4870X2 3d mark 03 bugged score thread

[dinos22]

it's probably vista display driver recovery

i've seen this on Sony laptops we sell where the driver flips out a little and throws an error that there was an error and it's been fixed

incidentally those Sonys had nvidia GFX cards heh

[massman]

So it's VPU recover; besides disabling VPU recover in CCC, there's nothing else that we can do, I reckon. FM doesn't detect the VPU recover feature being enabled or disabled, so we'll have to rely on scaling predictions and honesty.

Andre said he is able to control the bugs, would that make him capable of minimizing the effect of the bug so that his scores still look real?

[T_M]

It would be more useful like massman said that we actually get some solid info and discussion about how it is being controlled rather than everyone just laughing at bugged runs like the other million threads.

[.p4nzer]

it's probably vista display driver recovery

i've seen this on Sony laptops we sell where the driver flips out a little and throws an error that there was an error and it's been fixed

incidentally those Sonys had nvidia GFX cards heh

I think dinos is right.

It's vista recover. Watch my bug sample. Vista has restored the driver. There is a message on the screen.

Bye

[Kinc]

The bios I use do not have VPU recovery at all. It doesnt show up in CCC and I have never experienced it, if it hangs it hangs without recovery. But doesnt help ut to prove legal scores, is it visable in CCC? That VPU Recovery is present?

[massman]

Could someone make a performance scaling graph in which we can see the effect of the increase of CPU/GPU/MEM frequencies? Randomly claiming other people cheat because the FPS is too high (only by a few percent) is not how it should be.

[.p4nzer]

The bios I use do not have VPU recovery at all. It doesnt show up in CCC and I have never experienced it, if it hangs it hangs without recovery. But doesnt help ut to prove legal scores, is it visable in CCC? That VPU Recovery is present?

Probably it's not vpu recover. Vista recover the driver.

I can say that 'cause someone make it works with nvidia 280.

I don't remeber the thread but it's in this forum.

I think it can be done with almost all the cards and we can't say if a score is bugged or not.

Bye

Edit: I have found the thread...

http://www.xtremesystems.org/forums/...d.php?t=179719

Works even on nvidia. But the test dynamics are different. Gt2 is low and gt3 even higher.

[ZillaPct]

This is a problem kernel of Vista, The kernel Stops only in very few. VGA errors situations this not affect the functioning of the system.

How many times I have happened in VGA tests (Video Games Bench) Vista work correctly with black screen of vga failure driver and the benchmark end correctly with adeguate FPS report.

This bug i presume is repetible with 3D03 with any vga driver.

[Kinc]

I have a QX9770 running under ln2 now. Will make a move with over 1400 fps in gt2 so everybody can see that it is a real score with no black screen errors.

[AndreYang]

How do you bug the scores Andre?

I promise your score GT2/GT3 are bugg, too. Don't tell me you don't know that. It's IMPOSSIBLE GT2 1450+ at your speed.

I test over 10 times which are the same GPU/CPU speed as yours. Black screen will make the same score as your 160K. If no any black screen, the score is only 155K.

[Kinc]

I promise your score GT2/GT3 are bugg, too. Don't tell me you don't know that. It's IMPOSSIBLE GT2 1450+ at your speed.

I do not know how to bug scores in 03.

This is how it looks when I run: http://www.kinc.se/gt21420.wmv (right click save as)

[NBF]

There's no CCC option to disable VPU recover under Vista.

Also it's very hard to hang Vista overclocking GPU, it usually aborts 3d run and recover the driver.

[AndreYang]

I have the same bios as you so I dont have it

Please prove it why your score is BIG different from SF3D.

[Kinc]

Please prove it why your score is BIG different from SF3D.

Check the video (http://www.kinc.se/gt21420.wmv), that is the result I get. (you have to right click and "save as")

P5E64 Evolution
QX9770 C1 @ 5750 ish
2 pcs EAH4870X2 @ 872/1070MHz Quad-CFX (the test is in the way in the video but if you frezze frame you can see the clocks)
Vista 32bit SP1

If you wonder why I done put the complete test on tape thats because I cant fit that on my digital camera.

[AndreYang]

Check the video, that is the result I get.

P5E64 Evolution
QX9770 C1 @ 5750 ish
2 pcs EAH4870X2 @ 872/1070MHz Quad-CFX
Vista 32bit SP1

Why did you change cpu to Quad core and mobo? It's nothing proof for your 03. Sometimes, single run didn't prove black screen and bug. I also can do single run no black screen and the score is very high. Vista/32/64 are all the same, not only for Vista 64.

[CCC]

kinc run lower card clocks like stock card clocks and show us this scalibilty its already proven pushing cards to much casues these issues and here you are running at 870/1070, lol

cheating and not knowing your bugging it two diff things

rem the sm2 06' bug you had?

things happen, but for top guys to not be getting this running stock card clocks to say 800/900 mild but not to high card clocks shows ya something isnt right here

[DeDaL]

Vista SP1 32

128197




Here is my run with 780/900 on cards and 1112fps in GT2, why 1450 is inpossible ? at 870 on GPU and 6200 on CPU

[Kinc]

Why did you change cpu to Quad core and mobo? It's nothing proof for your 03. Sometimes, single run didn't prove black screen and bug. I also can do single run no black screen and the score is very high. Vista/32/64 are all the same, not only for Vista 64.

Becuse thats what I have on my table at the moment. Becuase im trying out 06 and Vantage with 4870X2. Thats is the score I get every single time.

It the movie isnt proof enough then what can we do? No more 3dmark 03 I guess.

Please be more calm about the issue Andre, I don't see why you are so agressive with me. I'm just trying to solve the issue so we can continue to use 3dmark 03.

[Kinc]

kinc run lower card clocks like stock card clocks and show us this scalibilty its already proven pushing cards to much casues these issues and here you are running at 870/1070, lol

cheating and not knowing your bugging it two diff things

rem the sm2 06' bug you had?

things happen, but for top guys to not be getting this running stock card clocks to say 800/900 mild but not to high card clocks shows ya something isnt right here

I just did a run with same cpu speed with 800/1000MHz. I will make it in to wmv and upload in a few mins.

EDIT: here is the video with 800/1000; http://www.kinc.se/800.wmv (right click and save as )

[Gautam]

The scores that joe_cool, Kinc and spl are getting are too consistent to be what I'd consider "bugged." Almost every run that they do has high GT2 FPS.

I think that this is a seperate issue from the 03 bugged runs. (But right now it's admittedly difficult to tell)

We need to find the reason that some have 8k-10k higher than others. Right now amidst all the confusion the easiest explanation to put to it is that it's a bug. I still think there's more to it.

I have personally gotten very low GT2 performance at times only when I didn't have my fans at 100% indicating some kind of throttling issue. When I got the low GT2's, there was no obvious giveaway. No artifacting, no black screen or anything. The only difference was that the FPS at the opening of GT2 was about 500 lower than normal.

[Gautam]

I just did a run with same cpu speed with 800/1000MHz. I will make it in to wmv and upload in a few mins.

EDIT: here is the video with 800/1000; http://www.kinc.se/800.wmv (right click and save as )

Okay, 80 FPS lower at 1360 FPS. Definite GPU scaling going on. At 5400MHz and 790/915 I get 1270. It seems like it just might be an issue with setting clocks, as hipro stated.

Can you do 800/900 as well please?

[Kinc]

Okay, 80 FPS lower at 1360 FPS. Definite GPU scaling going on. At 5400MHz and 790/915 I get 1270. It seems like it just might be an issue with setting clocks, as hipro stated.

Can you do 800/900 as well please?

Ah man you are killing me Just turned the system of (going to change board to a good old P5E64 WS Pro).

I will run more in the next days with E8600 and QX9770, will run at complete stock clocks. To show some scaling.

[.p4nzer]

So what's the bug? The low run or the high run?

The black screen bug works even with a single vga. I've proved that with my screen on the first page of this post (143k with 3.6 ghz e8500 and 875/1250 3870 single -.-').

So if we try different 3d03 runs on xp and vista with only 1 4870x2 we can understand if the bug is the low run or the high run. We can assume the 2 runs reach similar scores with same frequencies... We can assume the xp result is marginally higher. So we will find if the andre and co runs or the kinc, spl and so on runs is the bugged one.

Maybe we have found two problems and assume that is only one problem. Perhaps we have a real bug (black screen is of course a bug) and something else that we can call tweaking (if we consider that kinc made 161k with 6.2 gigs e8600 and andre made 160k with 6.4 we can say there's a tweak that we don't know).

Bye

[SF3D]

I have a request.

Please, do not submit any 03 results to HWbot with 4870X2. I think we'll have to remove all those scores from our database or we have to remove points.

Thank you!

[blind_ripper]

SF3D and what about the single card score ,can they be bugged 2 ?