4870X2 3d mark 03 bugged score thread

[cowie]

yes its weird sf3d but i believe your score is right for the cpu speed and the test scores are all in line sort of speak but then i see other scores at say 4700 mhz which have normal wings of fury fps for that cpu speed and nature fps then trolls lair and proxycon are higer than people running the benchmark with the cpu 1000mhz higher which isn't right, my own thoughts is that somehow its running these tests at 800 x600 in the background without 3dmark03 knowing about it, but maybe as you say the lower scores are bugged and the higher score ones without the black screen bug are right but i have 2 4870x2s and ran 3d03 at 5200mhz with gpus at 830/1000 and got 135k roughly at 4800 at 840/950 i got 129k someone else on my forum has a similar setup with a quad qx on ss cooling at 5100 mhz got 133k , so i just don't know , its a shame as this is the benchmark which the quad gpus shine, its a shame its now tainted with all this bug business

i seen a qf score yesterday thats bugged has gt2 gt3 way too high for clocks.
hwbot has got to look into it or it will become common place for bugged 03scores
matter of fact lower gt1 and gt4 then your score but higher gt2 gt3 and a good 8k more points overall

[[XC] Oj101]

Heh, heh, heh........I powned you all.....I moded a CRT doing 350Hz......

I beat you too, my old 14" CRT doesn't have the chip which relays info regarding the maximum resolution and refresh rates to the PC, so I can set whatever I like

[.p4nzer]

People here have tried with rivatuner opened to see if the cards go down of clock but if rivatuner doesn't see that?

I am refering to the fact that the thermal throttling cannot be showed 'cause it affects only pipelines (stream processors)... or something we don't know. Maybe the thermal sensors have an high differential between each other just like the intel core duo. I'm saying that 'cause your results are so weird... So we may test even that.

Has somebody a watercooling (ybris acs-g works on 4870x2) or pot that fits on these cards? If we ran tests at stock frequencies and cold temp maybe something will change...

Bye

[STEvil]

Yesterday again i'll test all day.
So, I can say - PSU is part of problem, maybe VGAs PWM problem(but why GT2/GT3, i cant uderstand, but i have some thoughts about it). When i change rails, score is rcovered sometimes(even on MSI VGA, but on Powercolor VGA is more repeatable).

I think bug is a low scores, not hight.

The PWM amperage is listed in rivatuner: http://www.xtremesystems.org/forums/...&postcount=106

If the clocks are bugging then the PWM amperage will be higher or lower depending on the clocks. Temperature of the cores and PWM will also influence PWM amperage. Inspection of these values would tell us over a bunch of runs..

[Marvin_The_Martian]

You mean like, uhm...

I'm not sure if RT can follow two gpu cores at once, but I think when you look at vreg amp draw it would show there if a core dropped out to?

this??

Surprised no one treid this already, whish I had the hw to test it for myself but I don't

[DeDaL]

Yesterday i tested single 4870 and CF and have same bug, and still GT2/GT3.
Restart, switching off PSU, changing rails is helping also.

[Marvin_The_Martian]

DeDal could you maybe try getting some SS with RT open showing amp draw during a bugged and non bugged run...??

[DeDaL]

SS

?
sorry for my missanderstanding

If you about freq dropping, i test with RT monitoring. With or without bug - freq is same.

[massman]

SS = screenshot

[Marvin_The_Martian]

?
sorry for my missanderstanding

If you about freq dropping, i test with RT monitoring. With or without bug - freq is same.

Amperage not clocks

See my post on 205 and two pages back

[CCC]

I can run 4870 cf or 4870x2 single all day long and never once to I get a low score 'bug' like I get with cfx

I have tried everything here people have done just to try and replicate it once and to no luck.

so either all my cfx 03's are fine and the rest are soemwnow screwed or for the life of me I cant get one 03 to run normal

[SF3D]

I can run 4870 cf or 4870x2 single all day long and never once to I get a low score 'bug' like I get with cfx

I have tried everything here people have done just to try and replicate it once and to no luck.

so either all my cfx 03's are fine and the rest are soemwnow screwed or for the life of me I cant get one 03 to run normal

Yes. It would be easier to accept that our runs are bugged, if there would be more users with "high" results. Now there is 3 people showing their results and few more who are not posting them.

If this bug rate would be 50/50, I could easily think, that my scores are bugged. Now it seems, that only the highest scores are having the GT2 / GT3 bug.

I think that we have too high hopes about 03 scaling. There must be some drop in scaling and for example 180K might be really really hard to get without bugs. Who will know the answer I quess time will tell.

[Gautam]

My runs are 100% consistent, and I fully believe now that Marcus' and spl's are too. (Both of their scores, and T_M's, and joe_cool's are perfectly consistent with each other) I used to get scores in line with you guys, but as soon as I updated to service pack 1, they are now always 5000 points higher. My scores are literally within 100 pts of T_M's at the same speeds.

I'm leaning towards this being an issue with the Windows build version, and certain updates, since downloading SP1 is what triggered the boost for me. (What kind of scores would you guys get if you didn't use SP1?) I have a feeling that if I used my DVD of Vista (an academic version which I got within the first week of release) and the SP1 standalone pack, I could reproduce these scores on any system in the world. I have already done so on two different Rampage Extremes and a P5E64 Evolution, fresh installs each time. There is really nothing special I even have to do. The refresh rate helps, but even without it my scores are still "high". Every damned time. And it is not a bug, because I've been running 03 absolutely non stop for the past two weeks (Jason knows), and my higher scores visibly have much higher framerates than lower ones. e.g. my opening of GT2 went up from ~1300 to ~1500 after SP1 at 4.5GHz, along with similar gains in each of the other 3 tests, so there is actual performance gain here. Not a black screen or a bugged FPS counter.

As I said earlier, I have always felt that this is some software issue, and not a matter of bugging. The black screen bug is indepedent, and a convenient excuse, but not the true issue at hand. All it's done is added to the confusion, and right now, we have to forget about it and ignore it. You guys are not bugged low, and the few of us with high scores are not bugged high. That would be like saying that people that used Win2K for 01 back in the day were "bugged high" because others used WinXP. The people with low scores seem to be able to run them consistently as long as they want, as can the people with high scores. Neither group is bugged, because there are two distinct groups, and each group shows a lot of consistency.

[T_M]

Has anyone tested with multi-PSU's to check if they are starved of power?
Tonight i found that my 1kW silverstone couldnt handle my E8600/CF4870X2/RampageExtreme/Waterpumps/Fans/Etc during Canyon '06 and high air clocks :s

[Gautam]

My single PCPC 1KW-SR has been fine up at 5.6-5.7 on DI. But for LN2 I always use two PSU's.

[massman]

To be honest, I'm fed up with this situation. It's very hard to call certain scores bugged and others not just because less people are able to repeat the good numbers. As the cause of the low or high bugged scores is not known, I tried to find out which of the two is bugged.

As we all know that 3DMark03 is all about the GPU power, I ran through the HWBot benchmark data and filtered out all results since the Ati 2900XT of the high-end cards. Note that some data, such as of the CFX (which indicated 3x CF) may not fit the graphs because of the lack of results).

First of all, I wanted to know how the different cards, going from 2900XT to 4870X2, scaled over time in different benchmarks. I left out 3Dmark01 and Aquamark because those two are too cpu biased.



In comparison to single 3870:

After that, it was time to have a look at 3DMark03 in detail, especially to see how the 3DMark subtests scale.







Now, the problem with these graphs is that they don't really show the exact performance scalingThe only thing these show that the newer you card, the higher your score and the more cards, the higher your score.

That's why I re-used the data and analysed the way they behave in comparison to a single version of the series.



Now the problem with this graph is that it doesn't show how the cards scale in terms of multiple cores, but in terms of multiple cards. That's why I drew another graph indicating the scalability over cores.



As you can see, there's something seriously wrong with the X2 CF scaling: GT2 and GT3 are way too low! Why?

The first graphs in this post show the scaling how it should be, because I used all the data from HWBot that was available. The performance scaling graphs are in other words based on thousands and thousands of results. The subtest detail graphs are only the top5 of each category. Why? Because those are the results that are questioned. Now, in the X2 CF category only the score of Spl is bugged high, which means that the four others are bugged low. The incorrect subtest scaling is therefor visible, unlike what we see in the single X2 category, in which most scores are in fact not bugged. In other words: we should indeed be hitting 180k with X2 CF, not 160k.



This graph shows the poor GT2 and GT3 scaling with X2 CF. This is the data of the top5 in hwbot.

Please know that scores made with the blackscreen bug are indeed bugged and shouldn't be made public!

[Gautam]

Are you using averages for those graphs or what?

Be careful in throwing around the term "bug".

[massman]

Yes, always averages. The graphs I use to show how the different cards should scale use averages of ALL results in the hwbot database. The graphs I use to show that the X2 CF top results are bugged low are average of the top5 scores in that particular category, not the average of all scores.

Please don't get all semantical on the term 'bug'. I can call it anomality if you like, it doesn't change the fact that I mean that the scores that are low are the ones that are abnormal, not the ones that are high.

[Gautam]

I'm not, the only thing is that when you use the word bug it makes people think that if they can rerun the bench enough times, they'll get a "normal" score.

Great info though...now the challenge is to narrow down what the cause is. For me it was SP1 but I guess it isn't for anyone else. :-/

[STEvil]

Maybe it is PSU related.

Remeber the G92's having issues with some multi-rail PSU's producing instabilities, low clocks, or triggering nVidia power sentry? Maybe something similar is happening here?

[T_M]

Has anyone tested with multi-PSU's to check if they are starved of power?

Anyone?

[Gautam]

Scores don't change at all here between 1 and 2 PSUs Tim.

[T_M]

Yes, but you arent "bugged" low

[Achill3uS]

when i tested cfx last week, used 2 psu coz with one (1KW ultra x3) sometimes ive got crash on low clocks, but it could handle the cards when i use only aircooling, so get an other for the cards (asus 750w) and after i could go easy higher near of my cpu's and card's limits, but it doesnt gave me bugged scores anytime used vista64 sp1, two psu or one haha, so my answer is yes, they are starve for more power...
here is my score: 143152, GT2 is also under 1300fps like Andre's non bugged results

[hipro5]

I had my FIRST "black screen bug" at 3.6GHz and ONE 4870x2 DEFAULT on WinXP SP2 so I understood what you guys are saying.....
I got 1245fps on GT2 and it happened ONLY AFTER I installed the intel inf file......
Can someone check it out please cuz I don't have that much time.....
I ALSO gained some fps with the intel inf.....