Time to flash another vendor vbios with proper VRAM voltage....

BTW, you can alleviate the issue by turning the fan to max in order to cool down the vram.

Can you upload the vbios dump here? Open GPUZ, BIOS version ---> click the arrow---> save to file.

I wanna check if the asus vbios flasher actually update your gpu vbios or not.

EDIT:
Please check using msi afterburner for memory voltage, is it still 1.6V?
MSI afterburner ---> settings ---> tick "unlock voltage control" and "unlock voltage monitoring".
If it still 1.6V (1600mV), then you have failed to update the vbios.

This post has been edited by JohnLai: Jul 17 2015, 01:31 PM

........so, your model has the latest vbios which reduces the VRAM clock to 1.5V installed properly....

and yet....the flickering still occurs....... According to you, reducing the VRAM clock somehow alleviate it....

We can't rule out the VRAM as source of flickering yet.

Hmm....Wanna try the opposite method?
Increase the VRAM voltage slightly to 1.52 or 1.55v.
Set the VRAM clock back to 1500Mhz.

My reasoning, 280x got 12 vram chips around the board. Perhaps the memory controller doesn't output 1.5V precisely for 12 VRAM chip.
Undervoltage also can cause flickering issue.

EDIT: Write too fast until need to correct so many times......

This post has been edited by JohnLai: Jul 17 2015, 07:40 PM

So....after you test with faster fan speed, does the flickering/artifacts still occur?

If it didn't occur, then it might be subpar hynix VRAM even though it is running at correct 1500mhz and 1.5V.

You can create a custom aggressive fan curve for faster fan speed.


MSI Afterburner --> Settings -->Fan -->Tick "Enable user defined software automatic fan control".
You should know how to set the curve . Quite intuitive interface.

edit : Once you click OK, remember to HIGHLIGHT "User Define" near the Fan Speed (%)




Wait a minute.....does vram slight increase in voltage to 1.52 or 1.55v fix the artifacting at 1500mhz?
Or because of the max fan speed?
Which one work for your gpu? Both?

I am getting confused here...

This post has been edited by JohnLai: Jul 18 2015, 01:05 AM

Good thing you use witcher 3 to test. This game is really good in stressing GPU component especially in city of Novigrad at ultra settings.
1519mV for VRAM should be okay (never got the precise output anyway)
If there isn't anymore artifact after 1519mV and fan speed of 80% applied......then you got lemon Hynix VRAM which is too subpar (prone to instability and overheating even with the rated specification setting being applied).

Since you have been playing with fan speed at high rpm, why not apply a bit of overclock for your gpu core clock as well?
A little guidance ;
GPU core clock set to 1170Mhz (1225, 1188 or even lower 1160, for gpu core clock = the higher, the better) with gpu core voltage set to 1300mV (or 1265, for voltage, the lower the better).
Power Target set to 120% (aka +20%)
I only suggest this because you have been setting a high gpu fan speed at 80%.

The only downside of high fan speed = dust problem. Remember to keep fan and heatpipe fins clean.

Do more test with witcher 3. (Technically, for stressing purpose, unigine or furmark tends to be used)

EDIT:
Remember to report back if the gpu core OC is stable and which value you use.
Maintain the vram clock at 1500mhz and vram voltage at 1520mV (1519). Fan speed at 80%.
Stress the GPU to 100% usage, check for any artifacting/crashing/blue screen and report back the maximum temperature.

If the GPU OC is stable, I got final testing for you to do.
By the way, you didn't tell us your power supply unit model?

This post has been edited by JohnLai: Jul 18 2015, 10:43 AM

Actually, he said he already update to latest Vbios which reduces vram clock from 1600mhz to 1500mhz and vram voltage from 1.6V to 1.5V, but the flickering still happen.
That why I propose him to increase the VRAM voltage slightly. Of course, between 1550mV and 1520mV, of course 1520mV is a better choice.

The gpu core overclock is another matter........

Good PSU, can rule out unstable pcie voltage supply.

Well..... the tweaking is up to you.

In case if you RMA your GPU, remember to double check the VBIOS version once you got a new one.
Later you got the batch with 1.6V + 1600Mhz vram = sure the vram dies sooner.
So, update accordingly before heavy gaming.

Eh,so how ar? You said you will test the gpu for more than 3 hours, still got artifacts?
EDIT: If only set the VRAM 1519mV with fan speed auto, still artifacting? Or fan speed 80% is needed?

This post has been edited by JohnLai: Jul 19 2015, 10:41 AM

Might as well as increase it to 1600Mhz and see if the artifacting still there. (Fan 70%, shouldn't have any issue with temperature)
There were reports on VRAM becoming unstable at low clock speed. Just dont change the VRAM voltage. Maintain it at 1519mV.
Some VRAM chips are weird in a sense the chip becomes more stable at higher clock than lower clock. Generally, the lower the speed, the more stable.


-.- flicker when browsing sites? What kind of flicker? Horizontal small lines flickering? Checker board pattern flickering? Large vertical line flickering?

This post has been edited by JohnLai: Jul 24 2015, 01:17 AM

Okay.....this is probably 2D application VRAM downclocking issue.

Hmm......not sure if msi afterburner got this functionality, I only know ASUS GPU Tweak can set the minimum vram clock speed for low GPU load (2D load).

Need to set minimum 2D VRAM clock of 300mhz or 600mhz (preferably 300mhz)
Cause by default 2D VRAM clock speed is 150mhz (some said this frequency is too low for the vram chip to function correctly or the memory controller can't response fast enough to supply voltage for the operation)

When the gpu memory controller steps up vram speed from 150mhz to 600mhz or even 1500mhz under load, the flickering occurs.
You could use asus gputweak, it has similar controls as msi afterburner.

The setting you have done in msi afterburner only applies for 3D heavy application (under heavy graphic load).

 GPUTweak_Manual_EN.PDF ( 2.41mb ) Number of downloads: 15


From the gputweak manual, page 18 and 19. Need to change to advanced mode for the 2D/3D option to be available.




This post has been edited by JohnLai: Jul 24 2015, 01:44 AM

????
The 2D VRAM clock is 600mhz by default?

I propose you to RMA the gpu as soon as possible.
Your GPU is practically lemon goods

With 2D vram clock at 600mhz, the flickering horizontal issue should not be happening in first place.
In other word = you got seriously faulty GPU. Not to mention your 3D clock issue.

So...you rather live with the faulty gpu than getting a properly functional gpu?

Logically, if you spend RM900+ for gpu, better get the most out of it even if it means paying extra for RMA.

I guess Fiji really 'overclocked' to maximum for its current 'stock speed'.
No wonder AMD locks down the voltage to hide it.


http://www.techpowerup.com/reviews/AMD/R9_..._X_Overvoltage/

Perhaps battlefield 3 is more optimized/stable/mature game than battlefield 4?

Or probably he wanna reduce cpu impact on gpu? Who know.......

Maybe battlefield 3 won't hit VRAM limit at UHD resolution than B.4?.....otherwise, more than 4GB = texture swapping time......

Gameworks impact? Such as clothes cpu based physx?

Well.....AMD did claim it is an 'overclocking' dream.....not 'performance' dream.

Example, Bulldozer and Piledriver CPU are good overclocker.....can generally overclocked around 4.3ghz to 4.8ghz easily, but the performance isn't there.

Same with Fury X.....

Last I heard, AMD was supposed to clock Fury X at 800Mhz and sold at initial price of USD800.

Appearance of GTX 980TI at USD650 changes the game.

AMD decided to hold the launch and clocked Fiji core higher to 1050Mhz and revised the launch price to USD649.
Thus, those who bought Fury X at USD650 have to thank Nvidia for the cheaper price.

Then again, if Fiji really launched at its original clock of 800Mhz, it would have reasonable performance/watt (just check at fiji undervolting result) with overclocking capability up to around 1100Mhz? (1100Mhz - 800Mhz = 300Mhz, so.....yeah...overclocking dream indeed, too bad this plan doesnt pan out).

Hmm....it is either Nvidia GTX970 3.5GB or AMD R9 390.

Dont get AMD 380. It is an old design without any support for better Powertune/ZeroCore and Variable Refresh Rate support.

R9 390 = Better idle power management than 380 (use more power than gtx970), 8gb of VRAM (sure can put a lot of gta V and battlefield texture cache), support for Freesync.
GTX 970 = overclocking beast with 3.5gb vram, support for Gsync. 8bit HEVC support in hybrid DXVA.

In case if you plan to purchase monitor with variable refresh rate in future, you need to make the choice for either freesync or gsync video card now.
I can see you are the type that purchase an expensive goods for long term usage. So, choose which variable refresh rate(VRR) video card to match with your future VRR capable monitor.

Opps, my mistake. I confused 370 with 380. Thanks for pointing it out.
380 is technically advanced than 390, however, 380 just doesn't have the performance of 390 even with all those improvements >.<

AMD GPU really has the lowest idle power than competitors by 2-4watts as mentioned by @Unseen83

Yes, Nvidia GPU supports audio passthrough etc using HDMI.

In case you buy nvidia gpu = remember to set the color range 16-235 or 0-255 according to your display type (TV or Monitor) at the Nvidia Control Panel.

Note:Try to spend more for GPU with "Semi-passive" fan + heatpipe design

Example, when idling or temperature below 65'C, Asus GTX970 fans won't spin (less dust cleaning issue, longer fan durability?).

Heard the new ASUS 390 also has the semi-passive design as well.

Maybe he bought a second hand gpu without knowing it.
The first buyer probably flashes the vbios to unlock 290 to 290x.

Instabilities? Those parts are disabled for a good reason......


http://www.overclock.net/t/1443242/the-r9-...x-unlock-thread

Read all pages if you wanna know what kinds of instabilities that might appears........

If you bought the gpu from eBay, Alibaba and etc......there is high probability the card was used for bitcoin/litecoin/anykindofcoin mining activity.

Short story = gpu used for mining virtual coin tends to die quickly.

Labuan? Now everything also taxed with GST. Some electronics even got double-taxed.

The only way I can think......is to buy the product online, declare it as 'other form of goods ' and hope the goods won't arouse custom officer suspicion........

This post has been edited by JohnLai: Aug 3 2015, 11:20 PM