Saphire Platinum X800 XT at 500Mhz gpu core?

[SunTzu69]

Apparently, Saphire is selling AGP 8X X800XT Platinum Edition cards that are not what the packaging suggests. I have confirmed this on two websites where people have received boxes with a Platinum sticker on it, after having purchased what they thought was a good deal on the card, and what they actually received was a plain X800XT clocked at 500MHZ Gpu core.

There is alot of frustration on the part of some of these customers. To add to the confusion, some of the boxes contain blue pcb cards with the Saphire logo, and others a red ATI card with the Minotaur logo. I do not understand exactly what is going on, but be warned that if you order a Saphire X800XT PE, you might get a card that can't even overclock to 520Mhz !!

If anybody has any additional information please post it here. Links below:

http://www.ncix.com/forums/index.php?mode=showthread&msg_id=597811&threadid=597811&forum=101&product_id=12417&overclockid=0#msg597811

http://www.3dchips-fr.com/News/commentaires/200409151_0.html

Edit: Yet another link that seems to prove this !!!
http://www.neoseeker.com/resourcelink.html?rlid=86055

[dreIU]

where are peopole buying these cards from?

[SunTzu69]

They seem to be showing up all over the place !! Just recently.. it seems like ATI has attempted to solve the lack of success in producing the PE by issuing these new AGP 8X X800XT's. The only thing wrong with that IMO is that at least one manufacturer (Saphire) seems to be puting the PE label on them although they don't run at the same clock speed, and will not go near PE overclocking speeds.

Very discomforting.

From the last link posted above:

" Its seems as though the board’s that failed Platinum Edition specification were being dumped into an X800 PRO specification – now, though, ATI are relenting on their earlier product definitions and instead of moving these products into the PRO specification they are moving them into a high specification X800 XT (500MHz core, 500MHz RAM) specification.

Here we’re looking at one of the first of the AGP ATI Radeon X800 XT’s to hit the market, made available by Sapphire.

...

The X800 XT specification follows closer to the X800 XT Platinum Edition specification in terms of core configuration, with the full 16 pipelines enabled and only 20MHz lower core speed, however its memory specifications are closer to the PRO with a memory speed 50MHz higher, but 60MHz lower than the XT Platinum Edition.

Unsurprisingly the Sapphire X800 XT board bears an uncanny resemblance to the reference X800 XT Platinum Edition and its very difficult to spot any visual differences between the boards. This is probably all to be expected as all the X800 boards are made by one manufacturer at the moment, with the main board vendor differences lying in the cooler they use - in this instance Sapphire are using ATI's reference cooler as well, just with a Sapphire sticker placed over ATI's normal image.

As is presently ever the case with any graphics board that uses GDDR3, the memory used is Samsungs K4J55323QF GDDR3 RAM. One thing of note is that whilst the specification of the X800 XT are 500/500, which would mean that 2ns memory would be within specification, the board has 1.6ns RAM which would be good for up to 600MHz, according to Samsungs ratings - this further indicates the board was built for Platinum Edition specifications, but didn't quite manage that."

Having read through most of the review, it seems that we can no longer say that all AGP 8x X800XT's are Platinum Editions. In fact we might be seeing more non-PE's hit the market in AGP format than the PE's. We will likely pay less for these cards, and there will be in stock quite a bit more than the PE's, but they lack somewhere from 5%-10% of the real PE's performance.

More importantly, beware of the Saphire PE's !! Some seem to actually be these X800XT's. NOT THE SAME THING...

[KnownKiller]

I just recently installed the new Saphire X800XT non PE and it says in the device manager and the control panel that its a X800?? Is that whats its supposed to show? Not X800XT?

Kinda worried since i took the heatsink off to add my water block and it even shows X800 on the GPU.

Bought it from newegg..

P.S. In the controll panel it show that 16 pipelines are enabled and the defualt speed looks correct at 500/500. I can overclcok it so far on the first night at 542/606.

[Free_Loader]

That should be okay to me because the X800Pro I got did well but not as well as you are on the preliminary run....look at sig!

[micamica1217]

SunTzu69,

yes I've noticed the same lower clocks from some replys that I got on the matter....
it's a shame that ATI would lower the clocks slightly.

yet the results from real world gaming between the two is sooo close that I would not think more then a moment on it.
in fact, once OCed, it's the same.

mica

[OBLIVIONLORD]

Bout the same as paying for a 3.0 P4 and getting a 2.8ghz instead. Even though it can clock to the same speed with near flawless results is not the point. The fact that your paying for hardware that is not up to spec as stock speed which is stated on the box itself. That gets me going.

[hUMANbEATbOX]

Bout the same as paying for a 3.0 P4 and getting a 2.8ghz instead. Even though it can clock to the same speed with near flawless results is not the point. The fact that your paying for hardware that is not up to spec as stock speed which is stated on the box itself. That gets me going.

except your not paying for performance you aren't getting. xt's SHOULD be cheaper than xt-pe's, its just a shame sapphire is mislableing these cards as PE's in some cases, if those claims were in fact true.

i for one am now in the market for a x800XT instead of a pro. this works out well for me. ;)

[SunTzu69]

SunTzu69,

yes I've noticed the same lower clocks from some replys that I got on the matter....
it's a shame that ATI would lower the clocks slightly.

yet the results from real world gaming between the two is sooo close that I would not think more then a moment on it.
in fact, once OCed, it's the same.

mica

My only concern, being in the market to purchase one of these cards, is that the Platinum Edition cores do have more overclocking potential. ATI has stated this and I did read it online today. If fact, sites such as NCIX.com are indicating that there is no garantee that the X800XT will even match PE speeds.

I am not saying you are wrong, because I have no real experience in attempting this, but I would not assume that they are the same once OCed ATM.

[SunTzu69]

except your not paying for performance you aren't getting. xt's SHOULD be cheaper than xt-pe's, its just a shame sapphire is mislableing these cards as PE's in some cases, if those claims were in fact true.

i for one am now in the market for a x800XT instead of a pro. this works out well for me. ;)

You are right. This could be a great opportunity for many to get their hands on a slightly lesser card (But still fantastic in comparison to the X800 Pro) at a lesser price and better value. The only danger is all of the people who will be ordering what they believe is ATI's fastest card at a lower price, when it is not the fastest. And this is especially true if some manufacturers are really puting a "Platinum" sticker on some of these boxes.

[Sentential]

Well ATi is getting hammered by the 6800U and GT. Since their yields are so bad at the moment, they have no other choice. However, despite these cards being inferior, for alot of people its great cuz the 16 pipelines are functional..

Hell I'd buy one.

[felinusz]

However, despite these cards being inferior, for alot of people its great cuz the 16 pipelines are functional.

Hell I'd buy one.

I second that - if Sapphire hadn't gone and bungled up with the 'PE' stickers that shouldn't be on the boxes, there would be no bad vibe about these cards at all, look at it this way:


~ Yields of the Samsung RAM the XT-PE uses are terrible, meaning ATi cannot produce enough fully-functional XT-PE cards to satisfy the massive demand.

~ This XT non PE card will effectively quench the demand somewhat, while also offering a lower cost, *in stock*, and high performance option, which is something ATi (as well as the market) needs right now - people *will* buy these.

~ This card is essentially an X800 Pro equipped with slightly underperforming 1.6 ns RAM that's been pipe modified for you; what with the apparant hit-or-miss unreliability of VIVOs pipemodding, this is a welcome thing for me to see - I'd rather buy one of these then a VIVO Pro that may or may not pipemod for me.

~ I'd be willing to bet cold hard cash that after a VDDQ and VDDR volt mod, the RAM on these will fly. Let's wait and see if this is the case - at least it isn't 2.0 ns RAM.


One thing of note is that whilst the specification of the X800 XT are 500/500, which would mean that 2ns memory would be within specification, the board has 1.6ns RAM which would be good for up to 600MHz, according to Samsungs ratings - this further indicates the board was built for Platinum Edition specifications, but didn't quite manage that.

That article should add that these cost less then an XT-PE, which isn't even in stock anywhere. :beer:

[hUMANbEATbOX]

Well ATi is getting hammered by the 6800U and GT. Since their yields are so bad at the moment, they have no other choice. However, despite these cards being inferior, for alot of people its great cuz the 16 pipelines are functional..

Hell I'd buy one.


its not ATI'S yields that are low afaik. they don't make the ram.

although i'm confused about 1.6ns ram being on regular XT's. why?? not that i would complain that i didn't get 2.0ns.

[felinusz]

I would *guess* that these are cards which failed to live up to XT-PE specifications, and didn't pass ATi's 'rigorous' quality testing.

As with most anything computer-hardware related, yields are probably to blame here.

Since this Samsung RAM has such a small viable fabrication yield, I'm sure there's plenty of perfectly useable RAM produced that just doesn't live up to the spec of 600 MHz, and which fails the highest level of quality testing accordingly.

It's comparable to an Athlon XP Barton Core 3200+, and an Athlon XP Barton Core 2500+. The two use the same core, but one (as far as we know anyways) failed to meet with the testing requirements of a 3200+ chip, and was packaged as a slower 2500+ processor instead.

This is likely the same thing happening, just with video card RAM ;)

The thing to remember is that these parts are still useable - quality testing is often just checking that a piece of hardware can hold it's specified stock speed with stability throughout a range of different conditions (i.e. very high temperatures). If a piece of hardware can't handle quality (temperature) testing, it can't be sold; your average computer user has a very hot case with no airflow that is never dusted inside, or even opened. If high temperatures caused instability at stock speeds, a company like ATi would have a miserable time dealing with all the complaints and RMAs.

With proper cooling, and care (and a voltage modification), a piece of hardware that couldn't hold it's specified speed at 70 degrees celcius under load, during quality testing, (X800 XT) will overclock and run very nicely for the enthusiast user, although theoretically not as well as a piece of hardware that passed testing (X800 XT-PE).

The above is 100% speculation; it is what I personally think to be the case, but may be incorrect.

[Sentential]

I think its more an issue of them avoiding GC16 to try and get production numbers up. We will see how effective this is.

[KnownKiller]

Can anyone here answer my question??


I just recently installed the new Saphire X800XT non PE and it says in the device manager and the control panel that its a X800?? Is that whats its supposed to show? Not X800XT?

Kinda worried since i took the heatsink off to add my water block and it even shows X800 on the GPU.

Bought it from newegg..

P.S. In the controll panel it show that 16 pipelines are enabled and the defualt speed looks correct at 500/500. I can overclcok it so far on the first night at 542/606.

[Sentential]

Can anyone here answer my question??
Yea, you can clock it as far as you want so long as its stable and doesnt artifract.

[micamica1217]

My only concern, being in the market to purchase one of these cards, is that the Platinum Edition cores do have more overclocking potential. ATI has stated this and I did read it online today. If fact, sites such as NCIX.com are indicating that there is no garantee that the X800XT will even match PE speeds.

I am not saying you are wrong, because I have no real experience in attempting this, but I would not assume that they are the same once OCed ATM.


again, I have to agree with you to some degree....as my original reply was way to vague.

let me fist start by saying that while the original x800xt(pe) cores might be able to OC more then the newer ones, they still have one problem that most peeps don't know about.
they are memory bandwith limited...by alot.

that's why a 12 pipe x800pro, once OCed to say x800xt(pe) speeds, are so close to a x800xt(pe) in performance in real world gaming.
only once heavy SM2.0 is used, does the x800xt's break free from the pro's.
(and even then they can be close at times.)

(note the last link you gave for beyond3d....
the toxic x800pro with it's OCed memory (the core is not OCed "out of the box"), has far less problems almost keeping up with the x800xt and x800xt(pe).
in fact, the toxic almost has the same memory bandwith as the newer x800xt)

what we are not seeing is a pro card OCed to x800xt/x800xt(pe) speeds.
that is somthing that I have in my rig now....as my pipeline mod failed.
540/550 are my speeds with my card (sammy 2.0 mem).

yet, what realy worries me is the poor OCing of the 1.6 mem chips that come with the x800xt/(pe)'s...
now they suck in OCing compared to the 2.0's
you should feel lucky if you reach 560 speeds for the xt and 580 for the xt(pe).
and if you get slightly higher, don't expect a huge performance boost....
as the extra pipes realy need somthing like 620 speeds to realy fly (and more will only fether help the bandwith limitations).

even volt modding the card, doesn't seem to help OCing the memory that much, and this is the part that needs OCing the most.
the x800xt/(pe)'s would realy fly if there realy was memory fast enough to handle such a chip....to bad there isn't.


now as for saphire miss labeling the box....they have done it in the past, and I still don't like it.

mica

[Sentential]

Mica, you have the IDs wrong.

GC20 = 2.0ns = 1100mhz = GT / PRO
GC16 = 1.6ns = 1200mhz = XT / Ultra / (VIVO PRO)

GC20 ocs like ****, not 1.6, unless ATi is continuing the tradition of using ****ty cheap RAM.

[micamica1217]

Mica, you have the IDs wrong.

GC20 = 2.0ns = 1100mhz = GT / PRO
GC16 = 1.6ns = 1200mhz = XT / Ultra / (VIVO PRO)

GC20 ocs like ****, not 1.6, unless ATi is continuing the tradition of using ****ty cheap RAM.

I don't have the I.D.'s wrong....any 1.6ns chip that can at times not even reach 600 speeds is what I call a poor OCer.
and if the better ones can only OC to say 590 or so, then yeah, the 2.0ns chips are better OCers even if they don't always reach the same speeds.

I've seen 1.6ns chips OC to 600 without vmods....
I've seen 2.0ns chips reach 580 without vmods.

but a better way of looking at this is what most of us will get when OCing with out mods or extra cooling.
the x800xt(pe)'s seem to get about 580 from 1.6ns chips...4% OC from stock.
the x800pro's seems to get about 554 from 2.0ns chips...23% OC from stock.

I am only talking about ATI cards, as I only have few 6800ultra OCing results as of now.

mica

[Sentential]

I don't have the I.D.'s wrong....any 1.6ns chip that can at times not even reach 600 speeds is what I call a poor OCer.
and if the better ones can only OC to say 590 or so, then yeah, the 2.0ns chips are better OCers even if they don't always reach the same speeds.

I've seen 1.6ns chips OC to 600 without vmods....
I've seen 2.0ns chips reach 580 without vmods.

but a better way of looking at this is what most of us will get when OCing with out mods or extra cooling.
the x800xt(pe)'s seem to get about 580 from 1.6ns chips...4% OC from stock.
the x800pro's seems to get about 554 from 2.0ns chips...23% OC from stock.

I am only talking about ATI cards, as I only have few 6800ultra OCing results as of now.

mica

Well I belive that ATi has the timings in the BIOS incorrect. Because over on nV's side all the 1.6ns Ultras come close to 1.3ghz using that RAM when Oced. Most of us can get damn close to 1.2ghz on GC20

Ive seen people with 1250mhz+ On GC16. So Im not sure wtf ATi is doing wrong. Probably because it isnt heatsinked properly (and yes it does need it)

[micamica1217]

Well I belive that ATi has the timings in the BIOS incorrect. Because over on nV's side all the 1.6ns Ultras come close to 1.3ghz using that RAM when Oced. Most of us can get damn close to 1.2ghz on GC20

Ive seen people with 1250mhz+ On GC16. So Im not sure wtf ATi is doing wrong. Probably because it isnt heatsinked properly (and yes it does need it)

I can't speak about the timmings, as I won't change mine till I learn more.
they may just be really too agresive, and need to be lowered...easy to do with ATItool.
yet I agree that added cooling is realy needed at high speeds.

and no, the ATI cards are not cooling the ram at all with the stock cooler.

any added links to mem timmings would be nice dude.

mica

[Sentential]

I can't speak about the timmings, as I won't change mine till I learn more.
they may just be really too agresive, and need to be lowered...easy to do with ATItool.
yet I agree that added cooling is realy needed at high speeds.

and no, the ATI cards are not cooling the ram at all with the stock cooler.

mica

Well there in lies the anwser ;). If you notice nV has a heafty cooler for their RAM. In fact stock cooling is so good on nV that most of us are just adding AS5 and buying a copper replacement cooler. The stock cooling is almost as good as the Silencer, and with the new copper replacements that eVGA is selling, it should be equal or better.

[KnownKiller]

Sorry this question i ment..

I just recently installed the new Saphire X800XT non PE and it says in the device manager and the control panel that its a X800?? Is that whats its supposed to show? Not X800XT? Even in 3dmark it says X800..

[felinusz]

If it has all 16 pipelines unlocked by default, and has stock speeds of 500/500, then I would think that you're fine, and you've got the X800 XT that you payed for. It doesn't say "X800 Pro", and no "X800" card exists, so I'd assume it's a typo, or perhaps intentional to prevent confusion with XT-PE cards.

even volt modding the card, doesn't seem to help OCing the memory that much, and this is the part that needs OCing the most.

That's partially because DDR III *needs* a VDDQ overvolt in tandem with a VDDR overvolt to see any gains. Raising the VDDR by itself won't do much at all, which is what a lot of people are trying, leaving them dissapointed with their voltage mod.

Mica, you raised a very interesting point there about the memory, the poor overclocking results from the Samsung 1.6 ns seem to imply some very serious problems with GC16 yeilds.

Is Samsung the only fabricator of 1.6 ns memory?

[Sentential]

Is Samsung the only fabricator of 1.6 ns memory?
Correct, however it is a COOLING issue. Us 6800 guys are seeing damn near 1.3ghz on GC16. The stuff that is in circulation isnt poor OCer, its just that people do not take the proper measures to cool very very hot RAM.

Its not hot at stock, but once you OC it, boy does it heat up

However I will say that Samsung is having a hellova time trying to make GC16, so Im not surprised if ATi is buying the cheap / non-binned RAM.

I know for a fact that none of us on the 6800 camp is haveing any trouble getting excellent OCs on both GC16 and 20. Although the early batches of Ultras had issues, this was later corrected with a BIOS update direct from nV to soften the CAS latency, to I believe 5.

[micamica1217]

That's partially because DDR III *needs* a VDDQ overvolt in tandem with a VDDR overvolt to see any gains. Raising the VDDR by itself won't do much at all, which is what a lot of people are trying, leaving them dissapointed with their voltage mod.

Mica, you raised a very interesting point there about the memory, the poor overclocking results from the Samsung 1.6 ns seem to imply some very serious problems with GC16 yeilds.

Is Samsung the only fabricator of 1.6 ns memory?

yes, again I have to agree.....
here are some results from Kraig's x800xt(pe) as told by ViperJohn's notes:

(3) Core produced no artifacts that I could find ahead of lockup.
(4) 3DM01se Nature torture loop run at 630/624 clean and as typical produced the highest sustained core
and memory temperatures the card will ever experiance at the voltages and clocks run.
(4) Core die temp bounced between 44 and 47C during 3DM01se Nature torture loop w/ 29.5 - 30.5C air
temp at GPU cooler fan intake running for a 32 - 33C temperature reduction from stone stock. Core
temp during 3DM03 Mother Nature torture loop at 627Mhz ran just 40 - 43C at the same GPU fan air
intake temp indicated above.
(5) Memory did respond to Vdd changes but only when a corresponding increase in Vddq went with it.
Most unusual. These new DDR-III chips have very different characteristics and response to voltages
changes than previous DDR-I and DDR-II chips.
(6) Memory definately liked Vddq running higher than memory Vdd (against Samsungs spec) and the way
ATI had the stock memory voltages set as well.
(7) Memory chip temperture increased dramatically with increase voltages. Even with SuperSinks the
memory chip temp hit 46.3C during the Nature torture loop at 624Mhz is an open case for a temp
reduction of 20 - 22C from stone stock. A card top Active Memory Cooler brought the chip temp down
to 37.1C during the torture loop for a chip temp reduction of 29 - 31C from stone stock. I pity the poor
guy that slaps a set of Tweak Monsters on these chips and then leans on the voltages!!!
(8) Memory OC is still disappointing to me. The increased Vdd and Vddq got the memory OC as a whole
up nicely with up to 624Mhz clean in many tests including 3DM01se Nature. 3DM01se Lobby low and
high resolution are the Achilles heal where a memory wedge would consistantly be throw at a couple
specific locations with a memory clock above 591Mhz. Same with 3DM01se Nature at 627 memory
clock. These artifacts were always in exact same place everytime. This occured in other tests as
well just at different memory OC levels. This is indicative of one weak chip, with memory addresses
that have problems with certain bit patterns, in the set of chips this card was built with (luck of the
draw factor) rather than the memory as a whole.
(9) Use the memory OC's as a guide here. You will probably find the clean memory OC is well into the
600's clean with many or most games.
(10) Test computer speed/memory bandwidth is really bottle necking the modded card. Quite literally sticking a cork in it's exhaust pipe! Card would bench much better modded in a real fast P4 or AMD 64 system.


if you did a quick search for Krag's card, you'll see that it was also modded with extream cooling...not just the voltage mods.
and the results are still not that good.
the 1.6ns memory still seems to be poor as for OCing.
if you have to go to this extent just to push 600 memory to 624....that's not that worth it to me or most people.

Sentential,

by all means, I'm sure many people with nVidia cards are getting great OCs on thier memory.
and everything I'm about to say, is more aimed at me and more intence OCers.
---------

please understand that I could OC well past my everyday OC of 540/550....
but I must get clean or bug free results at all tests at such OC for me to accept it.
in other words, if I pass all tests at my OC, yet in one game or application I get some type of corruption, then I back down my OC and keep it at that.

for example, I know of quite a few peeps that state that thier max OC is bla/bla, yet when playing Doom3, they have to lower the OC or even run at stock speeds with thier card.
(maybe even opening the case door to play Doom3, with the max OC in place.)
to me, that's not a stable OC and should not be counted as the results of what is good,fine or "I know for a fact that none of us on the 6800 camp is haveing any trouble getting excellent OCs on both GC16 and 20".

so you say that in the NV camp, most people are doing great after the new bios....
just how well do you trust the results they are talking about?
would you say they test as much as you, me, or maybe ViperJohn?
do you trust that they are keen in spotting a green "dot" in one or two small places in 3dmark03...and then backing down the OC due to the mem OCed to far?
better yet, do you even think they are watching the test runs as close as you or I?

you see, I trust very few peoples results, as most just want to brag about thier cool OCing skillz...lol

maybe I'm just a little harsh when it comes to OCing, as even 10 hours of prime stable, is not stable enough for my CPU testing.

mica

[Sentential]

I dont think the blame lies with Samsung as much as it should lie with ATi. I know that on my card that I can play and bench using 1.2ghz with no visual artifracts, except after several intense hours of play.

Also I will say that my results are legit. I can play Doom3 and Farcry for hours without ANY sort of artifracting at 412 | 590. I am not sure why nV's RAM is consistantly higher, but something is amiss.

I think its more a case of RAM timings or binning perhaps. ATi has a track record of using poor RAM. nV tends to buy the very best RAM and perhaps this might be what is going on. Also one other thing to consider. nV uses single sided RAM. Maybe that makes a difference?

[Sentential]

Once more thing. Most of us nVers are using coolbits. If the OC is unstable it WILL NOT allow you to overclock further. So those people using coolbits, their results are legit, because software fail-safes prevent allowing unstable OCs.

[micamica1217]

Once more thing. Most of us nVers are using coolbits. If the OC is unstable it WILL NOT allow you to overclock further. So those people using coolbits, their results are legit, because software fail-safes prevent allowing unstable OCs.

ROTFLMAO...not at you sentantial, but at your thoughts on coolbits.
how would coolbits know if your getting just one rare artifact in a game or benchmark?
and what fail-safes does the new coolbits have?

man, it's been almost two years since I used coolbits...what's so new besides the crappy defalt OCing feature it has?

mica

[Sentential]

ROTFLMAO...not at you sentantial, but at your thoughts on coolbits.
how would coolbits know if your getting just one rare artifact in a game or benchmark?
and what fail-safes does the new coolbits have?

man, it's been almost two years since I used coolbits...what's so new besides the crappy defalt OCing feature it has?

mica
No actually it does. If you run an artifract tester or stress the card prior using coolbits it is really accurate. It physically wont let me OC past 1.19 (sometimes 1.20) without forcing me down saying "Warning cannot run at specified clock speeds" So far it has NEVER been wrong. Every speed that I have access to is 100% artifract free and stable

Thats the nice part about it. You can detect optimal settings then bump it up to its max. Its pretty simple actually, much easeir than the ATi tool

[felinusz]

but I must get clean or bug free results at all tests at such OC for me to accept it. in other words, if I pass all tests at my OC, yet in one game or application I get some type of corruption, then I back down my OC and keep it at that.

for example, I know of quite a few peeps that state that thier max OC is bla/bla, yet when playing Doom3, they have to lower the OC or even run at stock speeds with thier card.
(maybe even opening the case door to play Doom3, with the max OC in place.)
to me, that's not a stable OC and should not be counted as the results of what is good,fine or "I know for a fact that none of us on the 6800 camp is haveing any trouble getting excellent OCs on both GC16 and 20".

The real clincher is that an unstable overclock is no good to *anyone*, least of all the person who's running it.

The issue has been beaten to death several times, but there are still (and probably always will be) those who are perfectly happy running an unstable clock on their hardware. My take on it is that if it isn't tested 24+ hours in Prime95 Torture, and 3Dmark01 Nature, it can't truly be called stable.

I don't think it's fair to direct this solely towards those overclocking nVidia video cards though, as lots of people using ATi cards do this as well, probably just as many - most without even realizing it.

If the overwhelming majority of nVidia memory overclocks are better than comparable (same RAM) ATi memory clocks, it probably has some bearing on what's actually going on.

I think that ViperJohn said it when he mentioned that the X800 XT-PE card he was working on probably had one slightly bunk RAM chip, causing problems at specific points in several benchmarks where a large chunk of information is being processed.

[Sentential]

I thinek that more whats going on to be honest. If you recall nV had similar results with the 5900s. They went alot farther, granted they used different memory, but they still OCed further.

[OBLIVIONLORD]

Instead of arguing about the current topic why not argue to ATI and Nvidia for not having the freakin 6800U/X800Xt-Pe on the shelves but, have to wait for the middle of October to hopfully get one at retail price. ARGGGGG