3DMark Sky Diver Benchmark Thread

[Geeman1979]

Spare a thought for the poor Titan owners, at least you can see the bench running.

You do if you use the 335.23 driver. Also no SLI support! for me anyhow.

 

[Kaapstad]

You do if you use the 335.23 driver. Also no SLI support! for me anyhow.

Same here using 337.88 drivers.

 

[humbug]

It's a nice looking bench. Firestrike is rather bland in comparison i think.

I though i killed my GPU yesterday on this, it was running soooo well at 1200Mhz, i though i would give it a shot at 1250, again it was going well, until it got half way through the combined test, artefacts, flicker and Black Screen....
Ok so, hard reset, back into windows, within seconds flicking and Black Screen on the Desktop... oh no, hard reset... same thing over and over again

Boot into safe mode, unistall all overclocking apps, no, didn't fix it

Back into safe mode again to unistall the drivers.
Running fine without drivers, ok. reinstall drivers. All is back to normal. Phew

I think what happened is the drivers monitoring the GPU didn't like what was going on and set the volts to 0mv during the bench.

This is the first time the Drivers have intervened like this on this GPU, but not the first time ever as it did it to me on the 7870XT when i ran that really hard during benching, it did the same thing in setting the volts to 0.

The trouble is because this one has a factory overclock it has +50mv on it out of the box, with 0mv like that it black screens on the desktop.

I don't know how i feel about the drivers behaving like big brother monitoring my GPU and interfering like that, they may well have saved its life, or just meddling.

Either way, i #### bricks.

 

[Kaapstad]

I though i killed my GPU yesterday on this, it was running soooo well at 1200Mhz, i though i would give it a shot at 1250, again it was going well, until it got half way through the combined test, artefacts, flicker and Black Screen....
Ok so, hard reset, back into windows, within seconds flicking and Black Screen on the Desktop... oh no, hard reset... same thing over and over again

Boot into safe mode, unistall all overclocking apps, no, didn't fix it

Back into safe mode again to unistall the drivers.
Running fine without drivers, ok. reinstall drivers. All is back to normal. Phew

I think what happened is the drivers monitoring the GPU didn't like what was going on and set the volts to 0mv during the bench.

This is the first time the Drivers have intervened like this on this GPU, but not the first time ever as it did it to me on the 7870XT when i ran that really hard during benching, it did the same think in setting the volts to 0.

the trouble is because this one has a factory overclock it has +50mv on it out of the box, with 0mv like that it black screens on the desktop.

I don't know how i feel about the drivers behaving like big brother monitoring my GPU and interfering like that, they may well have saved its life or just meddling.

Either way, i #### bricks.

Drivers for both NVidia and AMD are going to be a bit quirky for the next few weeks until they bring them up to date for this bench.

 

[setter]

Hopefully then we'll see a bit more performance. Changing from 337.50 betas to the latest whql's netted some nice gains for me.

 

[humbug]

Ill see about droping my ram speed down from 2400 to 1600, cant see it making an awful difference though.

Actually could you? that would be interesting.

 

[setter]

Heres the results from running ram at a lower speed. Same cpu clock and gpu at 1189/3305.

1600mhz ram speed
26567

Graphics Score
36014

Physics Score
11888

Combined Score
23824

2400mhz ram speed
27549

Graphics Score
36014

Physics Score
12825

Combined Score
26474

Quite a drop that i wasnt expecting tbh. In games ram clocking makes virtually zero difference as ive ran this kit at 2600mhz.

 

[Uncle Petey]

where's all the 7970's I feel lonely

Hey, I tried

 

[LtMatt]

Quite a drop that i wasnt expecting tbh. In games ram clocking makes virtually zero difference

Sometimes it makes a difference. More so in multi gpu i expect though.

2700k 4.4ghz
1600mhz 9-9-9-27-2T - stock seconds
290 crossfire @975/1250 stock

Spoiler

2700k 4.4ghz
2133mhz 9-10-10-21-1T - tight seconds
290 crossfire @975/1250 stock

Spoiler

2700k 5.27ghz
2133mhz 9-10-10-21-1T - tight seconds - BCLK 102.5
290 Crossfire 975/1250 stock

Spoiler

 

[setter]

Your spoiler tabs no worky mate.

 

[LtMatt]

Your spoiler tabs no worky mate.

Cheers, corrected.

 

[setter]

Nice un mate, as your running xfire at a higher res it will help i suspect, and 16gb too iirc. I fancy adding another 8gb but finding c10 vulcan reds is extremely hard now.

 

[LtMatt]

Nice un mate, as your running xfire at a higher res it will help i suspect, and 16gb too iirc. I fancy adding another 8gb but finding c10 vulcan reds is extremely hard now.

That was at 1080P but yeah. I wonder if it would help more at lower res because of the cpu bottleneck tbh.

 

[Kaapstad]

Sometimes it makes a difference. More so in multi gpu i expect though.

Not normally, in fact for games if you go too high on the RAM it actually reduces fps slightly. I think this could be where the CPU has to drive the RAM very fast (even though it's not needed) instead of driving the game.

I have tried several games with different RAM speeds and there is not much in it from 1600mhz up to 2400mhz, 2133mhz seemed to give the best performance but it was barely measurable.

If you look at setters results it is very similar, if you think of the Graphics score as similar to games you can see no difference running the memory @1600mhz or 2400mhz.

On the other hand if the CPU has to do some number crunching as in the Physics and combined tests then the extra RAM speed helps.

 

[LtMatt]

Not normally, in fact for games if you go too high on the RAM it actually reduces fps slightly. I think this could be where the CPU has to drive the RAM very fast (even though it's not needed) instead of driving the game.

I have tried several games with different RAM speeds and there is not much in it from 1600mhz up to 2400mhz, 2133mhz seemed to give the best performance but it was barely measurable.

If you look at setters results it is very similar, if you think of the Graphics score as similar to games you can see no difference running the memory @1600mhz or 2400mhz.

On the other hand if the CPU has to do some number crunching as in the Physics and combined tests then the extra RAM speed helps.

I should test some more games. I might do that if i can be bothered.

 

[setter]

For ram, c9 @2133mhz seems to be the sweetspot on ivybridge/sandybridge. On haswell, c10 2400mhz. Cant comment on sbe as ive never used those. Though i believe the new Z97 boards are much better for ram clocking than Z87. I wont be bothering with one though as 4790k runs in Z87.

@Matt, ahh 1080, assumed you were running 2560x1440.

 

[whyscotty]

I should test some more games. I might do that if i can be bothered.

Lazy git

 

[LtMatt]

Lazy git

It's a lot of work.

 

[Kaapstad]

I should test some more games. I might do that if i can be bothered.

What would be interesting is what difference RAM speed makes when connecting to a busy server, I suspect it could make quite a difference.

Now all we need is a real BF4 pro to test this.

 

[humbug]

Heres the results from running ram at a lower speed. Same cpu clock and gpu at 1189/3305.

1600mhz ram speed
26567

Graphics Score
36014

Physics Score
11888

Combined Score
23824

2400mhz ram speed
27549

Graphics Score
36014

Physics Score
12825

Combined Score
26474

Quite a drop that i wasnt expecting tbh. In games ram clocking makes virtually zero difference as ive ran this kit at 2600mhz.

1600mhz ram speed
Physics Score
11888

2400mhz ram speed
Physics Score
12825 (+8%)

Thats about what i thought, thanks setter