Lower than normal Valley Score

[Archangel1994]

So I got my GTX 970 and was thrilled to have it. It looks awesome, does awesome, etc. So I thought I would look up scores that other got using a GTX 970 and doing Valley benchmark 1.0. The first one I tried was 1920x1080p ultra graphics without AA. The reviews/others all showed about 4300 points but mine scored 3400. Then I tried the one JayztwoCents did, Extreme HD preset 1080p full screen. He got 2700 on 1520Mhz. I got 2290 on 1500Mhz...

 

Why are my scores significantly lower than others? I asked before whether there would be a bottleneck and everybody said no which was evident from the fact that my GPU was at 99% usage most of the time. Specs are in my sig. 

[SeanBond]

I definitely wouldn't be worried as long as you are getting the right numbers in actual games and other benchmarks. Couldn't imagine what would be the cause though.

[Enderman]

Do you have any background programs running?
Try running the benchmark while in safe mode.

[nj4ck]

probably because you're on a sandy bridge i5 at stock clock and most of the testers use socket 2011 i7s in their test rigs. 

 

its not like you have a bottleneck, but valley does stress the CPU more than many games do.

[Archangel1994]

probably because you're on a sandy bridge i5 at stock clock and most of the testers use socket 2011 i7s in their test rigs. 

I thought Valley was supposed to be a GPU stress test, not a CPU stress test.

[Aniallation]

I thought Valley was supposed to be a GPU stress test, not a CPU stress test.

It will still benefit from more CPU power

[SeanBond]

I thought Valley was supposed to be a GPU stress test, not a CPU stress test.

It is, but if your gpu was not running a constant 99% you probably got slowed a bit by your cpu. 

[Archangel1994]

It is, but if your gpu was not running a constant 99% you probably got slowed a bit by your cpu. 

With the exception of 1 or 2 scenes the GPU was almost constantly at 99%

[mvitkun]

I thought Valley was supposed to be a GPU stress test, not a CPU stress test.

It is, but it still benefits from a faster CPU like anything else would.

 

That said you could try raising the voltage on your 970 (not power target/limit) a bit to see if it does anything for your scores.

[Nico123473]

Not all GPU are the same, there might be some performance differences between cards, not all overclock the same for example

[SeanBond]

With the exception of 1 or 2 scenes the GPU was almost constantly at 99%

That is weird, do you have any games you can benchmark with? Metro, Tomb raider, Bioschock ect?

[Archangel1994]

Not all GPU are the same, there might be some performance differences between cards, not all overclock the same for example

Not the same is fine but about 20% score difference is too much of a difference really.

[nj4ck]

I thought Valley was supposed to be a GPU stress test, not a CPU stress test.

well, it is, but it still uses CPU, just like any other 3D application out there. When running ultra HD preset, the four cores on my 4670k @ 4.4GHz get up to around 60% constant, your CPU at stock clock may or may not be struggling with it, you'd have to check that yourself. I also noticed you're running 1333MHz RAM, not sure if that makes a difference (I don't think it does), but there's another possibility.

[MountainDrew]

I got a score of 2656 overclocked by a bit on my Gigabyte G1 GTX 970 on Unigine Valley (Extreme HD Preset).

 

Unigine benchmark results3.htm

 

I don't know why yours is so low.  Could be the lower memory frequency, or that I have an i7?

[Archangel1994]

I got a score of 2656 overclocked by a bit on my Gigabyte G1 GTX 970 on Unigine Valley (Extreme HD Preset).

 

Unigine benchmark results3.htm

 

I don't know why yours is so low.  Could be the lower memory frequency, or that I have an i7?

What's your core and memory frequency?

[Bogica]

Ignore benchmarks and do real world performance, see a review and set up a game with the exact same settings as the reviewer did and if the FPS is close to that in the review, you are good to go.

[MountainDrew]

What's your core and memory frequency?

My 4770k is clocked at 4.2 Ghz and my memory has always been at 1600mhz.

[Archangel1994]

My 4770k is clocked at 4.2 Ghz and my memory has always been at 1600mhz.

I meant your GPU

[MountainDrew]

I meant your GPU

Oh Sorry  :rolleyes:

 

[Archangel1994]

Oh Sorry  :rolleyes:

 

GPU Z GTX 970 OC.gif

Hmmm, your GPU memory is clocked higher than mine. Will re-bench with higher mem clock speed

[KamiKatze]

I have a 970 in the mail, ask me next week when it arrives. I'll follow this topic. Check my sig, if that doesn't perform differently, there must be a problem with those reference benchmarks.

Edit: nope, can't follow the thread. Somehow.

[Archangel1994]

Oh Sorry  :rolleyes:

 

GPU Z GTX 970 OC.gif

SO I just redid it with a memory offset of +200 (Effective +400 MHz on the memory) and the score went up to 2370. Still a bit too low. I should try to find someone with an I7 and high speed memory (1600/1866)

[MountainDrew]

Hmmm, your GPU memory is clocked higher than mine. Will re-bench with higher mem clock speed

Yeah, I went pretty aggressive on the memory since valley seems to like higher memory clocks.

 

If I go too high on the GPU clock I get artifacts and valley closes down (1550mhz and up causes this).

 

My temps are fine though.

[Archangel1994]

Yeah, I went pretty aggressive on the memory since valley seems to like higher memory clocks.

 

If I go too high on the GPU clock I get artifacts and valley closes down (1550mhz and up causes this).

 

My temps are fine though.

You know I really just noticed that mine is running PCIe 2.0 x16 (Knew that) and yours is running PCIe 3.0 x16. That cannot possibly be it right? PCIe 2.0 x16 is PCIe 3.0 x8 which should be more than enough

[Archangel1994]

Yeah, I went pretty aggressive on the memory since valley seems to like higher memory clocks.

 

If I go too high on the GPU clock I get artifacts and valley closes down (1550mhz and up causes this).

 

My temps are fine though.

Most interesting, did a pretty aggressive OC of +400 offset on the memory clock and now the score is 2425. Still though not quite as close as I want it to be to what Jay got or you (I want a minimum of 2600 really). Could be limited by the I5. Planning a broadwell upgrade when it comes out, that might help.