a cheat. right?

That is built right into the control panel of the driver with no access to other settings, not a cheat, a tweak. Using Rivatuner/ATI Tool to access other levels, even though they are part of the driver, to me is circumventing what is built in by ATI/Nvidia, which makes it unethical (there, I will change what I call it)

[RyderOCZ]

like others (and me) already said multiple times now; LOD doesn't matter much in recent 3D BMs; so that's a positive trend;)

Yes it does not seem to affect the newer bench's much if at all, but we still bench the old ones and count them as meaning something.

[Massman]

LOD tweak has never been tagged as unethetical or "cheat" untill I've read this thread; and that tweak has been around for more than 5 years.

 

Not entirely true, lods have been questioned a lot in the past. XS has had lots of threads regarding the use of lods when Ati users didn't have the possibility to alter the level of detail.

[Steven86]

editing the registry != cheat. it stops there. really.

 

being short minded?

 

what does Team japan, OPB, etc do with windows my friend?

indeed.

[Massman]

Enlighten us without bashing a person or community

[Steven86]

Enlighten us without bashing a person or community

 

a question is not bashing in my dictionary

[Steven86]

they do "indeed" with windows. nice. What is "indeed".

http://www.google.com/search?q=windows+indeed+registry

 

nothing found.

 

Lol telling me about attidude

hey whatever man

[SF3D]

I think Troman speaks about this Windows tweak:

 

...only part of it here...

 

[boot loader]

timeout=30

default=multi(0)disk(0)rdisk(0)partition(1)\WINDOW S

[operating systems]

multi(0)disk(0)rdisk(0)partition(1)\WINDOWS="Micro soft Windows XP Professional" /fastdetect /onecpu

 

That is not cheating, cause you just lose other core / cores.

 

Level of detail tweaks are not cheating as long as everyone can use it.

[Massman]

[boot loader]

timeout=30

default=multi(0)disk(0)rdisk(0)partition(1)\WINDOW S

[operating systems]

multi(0)disk(0)rdisk(0)partition(1)\WINDOWS="Micro soft Windows XP Professional" /fastdetect /onecpu

 

Shouldn't this be: /numproc=1 ?