XFX R9 280x DD Crossfire Artifact issue

[Megatron]

You're not showing max values in CPU-z and if u click the top left of GPU-z u have additional options, one is to display ASIC. Its there so you might aswell use it. So one card is still a bit hotter than the other as I can see.

You are in crossfire but you say "the video card" as in singular, are you trying to overclock both to the same overclock or are you disabling one?

Also u don't mention wether or not you used a shim which was my recommendation, the temps look good with the G10 attached but either way I'd use a shim to be sure contact is optimal.

As for my recommendations for further overclocking, I don't have the full info and also I've never done a multi gpu setup. Perhaps find your stable clocks for each card separately then take it from there.

On your hotter card the VRM temps are in the 70s but I can see that it was previously slightly hotter from the graph next to the number. Were the VRMs cooler when the G10 was fitted?

Since this thread was initially about heat based instability what made you decide to overclock? Do u need the extra power? I've not seen a screenshot with the G10 attached (of both GPUs). Once I can see that then I could make a more informed suggestion.

I believe your initial glitching was due to memory instability. I'd suggest overclocking core first alone then memory. You might find a hotter core makes the memory glitch, you could test this by underclocking the memory with same GPU clock.

Also why did u remove the G10 bracket? It can only aid your overclocking, and prolong card life.

 

[faizan]

Lol forgot.

I do mean video cards, crossfire is enabled and the option is set in MSI Afterburner to synchronize the clocks.

No I did not use a shim, didn't need to. The die/core is elevated higher than the surrounding parts.

I only have the G10 on one card at the moment, didn't remove it.

The VRM's do get to 80 on 1 card (one on stock cooling), but for the card with G10 bracket, it is doesn't hit 70.

I wanted to overclock from the get go but I was stuck with the artifacts so I wanted to fix that issue first.

Also, I noticed when I ran unigine again, my score dropped to 1476...do you know why there would be a fluctuation like that? I scored 1575 before on the same clocks. 1090/1625. Are the cards downclocking or something??

Spoiler

 

[Megatron]

I don't see any jitter in the GPU clock graph, a reboot may fix it. Your G10 card has 38% for fan speed, is the fan on the G10 bracket connected to the fan header? I posted a link before about how I'm controlling my fans. I replaced the NZXT 92mm fan with a SST-FW91 its got high static pressure good CFM and operates from 1500-2800rpm, with my fan curve my VRMs are 49'c. That is in heaven benchmark with core voltage set to 1250mV. I've got a copy of Sapphire Trixx which seems to overvolt past 1300mV:
http://www.overclock.net/t/1196856/...0-7970-7990-owners-thread/17260#post_18441904

Careful though don't use it in combination with furmark. Also try to change your memory voltage (check in Afterburner), doesn't work for me.

Might be of use:
http://www.overclock.net/t/1265543/the-amd-how-to-thread

Looks like you are doing OK with the overclocking.

Edit:
You say u want another G10, this is where I read about Corsair having shorter hoses:
http://www.overclock.net/t/1487012/official-nzxt-kraken-g10-owners-club/40#post_22213977

Also was looking for some guide on your RAM:
http://www.overclock.net/t/1268061/ocn-ram-addict-club-gallery/5160#post_23840201

Obviously only overclock 1 thing at a time till your sure its stable.

 

[faizan]

I have noticed something else...I don't know since when this has been happening but the non G10 card sometimes just locks to 850MHz..I checked the vram temp, the core temp, its not that much..vram go up to 80c, and core around 70c. And I think that is why I get fluctuations in the scores, because the core sometimes locks onto 850MHz for 1 card and the other one is working at whatever clock speed I have set.

 

[Megatron]

I remember one user said that he had a similar issue and came back to say that it only occured when he didn't have his monitor powered on before his PC. YMMV

 

[Ferrum Master]

The 280X needs a shim!

 

[Megatron]

I've seen a powercolor 280x and an Asus 280x with heatsink off and the Powercolor heatsink had a build in shim whereas the Asus was just flat (a reused cooler I guess). I used the Asus with stock cooling, it seemed to work ok, however when I fitted the G10 I used a shim.

The two GPUs had the same gpu crush protector which to the naked eye looked flush.

 

[faizan]

I found out what it was lol.

You guys could experiment this too, don't know if it'll do the same. I read on the futuremake forum (http://community.futuremark.com/for...3-resets-my-GPU-clock-frequencies-when-opened) about someone complaining about this same issue. Someone on there said that for some reason AMD drivers think the 3dmark is playing back video in the 3dmark menu and so the card just locks the clock as it is just video playback. So I closed 3dmark and voila, problem gone.

My 280X doesn't need a shim. I even ran furmark, the temps have never gone above 56c on the one with kraken installed. Maybe it depends on the revision of the card. Those temps, of 70c and 80c vrm I mentioned, were on the other card. The kraken install is fine, I haven't had issues with that. Sorry for the confusion

 

[Ferrum Master]

The two GPUs had the same gpu crush protector which to the naked eye looked flush.

The shim is needed because of the other components that push back the cooler. So the memory with pad pushes back the cooling monoblock, ASUS does not cool the RAM ic's with the main structure. Some water cooling heads suffer from this fact, so a shim is used.

The question with G10 is... for what kind of torque it is really calculated... I use the shim even on mine GPU, as mine air cooler started to bend and artifacts occurred. Shim is needed for some EK smaller water heads too. Also it adds protection.