Do we really? It has not happened yet in games as far as I know, and from the explanation AMD gave it will only happen in artificially created load scenarios (not even GPGPU stuff), hence I am perfectly happy with it.Quote:
Originally Posted by [XC] riptide
I think perhaps you should have done a little more research before putting one a third party cooling solution.Quote:
Originally Posted by truehighroller
I think that perhaps you should have done some more reading before calling every one idiots. I have been having this random crash issue for awhile now and believed it to be power related so I bought an add in booster and the heatsinks and the issue still rears it's ugly head every now and then.Quote:
Originally Posted by Frodin
I think that my first 4870 1GB card is the culprit because it is getting to hot VRM wise. I did research and not to many people seemed to be complaining that much and the people that were usualy stopped posting after some one told them to put a sink here or there. Probably because it has kept the issue at bay a little more like in my situation to where they were just like ahh f it and left it at that.
I how ever would like no crashes at all because of this over heating issue with the damn VRMs. I have been in the scene for awhile now and do have a decent amount of knowledge when it comes to overclocking and tweaking and building PCs so don't treat me like I don't.
Tetedeiench, are any 4850's affected? I mean they are just 4870's without GDDR5...I've just updated to 3.1.0 from 3.0.1 today and I'll run it 4850 stock, (625/1986) then 680/2100 (my ASUS top factory default) then 715/2150 which is my current overclock to see what works and doesn't work.
I did not call you or anyone else an idiot, I merely suggested that you could have done some more research before changing the cooling solution. It never hurts to do a little more research. There was a huge thread about VRM cooling in the water cooling section just after the 4870 was released (around last summer?), and I remember reading a thread in the air cooling section as well about someone having trouble with overheating with the ramsinks that came with a Thermalright T-rad. Largon suggested on the previous page that your VRM cooling was inadequate, no, actually, he said they were terribly inadequate:Quote:
Originally Posted by truehighroller
Quote:
Originally Posted by largon
I posted a thread once about vrm temps on my GTX280 which as I've mentioned uses the same vrms and Unwinder replied saying that the upper limits of these vrms are around 150c. Even on this stress test which is just crazy you are pretty far away from that. So, I doubt temps are your problem. Your vrms could be cooler but theyre not dangerously hot. What temps do you see while gaming? You may just have a bad card, it happens but your crash is not related to this "issue".Quote:
Originally Posted by truehighroller
Frodin- Do you have any links to those threads. They sound like a good read, thanks.
Quote:
Originally Posted by BababooeyHTJ
I will have to start digging deeper then I guess and will pull Logs and see exactly what the issue is, I very well might have a card going bad. I will play some games and keep logging enabled that way when everything crashes I can look at what the temps were right before the crashes. I would like to see those post as well. The one about air cooling might very well be the one I was looking around on already. Does any one know how to use the original plates that come with the coolers for cooling these things?
Because it's being done using standard API calls, meaning anybody could do it !! So IMO it's a design flaw when the card can't handle 100 % load at default clocks.Quote:
Originally Posted by Shadov
Ya. But game engines change. More powerful. People code graphics on little apps all the time. New GPGPU stuff is rolling off all the time. It annoys me because it LIMITS the way a developer can optimize his code for this particular card, IF he had a code similar to the one we see here. The card should never fault because of a cheap power design, ON STOCK CLOCKS AND VOLTS.Quote:
Originally Posted by Frodin
What is encouraging for all of us in the future is AMD's acknowledgment, and addressing of the issue, however unique it is.
You said it was when it reaches the 82a/83a line, which I said is not true because I did reach above 83 and it functioned normally. Thats besides the point now :)Quote:
Originally Posted by Tetedeiench
So according to ATI, if this is a protection measure on the reference design, but does not exist on non-reference designs, then how the heck do we know that those components won't eventually blowup? I mean do we know that the AIB's are not using reference components (besides the asus super)?
Seems a bit risky.. someone willing to run occt for a few days on their non-reference to see what happens?
On second thought, what ATI said makes sense. It is no practical scenario where you are stressing all the units including texturing and computing units at the same time... thats like having a game scene that needs no I/O, doing ridiculously inefficient calculations in the back to use up all the shaders...
No meaningful calculation can max the GPU at 100% in all components 100% of the time. Unless it is an infinite loop with no purpose.
I agree that its nice that ATi is going to address the issue.Quote:
Originally Posted by [XC] riptide
BUT IMHO its not an issue, ATi/AMD has said it themselves, the only way this could ever happen is if you deliberately go out of your way. This has nothing to do with "optimizations" that game makers may try and use. Standard API calls or not, its a power virus pure and simple
Using that same argument I can say every harddrive ever made has a design flaw because you can format it(using "standard", built in tools) and lose all data. :rolleyes: The only way a HDD will become formated is if I go out of my way, and well, run a format on it.
This.Quote:
Originally Posted by ahmad
Same is true of CPU's, just because the task manager and Prime95, Linpack, OCCT, etc, etc, says your using 100% of the CPU, doesn't mean your ACTUALLY using 100% in all CPU die components 100% of the time, simply that your using up 100% of the CPU's time.
I'm happy that this is how it turned out. I was kind of expecting AMD to ignore the problem. But perhaps the same mindset that is working for their processor division ( interfacing with the enthusiast community ) has spread to other divisions. Even though overclockers are a small market segment, and the problem is an extreme corner case, they still listened and are redesigning the chip and board. I would suggest that far from the blunder that many have claimed, this could actually bode well for their reputation.
+1 to that. Lets see what the R8xx series will do :).Quote:
Originally Posted by Solus Corvus
the cards can never reach 83 A in games ???
HA i say HAAAA
my Gainward 4870 reference already goes to 72A in crysis at 1600x1200 no aa no af catalyst driver defaults 9.5 after about 5 minutes
and it crashes after 30 mins -1 hour ...seems to be very random maybe it spikes to 83 A causing it to crash
http://img35.imageshack.us/img35/2949/52748904.th.jpg
don't use the riva osd ingame because you will get false data .let riva monitor your card in the background
now imagine i used a higher rezolution like 2500 ....it would crash wouldn't it ?
i say HAAA again
Since it takes so long to happen, it doesn't really sound like the same issue. RMA your card.
i just ran the crysis gpu benchmark and this is what i got
http://img132.imageshack.us/img132/8...ysisgpu.th.jpg
the 72 Amps that i got were during actual gameplay using the carrier level fighting the huge alien ships
why do my clocks go from 2dto 3d mode every damn second i don't see that happening on your card
damn you ati or gainward or whoever fault it is XD
i'm so tired of this problem i just want to play my games without crashing QQ
all my crysis runs were made without AA or AF using catalyst 9.5 default settings and yes i notice a higher amperage without the AA in your screenshot
i also notice that my clocks go from 2d to 3d mode every damn second i don't see that happening on your card !!
can you tell me if your card is reference 4870 512 gddr5 and who produced it ? mine is gainward made
could you please save your video bios file using gpu-z 0.3.3 (latest version) so i can try it on my card ? maybe it will make it stop swithing from 2d to 3d mode ...maybe that is the problem
all reference cards bios should be compatible with one another
here are some pics of my card
http://www.netraid.com.my/images/Gainward-HD4870.jpg
http://www.ixbt.com/video3/images/ga...4870-front.jpg
http://www.ixbt.com/video3/images/ga...-scan-back.jpg
Is this the HSF combo you are using with your PC? I was hoping to see an actual pic of your video card. In any case I've noticed that when dust build up at/near the vrm area of the video card it not only creates more heat but caused my video card to use more power. On a hunch I decided to clean that area out. I removed the front plate and took a can of air and sprayed in and around the vrm area. Doing so created a plume of dust. After air cleaning I noticed that my temps were lower and power consumption had went down to what you see today. If you are also experiencing the same problem this could help. I would suggest doing that. I would never recommend using bioses for situations like this so it really serves no purpose at this time. Which draws me to ask if you used or alter the bios in the past?Quote:
Originally Posted by bluedevil
For me, I've never seen the power draw you've shown when playing Crysis. So, it's either some sort of heat build up or a defective card as others have suggested to you. If the former does not work RMA the card. If you card is 2D/3D switching then remove all web browsers from your desktop and see if it stops. Better yet just create a 2D profile, add the icon to the desk top and click on it and see if that stops the 2D-3D switching.
I got powercolour 4890 I think it has a shutdown vrm temp at 95-110*c. I notice some people are doing over 125*c in vrm. Has anyone got 1000mhz/1200mhz bios try to update the bios seem if the vrm temp limit is gone?
http://img188.imageshack.us/img188/434/occt1.th.jpg
Eastcoasthandle
yes that is what i'm using and sorry but i don't have a camera to take any photos :(
i don't want to remove the heatsink because i have no paste to reapply and from what i have read the heatsink comes all off in one piece exposing the core and memory and than you have to remove the plastic to expose the actual part where the dirt gets in .
also i don't want my screwdriver to leave any marks on the screws because the shop could say that i messed with the card
i closed all the applications i could and that didn't stop the 2d/3d switching :(
please dude give me your bios i know how to use RBE to look in it so i can spot the differences ,ATIWinflash to flash it if i decide to and if something goes wrong i can make a autoflash disk using atiflash360 to unbrick it from dos or use a second card to flash the first one :D
Your other postQuote:
Originally Posted by bluedevil
Something is not right I've only seen 50A on mine (using AA).
http://i11.photobucket.com/albums/a1...is_GPU4xAA.jpg
http://i11.photobucket.com/albums/a1...Crysis_GPU.jpgQuote:
Originally Posted by bluedevil
This is without AA. See the trend? Your frame rates are a tad lower. Your vrm temps are higher and power consumption is greater.
Did you provide a pic of your video card? I like to see what it looks like.
I can assure you it's not the bios. I just ran the GPU test again and got this:Quote:
Originally Posted by bluedevil
http://i11.photobucket.com/albums/a1...is_1600Pt2.jpg
I thought about it and wanted to know if I was averaging 55A. So far it's not. I honestly think you should get yourself some TIM, open up the HSF and clean out any dust that may have accumulated in/near the vrm area.
I think he has flashed it before yes because I had that same issue when trying to mod my BIOS back in the day when I first got the one I think is causing issues and it did exactly this, the whole switching from 2d to 3d crap. I ended up finding a BIOS for a reference card that was clocked higher and flashed them both to it and the issue went away.Quote:
Originally Posted by bluedevil
That's what I was thinking as well. And do recall that specific issue when flashing. Perhaps you can help him by providing him a bios that could work with his video card? Maybe PM him and tell him exactly how you flashed the bios, etc? Because you and I don't have the 2D/3D switching problem. It's a thought...Quote:
Originally Posted by truehighroller
Side not:
I already told him to create a 2D profile using CCC and create an icon for the desktop. Once he clicks on it that should stop the 2D/3D switching (for now).
I mean this in the neutral way, but before you say 'it could happen', you should undertake a much deeper understanding of the GPU, the code that runs on it, and the parallels it draws to OCCT's code.Quote:
Originally Posted by [XC] riptide
I don't know much myself either, which is why I didn't chime in with my opinion on whether this could ever happen in real life.
Hey look at that...
The experts basically said what I have been saying all along.
GJ Tet.
How do you figure?Quote:
Originally Posted by LordEC911
Tetedeiench excellent work on this issue, i have one question though, is there a way to generate even more load on the gpu than occt currently does? or is the current occt test the best way to get maximum stress on the gpu?
reason i ask is that when i do overclocking on my cards or any hardware i find its always best to run the most stressful tester there is since it makes fault finding take less time.
so is occt the gpu equivalent of linpak? or is it still possible to get even more stress created?
thanks
:rofl::rofl::rofl::rofl: First time in a while that I actually laughed out loud while reading a forum :up:Quote:
Originally Posted by LordEC911
After reading approx. 1/5 of the posts, I have given up reading all of them, but what I have seen is, that many people say the same: For my use, this is no problem, so WTF, who cares!
My opinion on the other hand, as an Engineer, is totally different.
If you can use software to stress the hardware so much, that it can't take the beating, then the producer has made a design flaw.
What good is it to have a V8 twin turbo engine, if the darn thing cannot get any air to the turbo's, and thus want get enough combustion air? No good.
This is the same that OCCT is doing: Pedal to the metal, but hey, there is not enough air, so the engine will become flooded, and dies.
Major problem this OCP, but maybe the guys at DAAMIT has done it nowing the problem, since there plans was to send out 4970 as an overclocked 4870, and thus limmiting the end consumer in making that clock them self.
Ok I think my issue was different because, I just pulled off my sinks on the little things behind the big things?
I copied this pic from here so bear with me here but I am referring to the smaller of the five chips that are circled here. Part of, one of them came off with my sink. Not to much but, I didn't dare plug it in again.
http://img194.imageshack.us/img194/9339/4870h.jpg
I will be hitting up Microcenter this week for a swap it looks like. I didn't even touch my other one yet. I was scared :p:.
once again
i love nvidia =)
I swear there are so many horrable car analogies in this it's rediculous and I laugh every time.
Is that because the fastest nV card is much slower than a HD4870 in this benchmark?Quote:
Originally Posted by xlazyx
he probably likes his nvidia cards since they dont crash in this benchmark. :p:Quote:
Originally Posted by largon
this makes me wonder now about my old ati card. i always blamed ati drivers for being crap and unstable, now im beginning to wonder if it was the weak voltage regulation and not the drivers.Quote:
Originally Posted by bluedevil
Haha, yeah, no kidding. I'm no engineer, but to me, this program crashing some of the cards in question, is more like trying to take your V8 twin turbo engine, and put it in an airplane. Why? Well, because we can, and people may someday want their cars to fly, so they should be prepared and able to do that if the need arises. I think it is just as valid as most car analogies.Quote:
Originally Posted by BababooeyHTJ
AMD developed a great GPU (probably spent a lot of money doing so) i mean look how easy it is to code for it and how fast it can be made to use all of the 800 shaders (Tetedeiench said he didn't do anything that special in the code ) but the guys who designed the pcb decided to cut corners to make an extra buck and hoped that none would find out. Now we know why all the vendors offer non reference cards with redesigned pcb's.
I would have gladly payed the 10-20 $ for a extra vrm chip to make shore my card can handle anything.
I fell like i've been scammed and i have lost my confidence in AMD because of that .
Shame on you PCB design team shame on you.
what makes things even worse is that recently amd lashed out at i believe palit/gainward for making non reference design cards that had more voltage regulation phases on them because they didnt wait for the main tiers of ati cards like sapphire etc to bring out thier non reference design boards first. bad amd.
My 3x 3870 OCed run the test just fine. :DQuote:
Originally Posted by Cyber-Mav
I haven't lost my confidence in ATI/AMD due to this. I will most likely buy the new x2 when it comes out.Quote:
Originally Posted by bluedevil
LordEC, I guess you're right about the power-virus thing.
"Even if the problem appearing is very unlikely, nothing tell it won't happen again. And this is a door open to viruses and such malicious softwares."
"AMD indicates that this problem will be fixed in later GPUs."
Quoted by the French reviewer, Translated by Tetedeiench
Meaning, it is possible for others to create "power-virus" programs that boot when windows is loaded, for a result of kee-rash!!! Those with NV cards probably do not have to worry! :D And also those who buy future ATI cards wont have to worry either.
(Hey, I'm not opening a can-o-worms here, if you try to turn this into a chat-zoo of debate, then it's you who opened it.)
lol .... scammed HAHA.....Quote:
Originally Posted by bluedevil
you do realize they DIDN'T do this on purpose...
It was an accident...
even the OP said he's sure if he got the right calculations for Nvidia's card's they probably will do the same thing.... so don't feel scammed as that's just stupid....
they made an error in having the OCP trigger from the Vrm's instead of the Core.... hardly a bad thing...
especially the only way your going to get this problem is to use the program...
which whats the point.... doesn't happen during folding or games.... soo its just you being petty about it to "loose confidence". -sigh-
A car analogy posted my member on another forum.
Quote:
Originally Posted by layte View Post
Probably because the only time you actually can load the GPU up in such a fashion is using an artificial tool that has no baring on any real world situation. I hate to use a car analogy, but if I sat my car in neutral and kept my foot on the accelerator something would fail sooner rather than later, why did they make a car that the user is not allowed to use at 100%. A similar situation to what we have here.
It could be worse, the GPU's could melt themselves instead of just failing.
They didn't say anything about OCCT being a power virus, as LordEC911 did. They are saying that malware could take advantage of the issue, not that OCCT is an example of malware.Quote:
Originally Posted by Bo_Fox
Heck, I just hope the code he is working on now, that might let us Nvidia boys get a few more FPS is successful. :)
worst analogy ever..Quote:
Originally Posted by Final8ty
A little more explanation would be appreciated.Quote:
Originally Posted by STEvil
you should read my other postsQuote:
Originally Posted by Jamesrt2004
this does happen in games with my card and i have found no other explication.
You could just have an outright faulty card.Quote:
Originally Posted by bluedevil
Then there is probably something else wrong.Quote:
Originally Posted by bluedevil
must be very specific. when you were running the game that did this to you were you monitoring the gpu current pull at the time? what did it peak at?Quote:
Originally Posted by bluedevil
His might be like mine where that chip is weak on one part of it and it is causing his flakyness. I have never had a chip piece come off with my heatsink like that, ever, it had to have been a faulty chip from the get go and this issue had haunted me forever so... I will establish if my system is more stabe now that, that card is out of the equation tonight when I get home and get to actually play some with the one that is left in it right now that I didn't want to touch until I get the other one back from RMA.
none ever reads my posts it seems ...oh wellQuote:
Originally Posted by Cyber-Mav
my Gainward 4870 reference goes to 72A in crysis at 1600x1200 no aa no af catalyst driver defaults 9.5 after about 5 minutes using the carrier level at the end of the game
and it crashes after 30 mins -1 hour ...seems to be very random maybe it spikes to 83 A during certain shader intensive scenes causing it to crash ... no artifacts no texture bugs just restarts the pc
http://img35.imageshack.us/img35/2949/52748904.jpg
don't use the riva osd ingame because you will get false data .let riva monitor your card in the background if you want to see your true Amps.
i don't know what to use to see the exact amps at the time of the crash because when it happens my pc restarts and i guess riva tunner does not have time to record the last value
Actually I did read your posts however you are not willing/able to open up your video card because you don't have thermal paste. :shrug: Because you haven't gone in depth you really don't know what the problem is at this time.Quote:
Originally Posted by bluedevil
You don't see anything wrong with your GPU drawing 25a at desktop? Your card is bad, it happens with ATI and Nvidia. I've had 1 bad GTX260 and a 280 fail on me, not to mention all of the other hardware that has failed on me.Quote:
Originally Posted by bluedevil
Somethings not right there, I get around 50A/3d (Crysis) 9.5A/2d on an X2@850/1000Quote:
Originally Posted by bluedevil
What is being done here would be akin to a WOT burnout or drag race, not sitting still while holding the throttle all the way open.Quote:
Originally Posted by Final8ty
Either way if what I buy has not been advertised as being incapable of performing at full functionality under certain circumstances and it does not perform at full functionality under certain circumstances then it has been falsely advertised.
Its more like drag race yes & we know what can happen to the engines on them.Quote:
Originally Posted by STEvil
Peak performance does not have a longevity rating most of the time.
Sounds like your card is faulty, RMA it. This "defect", if you can call it that, only appears in a certain set of situations, which from reading your posts, have little to do with it.Quote:
Originally Posted by bluedevil
The VRM's heating way up is a known issue with non refrence coolers, check out old water cooling posts on the matter.
I'll say it once more, I highly doubt its the issue that this thread is based thats causing your crashes. You have a faulty card, instead of arguing, RMA it. Hell, it might not even be the video card, it may be something else in your system, like the memory(ram).
All I'm trying to say is, don't jump to conclusions
Drag racing isn't covered in most factory warranties. If you break something while racing your car don't expect them to cover it.Quote:
Originally Posted by STEvil
Car analogies are inaccurate at best. In fact any analogy is unneeded, the issue can be described in simple terms that anyone can understand: It is a case of a series of video cards that won't run a certain program. It's that simple. There is no need to turn it into a hypothetical situation in an unrelated field just to describe it. The worst part is that the people making the analogies ( ie. probably geeks ) are describing a problem ( a hardware/software issue) that their peers ( other geeks ) can understand by relating it to something that neither of them understand very well ( cars ) - which is obvious by looking at the absolutely horrible car analogies used regularly.
I'll give you an analogy: Take your car, turn on the AC, the heat, the rear defogger (essential), your wipers at full speed (don't forget to spray), your interior lights, your fog lights, headlights, rear lights reverse lights, signal lights, hazard lights, roll up/down the windows, lock and unlock your doors (if not power, you may need assistance), your radio at full blast, while at the same time pushing the engine to maximum revs (pedal to the metal), while braking constantly, and turning the steering wheel from left to right continuously. Oh and I forgot to mention, to stress the bearings and transmission, you have to be going at the car's top speed to do this.Quote:
Originally Posted by STEvil
First of all, if you attempt this you will look like an idiot. Second, I can almost guarantee you something is going to either break or blowup if you do actually manage to pull this off. And third, some of these things are impossible to do at the same time, like turning on AC and heater at the same time (so that it blows both cold and hot air in all areas of the vehicle).
Do car manufacturers stress test their vehicles like this? If they did such a car company would have gone under a long time ago...
If your car does not function at "maximum capacity", does that make it's design flawed? Or just the user subjecting the car to abnormal and unusual tests?
Next time you are buying a car, I want to see you testing it like this. If you don't I am going to call you a hypocrite.
Get the word "impossible load under normal circumstances" through your heads people. If your video card fails in any games or normal apps (like your headlights don't work), take it back and get it replaced. Otherwise, life goes on.
ATI made the mistake of letting you turn your AC and heat at the same time. Nvidia doesn't let you do that. Which is why nvidia cards cannot be stressed like ATI cards.
Nicely put. :)Quote:
Originally Posted by ahmad
That's well put right there :up:Quote:
Originally Posted by ahmad
yup.. but have you ever seen a GPU with a longevity rating at all or even disclaimers that it cannot run at full peak performance at certain settings? I've not seen it for CPU's nor GPU's... nor PSU's either (well, probably a few..)! For a PSU you can be a bit more lenient.. darn wattage ratings :/Quote:
Originally Posted by Final8ty
agreeQuote:
Originally Posted by Solus Corvus
I'd agree with you if we were talking about average consumer cars which obviously cannot perform to full performance specifications for extended periods and 99% of the time will not be required to do so unless purchased for that exact purpose (in which case it was custom built..).Quote:
Originally Posted by ahmad
However we're talking GPU's which are produced for the enthusiast/high consumer market (has anyone tested lower-end GPU's?) where the video card is advertised to run at a specific speed (gpu clock/mem clock) at a specific voltage (gpu/mem/etc) while performing its functions. Should a game, games, or GPGPU program in the future use the GPU to the fullest (or near) extent then is it the programmers fault for making use of the hardware performance he/she was given (ie: deliberately making it unstable or having a better sample of the hardware than lots of others due to multiple PCB designs) or is it the fault of the manufacturer for not telling the vendors that they have limits and/or for not publicizing the limits?
To me its quite simple and its why you see AMD/ATi have acknowledged the issue.
AMD and ATi have acknowledged the issue because of ....Quote:
Originally Posted by STEvil
"To justify themselves, AMD insist on the fact that that no practical case will charge the GPU as much. You have to want it to be able to do it, and use a very specific code." -snipped from OP's reply
sounds pretty close to your rationalization regarding the average consumer cars..
honestly find a game or gpgpu proggy that can actually cause this issue, then maybe there would be a rightful discussion. No, OCCT does not count.
i hope HIS HD 4890 is out of this design flaw >.< , else i got screwed !
why are ppl still talking about the 3d scene when AMD already acknowledged the design flaw !
will they do an upgrade option to return the card and get the new one ?
Okay, let's stop the ridiculous car analogies. I made one myself, i'm ashamed i did :D
Yes, you have to use specific code (err, if by specific code they mean "use alot of alpha blending and DX9 shaders", well, so be it) for it to trigger, but it can be triggered. That's the important part. Wether it be rare, or unlikely, or not corresponding to YOUR personal usage of your card, it doesn't matter : it can be triggered.
However, remember that this protection is a last resort protection. You should NOT be able to trigger it in normal cunditions, and especially not at stock speed, with regular API calls. In a normal protection, the GPU should have reduced its frequencies, not forced you to reboot. Thus, the obvious design flaw.
Next, once again, i'd say for the 10th time, the "normal application" part. Who did say that 3D cards were limited to gaming ?
I mean, i am presented with a 3d API to draw 3d, 3d capable hardware compatible with that API... yet, everybody says its limited to games ?
Come on guys, you may have already ran the demos from Farbrausch (remember the 64k demo fr08_Final ?). That's not a game. That kicked ass. You may have enjoyed it. Nvidia and AMD themselves are making demos just as my donut, to demonstrate the use of a particular function : sometimes there's alot of geometry, sometimes there isn't. Just download DX9 SDK, or Nvidia Cg SDK, and you've got your small applications that are completly valid, made by the cards manufacturer, they are not game, yet i believe they are perfectly valid applications, because they are made by Nvidia, Microsoft, AMD themselves. Just look for demos on google (Nvidia demos, AMD demos, etc). You'll see my point. Are those demos "unvalid apps" because you cannot kill aliens ?
So this point is out : you can't play the GPU:3D furry scene as you wish, but that doesn't mean it is not a valid application. 3d = 3d scene, whatever this scene is. It is true that for end users, 3d=games, because they mostly remember their last Fallout3 or Crysis game session.
It is a marketing decision, IMHO, made by Nvidia and AMD, to cut costs, to build cards that work on the assumption that a 3d scene will have alot of geometry, and thus alot of idling on the GPU side. That does not make applications such as OCCT unvalid because they do not have alot of geometry, yet make use of advance functions. The decision which lead to this kind of situation is unvalid. I'm following closely to API specifications, made by microsoft, and those companies signed and agreed to follow them. Seems they failed to do so. Is it my fault ?
So please, let's stop this completly sterile debate. Yes, you do not play OCCT often. Yes, you will never encounter the great furry god of destiny in your latest game that will cause your favorite AMD4870 card to crash. Yes, it is very unlikely to crash in any game in the future.
And remember, i'm not saying "ditch all your HD4870s". I'm saying "they have a design flaw that can impair their functioning in some cases". Those cases were enumerated iby professionnal websites, which was what i wanted from the beginning, much better than i did. And that's why i didn't make any news about it on OCCT's website : that's not MY job. It's theirs.
Anyway, this debate has been raging on for 27 pages now, trying to make OCCT look like either a virus, or an unvalid application that does not matter. The fact that one app can trigger the thing, no matter what the application is, is the thing that is important.
AMD acknowledged the flaw, they're taking steps to correct it, and even if they're minimizing it a bit, the very fact that the steps they are thinking of taking go as far as designing a new reference board and changing the way their new GPUs will handle such cases indicates that this matter is not something that does not matter and can be ditched easily "because it's not a game". They are taking it seriously. Why not us also ? I am very happy with the way they are taking the matter :) It is a great way of reacting to the problem, i'd say.
Here, AMD's actions are more meaningful than their sayings. The very facts their steps may be so huge towards solving this matter indicates it is important.
Tetedeiench i have a question for you, can you make occt generate even more load on the gpu than it currently does? or has occt reached the stage where it already makes a gpu work to its theoretical maximum?
You can always make something more optimized, so i'd say yes, it is always possible to find a more optimized combination.Quote:
Originally Posted by Cyber-Mav
I'm working on making the GPU test more optimized for Nvidia cards at the moment, and i do think that'll have to do with a new algorithm, a new test. That's going to take some time though, and when i say some time, that mean months, at least. Weeks at best, but i doubt it.
I still don't know wether it is possible to come up with something that's pushing the Nvidia cards more. Anyway, work in progress ;)
My card that is left in my system still is locking up even at shader level 0. I need to look at that post about what other little miscelanious chips to cool, once I get my RMAED one back that is. Apparently there are a couple of small miscelanious chips that need to be cooled as well and I am thinking I am not cooling them at all so, that is probably my issue here. Good thing is that if it wasn't for this post I would have never figured this out and yes, I want one of these new designed GPUs as an offering for this mistake on their part.
let me know if/when you have the new test ready, i cant wait to hammer my card till it busts :DQuote:
Originally Posted by Tetedeiench
With furmark, OGL3, my VDDC current goes to 84A and stays there. Yet I don't experience any lockups or black screen thing. This flatly contradicts what OCCT programmer said.
Look at this GPU-Z log file. Sorry for the high temps. I turned off the fans to heat this up, yet it did not crash. My CPU temps are 77/72C. You get an idea how hot the water is :DQuote:
They basically crash because they can't handle the load. Early testing shows the VRM cannot supply enough power to the card. That's the only thing i could dig until now, with limited testing means. They seem limited to 82/83A. Please help us dig more !
Code:Date , GPU Core Clock [MHz] , GPU Memory Clock [MHz] , GPU Temperature [°C] , GPU Load [%] , GPU Temp.(DISPIO) [°C] , GPU Temp.(MEMIO) [°C] , GPU Temp.(SHADERCORE) [°C] , VDDC Current [A] , VDDC Slave #1 Temperature [°C] , VDDC Slave #2 Temperature [°C] , VDDC Slave #3 Temperature [°C] , VDDC [V] ,
2009-06-09 09:43:57 , 840.0 , 1100.0 , 52.0 , 98 , 51.5 , 58.0 , 52.5 , 84.0 , 73.0 , 74.0 , 77.0 , 1.2625 ,
2009-06-09 09:43:58 , 840.0 , 1100.0 , 51.0 , 98 , 51.5 , 57.5 , 52.5 , 83.3 , 73.0 , 74.0 , 76.0 , 1.2625 ,
2009-06-09 09:43:59 , 840.0 , 1100.0 , 52.0 , 99 , 52.0 , 57.0 , 53.0 , 83.3 , 73.0 , 74.0 , 77.0 , 1.2625 ,
2009-06-09 09:44:00 , 840.0 , 1100.0 , 51.0 , 99 , 51.5 , 58.5 , 52.0 , 83.3 , 73.0 , 74.0 , 77.0 , 1.2625 ,
2009-06-09 09:44:01 , 840.0 , 1100.0 , 51.0 , 99 , 52.0 , 58.0 , 52.5 , 84.0 , 73.0 , 74.0 , 77.0 , 1.2625 ,
2009-06-09 09:44:02 , 840.0 , 1100.0 , 52.0 , 99 , 52.0 , 58.0 , 52.5 , 83.3 , 73.0 , 74.0 , 77.0 , 1.2625 ,
2009-06-09 09:44:03 , 840.0 , 1100.0 , 52.0 , 99 , 52.0 , 58.0 , 53.0 , 83.3 , 73.0 , 74.0 , 77.0 , 1.2625 ,
2009-06-09 09:44:04 , 840.0 , 1100.0 , 51.0 , 99 , 51.5 , 58.5 , 52.5 , 84.0 , 73.0 , 74.0 , 77.0 , 1.2625 ,
2009-06-09 09:44:05 , 840.0 , 1100.0 , 51.0 , 99 , 51.5 , 57.5 , 52.5 , 84.0 , 73.0 , 74.0 , 77.0 , 1.2625 ,
2009-06-09 09:44:06 , 840.0 , 1100.0 , 52.0 , 99 , 52.0 , 58.0 , 53.0 , 83.3 , 73.0 , 74.0 , 77.0 , 1.2625 ,
2009-06-09 09:44:07 , 840.0 , 1100.0 , 51.0 , 99 , 51.5 , 58.0 , 53.0 , 83.3 , 73.0 , 74.0 , 77.0 , 1.2625 ,
2009-06-09 09:44:08 , 840.0 , 1100.0 , 52.0 , 99 , 53.0 , 58.0 , 53.0 , 84.0 , 73.0 , 74.0 , 77.0 , 1.2625 ,
2009-06-09 09:44:09 , 840.0 , 1100.0 , 52.0 , 99 , 52.0 , 59.0 , 53.0 , 84.0 , 73.0 , 74.0 , 77.0 , 1.2625 ,
2009-06-09 09:44:10 , 840.0 , 1100.0 , 51.0 , 99 , 51.5 , 59.0 , 53.0 , 84.0 , 73.0 , 74.0 , 77.0 , 1.2625 ,
2009-06-09 09:44:11 , 840.0 , 1100.0 , 52.0 , 99 , 52.0 , 58.5 , 53.0 , 84.0 , 73.0 , 74.0 , 77.0 , 1.2625 ,
2009-06-09 09:44:12 , 840.0 , 1100.0 , 53.0 , 99 , 52.0 , 59.0 , 53.0 , 84.0 , 73.0 , 74.0 , 77.0 , 1.2625 ,
2009-06-09 09:44:13 , 840.0 , 1100.0 , 52.0 , 99 , 52.0 , 58.5 , 53.5 , 84.0 , 73.0 , 74.0 , 77.0 , 1.2625 ,
2009-06-09 09:44:14 , 840.0 , 1100.0 , 52.0 , 99 , 52.0 , 59.0 , 53.0 , 84.0 , 73.0 , 74.0 , 77.0 , 1.2625 ,
2009-06-09 09:44:15 , 840.0 , 1100.0 , 52.0 , 99 , 52.5 , 58.5 , 53.5 , 84.0 , 73.0 , 74.0 , 78.0 , 1.2625 ,
2009-06-09 09:44:16 , 840.0 , 1100.0 , 52.0 , 99 , 52.5 , 58.5 , 53.5 , 84.0 , 73.0 , 74.0 , 77.0 , 1.2625 ,
2009-06-09 09:44:17 , 840.0 , 1100.0 , 52.0 , 99 , 52.0 , 59.0 , 53.5 , 84.0 , 73.0 , 74.0 , 78.0 , 1.2625 ,
2009-06-09 09:44:18 , 840.0 , 1100.0 , 52.0 , 99 , 52.5 , 59.0 , 53.0 , 84.0 , 73.0 , 74.0 , 78.0 , 1.2625 ,
2009-06-09 09:44:19 , 840.0 , 1100.0 , 52.0 , 99 , 52.5 , 59.0 , 53.0 , 84.0 , 73.0 , 74.0 , 78.0 , 1.2625 ,
2009-06-09 09:44:20 , 840.0 , 1100.0 , 52.0 , 99 , 52.5 , 59.0 , 53.0 , 84.0 , 73.0 , 74.0 , 78.0 , 1.2625 ,
2009-06-09 09:44:21 , 840.0 , 1100.0 , 52.0 , 99 , 52.0 , 59.0 , 54.0 , 84.0 , 73.0 , 74.0 , 78.0 , 1.2625 ,
2009-06-09 09:44:22 , 840.0 , 1100.0 , 52.0 , 99 , 51.5 , 58.5 , 53.0 , 84.0 , 73.0 , 74.0 , 78.0 , 1.2625 ,
2009-06-09 09:44:23 , 840.0 , 1100.0 , 52.0 , 99 , 52.5 , 59.0 , 53.0 , 84.0 , 73.0 , 74.0 , 78.0 , 1.2625 ,
2009-06-09 09:44:24 , 840.0 , 1100.0 , 52.0 , 99 , 52.5 , 59.0 , 52.5 , 84.0 , 73.0 , 74.0 , 78.0 , 1.2625 ,
2009-06-09 09:44:25 , 840.0 , 1100.0 , 52.0 , 99 , 52.0 , 58.5 , 52.5 , 84.0 , 73.0 , 74.0 , 78.0 , 1.2625 ,
2009-06-09 09:44:26 , 840.0 , 1100.0 , 52.0 , 98 , 52.5 , 58.0 , 53.0 , 84.0 , 73.0 , 74.0 , 78.0 , 1.2625 ,
2009-06-09 09:44:27 , 840.0 , 1100.0 , 52.0 , 98 , 52.5 , 59.0 , 53.5 , 84.0 , 73.0 , 74.0 , 78.0 , 1.2625 ,
2009-06-09 09:44:28 , 840.0 , 1100.0 , 52.0 , 99 , 52.0 , 59.5 , 53.5 , 84.0 , 73.0 , 74.0 , 78.0 , 1.2625 ,
2009-06-09 09:44:29 , 840.0 , 1100.0 , 52.0 , 99 , 52.5 , 59.5 , 52.5 , 82.6 , 73.0 , 74.0 , 78.0 , 1.2625 ,
2009-06-09 09:44:30 , 840.0 , 1100.0 , 52.0 , 99 , 52.5 , 59.5 , 53.0 , 84.0 , 73.0 , 74.0 , 78.0 , 1.2625 ,
2009-06-09 09:44:31 , 840.0 , 1100.0 , 52.0 , 99 , 53.0 , 59.0 , 53.5 , 84.0 , 73.0 , 74.0 , 78.0 , 1.2625 ,
2009-06-09 09:44:32 , 840.0 , 1100.0 , 51.0 , 99 , 53.0 , 59.5 , 53.0 , 84.0 , 73.0 , 74.0 , 78.0 , 1.2625 ,
2009-06-09 09:44:33 , 840.0 , 1100.0 , 52.0 , 99 , 52.5 , 59.5 , 53.5 , 84.0 , 73.0 , 74.0 , 78.0 , 1.2625 ,
2009-06-09 09:44:34 , 840.0 , 1100.0 , 53.0 , 98 , 52.5 , 59.0 , 53.0 , 81.3 , 73.0 , 74.0 , 78.0 , 1.2625 ,
2009-06-09 09:44:35 , 840.0 , 1100.0 , 52.0 , 98 , 52.0 , 59.5 , 53.5 , 84.0 , 73.0 , 74.0 , 78.0 , 1.2625 ,
2009-06-09 09:44:36 , 840.0 , 1100.0 , 52.0 , 98 , 52.5 , 59.5 , 53.0 , 84.0 , 73.0 , 74.0 , 78.0 , 1.2625 ,
2009-06-09 09:44:37 , 840.0 , 1100.0 , 52.0 , 98 , 52.0 , 59.5 , 54.0 , 84.0 , 73.0 , 74.0 , 78.0 , 1.2625 ,
ummm... did you also read where it only works with OCCT ???
did you also read that OCCT stresses the gpu more than furmark
did you also read that its only for refrence designed 4870, 4890
did you also read that ATI knows about the problem, and has CONFRIMED that there is a problem with the cards that only have 3 vrms
I don't understand the hostility. Flaw is a flaw whether it appears in extreme conditions or during everyday desktop usage. It is always better not to have any flaws no matter how rare it occurs. I personally thank the programmer who wrote OCCT for pointing out this problem.
I'd love to provide more data by testing out my 4890s but I can't run the program on Win7-64. Is there a solution for that?
Yes, but occt causes it the VDDCs to draw 82-83A. This is why it crashes, or at least that's how I interpret the original post.Quote:
Originally Posted by skycrane
Then that would be a core problem, not a power design problem.
Yes, I am using a referenced designed 4870Quote:
it is *NOT* related to Temperature, but to a weakness in the Power Supply stages of the Reference design HD4870/4890 cards from AMD. We don't know yet if it is the VRM themselves, or the OCP that triggers.
No, till now.
Was that referring to me :D. I wasn't being hostile, just stating my results. Afterall, the OP wanted for thoughts, and wanted him to dig more!Quote:
Originally Posted by SamHughe
Quote:
What's the point ?
...
They seem limited to 82/83A. Please help us dig more !
Actually it could indicate either, but given how many cards stop at about 82/83/84A (slightly different per card hence why not all reference cards exhibit the issue identically) it is most likely an OCP issue (ie: voltage supply, not core).Quote:
Originally Posted by Blazing fire
It wasn't about you. Just my opinion about the general attitude. Nothing personal.Quote:
Originally Posted by Blazing fire
Oh ok! Actually, I totally agree with you.
Hey all.
I just tried this with my XFX HD4870 XXX.
At stock freq i get 75A on the VRM. When overclocked to 850 i get the VRM to 81A and the screen goes black but comes back online after a few sec.
hey Tetedeiench any luck on getting a version of occt made that will bust my gtx260? :cool:Quote:
Originally Posted by Tetedeiench
I have the same problem with my HD4870 but with FurMark, after 2 mins of running FurMark, VDDC last reported 83.5 Amps, the screen goes black and the D601 red LED lit - critical Core power fault => so I'm pretty sure this is ATi/AMD's defect, not Tetedeinch's fault.
/If someone say "This test is unnecessarily stress my vid since my games play fine", I think they just can't bear having a defective card. Imagine in the very near future, when some developer write a game or a GPGPU app that can pust your GPU to its limit and you won't able to run them because of ATi's fault, what do you think :P
won't happen. ati devrel will show them a better way to do whatever they are doingQuote:
Originally Posted by David-Duc
Nice test. I bailed about 25 seconds in when my 4830 hit 82C... it loads about 60 in heavy gaming.
Please define "better" in this contextQuote:
Originally Posted by W1zzard
Also on a seperate note, how do you guys explain that the same cards that crash at 82A on setting 3, do not crash at 83A on setting 0?
.
Maybe im stupid here now, but is 4890 still a good buy?
My Toxic 4890 turned black right after the test start and im waiting for my second card now.
But this is making me a little nervous about going that way, or shouldnt i be nervous?
When i taked the test but without fullscreen i got 93.66A, the phase-temp was on 135:P, and the cooler on Toxic is pretty good and i have never seen 75 on the core before now:P with the fan on 100%.
some of those cards are unusable to play occt. so if you play occt often you may want to get another card, it will handle all other games just fineQuote:
Originally Posted by The Water Dog
hehe, i dont often play occt:P
But i think is weird that on fullscreen it goes black and without the fullscreen it is okey, it was on 93A in the test too and nothing wrong there.
What so special about fullscreen and not? stupid question, but just so i know.
Did some testing becouse Toxic is 960\1050 from the box, if i sett it down to "stock" 4890 then everything is okey. on 920\1000 it starts but after 30 sec. it goes black.
fullscreen is more demanding than windowed
thats why i said maybe im stupid now, sometimes i take things a bit too seriously:hitself:
But the 82amp. barrier is not the issue here.
Failed on me^^ 4870 CF. But tbh, this test = CARE. If it doesn't show this error in ANY other ("non-synthetic", if that's a word) applications it works 100%.
For info to OP i got Sapphire 4890 Toxic, on 960\1050 it goes black straigt away, setting down to "stock" 4890 is was fine and all up to 900\1000. From 920\1000 it goes black and have to reboot the computer.
All this with 1,4v( stock on Toxic) and where over 93a.
Would you mind stopping the threadcrapping? I think we've discussed the "I don't care" vs "I do care" position enough already, stirring the same pot again doesn't get us anywhere new.Quote:
Originally Posted by CandymanCan
my HIS 4890 reference with stock 850mhz dosnt crash
I just ran the GPU test on a pair of GTX 280s SLI for about 10 mins (clocks 712/1512/1242 - voltage 1.185). Full screen 1920x1200 settings. Max current draw was just below 30A on both cards. Max frames per second 144. Both cards are water cooled. Temps of GPU0 max 56C and GPU1 53C. Ambient temp was ~23.8C. During gaming temps usually top out at 15C above ambient.Quote:
Originally Posted by LordEC911
Keep in mind different settings should be used in order to fully load Nvidia chips with this test. They were posted earlier in this thread. Also was VSync off?Quote:
Originally Posted by cx-ray
People are always looking for CPU's with the lowest VID, or trying to find a D0 i7 CPU rather than C0. Are these people stupid too? Sure a C0 stepping with a high VID will work fine and never fail in a video game, but people from XS demand more from their hardware.Quote:
Originally Posted by CandymanCan
Not sure where to find the VSync off or even if it matters as the fps is already higher than the standard 60hz of an LCD. So, it doesn't look like it's sync locked.Quote:
Originally Posted by zalbard
New settings I just tried VSync forced of in the Global Settings tab of my driver (didn't see a difference).
13:27 min. with Shader Complexity settings 3, FPS ~83, GPU temps 56C and 53C, current draw still about 30A. Ambient temp ~23.4C.
I tried to search for other settings in this thread. The above was the closest I could find for Nvidia cards. If you know something better let me know and I'll test again.
as long as this doesn't happen in games
Does for me :(Quote:
Originally Posted by skugpezz
DoW II - max settings on boss battles = crash
dosent think that have nothing to do with this, that can have many other caouses.