trying to understand video card overclocking

[primetime]

I have noticed several times getting a higher score in FIRE STRIKE 1.1 using actually lower gpu clocks.....i have checked the log files with gpuz for valley and heaven benchmarks and the clocks don't appear to be changing or throttling...but its a little harder to tell with 3dmark since it changes screens and benchmarks and of course the clocks will throttle down like there supposed to in between the different tests...

why does it score lower with higher clocks even if gpuz log shows no throttling or is it something else all together that brings the score down?....truthfully im just trying to get the most i can from what i have is all...not like im lagging or anything or complaining.....just trying to understand it completely....and truth be told my 7970 is the weakest link in my system so im just trying to get the most from it ...without damaging it of course...for whats its worth i just put new past on and temps are staying at or below 70 for the most part now...so gpu temps should be ok....the voltage regulators on the other hand don't even have heat sinks on them....if i'm seeing them correctly...i guess the vrms are just overheating and making it very difficult to find the right overclock

 

[psyclist]

The memory has built in ECC now, so as you clock over and above it maximum, you are doing more harm than good. Core introduces artifacts and texture screw-ups that slow things down as well. Bring things down until you reach your max score and then test that in BF4 or something equally as stressing

 

[primetime]

will do bud...thanks....i was kinda thinking it might have been related to the voltage regulators being over heated....but what i did was keep running tests increasing till it dropped in score then backed off...will do a Thorough test in bf4 now to double check stability and look for any kind of artifacting

 

[Nenu]

As stated by psyclist it will be down to corruption being detected and calculations/copying being repeated, which holds up further processes.
When you slow the card down, the corruption reduces so there is less duplication of calculations/copying and less holding up of processing.

Corruption can be caused by many things, not just overheating vrms.

Run GPU-z and MSI Afterburner to monitor the card.

 

[primetime]

As stated by psyclist it will be down to corruption being detected and calculations/copying being repeated, which holds up further processes.
When you slow the card down, the corruption reduces so there is less duplication of calculations/copying and less holding up of processing.

Corruption can be caused by many things, not just overheating vrms.

Run GPU-z and MSI Afterburner to monitor the card.

that i am doing already...is there something i should be looking closely at? the clocks/mem seem to stay the same..so can i assume its not throttling?i am also recording the the logs with gpuz to look over for any throttling,( or something out of the ordinary) but as of yet have seen none,,,,what else should i be looking closely at?

Is there any benefit to raising memory speed much if at all...so far i have not seen any improvements doing so, at least not yet....thanks for the responses...my last cards was nvidia geforece 280...so things have changed quite a bit since then...and truthfully those puppy's didn't overclock at all lol at least not enough to make a hill of beans

keep the info coming guys thanks...as i said the card works perfect...just trying to get what ever it has to give since in bf4 it uses 100% gpu and about 20 % cpu...might as well get all she got's before i say fuck it and splurge on a 290x

 

[primetime]

you were right being 3dmark , valley, heaven stable means squat....bf4 had to drop clocks right off get go to 1070...may get it to go a bit higher...well see

 

[Nenu]

that i am doing already...is there something i should be looking closely at? the clocks/mem seem to stay the same..so can i assume its not throttling?i am also recording the the logs with gpuz to look over for any throttling,( or something out of the ordinary) but as of yet have seen none,,,,what else should i be looking closely at?

Monitor GPU and vrm temps.

Is there any benefit to raising memory speed much if at all...so far i have not seen any improvements doing so, at least not yet....thanks for the responses...my last cards was nvidia geforece 280...so things have changed quite a bit since then...and truthfully those puppy's didn't overclock at all lol at least not enough to make a hill of beans

If you dont see any benefit then no.

 

[primetime]

Monitor GPU and vrm temps.

i have no way to Monitor vrm temps...its a non standard 7970 card....if i had to do it all over...i would skip non standard cards....cause you know they burning hot with no heat sinks

 

[Vegas P11]

Don't Fermi and Kepler cards throttle when Heaven, Valley are detected? I remember reading that somewhere in a review and it was something nVidia did to prevent frying cards.

 

[primetime]

well the card in question is tahiti what im leaning is it might pass 100 times of 3dmark but fail first time in bf4 on bad seting

 

[aphasia]

i have no way to Monitor vrm temps...its a non standard 7970 card....if i had to do it all over...i would skip non standard cards....cause you know they burning hot with no heat sinks

have you tried HWiNFO64. this grabs all the relevant GPU monitoring data (core temp, VRM temps, clocks, voltages, utilisation, etc)

*sorry about pic size, linked direct from HWiNFO site.

worth a try even tho your card isn't a reference model.

quick note, not a good idea to run a overclock tools like 'MSI AfterBurner' with other monitoring tools like AB. AMD cards don't like having their senors polled at the same time by different monitoring programs. set the clocks in AB and close, then open HWiNFO to monitor. i know with my 7970, it can cause the screen to flicker or in some cases cause the out-put to completely corrupt the screen requiring a reboot.

 

[primetime]

have you tried HWiNFO64. this grabs all the relevant GPU monitoring data (core temp, VRM temps, clocks, voltages, utilisation, etc)

been using it for quite some time with afterburner/riva tuner server for awesome osd ability....but mine wont see vrm temps....i should probably purchase some low profile heat sinks and attach them to the vrms to be on the safe side...i think it would still clear underneath the stock heat sink

 

[aphasia]

been using it for quite some time with afterburner/riva tuner server for awesome osd ability....but mine wont see vrm temps....i should probably purchase some low profile heat sinks and attach them to the vrms to be on the safe side...i think it would still clear underneath the stock heat sink

damn that sucks, you would think a non-reference board would at least give you the basic sensor probes.

there is a possible alternative, use temp diode probes (a couple of bucks each at most). your motherboard would of course need to support the 2 pin leads. if not there are some after-market stuff that you can use to monitor your GPU VRMs...

Huhushop(TM) PC USB Powered Thermometer Temperature Sensor Data Log
http://www.amazon.com/Huhushop-TM-Powered-Thermometer-Temperature/dp/B0054U4YKI

FrozenCPU Temperature Probe Cable - 2-Pin - 25"
http://www.frozencpu.com/products/1..._-_2-Pin_-_25.html?tl=g12c133s492&id=9wRcTqrZ


Phobya Illuminated Thermometer with Digital Display - Blue Blacklight (71013)
http://www.frozencpu.com/products/2...cklight_71013.html?tl=g12c133s492&id=9wRcTqrZ

 

[misterbobby]

Don't Fermi and Kepler cards throttle when Heaven, Valley are detected? I remember reading that somewhere in a review and it was something nVidia did to prevent frying cards.

That is not true.

 

[psyclist]

Yeah if you can fit an aftermarket heatsink on your VRM's that would be a good idea. They are good up to a fairly high temperature though. I run mine up to 90, but I do believe they will run to 125 max Obviously not a good idea, but they are fairly tough. Have a look around the net to see what others have done with your specific model. I know Geild makes one for the 290X that might fit your card as well. Most cards are a standard width PCB.

Ive got some high performance thermal pads coming, made by fujipoly and sold by FrozenCPU. Trying to bring my VRM temps down as well so I can clock the card up higher! Overclocking is fun, a bit of trial and error, but you will eventually find your cards sweet spot.

 

[primetime]

Yeah if you can fit an aftermarket heatsink on your VRM's that would be a good idea. They are good up to a fairly high temperature though. I run mine up to 90, but I do believe they will run to 125 max Obviously not a good idea, but they are fairly tough. Have a look around the net to see what others have done with your specific model. I know Geild makes one for the 290X that might fit your card as well. Most cards are a standard width PCB.

Ive got some high performance thermal pads coming, made by fujipoly and sold by FrozenCPU. Trying to bring my VRM temps down as well so I can clock the card up higher! Overclocking is fun, a bit of trial and error, but you will eventually find your cards sweet spot.

i did finally get occt working correctly by ding an image restore....what kind of clocks are you getting?

[/URL][/IMG]

 

[primetime]

as seen here it can throttle : 878 gpu

[/URL][/IMG]

haven't saw it in games yet though

 

[psyclist]

They built into the 290X to detect when programs like this are used, Furmark, OCCT etc...They are labelled as power virus's and draw way more power and generate more heat than any game ive come across ever. So the cards automatically throttle when it senses you loading these up...I dont use em for stress testing.

I was at 1200/6400 @ 137mv 50% power target when it was cooler weather, now that summer temps have arrived ive downclocked to 1150/6200 at +100mv and 30% powertarget. I have ordered a G10 bracket and have a spare Antec 620 AIO cooler and VRM heatsink that im going to mount I think. Then should really be able to push the card.