MSI GTX 670 Power Edition Twin Frozr 4!

with afterburner 221, the core voltage slider is greyed out for me. either I'm stupid or the power edition is capped at 1.175v like every other gtx 670.
please tell me I'm stupid lol

If its "'greyed out", then my guess is you didn't check the "unlock voltage" box in the options

but i already got word from another member on another forum...he tested it on his MSI, and he says voltage is still locked with sliders maxed
 
If its "'greyed out", then my guess is you didn't check the "unlock voltage" box in the options

but i already got word from another member on another forum...he tested it on his MSI, and he says voltage is still locked with sliders maxed



Custom unlocked BIOS might help. Any way to edit the BIOS?
 
I was also sad to find that moving the voltage slider didn't have any results, but a very moderate +116 core gave me a boost of 1292MHz running 3DMark11. I am still VERY early in my testing phase, but so far this card (at stock speeds) is more than double my previous xScore (with a 560ti). Also, I seem to be topping out at 105% Power with the limit slider set to max in Afterburner.

Much more testing to do, but so far I am a very [H]appyCustomer.
 
1300+ anything is a great OC isnt it? Either these cards are great or you got a premo card. I will find out and report back my experience with mine tomorrow which is when it arrives. My current plan is at stock to run Heaven for 30-60 minutes max everything, then play BF3 for a while. If (hopefully, i pray) everything goes perfect, i will start OC'ing and see how far my specific card can go.
As far as technique to OC the card....do i just raise the power limit to like 110% then slowly raise the core clock 10MHz at a time with Afterburner, running Heaven for 30 minutes every time i raise it? Havent OC'ed a GPU before.

yeah basically set the max power limit in MSI AF....its not going to hurt anything, it just basically saying, the card can go up to x amount of power limit before it can't go more....most aftermarket cards can do 115-145% ish power limit....problem is, it becomes irrelevant because stock 1.175 volts only allows so much overclocking, and drawing not so much power

and yes i would say loop heaven for an hour...then go ahead and play a few hours of bf3....IMO, I always go for the "top down" method....instead of the "bottom up"....I would shoot straight for 1300mhz and see if it crashes immediately or within 10 minutes....if it doesn't, then crank it up 10mhz :)
 
OKAY. So I'm stable roughly around 1316mhz Core and 7.29ghz Memory. Got a score of 1266 @ 1080p on Heaven 3.0 w/ everything max. Temps and Fan speed aren't too bad at this point either, especially considering the room I'm in is almost 80 degrees Fahrenheit. Inaudible below 50% fan speed. 50 - 60% is still pretty quiet for a video card. Above that and I'd say it's as loud as my previous 6950 sapphire toxic but the fan never got that high during testing. I think it's pretty safe to say now the twin frozr coolers are pretty reliable in terms of cooling and noise level.
Here's the stats: At stock OC speeds, i was getting 44.9 average fps on heaven. I was expecting a greater increase in performance for a >1300mhz OC. You guys think the score is a little low for these clocks?

1ststableoc1080p.png
 
you need to lower your memory and run it again. I get a higher score than you with just 1277mhz gpu and 7030mhz memory.



 
Last edited:
almost forgot lol:
i5 2500k @ 4.5ghz
msi gtx 670 power edition
8gb g.skill ram 1600mhz
120gb kingston hyperx ssd + 500gb hdd
600W psu
Everything in a silverstone TJ08-E

hope that helps. Also, I just noticed something weird. Can anyone else confirm this with their 670 power edition card or is it just mine? GPU-z measures the boost clock 19mhz higher than advertised on newegg. Lucky draw or typo?
lolk.png
 
the official boost clocks could have changed but really it does not mean much. btw I hope you saw my previous post. knock that memory down to 7000mhz and run it again and your score may go up. there comes a point where ocing gddr5 stops helping and even can start hurting performance slightly.
 
the official boost clocks could have changed but really it does not mean much. btw I hope you saw my previous post. knock that memory down to 7000mhz and run it again and your score may go up. there comes a point where ocing gddr5 stops helping and even can start hurting performance slightly.

Quoted for truth.
 
what you're talking about i think happened to me at +700mhz offset. ocing the memory stopped producing gains and created artifacts at > +700 offset. every incremental increase in the offset prior, however produced fps gains. so i slowly brought it down to +640. But I will try your recommendation tomorrow, calling it a night. thanks
 
what you're talking about i think happened to me at +700mhz offset. ocing the memory stopped producing gains and created artifacts at > +700 offset. every incremental increase in the offset prior, however produced fps gains. so i slowly brought it down to +640. But I will try your recommendation tomorrow, calling it a night. thanks
yeah try again tomorrow as there is no way I should be getting a better score than you with slower core and memory speeds. also did you ever mention drivers? you need to be using the 304.48 drivers if you are not already doing so.
 
http://forums.guru3d.com/showthread.php?p=4345545

hey guys, someone over at OCN linked me to this guru3d thread....MSI apparently stepped in and says there will be voltage control for the 670 PE...not sure how legit it is though...the guy sounds like he can barely speak english?

Well most MSI reps can't speak english very well lol. You ever tried their CS? Might need a translator.

Love those guys though, the lightning lines are always awesome.
 
Sigh... right now these companies (Galaxy and MSI) are blatantly false advertising, language barrier or not. They know what voltage control is. It's not like there aren't any Asian owners of these cards who are asking the same thing.

Disappointed, again.
 
you need to lower your memory and run it again. I get a higher score than you with just 1277mhz gpu and 7030mhz memory.




Its not the memory most likely but the benchmark he's using. Heaven is terrible for benching, its gives wildly inconsistent results. If you run that benchmark 5 times even at stock speeds, you'll see different results. Heaven is a fantastic stability tool but a terrible benchmark tool.
 
what you're talking about i think happened to me at +700mhz offset. ocing the memory stopped producing gains and created artifacts at > +700 offset. every incremental increase in the offset prior, however produced fps gains. so i slowly brought it down to +640. But I will try your recommendation tomorrow, calling it a night. thanks

You'll see solid gains on your memory OC until it just fails. Heaven is a just a piss poor inconsistent benchmark. Pick ANY OTHER benchmark, in-game or synthetic and you will see consistent gains from overclocking your memory until it won't pass the benchmark anymore and crashes.
 
I just got mine unboxed. Here are a few quick pics:

IMAG0015.jpg

IMAG0018.jpg


Haven't ran any tests yet other than playing with the fan speed. It's silent at 45% and below inside my Temjin TJ08-e case and somewhat audible inside the case at 50% but not at all distracting. As the fan ramps up to 60% and above, the pitch increases and becomes more noticeable. It is 75 degrees Fahrenheit inside my house right now (23.9 celsius) and the card is idling at 30 degrees celsius. The box clearly says it supports voltage adjustments, so I am downloading afterburner 2.2.2 and gearing up for some tests!
 
Its not the memory most likely but the benchmark he's using. Heaven is terrible for benching, its gives wildly inconsistent results. If you run that benchmark 5 times even at stock speeds, you'll see different results. Heaven is a fantastic stability tool but a terrible benchmark tool.

It varies a few FPS depending on processor and processor clock speed. But I think it gives pretty consistent results using the same system. I would not expect anything varying more than 1 frame, or even half a frame on that benchmark.

For the 600-series maybe what you're experiencing is due to throttling of the cards above 70C. If you can keep a stable boost clock, then it will give consistent results.
 
definitely not throttling due to temperature. if you look at the vitals, my card didn't exceed 65 degrees during heaven
 
Well most MSI reps can't speak english very well lol. You ever tried their CS? Might need a translator.

Love those guys though, the lightning lines are always awesome.

they're technically telling the truth. you CAN adjust voltage, just not above the 1.175v cap. That would be termed unlocked voltage
 
You'll see solid gains on your memory OC until it just fails. Heaven is a just a piss poor inconsistent benchmark. Pick ANY OTHER benchmark, in-game or synthetic and you will see consistent gains from overclocking your memory until it won't pass the benchmark anymore and crashes.
I watched framerates dip in the AvP bench before my memory was clocked high enough to crash.


Its not the memory most likely but the benchmark he's using. Heaven is terrible for benching, its gives wildly inconsistent results. If you run that benchmark 5 times even at stock speeds, you'll see different results. Heaven is a fantastic stability tool but a terrible benchmark tool.
I can run that bench 10 times in a row and it is never more than .1 or.2 difference. in this case I ran it twice and got the exact same framerate. he has the same cpu as me and is at 4.5 where I am at 4.4. he also has the same memory speed so overall there is nothing that from a hardware standpoint should make him get a lower score.


EDIT: for metro 2033 I get 40.77 fps with my memory at 7160 but 42.24 fps with it at 7030. that's just a quick example showing memory ocing can cause slowdowns at some point.
 
Last edited:
yeah try again tomorrow as there is no way I should be getting a better score than you with slower core and memory speeds. also did you ever mention drivers? you need to be using the 304.48 drivers if you are not already doing so.

so I dropped the memory to 1700mhz like you suggested and performance dropped as expected. I even matched my clocks to your clocks just to see if I'd get a similar score and it still came out lower than yours. Are you using the new beta driver? I think im going to update to that and try again
 
so I dropped the memory to 1700mhz like you suggested and performance dropped as expected. I even matched my clocks to your clocks just to see if I'd get a similar score and it still came out lower than yours. Are you using the new beta driver? I think im going to update to that and try again
its the driver then because the 304.48 is supposed to help out in Heaven. that is why I asked if you were using it. you need the 304.48 anyway as it fixes the game stuttering issues too.
 
I can also confirm that voltage cannot currently be raised above 1.175. Also MSI afterburner 2.2.2 does not have options to adjust the vram or PLL voltage.
 
I hear MSI's triple overvoltage promise will be fulfilled in a new release of Afterburner
 
Damn you MSI!!! LOL :D

A card with the potential to rule them all, and what do they do.

They give us our dream date with the girl of our dreams who's hot and ready, and they only let us put the tip in, with a promise to go further later!! :D Whyyyyyyyy :confused:
 
Damn you MSI!!! LOL :D

A card with the potential to rule them all, and what do they do.

They give us our dream date with the girl of our dreams who's hot and ready, and they only let us put the tip in, with a promise to go further later!! :D Whyyyyyyyy :confused:

HA Just the tip. Nothing worse than a damn tease.

If Im lucky enough to snag a lighting Im sure I'll feel the same way.
 
Yeah, maybe msi can put 2 blue dots over the i on their logo to signify the blue balls they give their customers at times LOL

1328883568.jpg
 
Just put my 670 PE in my PC and am currently running Heaven. I can say this, it is significantly less "silent" than the 670 DirectCU II Top i had. When the Top would run Heaven for a hour and not crash, and it would get up to 63C degrees, fans at 50% or w/e, it was literally silent. This things not near that. If anything id say a little noisy. Disapointed by that.
Will update with results of hour long Heaven benchmark, also with the results of my attemps to OC.

Update: Technically my fan profile was set quite aggressively. I was at 59C with 67% fan speed about 10 minutes into the Benchmark. I made the fan profile more conservative, however i might point out, i tested the Top's fan noise by listening to it when it was at 60%, 70, 80, 90, 100...and from what i remember it at 100 wasnt as loud as the 670 PE at 75%.


UPDATE 2: Okay so i just ran about 45 minutes of Heaven, and a match of BF3 on Ultra everything 1080P (Everything totally stock on the GPU side, not a thing changed). Here are my results:

Power: % Max 87
GPU Temp: Max 59
GPU Usage %: Max 99
Fan Speed Max: 54%
Core Clock: Max 1202MHz
Shader Clock: Max 2404MHz
Mem Clock: Max 3005MHz

Okay so heres what im wondering and thinking: So i look at these numbers and they seem low. The temp seems low, the Power Max seems low, the Core Clock seems like it could be higher. What does it take for my Power % to be 90%+? Isnt the memory clock supposed to be 6008MHz? Why is mine maxing at 2404MHz?
 
Last edited:
I hear MSI's triple overvoltage promise will be fulfilled in a new release of Afterburner


That is what I read in the MSI Lightning GTX 680 thread.

1.3v will be a nice boost if the VRM design is incredibly well designed and efficient.

This card should really distinguish itself with over clock ability once the 1.3v Afterburner unlock kicks in.
 
Last edited:
Just put my 670 PE in my PC and am currently running Heaven. I can say this, it is significantly less "silent" than the 670 DirectCU II Top i had. When the Top would run Heaven for a hour and not crash, and it would get up to 63C degrees, fans at 50% or w/e, it was literally silent. This things not near that. If anything id say a little noisy. Disapointed by that.
Will update with results of hour long Heaven benchmark, also with the results of my attemps to OC.

Update: Technically my fan profile was set quite aggressively. I was at 59C with 67% fan speed about 10 minutes into the Benchmark. I made the fan profile more conservative, however i might point out, i tested the Top's fan noise by listening to it when it was at 60%, 70, 80, 90, 100...and from what i remember it at 100 wasnt as loud as the 670 PE at 75%.


UPDATE 2: Okay so i just ran about 45 minutes of Heaven, and a match of BF3 on Ultra everything 1080P (Everything totally stock on the GPU side, not a thing changed). Here are my results:

Power: % Max 87
GPU Temp: Max 59
GPU Usage %: Max 99
Fan Speed Max: 54%
Core Clock: Max 1202MHz
Shader Clock: Max 2404MHz
Mem Clock: Max 3005MHz

Okay so heres what im wondering and thinking: So i look at these numbers and they seem low. The temp seems low, the Power Max seems low, the Core Clock seems like it could be higher. What does it take for my Power % to be 90%+? Isnt the memory clock supposed to be 6008MHz? Why is mine maxing at 2404MHz?

Are you getting those values off of the graphs from an overclocking app? 6008mhz is the effective memory clock. You just have to take the value you have up there and x2: you get 6010mhz which is about right for a stock gtx 670. And it's a good thing that your power percent is not at 90+ yet. It means you still have a good amount of headroom to overclock. It's not until you begin reaching the upper limits of your core and memory clocks that you start seeing power % going slightly above 100%. So basically you can use your power percent as a marker to gauge how much overclocking headroom you still have left
 
Are you getting those values off of the graphs from an overclocking app? 6008mhz is the effective memory clock. You just have to take the value you have up there and x2: you get 6010mhz which is about right for a stock gtx 670. And it's a good thing that your power percent is not at 90+ yet. It means you still have a good amount of headroom to overclock. It's not until you begin reaching the upper limits of your core and memory clocks that you start seeing power % going slightly above 100%. So basically you can use your power percent as a marker to gauge how much overclocking headroom you still have left

Oh okay awesome. So i guess that means with my Top at 98% Power Max, there wouldnt have been much headroom compared to my low 87%. Oh okay, that makes sense haha (the x2=6010MHz). Im reading it off MSI Afterburner btw.

Okay so for OC'ing. Do i want to raise my Power Limit to say 110% or something right off the bat? Or start by raising the Core Clock to say 1300MHz and run Heaven and see if its stable. If not, look at the MSI AB readout, and then raise the Power Limit and re-Heaven and see if it crashes or not?

UPDATE: Set ABurner to Power Limit 110% and Core Clock +103 which created 1305MHz Core Clock and 86 Power % Max, but failed Heaven after a few minutes. Any recommendations?
 
Last edited:
My PE boosts to 1229 out of the box with no change in settings, but even with a maxed out power slider, I'm only able to do +40 on the core, for a max boost of 1269 in heaven. Memory was pretty good though, I'm able to hit 7200mhz (+600) with continued performance scaling in Heaven.

I look forward to afterburner getting updated to support the voltage tweaks. I'd like to get 1350 on the core for every day use.

At custom fan settings ramping up to 50% max, heaven 3.0 hits 67 degrees Celsius with room temps at 28.8 Celsius (75 degrees Fahrenheit). In game benchmarking, temps are maxing out at around 63 Celsius. 50% is barely audible when inside my Temjin tj08-e case. If the computer's speakers are turned on to any appreciable noise level, I cannot hear the video card.
 
Oh okay awesome. So i guess that means with my Top at 98% Power Max, there wouldnt have been much headroom compared to my low 87%. Oh okay, that makes sense haha (the x2=6010MHz). Im reading it off MSI Afterburner btw.

Okay so for OC'ing. Do i want to raise my Power Limit to say 110% or something right off the bat? Or start by raising the Core Clock to say 1300MHz and run Heaven and see if its stable. If not, look at the MSI AB readout, and then raise the Power Limit and re-Heaven and see if it crashes or not?

UPDATE: Set ABurner to Power Limit 110% and Core Clock +103 which created 1305MHz Core Clock and 86 Power % Max, but failed Heaven after a few minutes. Any recommendations?

You want to give yourself as high of a roof as you can so the gpu doesn't throttle for any reason. That means set your power target, or i guess power limit in your case, to max and set up a fan profile so that temps don't go above 70 degrees during OCing. Also, pegging the voltage to the max, 1.175v, is a good idea. Now rerun Heaven at those clocks again and see if that stabilized it. If not, decrease clocks incrementally until you do get stable. You're going to have to find a balance between Core and Memory. Do one first and then the other. I think just about everyone uses gpu-z for reading max clocks. good luck. I ended up with 1316mhz core and 7.27ghz memory. There hasn't been enough reports out there for the 670 power edition to judge if that's high or just normal for this card.
 
My PE boosts to 1229 out of the box with no change in settings, but even with a maxed out power slider, I'm only able to do +40 on the core, for a max boost of 1269 in heaven. Memory was pretty good though, I'm able to hit 7200mhz (+600) with continued performance scaling in Heaven.

I look forward to afterburner getting updated to support the voltage tweaks. I'd like to get 1350 on the core for every day use.

At custom fan settings ramping up to 50% max, heaven 3.0 hits 67 degrees Celsius with room temps at 28.8 Celsius (75 degrees Fahrenheit). In game benchmarking, temps are maxing out at around 63 Celsius. 50% is barely audible when inside my Temjin tj08-e case. If the computer's speakers are turned on to any appreciable noise level, I cannot hear the video card.

Hey can you do a heaven benchmark with max everything @ 1080p? curious as to what score you get

Also, I just wanted to say, I have the same case!
 
Last edited:
Well with Power limit set to max (114%) and Core Voltage set to max (+100) and the Core Clock set to 3000MHz, it failed in a few minutes. I then downclocked the Core Clock to 1271MHz and ran it for about 15 minutes and no crashes. Obviously i will run it longer and see if i can move it up from there as well, but it seems my card is not destined to be a 1300MHz card.

Interesting my Power % never goes higher than 88%.

Update: Just fiddled around a bit OCing and ran this setup for...3 hours without a problem:

Power: 90% Max
GPU Temp: 62C Max
Core Clock: 1271MHz Max
Shader Clock: 2542MHz Max
Memory Clock: 3005MHz Max

Voltage: +0
Power Limit: 100%
Core Clock: +69

Weird that it can do that (1271MHz) with stock Power Limit, but when i up the power limit to max, it cant do 1300MHz.
 
Last edited:
I watched framerates dip in the AvP bench before my memory was clocked high enough to crash.


I can run that bench 10 times in a row and it is never more than .1 or.2 difference. in this case I ran it twice and got the exact same framerate. he has the same cpu as me and is at 4.5 where I am at 4.4. he also has the same memory speed so overall there is nothing that from a hardware standpoint should make him get a lower score.


EDIT: for metro 2033 I get 40.77 fps with my memory at 7160 but 42.24 fps with it at 7030. that's just a quick example showing memory ocing can cause slowdowns at some point.

Cool story bro, well my two Gigabyte's are special then because I don't see any of that. Although the Heaven benchmark does give inconsistent results even at stock clocks which is why I don't use it for benching. You never admit to being wrong about anything though so I don't even know why I bother discussing anything with you. Clock your memory to whatever works best for you which obviously sounds lower than my Gigabyte's. I'm very happy that my Gigabytes show solid gains until the app crashes. I'm about 110% positive I've done more benching and testing than you have on your card since I've had mine longer and tested two of them and yet neither of my cards are showing any negative gains from overclocking the memory, and it was the same way with my AMD 7850 and 7950 as well, guess those two cards were special also.
 
It varies a few FPS depending on processor and processor clock speed. But I think it gives pretty consistent results using the same system. I would not expect anything varying more than 1 frame, or even half a frame on that benchmark.

For the 600-series maybe what you're experiencing is due to throttling of the cards above 70C. If you can keep a stable boost clock, then it will give consistent results.

My Gigabyte WF 670 cards don't throttle because they never exceed 60c using Heaven.
 
Back
Top