BTW, not trying to give you a hard time or anything, just asking/waiting for evidence that supports what you're saying. It's kind of hard to take your word on it when not 4 days ago you didn't understand the difference between Fermi and Kepler and thought that E@H performance was based upon GPU clock speed. Understand that it's one thing to question or research, it's another to make hard and fast claims about something that you admittedly understand little about.
That's ok, a lot of us don't know the intricate details of what makes all this tick. But... when you couple that with the fact that almost every recent post you make is with regard to either brand touting or bashing, then it looks like there's a bit of bias in your perspective.
No matter, the performance and evidence will show over time.
yes, it's not claimed 56-57C, because there 57 was when 7970 was single, but now there two of them near each other, and i have apply tiny overclocking to them, so the temperature raised a bit. and yes, i have opened my window for a better ventilation too.
you didn't understand the difference between Fermi and Kepler
i still believe kepler cards did not unleash their potential due to strange driver behaviour. for example: single 660ti crunch 2wu for a about 2200 seconds i.e. 1100 per wu and not faster. and nothing you can do about it.
but! if you set 660ti as a second gpu with a primary 680 for example (in my case) SUDDENLY it became crunching 2wu for 1900 second. if you don't believe me, than you may ask Jeroen. his 680 crunch faster if there a gtx 8800 installed second to 680.
so what do we must conclude? at least 2200 seconds for 2wu is not top performance of 660ti, it is clearly what it can run faster, but no one of us know why it does not. so my touting roots in this simple facts.
if this magic is a key difference between Fermi and Kepler, thank you are right and i don't understand difference between Fermi and Kepler.
You need to check sensors with a different application because they are not being reported correctly. Unless your ambient temperature is 10C, there simply is no way. Try a different application, Psensor for example.
I've had different apps report differently myself. One said my cpu was running 44C, another revealed it running at 70C. Simply put, you need to use another app to check temps because there is no way a 7970 under 98% load runs at 59C. A lot of people are lucky to get that at idle.
Here's a real example.
37C for cpu right?
Wrong. Here's the real temp
Quote:
i still believe kepler cards did not unleash their potential due to strange driver behaviour. for example: single 660ti crunch 2wu for a about 2200 seconds i.e. 1100 per wu and not faster. and nothing you can do about it.
Scroll down and read the answers. It is explained. It's based on architecture, not drivers, not clockspeed, not amount of RAM.
The Kepler cards blow Fermi cards out of the water with regard to games. If there was a driver issue that forced the Kepler to underperform, this would be apparent in games.
no it's not 10C, but, i think, 18C. man, you don't believe my crunching time, you don't believe my card's sensor, you don't believe my fingers when it report me: "heatsink of this 7970 is just warm, we didn't get burn". what else don't you believe concerning me? may i get full list? ) thank you.
why should i don't believe sensor if it report about 25C then card is idle? btw, 7970 consumes only 15W when idle. so, it's sensor reports real temp then idle, but lies when card is under load?
or i have to prove 25C by screenshot again? )
That's a blower card. Blowers, weather ATI or nVidia always run much hotter than non-blowers do assuming a well ventilated case (in a poorly ventilated case, a blower is less likely to cook the GPU).
[H]OCP, saw 53C for an overclocked 7970 ghz card; amusing because of the way XFX configured its fan profile it's actually running cooler than at factory clocks because the fan ramped faster than heat production.
Where did I say I didn't believe your crunching time?
What I know is two things...
1. You don't have a very good understanding of GPU performances.
2. You have an obvious brand bias.
That's all I know for certain. These two things lead me to doubt a lot of what you say.
It is actually you that are in disbelief. I show you links proving the architecture difference Fermi VS Kepler and you say its drivers. I give you links showing tested temps of your card and you say it's not true. I post pictures for you showing how two different apps can report different temperatures and you say that isn't the case. You don't even know why a 680 was at 705mhz. In short, you really don't have enough information to make hard and fast claims about this or that.
Difference here is, I'm not advocating a brand to anyone. I have no agenda. I have no need for people to agree with me, nor do I need people to commiserate with. I am not making any claims, I am examining yours.
The only thing I'm interested in is accuracy and truth. I have shown that some of the things you have proposed are not absolute facts and there is room for doubt. Try not to take things personal, which shouldn't happen if the only thing you're interested in is finding the facts. But I don't think you're interested in that because you refuse to look at (much less explain away) any contrary evidence or findings.
You must know that in scientific circles, when an idea or proposal is brought forth, it is up to the peers to try to falsify the claims. That is the duty of the community when examining any claim. If all you had were people that just took what other said on their word, we'd still be in the dark ages.
[H]OCP, saw 53C for an overclocked 7970 ghz card; amusing because of the way XFX configured its fan profile it's actually running cooler than at factory clocks because the fan ramped faster than heat production.
You know what, nevermind. I looked through this guy's posts and some of them are 3 years old railing against nvidia cards, meanwhile using them the whole time.
Where did I say I didn't believe your crunching time?
here: I don't completely believe the numbers you're getting
Quote:
What I know is two things...
1. You don't have a very good understanding of GPU performances.
agree, i didn't ever experience with cuda or opencl programming.
at this moment i just watching manufacturers specs for their gpus.
Quote:
2. You have an obvious brand bias.
do you mean i'm ati fanboy?
disagree, past two years my crunching cards was a set of nvidias gpus.
recenty i bought a new 660ti and was a little disappointed with it's performance and performance/price ratio. then i bought 680 and get same disappontment. then i had return 680 to the shop and have decided to try ati/amd and surpisingly got an amazing performance.
ati is faster in e@h and it is cheaper so that is all my "fanboyism".
Quote:
That's all I know for certain.
as we see 50% only.
Quote:
These two things lead me to doubt a lot of what you say.
It is actually you that are in disbelief. I show you links proving the architecture difference Fermi VS Kepler and you say its drivers.
then, please, explain the fact 660ti as well as 680 can perform faster under some not obvious condition? again, you may doubt my words, but you may ask Jeroen. he has those strange behaviour of his 680 too.
Quote:
I give you links showing tested temps of your card and you say it's not true.
what? where?
Quote:
I post pictures for you showing how two different apps can report different temperatures
yes, i see. 1st. different apps may make reading from different sensors.
2nd. ati app make reading from ati gpu, i have no reason to doubt it.
Quote:
and you say that isn't the case.
again, point there did i say so.
Quote:
You don't even know why a 680 was at 705mhz.
of course i don't know. why do you thing i'm asking? because i don't know why it is 705 instead of 1000.
Quote:
In short, you really don't have enough information to make hard and fast claims about this or that.
i have some facts from my working system and that is what i claim.
Quote:
Difference here is, I'm not advocating a brand to anyone. I have no agenda.
I have no need for people to agree with me, nor do I need people to commiserate with. I am not making any claims, I am examining yours.
The only thing I'm interested in is accuracy and truth.
Quote:
I have shown that some of the things you have proposed are not absolute facts and there is room for doubt.
i'm talking here about my system crunching time and gpu temperature. all WUs running time is recorded in stat, you can freely inspecting them. you ask for a temperature screenshot, i send it too you. if you have a doubt, then i can take my glass thermometer, put it inside 7970 as close to gpus as possible and make a photo of it's measurement.
BTW, not trying to give you a
)
BTW, not trying to give you a hard time or anything, just asking/waiting for evidence that supports what you're saying. It's kind of hard to take your word on it when not 4 days ago you didn't understand the difference between Fermi and Kepler and thought that E@H performance was based upon GPU clock speed. Understand that it's one thing to question or research, it's another to make hard and fast claims about something that you admittedly understand little about.
That's ok, a lot of us don't know the intricate details of what makes all this tick. But... when you couple that with the fact that almost every recent post you make is with regard to either brand touting or bashing, then it looks like there's a bit of bias in your perspective.
No matter, the performance and evidence will show over time.
yes, it's not claimed 56-57C,
)
yes, it's not claimed 56-57C, because there 57 was when 7970 was single, but now there two of them near each other, and i have apply tiny overclocking to them, so the temperature raised a bit. and yes, i have opened my window for a better ventilation too.
RE: you didn't understand
)
i still believe kepler cards did not unleash their potential due to strange driver behaviour. for example: single 660ti crunch 2wu for a about 2200 seconds i.e. 1100 per wu and not faster. and nothing you can do about it.
but! if you set 660ti as a second gpu with a primary 680 for example (in my case) SUDDENLY it became crunching 2wu for 1900 second. if you don't believe me, than you may ask Jeroen. his 680 crunch faster if there a gtx 8800 installed second to 680.
so what do we must conclude? at least 2200 seconds for 2wu is not top performance of 660ti, it is clearly what it can run faster, but no one of us know why it does not. so my touting roots in this simple facts.
if this magic is a key difference between Fermi and Kepler, thank you are right and i don't understand difference between Fermi and Kepler.
You need to check sensors
)
You need to check sensors with a different application because they are not being reported correctly. Unless your ambient temperature is 10C, there simply is no way. Try a different application, Psensor for example.
I've had different apps report differently myself. One said my cpu was running 44C, another revealed it running at 70C. Simply put, you need to use another app to check temps because there is no way a 7970 under 98% load runs at 59C. A lot of people are lucky to get that at idle.
Here's a real example.
37C for cpu right?
Wrong. Here's the real temp
shenanigans...
You need to go back and read my post here;
http://einsteinathome.org/node/196613&nowrap=true#119937
This has already been explained. There is no issue whatsoever with the Kepler driver which btw, is a unified driver. This link explains exactly why Kepler cards do not perform in crunching like Fermi cards do;
http://stackoverflow.com/questions/10765518/how-to-quantify-the-processing-tradeoffs-of-cuda-devices-for-c-kernels
Scroll down and read the answers. It is explained. It's based on architecture, not drivers, not clockspeed, not amount of RAM.
The Kepler cards blow Fermi cards out of the water with regard to games. If there was a driver issue that forced the Kepler to underperform, this would be apparent in games.
[img][/img]
no it's not 10C, but, i
)
no it's not 10C, but, i think, 18C. man, you don't believe my crunching time, you don't believe my card's sensor, you don't believe my fingers when it report me: "heatsink of this 7970 is just warm, we didn't get burn". what else don't you believe concerning me? may i get full list? ) thank you.
why should i don't believe sensor if it report about 25C then card is idle? btw, 7970 consumes only 15W when idle. so, it's sensor reports real temp then idle, but lies when card is under load?
or i have to prove 25C by screenshot again? )
RE: RE: my 7970 under
)
That's a blower card. Blowers, weather ATI or nVidia always run much hotter than non-blowers do assuming a well ventilated case (in a poorly ventilated case, a blower is less likely to cook the GPU).
[H]OCP, saw 53C for an overclocked 7970 ghz card; amusing because of the way XFX configured its fan profile it's actually running cooler than at factory clocks because the fan ramped faster than heat production.
http://www.hardocp.com/article/2012/10/30/xfx_double_d_hd_7970_ghz_edition_video_card_review/10
Where did I say I didn't
)
Where did I say I didn't believe your crunching time?
What I know is two things...
1. You don't have a very good understanding of GPU performances.
2. You have an obvious brand bias.
That's all I know for certain. These two things lead me to doubt a lot of what you say.
It is actually you that are in disbelief. I show you links proving the architecture difference Fermi VS Kepler and you say its drivers. I give you links showing tested temps of your card and you say it's not true. I post pictures for you showing how two different apps can report different temperatures and you say that isn't the case. You don't even know why a 680 was at 705mhz. In short, you really don't have enough information to make hard and fast claims about this or that.
Difference here is, I'm not advocating a brand to anyone. I have no agenda. I have no need for people to agree with me, nor do I need people to commiserate with. I am not making any claims, I am examining yours.
The only thing I'm interested in is accuracy and truth. I have shown that some of the things you have proposed are not absolute facts and there is room for doubt. Try not to take things personal, which shouldn't happen if the only thing you're interested in is finding the facts. But I don't think you're interested in that because you refuse to look at (much less explain away) any contrary evidence or findings.
You must know that in scientific circles, when an idea or proposal is brought forth, it is up to the peers to try to falsify the claims. That is the duty of the community when examining any claim. If all you had were people that just took what other said on their word, we'd still be in the dark ages.
RE: [H]OCP, saw 53C for an
)
oh, nice find Dan. That's pretty funny actually, can we say design flaw? Maybe not, pretty low temp at stock speed even.
Maybe OP has fans at 100%? Or has aftermarket fans added? Haven't even gotten to that though, guy won't even double check his own results.
You know what, nevermind. I
)
You know what, nevermind. I looked through this guy's posts and some of them are 3 years old railing against nvidia cards, meanwhile using them the whole time.
RE: Where did I say I
)
here: I don't completely believe the numbers you're getting
agree, i didn't ever experience with cuda or opencl programming.
at this moment i just watching manufacturers specs for their gpus.
do you mean i'm ati fanboy?
disagree, past two years my crunching cards was a set of nvidias gpus.
recenty i bought a new 660ti and was a little disappointed with it's performance and performance/price ratio. then i bought 680 and get same disappontment. then i had return 680 to the shop and have decided to try ati/amd and surpisingly got an amazing performance.
ati is faster in e@h and it is cheaper so that is all my "fanboyism".
as we see 50% only.
then, please, explain the fact 660ti as well as 680 can perform faster under some not obvious condition? again, you may doubt my words, but you may ask Jeroen. he has those strange behaviour of his 680 too.
what? where?
yes, i see. 1st. different apps may make reading from different sensors.
2nd. ati app make reading from ati gpu, i have no reason to doubt it.
again, point there did i say so.
of course i don't know. why do you thing i'm asking? because i don't know why it is 705 instead of 1000.
i have some facts from my working system and that is what i claim.