Is truform on the radeon 9700 working to full potential?

[Luminescent]

It seems that in games such as RTCW and Ut2003, there is a significant performance hit, of about 10fps between each level of tesselation. Is the truform system working in hardware, or is it being emulated throught software for now?

When you're running a game with all the IQ options turned up, at around 30fps, those 10fps really matter.

 

[KimB]

Well, UT2k3 uses a hell of a lot of polygons...doesn't it seem possible that the 8500 is just approaching its limit as far as poly throughput goes?

Oh, and you are using an 8500, right?

 

[Luminescent]

No, actually the 9700, only thing is I am very cpu and memory limited (xp 1900 and sdram), however I am wondering if it is working as it should on the processor.

 

[Nagorak]

It seems that in games such as RTCW and Ut2003, there is a significant performance hit, of about 10fps between each level of tesselation. Is the truform system working in hardware, or is it being emulated throught software for now?

When you're running a game with all the IQ options turned up, at around 30fps, those 10fps really matter.

If you're running about 30fps, then you need to be turning things off, not turning things on. 30fps in a game like UT2003 is not anywhere near being acceptable, unless you like the view looking up from your dead corpse.

 

[Luminescent]

Problem is that any of the other settings I turn up, graphically, don't really affect my performance as does the number of bots.

 

[Hellbinder]

i acctually have some input to this. I have conducted some benchmarks today with Trueform enabled and disabled in UT2003. Check this out.

Without Trueform.

dm-asbestos (botmatch)
18.089582 / 33.062317 / 117.786430 fps
Score = 33.072227

With TrueForm

dm-asbestos (botmatch)
26.952358 / 57.992764 / 117.227188 fps
Score = 58.005371

These Ran at 1024x768x32 with 4x FSAA and 16xaniso.

I have been repeating the tests with and without Trueform on all the Benchmark levels for about 2 hours now. Tweaking with Overclocking etc.. These results can be repeated on all the levels over and over again. Enableing Trueform (currently) hacks you for 25-30 FPS overall performance.

 

[Typedef Enum]

I think you have your numbers reversed, no?

 

[gkar1]

Well, UT2k3 uses a hell of a lot of polygons...doesn't it seem possible that the 8500 is just approaching its limit as far as poly throughput goes?

Oh, and you are using an 8500, right?

OMG good going reading the thread title lol

 

[KimB]

Yeah, sorry.

Well, then, it's most likely a driver issue if there is a significant performance drop on the 9700. That should certainly not be happening.

 

[demalion]

This really DOES look odd...could the difference be dedicated fixed T&L for 8500, emulated via programmable pipeline for 9700? Not as if the Truform scores are lousy, but I don't understand this behavior.

OK, maybe I'm just solidly and completely CPU/memory bandwidth bound on my PC 133 TBird system for the botmatches. I still don't get why I'd have no impact from enabling Truform for the benchmarking, and HB's 9700 system is being hit so hard.

Or maybe more polygons being used for the 9700 that isn't being enabled for my 128MB 8500?

Also, you didn't say what level of Tesselation was in use...my results were with 1 (I can't even imagine needing more, atleast at 1024x768).

Maybe try without SV, and with some of the tips about benchmarking D Vogel mentioned in the benchmarking thread in Games Talk?

I need to check my benchmark images for the botmatch, I guess...there is absolutely no difference between Truform on and off for the integer part of my fps result. Speculation, like DX 8 NPatch support being done on the CPU for the 9000 (known, I think) AND 9700 (the implication here) in current drivers , are what come to mind and I should atleast try to make sure my observation about my system are Kosher.

 

[demalion]

Hmm...where did you enable Truform? I've been editing the UT2003.ini file, and after doing that and watching the botmatch benchmarks closely, I conclude that in fact Truform is not being turned on at all. Certainly not to the degree I see in the game.

From this:

The way my benchmark results do not vary and HB's do is not an indication of a difference between Truform performance on the 8500 and the 9700.

I don't know how HB is enabling Truform for the benchmarks...editing ut2003.ini for me doesn't do it. Maybe he is editing default.ini? Or is there some Benchmark.ini (or equivalent) that I'm not aware of? Oh, wait, just found it by looking at the result file. MaxDetail.ini is what is being used by the benchmark program...but the .bat file doesn't seem to specify an .ini file, so probably uses your current settings. I'll add an edit in a bit with results from using my setting with Truform on and off and we can compare the hit.

Therefore, it seems (atleast in the temporary absence of my enabling Truform for the Benchmark to compare the hit):

The difference in hit between my system (nearly none) and these 9700 systems is likely due in this case to my slower system simply always being CPU limited (atleast for the in game variance, we'll see about the benchmark in a bit), and that this is below the level required to show a variance between Truform enabled or not.

EDIT: Ok, some benchmarking notes are posted. I split them out to avoid confusion. Also, an answer from HB would be helpful about how he benchmarked...I'll note that the default .bat files leave sound enabled.

 

[KimB]

default.ini just takes over when the ut2003.ini file is either deleted or outdated (i.e. after a patch).

Make sure you set the "UseNPatches" value to True and increase the tesselation above 1.

 

[demalion]

default.ini just takes over when the ut2003.ini file is either deleted or outdated (i.e. after a patch).

Thanks!

Make sure you set the "UseNPatches" value to True and increase the tesselation above 1.

Heh, my screenshots show I know about the first, and the 2nd is actually incorrect, a value of 1 does have an effect (as my screenshots also show...they are all using 1). Tesselation=0 is equivalent to no truform, not 1.

 

[demalion]

Benchmark results as per the post of mine slightly above.

Just one run each. I did visually verify that Truform was actually on. Too lazy for screenshots at the moment, but I will mention that the model you follow in this benchmark has the "twin tubes" I have mentioned are one easy way to spot Truform activation, so anyone with the applicable cards should not have trouble in verifying for themself.

.bat file, sound ON
dm-asbestos?spectatoronly=true?numbots=12?quickstart=true -benchmark -seconds=77 -exec=..\Benchmark\Stuff\botmatchexec.txt

Truform ON
5.515985 / 23.561359 / 41.656269 fps rand[25084]
Score = 23.581236

Truform OFF
7.618098 / 23.435379 / 42.130733 fps rand[25084]
Score = 23.449150

.bat file, sound OFF
dm-asbestos?spectatoronly=true?numbots=12?quickstart=true -benchmark -seconds=77 -exec=..\Benchmark\Stuff\botmatchexec.txt -nosound

Truform ON
5.082696 / 29.778038 / 51.217110 fps rand[25084]
Score = 29.790684

Truform OFF
6.895074 / 29.790787 / 51.602062 fps rand[25084]
Score = 29.803589

Well, I guess there is indeed no variance for my 8500, back to square one there.
EDIT: Hmm...well the minimum fps variance is fairly significant...I wonder when that low point is hit? When moving between the two lights projected through the moving fan seems the most likely spot.

Here is a comparison to the settings used for the results generated by using the "Benchmark.exe" program. I do wonder what "-UPT" does?

Benchmark.exe, sound OFF
dm-asbestos?spectatoronly=true?numbots=12?quickstart=true -benchmark -seconds=77 -exec=..\Benchmark\Stuff\botmatchexec.txt ini=..\Benchmark\Stuff\MaxDetail.ini userini=..\Benchmark\Stuff\MaxDetailUser.ini -nosound -UPT -1024x768

8.089665 / 29.832632 / 51.763012 fps rand[25084]
Score = 29.845421

 

[demalion]

Looking at the above, the theories that fit as far as I can understand are:

1) There is some sort of problem with the performance impact of Truform as exposed on the 9700, atleast if your expectations are based on the 8500. Whether this is because the drivers are treating the 9700 in the same way as the 9000, or some other factor, I'm not sure.

The CPU speed issue doesn't fit because (and I presume HB's With/Without labels are just switched at this time) HB's average results for Truform being activated are just too close to mine for a CPU that much faster...also, they are too close to my average for the performance advantage the 9700 should have. Then again, he doesn't say what physics setting he is using. Too many questions right now to be sure what his figures mean.

2) The game is responding to something about the 9700 and increasing polygon load out of proportion to what it is doing for my card. The net result is that Truform is having a much more significant impact because much more work is being added by enabling it.

We have the people here to answer the questions about each of these, with pretty much absolute certainty, and an even wider selection of people (i.e., everyone) to point out where I'm wrong in my conclusions and/or offer some conclusion I missed.

 

[Bambers]

What tesselation factor were you using? If you're only getting 23fps avg with truform off then I think your cpu is still the limiting factor.

I got this:

Max detials, software 3d sound, botmatch.

dm-asbestos - Truform OFF
3.476302 / 52.662483 / 112.634468 fps
Score = 52.689476

dm-asbestos - Truform ON, Tesselation factor 2
3.287766 / 50.946194 / 111.001900 fps
Score = 50.971107

dm-asbestos - 3x
3.059216 / 41.948940 / 106.156929 fps
Score = 41.956741

dm-asbestos - 4x
2.881504 / 30.175344 / 103.724976 fps
Score = 30.188593

I'm using the latest 6166 drivers and a DX9 beta, setting tesselation over 4 makes no difference and setting truform to always off in the drivers does disable truform so if it is in software the drivers can disable it there too.

The flybys werent affected by truform. It only seems to be applied to the characters.

 

[demalion]

Ah, so the mystery does seem to be because of the settings used...I guess Luminescent and Hellbinder are sleeping still :-/ . Thanks for seemingly indirectly answering my question about their settings, Bambers.

Reading Luminescent's post again, perhaps the question was about the results Bambers is posting. Perhaps the confusion is because of some misunderstanding of the Tesselation level...again, Tesselation of 1 is NOT off.

Bambers, read my posts again if that "23" question was directed at me. If it was a typo for "33", I guess we'll have to wait for HB. Your results do seem to indicate that the performance drop is an issue of their Tesselation levels.

 

[rubank]

...

The CPU speed issue doesn't fit because (and I presume HB's With/Without labels are just switched at this time) HB's average results for Truform being activated are just too close to mine for a CPU that much faster...also, they are too close to my average for the performance advantage the 9700 should have...

Are you running @ 4xFSAA and 16x aniso like Hellbinder?
I didn´t se you mention that, or did I just miss it?

Or did you miss it?

 

[demalion]

...

The CPU speed issue doesn't fit because (and I presume HB's With/Without labels are just switched at this time) HB's average results for Truform being activated are just too close to mine for a CPU that much faster...also, they are too close to my average for the performance advantage the 9700 should have...

Are you running @ 4xFSAA and 16x aniso like Hellbinder?
I didn´t se you mention that, or did I just miss it?

Or did you miss it?

I guess I split up my post too much, since two people have missed this now.

I'm running on an 8500, not a 9700. I'm providing reference for how much of a performance hit my 8500 takes compared to the 9700, which would be required to establish whether there was something "wrong" with the 9700 implementation.

With Bambers' data it seems pretty clear that the original issue with Luminescence and Hellbinder are due to having tesselation level set too high, which in turn may be related to a mistaken assumption, such as Chalnoth seemed to have for instance, that a "Tesselation=1" value is "off". See the thread I started for this info.

Hmm... <-this is the sound of me proceeding to refrain myself from swearing in response to my thread being moved to a forum it was not intended for, most likely due to the off topic posts of some in the thread. Does the discussion outlined in the first post in the thread have no bearing on whether a thread is moved or not?

Also, moderators, what happend to the "Moved->" text for when a thread gets moved...I hadn't realized it was missing, I had just thought no one was replying anymore.

Could it please be moved back? Aside from the off topic weaponry critique (atleast off topic for the 3d technology and hardware forum which is why I put it here), there is a lot of info there that warrants discussion here, or am I wrong?

 

[rubank]

No, I didn´t miss that you´re running an 8500 as opposed to a 9700, but you said HB:s scores are too close to yours taking just that and his faster CPU into consideration.
That´s why I wondered if you, like HB, are running @ 4xFSAA and 16xAF: otherwise your comparison (of your scores vs. HB:s) is invalid.

Just a simple question, deserving a simple answer.