GTX 980 thoughts

Yeah but the graphics score never lies.

Also i scored the same or similar with my 4690k which is a 4 thread processor.

an i7 920 isn't fast by todays standards anyways.

Graphics score is not pure graphics but mix of CPU clock , GPU speed and general memory/cache performance. In most 3DMarks, first tests use mainly 1-4 threads and base on high CPU clock to boost GPU result. In latest 3DM it uses more threads. Physics test base on CPU ( mainly threads, less clock ), memory and cache speed.

If your GPU is CPU limited in a test it's not meant to be, you have a very old system, older than you should anyways.

Eitehr way i find it hard to believe i'm outscoring him by over 3000 points.

Seriously, don't tell me something isn't wrong with his setup, this needs to be looked at.

There can be something wrong, I have nothing to compare right now so I won't tell. For sure drivers are sometimes acting weird and performance is sometimes dropping. I would reinstall chipset drivers and nvidia drivers to 344.16 and check again. 344.48 are giving weird scores. On Win7 it improved my results but on Win8 is slightly worse.

I jumped over 1K from going from 4c/8t to 6c/12t at the same clockspeed in Fire Strike...Its core heavy for the final score. That said, I am not sure how the GPU score reacts to cores and clocks. I would assume it would respond to clocks better than cores if there are more than 4c/4t available, but not sure.

It is worth checking the system, but I think the effect of HT/more cores/clock speed differences are discounted...

your combined score jumped, i doubt your graphics score jumped. I've seen guys with 13,500 or so where as i can only get 12,000 and the difference is they have , say, a 4790k instead of a 4690k like i have, but that's not what i'm referring to. There's NO WAY he should have such a low score unless something else is wrong. I just find it hard to believe right out the box that his graphics score is 3000 less than mine.

The Graphics score will go up slightly with more cores.

Its the nature of some of these 3dmark benches They are heavily core dependent.

Thats one reason i used to support Haven bench so much more then 3dmark.

as the score of 3dmark is heavily heavily dependent on cores - I remember benching my 7970 from 4 cores on a 2500k @ 4.7 ghz vs 8 cores on a 930 @4.0ghz and the 930 @ 4.0 had higher 3dmark11 scores.

In games the 2500k @ 4.7 Had (slightly) better fps overall , and my 3770k Mowed down both cpu's (in 3dmark) with 4c/8t @ 4.8 or 4.9ghz even though the entire time my 7970 was clocked the exact same and in games it wasn't much if any faster then the 2500k was.

I think if you wanna bench the GPU's head to head unigine heaven or the new version is still likely the best way to do that.

Seems like heaven ( or valley ) is the only benchmark which is repetable and base mainly on GPU performance. Other benchmarks like Catzilla are not using much CPU power but are not repetable what means you can see +/- 10% score difference at the same settings ( or even higher ).
All 3Dmarks are using CPU speed to boost 3D scores ( actually maximum FPS, not minimum ) and first tests in most 3DMark versions are single threaded. Later you see that first tests are better on something like even overclocked Pentium G3258 while physics and some other tests at the end are using more threads and perform much better on 6 core+ CPUs that can't make so high clock.

The same results on AMD CPUs in 3DMarks are really low because of low single threaded performance. You can't really see that in games minimum FPS on AMD is much lower but maximum FPS on Intels is always much higher in titles which are using 1-2 threads, even at lower clocked CPUs.

This is more than a statistical difference though. We're talking a huge difference in his combined score vs mine. Even if i downclocked my cpu to like 3 ghz it wouldn't score that low.

His score is similar to an overclocked 680 gtx. That's how low it is.

EVGA GTX-980 SC ACX 2.0 / I7 4790K BENCHMARKS!! (3DMARK11)