Why is AMD quiet about black screens on 290x?

Orangey · 13 Jan 2014 at 10:02

Mem ocing really helps with mining though.

LtMatt · 13 Jan 2014 at 10:03

Orangey said:
Mem ocing really helps with mining though.

Well i can't argue with that as i never tried it.

Rezident · 13 Jan 2014 at 12:52

old gamer said:
Let me know how you get on, i crash on the level after homecoming, i think its legend of something or other

Finished it with no other crashes, just the one blue screen I think. No black screens yet in any games, Bioshock Infinite and dlc has been rock solid.

old gamer · 13 Jan 2014 at 13:15

rezident what drivers are you on

K1LaW8tt5 · 13 Jan 2014 at 13:17

Marine-RX179 said:
The issue is however we don't really have direct access to voltage for the memory, and the biggest contributor to black screen is unstable memory clock.

I mean I bought one of the VTX unlocked 290x from OcUK, and the memory clock top out at overclock at 5600MHz max stable and any higher I'd would get black screen. I paid extra £60 without the free BF4 games (there were reference 290 with free BF4 at the time for just £299.99, making the reference 290 essentially only cost £280) for the extra shader cores/steam processor from the unlock. I mean it is a great card, however, a standard 290 with memory clock to 6000MHz+ would easily match my unlocked 290x which the memory can only clock to 5600MHz max, so in a way I don't know what I paid the extra £60-£80 for...

In Gibbo's 290 overclocking thread, he said with unlocked voltage, both Elpdia and Hynix memory will hit 6000MHz+ without problem, but from my personal experience, adjusting "core" and "aux" voltage don't seem to jacks for helping me to clock the memory clock higher. If I knew I'd be getting duff memory with my unlocked 290x, I'd probably have held off and save up and wait for a MSI 290x Lightning instead with some "proper" memory.

I'm getting mix feeling of happy with the performance of the card and disappointed at the same time

You paid a little extra for a 290 that was garanteed to unlock to a290x for less than the price of a 290x. Gibbos results are best case and he probably didnt test bf4 much which is where the memory fails most. My 290x will eventually blackscreen at anything over 1350(5400) on the memory yet i paid the full price of a 290x. You seem to have a huge sense of entitlement here.

Kaapstad · 13 Jan 2014 at 13:21

weldon855 said:
You paid a little extra for a 290 that was garanteed to unlock to a290x for less than the price of a 290x. Gibbos results are best case and he probably didnt test bf4 much which is where the memory fails most. My 290x will eventually blackscreen at anything over 1350(5400) on the memory yet i paid the full price of a 290x. You seem to have a huge sense of entitlement here.

Why does anyone want to or need to overclock the memory for gaming anyway, the gains are minimal. I am quite happy to use one card and no overclocking for gaming.

Silent_Scone · 13 Jan 2014 at 13:27

As above, not just on these but across the board GPU overclocking for performance gained in game is minimal. If you're at a stage where you're needing to OC your GPU in order to be able to maintain performance then you might as well upgrade or go multi-card. There is also nothing to state as far as I can tell that Gibbos clocks were playable clocks either.

Diagro · 13 Jan 2014 at 13:31

my 290 thats unlocked to an X wont do more than 1325 on the Memory. But TBH with two in Xfire i get 60fps in ANY game @ 2560x1600 so i dont see an issue.

I will sell them when the next gen cards are out anyway so for me meh who cares

Silent_Scone · 13 Jan 2014 at 13:33

No retailer can guarantee clock speeds regardless if they're flashing them or not. You can't soak test something like that as it's too subjective to things such as ambient temps or other components. Plus the memory on the R290 seems to be fairly weak across the board even best case when you consider the NV counterparts can overclock substantially.

K1LaW8tt5 · 13 Jan 2014 at 13:48

Kaapstad said:
Why does anyone want to or need to overclock the memory for gaming anyway, the gains are minimal. I am quite happy to use one card and no overclocking for gaming.

I agree, hence why I would be happy if I paid for a 290 and got a 290x regardless of how high the memory will overclock.

Gregster · 13 Jan 2014 at 13:49

So now there is custom cards which also suffer the black screen issue. Has AMD said anything about this?

jrodga2k5 · 13 Jan 2014 at 13:51

Gregster said:
So now there is custom cards which also suffer the black screen issue. Has AMD said anything about this?

Well luckily for me I haven't experienced the Black Screen issue so I'll just blissfully continue onwards on my fantastic experience of the card

LtMatt · 13 Jan 2014 at 13:53

I'll run some quick tests to see how much difference going from 1250 - 1550+ actually makes. I'm guessing no more than a couple of fps.

iFrag · 13 Jan 2014 at 13:54

Gregster said:
So now there is custom cards which also suffer the black screen issue. Has AMD said anything about this?

Bailey told me that I need to wait for the new 14.1 drivers before I RMA my card, must admit though, upping the core voltage has helped.

Baboonanza · 13 Jan 2014 at 13:54

Kaapstad said:
Why does anyone want to or need to overclock the memory for gaming anyway, the gains are minimal. I am quite happy to use one card and no overclocking for gaming.

It's actually highly game dependant. When I was overclocking the VRAM on my 290X I saw little gain in most things but comparatively large gains in the Sleeping Dogs benchmark (I couldn't say if they were linear or not, I didn't do the maths). It just depends where to bottleneck is and Sleeping Dogs seems to do something that other games don't. That game also really ups my VRM temps and I've noticed much higher than normal temps on the IceQ 7950 I used before too.

LtMatt · 13 Jan 2014 at 14:01

Baboonanza said:
It's actually highly game dependant. When I was overclocking the VRAM on my 290X I saw little gain in most things but comparatively large gains in the Sleeping Dogs benchmark (I couldn't say if they were linear or not, I didn't do the maths). It just depends where to bottleneck is and Sleeping Dogs seems to do something that other games don't. That game also really ups my VRM temps and I've noticed much higher than normal temps on the IceQ 7950 I used before too.

Yes Sleeping Dogs is the elephant in the room regarding this theory of ours. Overclocking the memory can make a difference in Sleeping Dogs because of the SuperSampling. It really saturates the bandwidth.

I've documented this previously. In fact when Kaap did some memory clocking testing i tried (but failed) to get him to ditch Tomb Raider in his 290x Vs Titan memory clocking tests and use Sleeping Dogs because it actually benefits greatly from a memory overclock.

LtMatt said:
Kaap you'd be much better off usng Sleeping Dogs for these tests. I recommend you re-run them using that. Memory makes no difference in Tomb Raider. Just not enough to saturate the bandwidth. Sleeping dogs on the other hand with Extreme...

I noticed a 10% performance improvement on my 7970 going from 1375mhz-1875mhz.

LtMatt said:

I haven't but seeing as you asked i just did a quick test. I only ran one game but will run a couple more later when i have time.

Sleeping Dogs - 1080p Extreme 1250 Core vs 1375 stock and 1875 OC'd.

I speculate the difference would be larger at a higher res (1440p+) although we might not see it as the pixel fill rate might hold it back somewhat.

Click to expand...

Source
http://forums.overclockers.co.uk/showpost.php?p=24133681&postcount=19

Silent_Scone · 13 Jan 2014 at 14:08

Tomb Raider also has Super Sampling doesn't it? If anything it shows that SD is poorly optimised.

LtMatt · 13 Jan 2014 at 14:14

MjFrosty said:
Tomb Raider also has Super Sampling doesn't it?

It does but its no where near as intensive as Sleeping Dogs.

As we did with HDAO, however, we take AA one step further in Sleeping Dogs. The “Extreme” anti-aliasing setting uses the compute horsepower of Graphics Core Next to do another anti-aliasing pass on the final frame, which will smooth out those last four pixels of aliasing we described in the example above. The resources required to drive the extreme setting are quite intense.

When all is said and done though, Sleeping Dogs’ extreme preset offers the highest possible anti-aliasing quality available to a graphics card.

Source
http://reader.mreotech.com/sleeping-dogs-gaming-evolved-and-you/

Silent_Scone · 13 Jan 2014 at 14:23

Can't find what the preset for medium and high actually translate to in terms of 'how much' SSAA the game is applying. If extreme is 4X that's the same as Tomb Raiders preset which is down sampling to 4k.

So it should put equal strain, and the differences between this and Sleeping Dogs should in theory deem similar results between the two cards, just with more improvements on this particular engine (Unreal 3.0 tech isn't it?) over Tomb Raider as Sleeping Dogs isn't the best optimised game really.

Wait, here it is within your link

As we did with HDAO, however, we take AA one step further in Sleeping Dogs. The “Extreme” anti-aliasing setting uses the compute horsepower of Graphics Core Next to do another anti-aliasing pass on the final frame, which will smooth out those last four pixels of aliasing we described in the example above. The resources required to drive the extreme setting are quite intense, so users of HD 7800 and HD 7700 Series GPUs might try the “high” preset (2.25x SSAA, no post AA) or the “normal” preset (post AA only).

So 'high' is 2.25x SSAA. Extreme doing it's final pass is equal to what then? Surely it's just 4X SSAA.

LtMatt · 13 Jan 2014 at 14:37

Tomb Raider gains from memory overclocking are minimal vs Sleeping Dogs.