4870X2 3d mark 03 bugged score thread

[NCspecV81]

http://www.xtremesystems.org/FORUMS/...74#post3417774

eva - there are good cards and bad cards.

Whats your MemIO temps on that card (GPU-Z and Rivatuner can show you)?

I don't believe this one bit. My cards do it randomly. I've had high and both with both cards, and even on the same boot.

[STEvil]

Yes, and there are good cards that will give good results often where bad cards do not.

[eva2000]

eva - there are good cards and bad cards.

Whats your MemIO temps on that card (GPU-Z and Rivatuner can show you)?

~34-38-43C at idle IIRC

http://www.xtremesystems.org/FORUMS/...74#post3417774

I don't believe this one bit. My cards do it randomly. I've had high and both with both cards, and even on the same boot.

interesting will give that a try

thanks

[NCspecV81]

Yes, and there are good cards that will give good results often where bad cards do not.

If this was the case, then bad cards would stay bad. Not randomly be good and bad. I can say without a doubt, that it's not this lame concept of good and bad cards. It's random at this point. I've booted between runs (3-5 minutes at most) and got high and low runs. If I had bad cards, then my score would always be low. I only have 2 cards, not some endless supply of hand picked X2's to pilfer through.

My cards have produced high and low runs. I've had both types of runs without changing anything other than just rebooting. That means the same clocks, fan speeds, orientation on the board, etc.

http://service.futuremark.com/compare?2k3=5975982

Would you consider that a high run? If it were the case were good cards are such, and bad cards are bad, how does it explain the randomness with the same cards? If you check the thread I linked to you will notice some scores I had to show some other variables in settings. I'll link you to an identical high run with the SAME cards in the SAME orientation, which can be found here: http://www.veoh.com/videos/v16462392Hr9DFged - I did not use the enable gpu scaling on this run. So approximately 10k of difference with the same cards.

[STEvil]

Could be just like CPU's. Using Q6600 G0's for example: some go high with low vcore, some dont. Some react to vcore differently under the cold, some dont.

My 4870X2 produces high runs very regularly. Lots of people in this thread have cards that produce low runs very regularly. Some have cards that swap almost 50% each time.

[RLM]

I don't think the memio temps are bad here, and my card still plays up in 03.

RLM

[STEvil]

I have about a 10c difference between DisplayIO/Shader and MemIO. Might it be a good idea to establish what the average spread is for everyone first?

[NCspecV81]

Could be just like CPU's. Using Q6600 G0's for example: some go high with low vcore, some dont. Some react to vcore differently under the cold, some dont.

My 4870X2 produces high runs very regularly. Lots of people in this thread have cards that produce low runs very regularly. Some have cards that swap almost 50% each time.

yes but you don't magically begin to have high cpu clocks with lower voltages in the same system, given all the extra variables are the same. I'm using the same card...the SAME CARDS to produce both high and low runs.

[RLM]

I have about a 10c difference between DisplayIO/Shader and MemIO. Might it be a good idea to establish what the average spread is for everyone first?

Look first off I'm not disputing your theories and I'm certainly no expert.

It seems to me something's a bit flakey with these cards. A component maybe that's just not up to scratch? but hey what do I know.

That screengrab I posted was from an other thread of mine on i4memory and related to something else. I didn't prepare it especially. Can you elaborate on what you mean by an 'average spread'.

RLM

[STEvil]

Look first off I'm not disputing your theories and I'm certainly no expert.

It seems to me something's a bit flakey with these cards. A component maybe that's just not up to scratch? but hey what do I know.

That screengrab I posted was from an other thread of mine on i4memory and related to something else. I didn't prepare it especially. Can you elaborate on what you mean by an 'average spread'.

RLM

Mine is 10c hotter if I leave the fan at stock (it never goes above ~28% fan speed). If I run it at 50% then 7-8c is average. This is in general use and 3d03 tests, ~28c ambient.

Does yours and other people's show the same factor of temperature increase for MemIO at similar settings or is it higher or lower (thus indicating a spread)?


NCSpec - I am also using only one card. Yes I agree that with a single CPU with all variables set constant there should be no change. However I was pointing out that not all parts of one line may perform the same. Possibly we are witnessing a symptom of the issue causing "good" and "bad" runs which may explain or enlighten why there are "good" and "bad" cards. Could be that maybe sideport is more than just a driver setting? Maybe there's a BIOS bug?

[NCspecV81]

Mine is 10c hotter if I leave the fan at stock (it never goes above ~28% fan speed). If I run it at 50% then 7-8c is average. This is in general use and 3d03 tests, ~28c ambient.

Does yours and other people's show the same factor of temperature increase for MemIO at similar settings or is it higher or lower (thus indicating a spread)?


NCSpec - I am also using only one card. Yes I agree that with a single CPU with all variables set constant there should be no change. However I was pointing out that not all parts of one line may perform the same. Possibly we are witnessing a symptom of the issue causing "good" and "bad" runs which may explain or enlighten why there are "good" and "bad" cards. Could be that maybe sideport is more than just a driver setting? Maybe there's a BIOS bug?

Indeed, I think as well there is something going on under the covers of bioses and drivers. I don't think its a heating/cooling issue on the plx or just a good or bad card. There's soemthing going on that's way more intricate than that.

[massman]

http://www.xtremesystems.org/FORUMS/...74#post3417774

I don't believe this one bit. My cards do it randomly. I've had high and both with both cards, and even on the same boot.

It's definitly not the setting you pointed out in that thread, because you don't have the symptoms of this issue, which are: GT2 and GT3 much lower. 30FPS is nothing.

If this was the case, then bad cards would stay bad. Not randomly be good and bad. I can say without a doubt, that it's not this lame concept of good and bad cards. It's random at this point. I've booted between runs (3-5 minutes at most) and got high and low runs. If I had bad cards, then my score would always be low. I only have 2 cards, not some endless supply of hand picked X2's to pilfer through.

Seems to me that you want it to be something else than good or bad cards. How can you be so sure about this issue when you're replying in a thread that contains 20+ pages of possible solution and yet no one but Hipro5 has been able to find a cure for it. That's just not right.

What the actual problem is, is at the moment unknown and the only current explanation is the concept of good and bad cards. Come up with a better theory or accept the current one, which actually gives a 100% workable solution. The only problem is that the solution currently involves money.

My cards have produced high and low runs. I've had both types of runs without changing anything other than just rebooting. That means the same clocks, fan speeds, orientation on the board, etc.

Mines has too, although I have found a certain point at which I can switch between runs, which has nothing to do with any form of overheating whatsoever. Overclocking the FSB helped me to go from low to high runs.

http://service.futuremark.com/compare?2k3=5975982

Would you consider that a high run? If it were the case were good cards are such, and bad cards are bad, how does it explain the randomness with the same cards? If you check the thread I linked to you will notice some scores I had to show some other variables in settings. I'll link you to an identical high run with the SAME cards in the SAME orientation, which can be found here: http://www.veoh.com/videos/v16462392Hr9DFged - I did not use the enable gpu scaling on this run. So approximately 10k of difference with the same cards.

If those two scores are actually done on the same system, frequencies, tweaks, ... than there's something wrong, but not this issue. This issue, as I already pointed out, deals with the GT2/GT3 being either very low or high (high = normal in fact), while you go from 126k to 138k where:

-GT1: 1160 -> 1300
-GT2: 1110 -> 1240
-GT3: 891 -> 961
-GT4: 904 -> 987

GT1 indicates a change in cpu frequency, GT4 indicates a change in gpu frequency. The increase in GT2 and GT3 can be explained by an increase of cpu and gpu frequency.

Could be that maybe sideport is more than just a driver setting? Maybe there's a BIOS bug?

The sideport function of the 4870X2 cards was apparently fused off the cards by Ati because enabling it caused cards to use much more energy, while the sideport performance scaling was very poor.

[STEvil]

The sideport function of the 4870X2 cards was apparently fused off the cards by Ati because enabling it caused cards to use much more energy, while the sideport performance scaling was very poor.

Who do I contact for a refund?

[NCspecV81]

It's definitly not the setting you pointed out in that thread, because you don't have the symptoms of this issue, which are: GT2 and GT3 much lower. 30FPS is nothing.



Seems to me that you want it to be something else than good or bad cards. How can you be so sure about this issue when you're replying in a thread that contains 20+ pages of possible solution and yet no one but Hipro5 has been able to find a cure for it. That's just not right.

What the actual problem is, is at the moment unknown and the only current explanation is the concept of good and bad cards. Come up with a better theory or accept the current one, which actually gives a 100% workable solution. The only problem is that the solution currently involves money.



Mines has too, although I have found a certain point at which I can switch between runs, which has nothing to do with any form of overheating whatsoever. Overclocking the FSB helped me to go from low to high runs.



If those two scores are actually done on the same system, frequencies, tweaks, ... than there's something wrong, but not this issue. This issue, as I already pointed out, deals with the GT2/GT3 being either very low or high (high = normal in fact), while you go from 126k to 138k where:

-GT1: 1160 -> 1300
-GT2: 1110 -> 1240
-GT3: 891 -> 961
-GT4: 904 -> 987

GT1 indicates a change in cpu frequency, GT4 indicates a change in gpu frequency. The increase in GT2 and GT3 can be explained by an increase of cpu and gpu frequency.



The sideport function of the 4870X2 cards was apparently fused off the cards by Ati because enabling it caused cards to use much more energy, while the sideport performance scaling was very poor.

you took everything I said out of context. the 138k was at a different frequency, which I won't deny. It was not a low bugged run at 4.6GHz to get 8pts shy of 139k, but the 4GHz runs I posted in another thread are low runs. However, that thread was simply pointing out the difference in score by forcing a certain amount of gpu scaling. I was hoping it would be able to shed some real light, and facilitate great minds to think about the possibilities using known facts on data with the low bugged runs. I NEVER mess with my PLX chip and my cores are mostly within a +/- 5C degree delta between each other on AIR, and my scores will vary between high and low runs. Thats why I suggest that the PLX chip is not the issue as other guys on extreme temps get both high and low runs without any plx heating.


However, just for s and giggles let me post another 4ghz run that's a high run that I have on video. Please explain the 10k in score on air with the SAME cards. I have done a similar run to the one I have posted in another thread at 119k w/ the forced gpu scaling and with the high run, and have netted 130k. Bad cards and an over cooled PLX doesn't seem to fit the bill with the data available on the table.

http://www.veoh.com/videos/v16462392Hr9DFged


same cpu; same cards; same system and settings:

[bench2die]

Do you guys find ncspecs runs of 143,xxx at only 4.7 suspicious? He is beating quads at 5.7-9 and duals at 6.2

[Gautam]

His run is fine, the others are bugged low. (Or were done with "bad" cards as the current working theory goes) This has been beaten to death in this thread. The proper score at 6GHz is in the 160k range, as was first shown by Kinc. (And heavily doubted even by myself at the time it was first done)

[dinos22]

no those scores look fine man

[George_o/c]

NCspecV81's run seem to be fine dinos ... the others are bugged low, just like Gautam pointed out

Hipro's pushing hard again, not with i7 though Using his Vcore+Vgpu maximizer with 4870X2s

[dinos22]

yeah thats what i said

[George_o/c]

LOL, I misinterpreted your post, sorry man

[Gautam]

Hipro's pushing hard again, not with i7 though Using his Vcore+Vgpu maximizer with 4870X2s

[George_o/c]

Yeah Gautam I suppose, we will see him back when he has fully tested his Vgpu+Vcore Maximizer with 4870 and 4870X2s ... Maybe first day of new year ? Don't know ...

[bartx]

I know, that this thread is a bit old, but I've got some new info. This problem is also on 4850/4870/4890 cards. So, in my opinion, it'll be related rather with the core, than with pcb issues. I've tested a few cards, my friend also tested one and here is how it looks:

4850 Gigabyte 512mb - always good score
4850 Asus 512mb glaciator from my friend - sometimes good score, but more often bugged
4870 sapphire 512mb which I had some time ago - always good score
4870 msi 512mb which I have now - sometimes good score, but more often bugged
4890 sapphire - always bugged score

I have absolutely no idea, how to make a good score on cards, which are able to make good/low score depends on run.

So, if it appears on a few cards with different PCBs, when only core is the same, I guess that it's related rather with the core.