Overclock.net banner
1 - 16 of 16 Posts

·
Registered
Joined
·
201 Posts
Discussion Starter · #1 ·
Hey everyone, I have a few questions whose accurate answers will help me a great deal.

1st Q: I want to know what are the potential dangers of voltage tweaking while overclocking??

I mean is temperature rise the only concern and I can rise the voltage gradually up to achieve more and more core and memory clock until a risky temperature is reached?

Or overvoltage itself is a dangerous thing? Like the gpu might be damaged or burned due to solely overvoltage despite having the temperature under control?

2nd Q: Any other things I should care about while playing with the voltage??

3rd Q: If I increase voltage, naturally the card might need to use more power. But may it even cross the official TDP?

Isn't TDP the official maximum power a product can take under worst case scenario?

4th Q: I have a HD 7850 which has no voltage control in AMD overdrive. So if that implies I'm not supposed to tweak voltage myself for this card then the official TDP might not include the extra power usage due to voltage tweak right??

5th Q: There is a power control settings in CCC which remains on 0% by default. What does it mean?

Is it calculated like this: For example, my card's TDP is 130 watt. So if I take the button to 10%, the TDP of the card now becomes 130 + (10% of 130) =143watt. In other words, the card now can draw a maximum of 143 watt if needed.

If I'm wrong with this, please give me the correct definition.
 

·
Waiting on New Hardware
Joined
·
4,478 Posts
overvoltage cause much quicker degradation of the silicon die than simply overclocking. adds the 2 together and u get double whammy. that said, today's silicon are very robust and can handle at least 0.1v above their default voltage. the max i'd ever go is 0.2v above, although i've seen people who went 0.3v above (under water cooling) and they're just fine.
you don't have to worry about voltage limit since you're overclocking a GPU in software so the manufacturer already set a 'safe' maximum. for your card it's 1.3v i believe
you need 3rd party programs, such as MSI Afterbuner, to be able to overvolt.
the power thing is basically TDP. if u move it to +20 it means that you can exceed the card's TDP by 20%. remember to do this when overclocking or overvoltaging
 

·
Iconoclast
Joined
·
32,305 Posts
1 & 2. Voltage, temperature, and current draw (determined by clock speed and load) all contribute to damage via electromigration.

Even if only one factor is increasing, rate of degradation is increasing. 60C with 1.1v will last longer at 800MHz than it will at 1Ghz, all other things being equal. You can easily destroy a part with excess voltage, even if temperatures or clock speeds do not increase. Indeed, the more voltage you use, and/or the further you OC, the colder you want the part to be.

100C may last 10 years at stock, but to get more than 12-24 months out of a heavily OCed and overvolted part, you may need to consider 60C your limit, for example.

3. The official TDP can be exceeded at stock, though generally only in very demanding apps. Overclocking and/or overvolting can dramatically increase power consumption. The TDP on my 6950s is 200w. The cards can each pull ~300w when unlocked and OCed to 950/1450 with 1.2v, in tests like OCCT. Even in gaming they can easily exceed the +20% TDP limit, and would throttle like crazy if I didn't have PowerTune totaly disabled via MSI AB.

TDP is absolutely NOT the most power the card can consume, nor the most heat it needs dissipated in a worst case scenario. It's exactly what it says it is, thermal design power. This can be defined differently by different manufacturers.

In the case of AMD and NVIDIA GPUs, TDP is an arbitrary limit concocted to be able to offer a part with the maximum performance in typical gaming scenarios while cheaping out on VRM and cooling as much as possible. Exceed TDP and you get throttled. Not reach TDP and you get a small turbo boost.

4. Even moderately more voltage would push most cards well past their stock TDPs, and could even result in throttling in games.

5. Your conclusion is largely correct. Stock TDP + the slider value is the maximum the card is allowed to draw before throttling.

+20% is usually sufficent for most OCs on stock voltage, at least in games.

However, the true value needed to prevent throttling on most AMD cards, if voltage and clocks are push to their reasonable safe limits on air, and you wanted no throttling even in worst case scenarios, would be something closer to +50-70%, which is not available on most cards, and is why I disable PowerTune.
 
  • Rep+
Reactions: Driimit and james8

·
Registered
Joined
·
2,785 Posts
Answer.... Poof...
smile.gif
 

·
Registered
Joined
·
201 Posts
Discussion Starter · #7 ·
@Blameless, thank u for the elaborate answers.
Quote:
Voltage, temperature, and current draw (determined by clock speed and load) all contribute to damage via electromigration.

Even if only one factor is increasing, rate of degradation is increasing. 60C with 1.1v will last longer at 800MHz than it will at 1Ghz, all other things being equal.
So according to u, OC itself degrades the longevity or quality of a gpu, no matter how small the degradation is? Because current draw will increase with increase of clock speeds and u said any of the 3 factors will lessen the lifetime of a gpu to some extent. So OC isn't 100% harm free in the literal sense even if done 'properly' ?

My stock voltage is 1.138 and maximum allowed in Trixx and other softwares is 1.225 which . So if i need to raise it to that point, how much am I damaging the gpu practically? james8 said in his reply to this thread: overvoltage cause much quicker degradation of the silicon die than simply overclocking. adds the 2 together and u get double whammy. that said, today's silicon are very robust and can handle at least 0.1v above their default voltage. the max i'd ever go is 0.2v above, although i've seen people who went 0.3v above (under water cooling) and they're just fine.

Now, 1.225 -1.138 =0.087 which is well less than 0.1. So how much am I hurting my gpu in a practical sense if I go to 2.225?

If at 1.138V my temperature at maximum load (furmark etc stress testing tool, not games) is 77 C, then what do u think should be the temperature I have to descend to at 1.225V if I need to keep the gpu longevity constant?? I know it's impossible to give an exact and objective answer to this q. So am wanting an approximate answer from ur overall experience.

Quote:
Even moderately more voltage would push most cards well past their stock TDPs, and could even result in throttling in games.
I guess by throttling u mean, the degradation of performance when the performance requires more power than the (stock TDP+power control modification) ? If the (stock TDP+power control modification) < the power required under load while the card is OCed/overvolted, performance degradation will occur?

My cards TDP is 130 watt. And as u may know the HD 7800series cards(or probably the whole 7000 series) are really power efficient comparing to cards of similar performance. So how much extra power u assume it might take under load if i increase the voltage by 0.087 (i.e. stock to maximum)??

My card has one 6-pin connector. So it isn't supposed to go over 150 watt. But AMD overdrive itself gives me the freedom to push the 'power control settings' button upto 50%. So power control settings actually is what u and I said, then AMD is allowing to provide 195 watt(130 + 50% of 130) to the gpu and giving only one 6-pin connector on the other hand?? That doesn't make sense.. It it were any other software, like unlocked msi AB which is giving me that 50% then it might have made sense. I think u got my point.

Thank you. Waiting for answers...
 

·
Iconoclast
Joined
·
32,305 Posts
Quote:
Originally Posted by Driimit View Post

@Blameless, thank u for the elaborate answers.
So according to u, OC itself degrades the longivity or quality of a gpu, no matter how small the degradation is? Because current draw will increase with increase of clock speeds and u said any of the 3 factors will lessen the lifetime of a gpu to some extent. So OC isn't 100% harm free in the literal sense even if done 'properly' ?
If there is power going to the hardware, it's degrading/wearing out. Even at stock, even underclocked. OCing simply accelerates the process (all other things being equal).

Generally, this degradation is slow to the point that other components on the PCB would fail long before the GPU itself, or the whole card would be long obsolete when the GPU did finally fail, or degrade significantly.
Quote:
Originally Posted by Driimit View Post

Now, 1.225 -1.138 =0.087 which is well less than 0.1. So how much am I hurting my gpu in a practical sense if I go to 2.225?
1.225 will likely last a while. 2.225 would kill the chip in seconds at room temp.
Quote:
Originally Posted by Driimit View Post

If at 1.138V my temperature at maximum load (furmark etc stress testing tool, not games) is 77 C, then what do u think should be the temperature I have to descend to at 1.225V if I need to keep the gpu longivity constant?? I know it's impossible to give an exact and objective answer to this q. So am wanting an approximate answer from ur overall experience.
Really hard to say. A lot of variability from part to part and a lot of factors.

http://en.wikipedia.org/wiki/Black%27s_equation

If I had to take a guess, I'd say you'd probably want to stay below ~70C to get similar longevity.
Quote:
Originally Posted by Driimit View Post

I guess by throttling u mean, the degradation of performance when the performance requires more power than the (stock TDP+power control modification) ? If the (stock TDP+power control modification) < the power required under load while the card is OCed/overvolted, performance degradation will occur?
PowerTune will actively reduce clock speeds when the TDP exceeds the set limit.
Quote:
Originally Posted by Driimit View Post

My cards TDP is 130 watt. And as u may know the HD 7800series cards(or probably the whole 7000 series) are really power efficient comparing to cards of similar performance. So how much extra power u assume it might take under load if i increase the voltage by 0.087 (i.e. stock to maximum)??
Not too much.

In a perfect world, you could multiply your voltage by the ratio of your OC and get consumption relative to stock, but increased heat, leakage, and things like the VRM being outside it's peak efficiency curve is going to skew things.

I would guess, very roughly, that going from 1.138v to 1.225v is going to increase power consumption by 10-15%, if clocks stay the same.
Quote:
Originally Posted by Driimit View Post

My card has one 6-pin connector. So it doesn't supposed to go over 150 watt. But AMD overdrive itself gives me the freedom to push the 'power control settings' button upto 50%. So power control settings actually is what u and I said, then AMD is allowing to provide 195 watt(130 + 50% of 130) to the gpu and giving only one 6-pin connector on the other hand?? That doesn't make sense.. It it were any other software, like unlocked msi AB which is giving me that 50% then it might have made sense. I think u got my point.
Thank you. Waiting for answers...
The manufacturer can set the TDP limit, and this will apply to overdrive as well as 3rd party utilities.

In practice, you can pull a lot more than 75w through a 6-pin connector with little risk.
 

·
Registered
Joined
·
201 Posts
Discussion Starter · #9 ·
Thanks a lot for your descriptive and specific answer. And it was lot more convenient to read with questions being quoted above the answers.

Quote:
1.225 will likely last a while. 2.225 would kill the chip in seconds at room temp.
2.225 was a typo. I was supposed to write 1.225. So u think going from 1.138 to 1.225(0.087 difference) will have very trivial impact on the longevity of the graphics card in a practical sense?? As u said PCB will in general fail long before the GPU or b4 the card being obsolete etc.

So practically I can go to 1.225 and increase the clock and core speeds accordingly without any 'real-world'(not theoretical) risk of hurting the longevity, right(provided I reduce the temperature a few degrees)??
Quote:
PowerTune will actively reduce clock speeds when the TDP exceeds the set limit.
What is PowerTune? The power control settings in CCC?
Quote:
In a perfect world, you could multiply your voltage by the ratio of your OC and get consumption relative to stock, but increased heat, leakage, and things like the VRM being outside it's peak efficiency curve is going to skew things.
What do u mean by the ratio of OC? Ratio of increased clock speed to stock clock speed? If so, Then which clock?
Quote:
The manufacturer can set the TDP limit, and this will apply to overdrive as well as 3rd party utilities.

In practice, you can pull a lot more than 75w through a 6-pin connector with little risk.
I know a lot more can be pulled through 6-pin. But Isn't 75 watt the standard practice due to safety etc reasons? That's why I asked that q. Giving 50% room in CCC for power control means enabling me to go as much as 195 watt. So to keep the standard same, I think AMD's supposed to provide another 6-pin. For the same reason as the company would give two 6-pin connector if the offical TDP of the card were barely more than 150 watt i.e. 155/160 watt.

Another question: Suppose I've got a specific 'sweet spot' for my card in terms of core speed and memory speed keeping the voltage to the maximum my card allows(1.225 V) and at those values the card remains stable under stress test, games etc.

Now is it necessary to reduce everything to stock( clock speeds and voltage) when I'm not gaming or stress testing, just doing normal things in windows or in idle condition?? I think it's useless to reduce it, right??
Because the graphics card only uses what it needs. In idle condition, the clock speeds and voltage remains same no matter how much I OC it.
 

·
Iconoclast
Joined
·
32,305 Posts
Quote:
Originally Posted by Driimit View Post

So practically I can go to 1.225 and increase the clock and core speeds accordingly without any 'real-world'(not theoretical) risk of hurting the longevity, right(provided I reduce the temperature a few degrees)??
Oh, you'll likely be hurting the longevity, but I would still expect the card to last years of on and of use.
Quote:
Originally Posted by Driimit View Post

What is PowerTune? The power control settings in CCC?
PowerTune is the sensorfirmware/driver technology that allows AMD to measure and throttle the power consumption of it's cards. The power control settings in CCC are a small part of it.
Quote:
Originally Posted by Driimit View Post

What do u mean by the ratio of OC? Ratio of increased clock speed to stock clock speed? If so, Then which clock?
Core clock. Memory clock matters, but to a much smaller extent.

If my card is 1GHz stock, and I clock it to 1.2GHz, it's going to pull at least 1.2x as much current (probably more in real world scenarios). If it's also 1v stock and I need 1.2v for the OC, that 1.2x as much voltage, obviously. Thus, in this case im looking at at least 1.44x (1.2 x 1.2) as much power consumption as stock (probably closer to 1.6-1.7x after increased leakage and less efficient power conversion is taken into account), as an example.
Quote:
Originally Posted by Driimit View Post

Now is it necessary to reduce everything to stock( clock speeds and voltage) when I'm not gaming or stress testing, just doing normal things in windows or in idle condition?? I think it's useless to reduce it, right??
Because the graphics card only uses what it needs. In idle condition, the clock speeds and voltage remains same no matter how much I OC it.
If the card is automatically dropping to a lower power state when it's supposed to, and only running your OC in demanding 3D apps, then no, there is no reason to manually disable the OC. If you have such power states disabled for some reason (some of my uses result in instability during power state transitions, so I force constant clocks and voltages) then it might be prudent to manually set lower clocks when you don't need maximum performance.
 

·
Registered
Joined
·
201 Posts
Discussion Starter · #12 ·
Blameless, thank you very much for your specific answers. I got useful information from there. Now I have two more queries.

Quote:
Originally Posted by Blameless View Post

If my card is 1GHz stock, and I clock it to 1.2GHz, it's going to pull at least 1.2x as much current (probably more in real world scenarios). If it's also 1v stock and I need 1.2v for the OC, that 1.2x as much voltage, obviously. Thus, in this case im looking at at least 1.44x (1.2 x 1.2) as much power consumption as stock (probably closer to 1.6-1.7x after increased leakage and less efficient power conversion is taken into account), as an example.
If the card is automatically dropping to a lower power state when it's supposed to, and only running your OC in demanding 3D apps, then no, there is no reason to manually disable the OC. If you have such power states disabled for some reason (some of my uses result in instability during power state transitions, so I force constant clocks and voltages) then it might be prudent to manually set lower clocks when you don't need maximum performance.
Q.1: If I'm sure I've more than enough power in my +12V rail, isn't it better just to turn the Power control settings slider to maximum(in my case +50%) the CCC allows? In that case, I'm increasing the maximum power the GPU can use when it needs and also it will will use that much power only if needed, right?

Q.2: My gpu has 130W TDP. Taking the power control slider to max(+50%) will make it 195W TDP(as u said earlier that's the way the calculation works when answering one of my q). Now if under extreme load the gpu needs to use more than 195 TDP (for increased overclocking, overvolting or whatever reasons), what will happen??
 

·
Banned
Joined
·
21,557 Posts
1 - Max it.

2 - You're never going to draw more than 20% more. Board only has 1 pcie 6pin, that's 75w from the 6pin + 75w from the slot. Sapphire is pulling their big numbers game with the +50 PERCENT OMG! In other words its a gimmick.
 

·
Premium Member
Joined
·
11,034 Posts
Quote:
Originally Posted by Driimit View Post

Blameless, thank you very much for your specific answers. I got useful information from there. Now I have two more queries.
Q.1: If I'm sure I've more than enough power in my +12V rail, isn't it better just to turn the Power control settings slider to maximum(in my case +50%) the CCC allows? In that case, I'm increasing the maximum power the GPU can use when it needs and also it will will use that much power only if needed, right?
Q.2: My gpu has 130W TDP. Taking the power control slider to max(+50%) will make it 195W TDP(as u said earlier that's the way the calculation works when answering one of my q). Now if under extreme load the gpu needs to use more than 195 TDP (for increased overclocking, overvolting or whatever reasons), what will happen??
1. The best approach would be to build custom profiles in sapphire trixx for gaming. Its always better to push extra voltage only when needed. Once you have arrive at the max stable OC clocks, voltages and custom fan speeds after tweaking and testing save those settings in a custom profile. While gaming load the custom profile from sapphire trixx and run the game. when you quit revert back to default settings. you could create a default profile in trixx too with stock clocks of your card which is 920 mhz core clock, 0% power option, auto fan.

2. To answer your second question a 130w TDP is the thermal design power. at stock speeds the card will draw lesser. here is a PCS+ HD 7850 (1 Ghz) drawing 112w peak power in Crysis 2 and 119w max power in Furmark . Stock HD 7850 at 860 Mhz draws 96 and 101 w respectively.

http://www.techpowerup.com/reviews/Powercolor/HD_7850_PCS_Plus/26.html

Which version of catalyst are you using. I am using 12.8 and I can only see +20% power control for my HD 6950. +20% has always been the limit even in sapphire trixx. means With +20% power control you are looking at 130 x 1.2 = 156w which matches the 150w power draw possible through a PCI-E 2.0 slot and a 6 pin PCI-E power connector. remember AMD has to guarantee the card works on a wide range of motherboards , both PCI-E 3.0 and lower and wide range of PSUs. So AMD would not draw more than 75w from the PCI-E 6 pin power connector even if the spec allows for overdraw.
 

·
Registered
Joined
·
201 Posts
Discussion Starter · #16 ·
Quote:
Originally Posted by tsm106 View Post

2 - You're never going to draw more than 20% more. Board only has 1 pcie 6pin, that's 75w from the 6pin + 75w from the slot. Sapphire is pulling their big numbers game with the +50 PERCENT OMG! In other words its a gimmick.
The 6-pin PEG connector is able to provide much more than 75 watt in real(each pin in the 6-pin PEG is rated up to 8A and 6-pin PEG has 2 pins(sometimes 3) connected to +12V rail which makes the whole PEG capable of providing 96+96=192 watt.

PCI Express Auxiliary Graphics Power Connectors

So taking the power control at 50% max(which the CCC allows for my card) will make the TDP 130+50% of 130=195 watt. So no more than 120 watt is going through the PEG in worse case scenarios. (Though I must say even with huge overclocking and reasonable overvolting I didn't need to go more than a mere +10% as I explained later answering raghu78)

Quote:
Originally Posted by raghu78 View Post

1. The best approach would be to build custom profiles in sapphire trixx for gaming. Its always better to push extra voltage only when needed. Once you have arrive at the max stable OC clocks, voltages and custom fan speeds after tweaking and testing save those settings in a custom profile. While gaming load the custom profile from sapphire trixx and run the game. when you quit revert back to default settings. you could create a default profile in trixx too with stock clocks of your card which is 920 mhz core clock, 0% power option, auto fan.
Yeah I know what you said but your answer isn't relevant to my question I'm afraid. It has got nothing to do with saving profiles and running games with that and I in fact do these. I said isn't it better to max out the power control settings(provided I'm sure I've enough power in my +12V rail)? Because the gpu will use that much power only when needed and with maxing it out, I'm just giving the gpu the privilege to take the extra power if it needs. That was my question.

Quote:
Originally Posted by raghu78 View Post

2. To answer your second question a 130w TDP is the thermal design power. at stock speeds the card will draw lesser. here is a PCS+ HD 7850 (1 Ghz) drawing 112w peak power in Crysis 2 and 119w max power in Furmark . Stock HD 7850 at 860 Mhz draws 96 and 101 w respectively.
http://www.techpowerup.com/reviews/Powercolor/HD_7850_PCS_Plus/26.html
Which version of catalyst are you using. I am using 12.8 and I can only see +20% power control for my HD 6950. +20% has always been the limit even in sapphire trixx. means With +20% power control you are looking at 130 x 1.2 = 156w which matches the 150w power draw possible through a PCI-E 2.0 slot and a 6 pin PCI-E power connector. remember AMD has to guarantee the card works on a wide range of motherboards , both PCI-E 3.0 and lower and wide range of PSUs. So AMD would not draw more than 75w from the PCI-E 6 pin power connector even if the spec allows for overdraw.
I'm using the latest non-beta version. 12.8 I guess. And the power control range depends on the card, not on the driver as my observation says. My 7850 OC edition has +50% max limit, the 6670 I had didn't have any power controller at all(used the same ccc version) and as u said your 6950 has 20% max power control.

And I've +50% power control not only in Sapphire Trixx, but in CCC too which means AMD officially is allowing me to do that(unlike going past 1050 MHz in core clock with the help of Trixx whereas maximum limit for CCC is 1050 MHz).. So maybe AMD or Sapphire(in my case, the manufacturer) is considering the extra wattage that can be provided through 6-pin PEG(upto 192 watt as I said above answering raghu78).

So the power control is about the particular model of the card or even about the manufacturer maybe(as I've not tested any HD 7850 except my Sapphire one), not about the driver.

However, I've OCed my card to upto 1150 MHz in core clock(with 0.042V voltage increase for stability) just by increasing the power control upto 10%. Taking it to 0% and keeping the clocks and voltage same throttles the card as I observed in benchmark... And as I've no serious plan to OC the card more(because it will increase the temp. further and 1150 from 920 stock is a 25% increase and even 33% if considered that the reference boards from AMD originally have 860 MHz stock. As rumored, The HD 7850 is an overclocking beast indeed.), so I find it safer to keep the power control around 20% to 30%(though 10% does the job for the current OC as I said)

Thank you tsm106 and raghu78 for your replies.
 
1 - 16 of 16 Posts
This is an older thread, you may not receive a response, and could be reviving an old thread. Please consider creating a new thread.
Top