Overclock.net banner

1 - 20 of 25 Posts

·
Registered
Joined
·
190 Posts
Discussion Starter #1
I posted this in the official 465GE owners club, but I have hijacked that thread long enough, and thought it would be good to get advice from people who don't own that card, and thus may not follow that thread.

Here is the setup:

GTX465 flashed to 470 with the standard BIOS +voltage control.

Furmark crashes at 70C-80C even with stock clocks. The higher the clocks, the lower temp the crash occurs.

With certain settings, the temps can climb quite a bit, 85C+ and not crash.

Settings that cause crash:

Stability Test "Extreme Burning Mode"
1280x1024
MSAA: None
~45FPS
Crash to black screen between 70-80C depending on clock

Setting that are stable:
Stability Test "Extreme Burning Mode"
1280x1024 or higher
MSAA: 4X
~30FPS or lower if higher res
85C for 15 minutes, no artifacts or crash.

Should I worry that it isn't stable without AA? Kombuster is stable at same res with no AA.

Maybe this is why they binned it as a 465? I have run games for over an hour totally fine.
 

·
Registered
Joined
·
190 Posts
Discussion Starter #3
Quote:

Originally Posted by LethalRise750 View Post
Well, when you add AA you're increasing the memory usage on your 465/470... Have you tried cranking the fan to 100% to rule out the memory modules or volt regs overheating?
Yeah, fan speed was 100%. I have put my case fans max and it doesn't make a consistent difference. It seems random as to when it crashes.
 

·
Registered
Joined
·
190 Posts
Discussion Starter #5
Quote:

Originally Posted by LethalRise750 View Post
Have you tried 3DMark Vantage to see if it crashes at all?
Yes, it ran fine. Heaven ran fine as well, but I don't think either of those programs produce the temps Furmark does.

Memory usage without AA is 158, with 4x AA is 192.
 

·
Not new to Overclock.net
Joined
·
77,827 Posts
Sparker:

The unlockable Golden Editions are just GTX 470s binned as GTX 465s and therefore tweaking is pretty much required in order to make these cards comfortable as GTX 470s. After all, why do you suppose they were binned as GTX 465s even though they have the GF100-275-A3 chip and 1280 MB of onboard memory? The reason why they stuck a GTX 465 BIOS on these cards is because they didn't pass Quality Control as a GTX 470. Fortunately, they're still good enough that we can make them work. It just requires a little bit of tweaking.

So, increase the voltage. If that doesn't work, increase it again. If you get up to like say 1.100V and it's still not making a difference, then lower the memory clock. Anything over 1800MHz is good! So 1800MHz or lower should be much more stable. As for the core clock, anything over 800MHz is good too. In some cases, even 775 MHz is pretty damn good.

So, keep tweaking. Now that you flashed it, you have to play the role of a manufacturer trying to make it stable as a GTX 470.
 

·
Registered
Joined
·
190 Posts
Discussion Starter #9
I'm tweaking quite a bit, I just don't know what causes what symptoms yet, but I did find a pretty obvious correlation.

The lower my volts, the higher my stable temp is. Using the same clocks and only changing the voltage let it go 4C higher before black screen.

The obvious issues here is that my goal in OC is then limited by how many volts I can do before it crashes.

Also, when testing out the lower volt ranges I have to lower the fan speed to even get the temps close to where they are in the higher volt range, which seems stupid because it is just adding another variable with fan speed.

How can I get higher temps at 100% fan?
 

·
Not new to Overclock.net
Joined
·
77,827 Posts
Quote:

Originally Posted by sparker View Post
I'm tweaking quite a bit, I just don't know what causes what symptoms yet, but I did find a pretty obvious correlation.

The lower my volts, the higher my stable temp is. Using the same clocks and only changing the voltage let it go 4C higher before black screen.

The obvious issues here is that my goal in OC is then limited by how many volts I can do before it crashes.

Also, when testing out the lower volt ranges I have to lower the fan speed to even get the temps close to where they are in the higher volt range, which seems stupid because it is just adding another variable with fan speed.

How can I get higher temps at 100% fan?
When testing with lower voltage ranges, why do you want temps close to where they are with a higher voltage? Also, why do you want higher temperatures at all? The lower the temperature, the better.

What are your current clocks, and what's your current full-load voltage?
 

·
Registered
Joined
·
190 Posts
Discussion Starter #11
Quote:

Originally Posted by TwoCables View Post
When testing with lower voltage ranges, why do you want temps close to where they are with a higher voltage? Also, why do you want higher temperatures at all? The lower the temperature, the better.

What are your current clocks, and what's your current full-load voltage?
Well, I don't want high temps. When I got the card unlocked, I did a search on how to overclock it. I found this guide, in which they max the voltage and find the highest clock they can reach with the highest allowed voltage. After I tweaked it a bit, I found I was using way more voltage than required because my card couldn't do more than ~825 core anyway and I assumed it was needing more voltage, but the reality is it didn't need anywhere near that much (1.087).

This led me to do a stability test in which I see it locking up at specific temps well below the supposed max range of ~100C. Seeing this made me think something was wrong, so I set out to figure out how to sustain higher temps without crashing, assuming the card are supposed to be good at higher temps.

So the reason I am trying heat the card up is to see what is wrong that it won't get as hot as other peoples' cards, and I really want is someone to say "don't worry about it if its stable" because I have no real-life need to have it this hot, but I am just trying to figure out why it won't run stable at temps other people will.

I have two settings I've been fine-tuning that perform similarly stability wise.

1.025v/800core/1800mem/80C max
.987v/775core/1700mem/84C max

Does it matter if my card isn't stable at 80C if it never reaches 80C?
Why can other cards go ~100C?
Why do I care if Furmark crashes it when everything else works?
What settings should I be running with Furmark anyway?
 

·
Not new to Overclock.net
Joined
·
77,827 Posts
Quote:

Originally Posted by sparker View Post
Well, I don't want high temps. When I got the card unlocked, I did a search on how to overclock it. I found this guide, in which they max the voltage and find the highest clock they can reach with the highest allowed voltage. After I tweaked it a bit, I found I was using way more voltage than required because my card couldn't do more than ~825 core anyway and I assumed it was needing more voltage, but the reality is it didn't need anywhere near that much (1.087).
Then it might be best to ask for overclocking help in the Golden Edition club because that guide you found is a guide for overclocking a reference GTX 470. Our cards were binned as GTX 465s, so we can not expect them to behave like a reference GTX 470 even though they function with a GTX 470's BIOS.

Quote:

Originally Posted by sparker View Post
This led me to do a stability test in which I see it locking up at specific temps well below the supposed max range of ~100C. Seeing this made me think something was wrong, so I set out to figure out how to sustain higher temps without crashing, assuming the card are supposed to be good at higher temps.
This is one of the reasons why it was binned as a GTX 465.
Otherwise, it would've been flashed with a GTX 470's BIOS and sold as such.


Quote:

Originally Posted by sparker View Post
So the reason I am trying heat the card up is to see what is wrong that it won't get as hot as other peoples' cards, and I really want is someone to say "don't worry about it if its stable" because I have no real-life need to have it this hot, but I am just trying to figure out why it won't run stable at temps other people will.

I have two settings I've been fine-tuning that perform similarly stability wise.

1.025v/800core/1800mem/80C max
.987v/775core/1700mem/84C max
1.025V is pretty low for this card with 800/1800. But, 80°C is an abnormally high temperature for this card. However, even having said that, some people have noticed that their Golden Edition runs hotter due to just needing either the heatsink tightened down better, or perhaps a better thermal paste application - or both.

Quote:

Originally Posted by sparker View Post
Does it matter if my card isn't stable at 80C if it never reaches 80C?
I think it's possible. I've seen others in the Golden Edition club say that once these cards get into the 80's, it's very easy for them to become unstable.

Quote:

Originally Posted by sparker View Post
Why can other cards go ~100C?
Because the GTX 470s we have weren't stable in the 80s and so that's why they were given a GTX 465s BIOS.

But now a few very, very lucky Golden Edition owners can get hotter temps without instability, but they are very lucky to have such good cards.

Quote:

Originally Posted by sparker View Post
Why do I care if Furmark crashes it when everything else works?
What settings should I be running with Furmark anyway?
I guess instability is instability. So if it can't run Furmark, then it's not stable. Have you tried 800/1800 with 1.087V? How about 1.100V?

But now that I've asked that, I'd like to ask a temperature-related question: where do you keep your system? Is it in a deskhole? I mean, how much clearance does it have on the sides, top, and back?
 

·
Registered
Joined
·
190 Posts
Discussion Starter #13
Quote:

Originally Posted by TwoCables View Post
I guess instability is instability. So if it can't run Furmark, then it's not stable. Have you tried 800/1800 with 1.087V? How about 1.100V?
Well then, where do you draw the line? Furmark goes up to 2560x2048 so should I be able to run at that res? I guess I am having a problem with the one-size-fits-all mentality about stability test. Maybe Furmark is the silver bullet, but I don't know enough to argue my case. It just doesn't seem like one program should be the end-all,be-all.

Quote:
I think it's possible. I've seen others in the Golden Edition club say that once these cards get into the 80's, it's very easy for them to become unstable.

But now that I've asked that, I'd like to ask a temperature-related question: where do you keep your system? Is it in a deskhole? I mean, how much clearance does it have on the sides, top, and back?
Thanks, this has made me feel a bit better.

I ran 800/1800 at 1v there was a few artifacts, so I bumped it up to 1.025. From there I tried up to 1.087(max for this BIOS(I think)) and there was no difference in artifacts, but with each voltage increase it would crash at a lower temp, and heat up faster, so I put it back down to 1.025.

The system is out in the open, it is an Antec 300 case and I think it has pretty good airflow, if the amount of noise generated is a good indicator...
 

·
Not new to Overclock.net
Joined
·
77,827 Posts
Quote:

Originally Posted by sparker View Post
Well then, where do you draw the line? Furmark goes up to 2560x2048 so should I be able to run at that res? I guess I am having a problem with the one-size-fits-all mentality about stability test. Maybe Furmark is the silver bullet, but I don't know enough to argue my case. It just doesn't seem like one program should be the end-all,be-all.
I guess I'm realizing that I just don't really know.
But, I'm using an EVGA-identifying BIOS which allows me to use EVGA OC Scanner. When I use the recommended settings (as recommended by OCN's experts), I got my overclock to pass the full 2 hours of the test, which is the maximum length of time the test will run.

After that, I played all of my games and didn't see any strange graphical glitches or artifacts. That was nearly a month ago now, and everything is still going strong. I'm currently using 765/1848 (1848 is the stock memory clock for my BIOS, and 701 is the stock core clock), but if I try 770, or 775 core, then I get a couple of strange, brief artifacts in certain areas of Mirror's Edge. But at 765, it doesn't do it, so I'm guessing it's maxed out now.

Quote:

Originally Posted by sparker View Post
Thanks, this has made me feel a bit better.

I ran 800/1800 at 1v there was a few artifacts, so I bumped it up to 1.025. From there I tried up to 1.087(max for this BIOS(I think)) and there was no difference in artifacts, but with each voltage increase it would crash at a lower temp, and heat up faster, so I put it back down to 1.025.

The system is out in the open, it is an Antec 300 case and I think it has pretty good airflow, if the amount of noise generated is a good indicator...

I think I can hear your rig all the way over here.


Anyway, then maybe this might be one of those cards that may require a tightening of the heatsink, or perhaps better thermal paste (or just a better thermal paste application).

It may even just require experimentation with other BIOS versions.
I actually had to do that before I found one that was noticeably more stable.
 

·
Registered
Joined
·
190 Posts
Discussion Starter #15
Quote:

Originally Posted by TwoCables View Post
I guess I'm realizing that I just don't really know.
But, I'm using an EVGA-identifying BIOS which allows me to use EVGA OC Scanner. When I use the recommended settings (as recommended by OCN's experts), I got my overclock to pass the full 2 hours of the test,
Hmmm... OC Scanner is running on mine, I used to have a EVGA 460 on this machine... I don't see a "recommended settings" button though, do you care to elaborate what they might be?

btw, is it some sort of a crime that my 460 now resides in an eMachine?
 

·
Not new to Overclock.net
Joined
·
77,827 Posts
Quote:

Originally Posted by sparker View Post
Hmmm... OC Scanner is running on mine, I used to have a EVGA 460 on this machine... I don't see a "recommended settings" button though, do you care to elaborate what they might be?

btw, is it some sort of a crime that my 460 now resides in an eMachine?

Omg, that poor GTX 460. That's brutal.


Anyway...
There's no button for it. I was referring to settings that were recommended by OCN's experts (OCN = Overclock.net). Here they are:



Once you use all of these settings, they'll stick. Except, "Unlock power draw" always reverts back to being disabled/locked/unchecked when this program is closed.
 

·
Registered
Joined
·
190 Posts
Discussion Starter #17
When I do it this way, the logo doesn't spin like the normal stress test. Is that right? It's still putting a hurt on the GPU, there's just no movement.

10 mins, 0 artifacts 800/1800 1.025V
 

·
Not new to Overclock.net
Joined
·
77,827 Posts
Quote:

Originally Posted by sparker View Post
When I do it this way, the logo doesn't spin like the normal stress test. Is that right? It's still putting a hurt on the GPU, there's just no movement.

10 mins, 0 artifacts 800/1800 1.025V
It is indeed normal. I don't know the purpose for it, but it's certainly normal. It's just like ATiTool: the normal 3D View animates the furry cube, but when you have it scan for artifacts, it doesn't animate the cube. It also place the cube into the first frame of the animation and it just sits there looking like it's not doing anything.
 

·
Registered
Joined
·
190 Posts
Discussion Starter #19
Quote:

Originally Posted by TwoCables View Post
It is indeed normal. I don't know the purpose for it, but it's certainly normal. It's just like ATiTool: the normal 3D View animates the furry cube, but when you have it scan for artifacts, it doesn't animate the cube. It also place the cube into the first frame of the animation and it just sits there looking like it's not doing anything.
Thanks, I'll have a report in 1:45
 

·
Not new to Overclock.net
Joined
·
77,827 Posts
Quote:

Originally Posted by sparker View Post
Thanks, I'll have a report in 1:45

Alright then. I'll probably be here. hehe
 
1 - 20 of 25 Posts
Top