MSI 290X Lightning

Jasseh · 16 Apr 2014 at 18:31

Hi 8 Pack,

I've just installed my replacement 290X (after the beta drivers killed my other last week) and figured rather than risk damage to the replacement card I would ask for your recommended settings for attempting to overclock without trying to go too high.

I will be sticking with the official drivers this time (13.12 installed) and MSI VGA Fan Control is showing all 3 fans are running correctly.

Without me changing any settings it looks like CCC has applied a 6.4% GPU overclock. I'm going to set this to 0 and await your advice. I really don't want what happened with the last one happening again

.

Thanks in advance,
Jason

Jasseh · 16 Apr 2014 at 19:23

Forget overclocking... Just got black screen'd so I've taken the 290X back out as a precaution (I don't want another one dying).

I opened up GPU-Z, CCC and MSI VGA Fan Control to monitor fan speeds and temps then ran 3DMark Fire Strike.

The test passed fine and scored 10100 but then I noticed the VRM temps being reported were strange.
GPU-Z reported lowest -148xxxxx and highest 393xxx with random numbers showing between.
MSI VGA Fan Control PWM temp was changing every second to numbers like 0, 39, 42, 143, 252.

I was about to post that information here to see if it was just reporting badly but then the black screen happened. The black screens were the first thing to go wrong with the previous card so I rebooted into the bios to make sure there were no artefacts showing in there then powered the PC off and switched back to my 7850 again.

I'm assuming something else in my PC is the problem rather than the card, could it be the power supply?

(This was running at stock with current non-beta drivers)

Avenged7Fold · 16 Apr 2014 at 23:03

PSU should be fine. Black screens are mostly down to memory unstable. Crazy temperature readings like that seem normal to me for most benchmark programs i use, so i really only trust GPU-Z temps. If the GPU-Z temps are fine, likely there is no problem with the cards sensors.

Clocking on these things are quite easy. I use trixx or MSI AB and start by slapping the Powertune to +50% off the bat, since i am pretty confident in my cards cooling. A lightning would easily take max power tune as well.

After that i bump the core up to 1050 or 1100 off the bat and test stability and temps in the games i usually play or using a looping gaming benchmark. If temps are reasonable and so is stability, keep bumping the core by 5-10 until you get driver crashes, artifacts or the temps seem scary. Driver crashes and artifacts are usually caused by unstable core, so putting the core voltage up deals with that.

Depending on cooling you can go up to +200mv on Trixx, which is plenty. Go slowly so you get a good feel for what clocks are stable at what voltages. Once you find core voltage and clock you are happy with, you can go onto memory. There is no dedicated memory voltage on my card (don't know if lightning has the option) but i found memory stability heavily linked with core voltage and it is for this reason why i suggest you find the right memory clock after the core clock and voltage is found. Black screens usually mean unstable memory and require you to drop it down some clocks.

For memory clocking, you can start off with fairly big jumps, not really any danger with setting the memory clock too high, it just black screens if it is too much. Since its a lightning, i would go for 1500 straight off with +50% power and then try 1600 if it works. After that increments of 15-25 until you get instability, then its time to dial it down.

Good luck with the GPU lottery!

Jasseh · 16 Apr 2014 at 23:18

Avenged7Fold said:
PSU should be fine. Black screens are mostly down to memory unstable. Crazy temperature readings like that seem normal to me for most benchmark programs i use, so i really only trust GPU-Z temps. If the GPU-Z temps are fine, likely there is no problem with the cards sensors.

Clocking on these things are quite easy. I use trixx or MSI AB and start by slapping the Powertune to +50% off the bat, since i am pretty confident in my cards cooling. A lightning would easily take max power tune as well.

After that i bump the core up to 1050 or 1100 off the bat and test stability and temps in the games i usually play or using a looping gaming benchmark. If temps are reasonable and so is stability, keep bumping the core by 5-10 until you get driver crashes, artifacts or the temps seem scary. Driver crashes and artifacts are usually caused by unstable core, so putting the core voltage up deals with that.

Depending on cooling you can go up to +200mv on Trixx, which is plenty. Go slowly so you get a good feel for what clocks are stable at what voltages. Once you find core voltage and clock you are happy with, you can go onto memory. There is no dedicated memory voltage on my card (don't know if lightning has the option) but i found memory stability heavily linked with core voltage and it is for this reason why i suggest you find the right memory clock after the core clock and voltage is found. Black screens usually mean unstable memory and require you to drop it down some clocks.

For memory clocking, you can start off with fairly big jumps, not really any danger with setting the memory clock too high, it just black screens if it is too much. Since its a lightning, i would go for 1500 straight off with +50% power and then try 1600 if it works. After that increments of 15-25 until you get instability, then its time to dial it down.

Good luck with the GPU lottery!

I've already decided to ask for a refund on the card, thanks anyway though. Before I RMA'd the card I bought last week it started out with just black screening but then quickly ended up being dead. After posting on the forums I ended up believing that was down to the 14.3 beta drivers causing problems with fans 1 and 3.

This time I haven't even tried overclocking it and I stuck with the official driver... Even if the black screen problem is just from unstable memory, this is at stock. It should be stable at stock, especially for a card that is supposed to be capable of so much more.

I'm not willing to try messing around with settings for this card and risk it ending up like the one I already RMA'd. Instead I will go for the refund and look for a different card to buy.

Oh and yes, the 290X Lightning has dedicated voltage settings for GPU, VRAM and AUX.
Temps for GPU in GPU-Z were fine, temps for VRAM were random (and clearly unrealistic) numbers.