open letter to ATi on gamers-depot

Status
Not open for further replies.
gotta ask why were the x800 pro vivo cores with 12 pipelines clockable higher then the x800xt pe but unlocked weren't able to go past the x800xt levels? The unlocked pipelines worked just fine at the x800 pro core clocks though. Core speeds have to do alot with the pipeline numbers because of......

12 pipelines draw less power than 16. When we apply your “random logic, anecdotal evidenceâ€￾ to this, it indicates that the ASIC’s are capable of going beyond 520MHz.

This is also the reason why ATi can produce a x800 regular at the same core clocks as the x800xl

This is the same reasoning behind the x700xt and why a x800 xl will have a max clock speed of around 400.

No, these aren’t’ related at all.
 
DaveBaumann said:
gotta ask why were the x800 pro vivo cores with 12 pipelines clockable higher then the x800xt pe but unlocked weren't able to go past the x800xt levels? The unlocked pipelines worked just fine at the x800 pro core clocks though. Core speeds have to do alot with the pipeline numbers because of......

12 pipelines draw less power than 16. When we apply your “random logic, anecdotal evidenceâ€￾ to this, it indicates that the ASIC’s are capable of going beyond 520MHz.

This is also the reason why ATi can produce a x800 regular at the same core clocks as the x800xl

This is the same reasoning behind the x700xt and why a x800 xl will have a max clock speed of around 400.

No, these aren’t’ related at all.

No they worked just fine up till you tried to go to the xt pe levels and beyond then they started artifacting because the other pipelines weren't able to go that high. Not that there wasn't enough power to them. Not enough power should cause a crash or a system lock up not artifacting. That means the pipelines have enough power but were not stable.

On off consist of at least 2 transistors or a gate that consists of a single FET that uses its capacitor to store the value for a short time. if there is too much voltage or too little voltage it might cause timing issues. And this would cause a hard crash not artifacting. This is for dram of course but it should be similiar to any type of silicon chip.

Supply and demand, there is/was a bigger market for Pros than there were faulty R420 cores. Same deal with NVIDIA, consumers want the 6800NU, but NV40 yields are improving so they bring out NV41 to satisfy that bracket.

But there are more then enough x800 pro's out there to begin with the demand was being met.
 
No they worked just fine up till you tried to go to the xt pe levels and beyond then they started artifacting because the other pipelines weren't able to go that high. Not that there wasn't enough power to them. Not enough power should cause a crash or a system lock up not artifacting.

Please show us your reasearch that bears this out. Also show us you extensive reasearch on the exact issues displayed for each and every modded PRO board.

Not all PE issues are related to the core as well, and some of the yield can be put down to board level issues, which is why they have changes supplier for R430 and R480 boards - timing issues can be caused by board issues as well.

Anyway, yes, of course there were many R420 chips that couldn't reach PE speeds, but that doesn't bear any relation to R430. All of this is more unrelated issues since we are talking about different chips on different processes on different boards - just because an R420 chip couldn't hit 520 when it can 500 doesn't mean that R430 won't go to 400MHz OK, especially given that this is the only clockspeed nin they are selling them at.
 
DaveBaumann said:
No they worked just fine up till you tried to go to the xt pe levels and beyond then they started artifacting because the other pipelines weren't able to go that high. Not that there wasn't enough power to them. Not enough power should cause a crash or a system lock up not artifacting.

Please show us your reasearch that bears this out. Also show us you extensive reasearch on the exact issues displayed for each and every modded PRO board.

Not all PE issues are related to the core as well, and some of the yield can be put down to board level issues, which is why they have changes supplier for R430 and R480 boards - timing issues can be caused by board issues as well.

Anyway, yes, of course there were many R420 chips that couldn't reach PE speeds, but that doesn't bear any relation to R430. All of this is more unrelated issues since we are talking about different chips on different processes on different boards - just because an R420 chip couldn't hit 520 when it can 500 doesn't mean that R430 won't go to 400MHz OK, especially given that this is the only clockspeed nin they are selling them at.



Well I was searching around the web this is what I found on this forum

http://www.beyond3d.com/forum/viewtopic.php?t=15811&view=previous

Exactly what I stated, Guden Oden seems have already tried lower clocks on a 9800 pro and by under clocking indeed did causes a hard crash and we all know that over clocked will cause a crash or artifacting.

http://forums.bit-tech.net/showthread.php?s=1aae8992176fff97dfaf7a451251a158&p=825660#post825660

x800 pro to xt pe flash hard crash, temperatures don't go above 80 C so its not a temperature issue. The unlocked pipes are bad are they really bad? or the clocks that are affected the pipelines? Its the voltage as this later quote from this guys experiment shows.

flashing it back to Pro and it works perfectly, even when overclocked to XT speeds. i also tried a BIOS that makes it a Pro, but unlocks the extra pipelines so Pro @ 16P 475/450 worked Pro @ 16P overclocked to 500/500 and 520/560 both worked (doesnt this make it an XT!?)

Flashed back to pro unlocked pipes lower voltage unlocked pipes worked fine.

Why is this happening voltage is affecting the pipeline performance causing a hard crash with the xt pe bios

http://www.driverheaven.net/showthread.php?p=426003

Samething here.

http://www.hardcoreware.net/forum/showthread.php?p=311162#post311162

This guys stating that sapphire x800 pro vivos artifact more then the other brands.

http://www.rage3d.com/board/showthread.php?mode=hybrid&t=33783530

another artifacting after sometime.

There are a few of these out there.


Oh I didn't know they changed suppliers thats interesting ok thats good.
 
As I thought. A limited number of forum posts is not any kind of sample to build and understand any kind of yeild and issue analysis. Some of these even go to contradict some of your earlier points.
 
digitalwanderer said:
Razor1 said:
This guys stating that sapphire x800 pro vivos artifact more then the other brands.
And mine works just perfectly with 16 pipes enabled at XT PE speeds, what does that tell us? :rolleyes:

my x800xt pe works just fine at 650/600 whats that tell us ?!?
 
jvd said:
digitalwanderer said:
Razor1 said:
This guys stating that sapphire x800 pro vivos artifact more then the other brands.
And mine works just perfectly with 16 pipes enabled at XT PE speeds, what does that tell us? :rolleyes:

my x800xt pe works just fine at 650/600 whats that tell us ?!?

You two have too much disposable income... ;)
 
waitaminute... you mean anecdotal evidence from an anonymous forum can't be used as the basis for statistical analysis? fuck! *throw current progress on thesis in the trash*
 
The Baron said:
waitaminute... you mean anecdotal evidence from an anonymous forum can't be used as the basis for statistical analysis? fuck! *throw current progress on thesis in the trash*
Don't be too hasty, it's just CMU....not like it's a real school or anything Baron. ;)
 
DaveBaumann said:
As I thought. A limited number of forum posts is not any kind of sample to build and understand any kind of yeild and issue analysis. Some of these even go to contradict some of your earlier points.


12 pipelines draw less power than 16. When we apply your “random logic, anecdotal evidenceâ€￾ to this, it indicates that the ASIC’s are capable of going beyond 520MHz.

This is why I wanted to see your reasoning behind this next quote why is this happening? Its not a power issue. This is from a guy on Rage 3d btw. I know there were 3 or 4 people on Hardocp that got the samething, but at the moment can't find the thread, still looking for it.

I hope this is usefull. I did a quick test calcullation between your 12 pipes & 16 pipes benchmarks.

The Vivo's are failed XT's, so you have a higher change that they won't oc as high as Pro's.

12 pipes pro ( 19.53 )

10741 - 550 - 30mhz
11327 - 580 - 33mhz
12030 - 616 - 36mhz


16 pipes ( 20.72 )

10774 - 520
11333 - 547
12018 - 580

Note: this calcullation is based on the same memory speed ( 450 ), and shows the difference of 12 vs 16 pipes & there speed.

As expected, the higher you clock the XT, the more the difference needed is to overcome by the pro. As you start with a 30mhz difference, this grows larger the more you speed up the XT.

But it also shows something intresting. People who have pro's with pipes they can't enable, but who are great oc'ers, can take on XT speeds widout problems ( core based ).

I'll say it again, this was just a test to see how XT rejected cards ( who bearly reach XT speeds ) handle vs Pro cards ( who may oc higher ). Some people may have VIVO's who may oc beyond XT default speeds.

If you read through the threads, you will see this has happend to more then one person. And the volt mod was never neccesary that xt pe bios included the increased volts. Also there were many people that were having artifacts in those threads too.

Way to be a gentleman Dave, another person and me stated the problems with using extra power or lower power for that matter would cause a hard crash, which you plesantly said was false, thank you for your confirmation.

limited number yes but will all these people be making this up? Am I just saying that its going to happen to every single ATi card, no but in all likely hood with ATi's track record possibly might happen with the x800xl now you've stated that the timing issues were also caused by the board and ATi has switched manufacturers, I didn't know that which I reciprocated was a good change.

Do you want more links I can give you more links

http://www.rage3d.org/board/showthread.php?t=33766704&page=8&pp=30

http://www.driverheaven.net/printthread.php?t=55180

http://forums.techpowerup.com/archive/index.php/t-630.html

http://www.xtremesystems.org/forums/showthread.php?t=46664

And as I said I can find more if you want.....

But how many overclocking forums are there? not a whole lot so you aren't going to get a huge number of threads.
 
digitalwanderer said:
The Baron said:
waitaminute... you mean anecdotal evidence from an anonymous forum can't be used as the basis for statistical analysis? fuck! *throw current progress on thesis in the trash*
Don't be too hasty, it's just CMU....not like it's a real school or anything Baron. ;)
heh, I'm getting the hell out of engineering anyway.
 
digitalwanderer said:
The Baron said:
waitaminute... you mean anecdotal evidence from an anonymous forum can't be used as the basis for statistical analysis? fuck! *throw current progress on thesis in the trash*
Don't be too hasty, it's just CMU....not like it's a real school or anything Baron. ;)

If you want to buy 30 vivo cards and test it out be my guest, because I don't :D but I'm going on what others have been saying. I don't see why even if its only 10 people that did these types of benches why the others wouldn't get the same results. Since you have a modded x800 give it a try flash it back, see what happens. JVD you might want to try this too. Since you have a great overclocking card. See what the max clocks for a x800 pro vivo are well don't need to go max but see if it goes past what the modded xt pe levels you have right now. Forgot to add in there make sure your voltage doesn't exceed the x800 pro levels too.
 
Razor1 said:
But how many overclocking forums are there? not a whole lot so you aren't going to get a huge number of threads.

And all of this is very limited information about another chip based on another process - in reality it bears little relation. If you really want to look at 130nm low-k vs 110nm then look at the minimum clockspeeds for the parts and see how they differ - there is a 20% difference between the minimum clocks on the 16 pipe versions, greater than the gains cited for low-k, that increases when you look at R480, which has some layout tweaks for further speed optimisation (and which do you think R430 is going to follow?).

Basically, we'll have to wait and see, but when you look at the indicitors that pertain specifically to R430, what do they suggest already?
 
DaveBaumann said:
Razor1 said:
But how many overclocking forums are there? not a whole lot so you aren't going to get a huge number of threads.

And all of this is very limited information about another chip based on another process - in reality it bears little relation. If you really want to look at 130nm low-k vs 110nm then look at the minimum clockspeeds for the parts and see how they differ - there is a 20% difference between the minimum clocks on the 16 pipe versions, greater than the gains cited for low-k, that increases when you look at R480, which has some layout tweaks for further speed optimisation (and which do you think R430 is going to follow?).

Basically, we'll have to wait and see, but when you look at the indicitors that pertain specifically to R430, what do they suggest already?

Well after this discussion I'm with you :) will have to wait and see but it should be just fine, I wasn't aware the baords were also the problem.
 
Razor1 said:
And the volt mod was never neccesary that xt pe bios included the increased volts.
I'm going to chime in here...

No ATI bioses on ANY of the Radeon line has any relation or affect to any of the regulated voltages on the card. This includes (but is not limited to) GPU, VDD, VREF and VDDQ voltage lines.

THE ONLY WAY TO MANIPULATE GPU, VDD, VDDQ OR VREF VOLTAGES ON A RADEON CARD IS TO PHYSICALLY ALTER THE CARD IN SOME FORM. Bios alterations do not change the voltage, nor does AGP bus voltage from your computer's bios.

Just thought I'd clear that up. The X800XTPE's run 1.4v GPU; my X800Pro Vivo was running 1.3v when it came to me. With a little pencil graphite and some squinting, I was able to bump my core voltage to 1.4v and was able to attain a "stock" speed of 520/560. Not much more granted, but certainly enough to be a real XTPE when it was running the real voltage.
 
Albuquerque said:
Razor1 said:
And the volt mod was never neccesary that xt pe bios included the increased volts.
I'm going to chime in here...

No ATI bioses on ANY of the Radeon line has any relation or affect to any of the regulated voltages on the card. This includes (but is not limited to) GPU, VDD, VREF and VDDQ voltage lines.

THE ONLY WAY TO MANIPULATE GPU, VDD, VDDQ OR VREF VOLTAGES ON A RADEON CARD IS TO PHYSICALLY ALTER THE CARD IN SOME FORM. Bios alterations do not change the voltage, nor does AGP bus voltage from your computer's bios.

Just thought I'd clear that up. The X800XTPE's run 1.4v GPU; my X800Pro Vivo was running 1.3v when it came to me. With a little pencil graphite and some squinting, I was able to bump my core voltage to 1.4v and was able to attain a "stock" speed of 520/560. Not much more granted, but certainly enough to be a real XTPE when it was running the real voltage.

Right, it wasn't the bios flash that increased the volts some of the x800 pro vivos came with 1.4 volts.

http://hardforum.com/printthread.php?t=803583
 
Status
Not open for further replies.
Back
Top