Strange Bottleneck?

[spaniel53]

I currently have an i5-4670K (stock clock---3.4/3.8GHz, 4c/4t), GTX 1070 (Asus ROG Strix OC), on Intel's DH87RL motherboard with 16GB of DDR3 1600 (4x4GB). I run my games at 1440p on my 1440p,144Hz Dell S2417DG. I did a full reinstall of windows 10 last month. Thermals are relatively low, the 1070 in games hits ~60C, the CPU stays around 50C.

EDIT: CPU usage in Unigine Valley is 50-60% and in PU: Battlegrounds is 85-100%.

 

PCPartPicker: https://pcpartpicker.com/user/spaniel53/saved/KCv3CJ

 

I recently suspected some bottle-necking when comparing my fps in certain games as compared with my friends (who have the same graphics card--GTX 1070). To test this, I ran Unigine Valley at a number of settings on my PC, had a friend do the same(He has Zotac's GTX 1070 AMP! Edition), then took my card over to his place and had him run them again on his PC with my card replacing his.

 

His PC: i5-6600 (4c/4t, 3.5/3.9GHz), Zotac's GTX 1070 AMP! Edition, 16GB of DDR4 2133 (2x8GB), with the MSI B150I Mobo (mini-ITX).

 

The test results listed are for Valley 1.0, DX11,1080fullscreen, ultra, 0xAA, Extreme Tess.

 

 

(And yes, I did run my benchmark at 1080, not 1440. I unplugged my 1440 monitor and only had my 1080 monitor connected to run the benchmark.)

 

The evidence seems to point to a CPU bottleneck, but my friend's CPU seems VERY close in performance to mine (4c/4t, 3.5/3.9GHz). Could it be a motherboard issue or even a RAM speed issue? Seems very unlikely as well...

 

I would appreciate any insight into this strange situation. I'll be quick to respond to any questions.

 

Thanks.

Edited May 17, 2017 by spaniel53

[SImoHayha]

21 minutes ago, spaniel53 said:

I

 

Well your friend has the better system it's clearly well there in your face. It's not a bottleneck issue at all, he's just got the next generation of power both the latest Intel CPU along with DDR4.

 

It would be comparing a stock 69 Camero to a modern day Camero, just because I add turbo to the 69 one doesn't mean that it will be faster compared to the new modern day Camero with turbo and say it might be a bottleneck! No, it's because the engine's and everything about the Camero is the next generation 

[typographie]

Are you really seeing 80-100% CPU usage in Unigine Valley? I just ran it using an i5-6600K and I get readings in the 40% range per core using your settings and with vsync off. I know I've got a slightly faster CPU but it's a graphics benchmark, it really shouldn't be hitting any CPU that hard. It sounds like you might have something running in the background eating up your CPU's attention.

[spaniel53]

21 minutes ago, Ramamataz said:

Well your friend has the better system it's clearly well there in your face. It's not a bottleneck issue at all, he's just got the next generation of power both the latest Intel CPU along with DDR4.

 

It would be comparing a stock 69 Camero to a modern day Camero, just because I add turbo to the 69 one doesn't mean that it will be faster compared to the new modern day Camero with turbo and say it might be a bottleneck! No, it's because the engine's and everything about the Camero is the next generation 

The clock speed are within 0.1GHz and the core/thread count are identical....two series later of almost the same CPU really has that big of an influence? Could the architecture improvement in IPC really be that big? I thought these last few generations have been mostly power saving and little speed improvements...

 

But I would believe the RAM speed hypothesis...I just thought I heard DDR3--->DDR4 isn't that big of a difference in one of Linus' videos.

 

But note that he actually DOUBLED my average fps...that seems pretty substantial for mostly very specs.

Edited May 17, 2017 by spaniel53

[SImoHayha]

1 minute ago, spaniel53 said:

The clock speed are within 0.1GHz and the core/thread count are identical....two series later of almost the same CPU really has that big of an influence? Could the architecture improvement in IPC really be that big? I thought these last few generations have been mostly power saving and little speed improvements...

 

But I would believe the RAM speed hypothesis...I just thought I heard DDR3--->DDR4 isn't that big of a difference in one of Linus' videos.

 

The smaller something gets the more powerful and efficient they make things.

[spaniel53]

17 minutes ago, typographie said:

Are you really seeing 80-100% CPU usage in Unigine Valley? I just ran it using an i5-6600K and I get readings in the 40% range per core using your settings and with vsync off. I know I've got a slightly faster CPU but it's a graphics benchmark, it really shouldn't be hitting any CPU that hard. It sounds like you might have something running in the background eating up your CPU's attention.

Sorry, I was pretty unclear there. CPU Usage is around 50-60% per core in Unigine Valley.

 

The 80-100% CPU figure was in PU: Battlegrounds.

 

I did my best to close all background programs when I did these benchmarks and checked task manager for any programs eating up resources and there were no significant offenders.

[spaniel53]

5 minutes ago, Ramamataz said:

The smaller something gets the more powerful and efficient they make things.

I still hesitate to believe that would up the AverageFPS from 68 to 113. But I'd love to hear a second opinion on this. You may be right.

[typographie]

6 hours ago, spaniel53 said:

The 80-100% CPU figure was in PU: Battlegrounds.

Unfortunately I don't own that game so I can't compare my own results. I could be wrong about this, but I've heard that game has some optimization issues causing it to sometimes run poorly even on machines it should run well on. Early Access, etc. etc. But that said, I haven't heard that it's unusually demanding of a CPU.

 

What is your GPU usage like in that game? Is it often dropping below 100%?

 

Edit: According to PCGW,

"nVidia optimisation: under manage 3D settings in the NVCP select the specific 3D settings for PUBG and change threaded optimisation from "auto" to "on" as many have reported this has increased their performance considerably." Try that?

 

6 hours ago, spaniel53 said:

I still hesitate to believe that would up the AverageFPS from 68 to 113. But I'd love to hear a second opinion on this. You may be right.

I'd tend to agree with you. The difference from Haswell to Skylake is a 5–10% improvement in performance at a given same clock speed. DDR4 can help a little in some games, but probably by just a few FPS. 

[spaniel53]

10 hours ago, typographie said:

Unfortunately I don't own that game so I can't compare my own results. I could be wrong about this, but I've heard that game has some optimization issues causing it to sometimes run poorly even on machines it should run well on. Early Access, etc. etc. But that said, I haven't heard that it's unusually demanding of a CPU.

 

What is your GPU usage like in that game? Is it often dropping below 100%?

 

Edit: According to PCGW,

"nVidia optimisation: under manage 3D settings in the NVCP select the specific 3D settings for PUBG and change threaded optimisation from "auto" to "on" as many have reported this has increased their performance considerably." Try that?

 

I'd tend to agree with you. The difference from Haswell to Skylake is a 5–10% improvement in performance at a given same clock speed. DDR4 can help a little in some games, but probably by just a few FPS. 

Yes, it is VERY poorly optimized, and probably isn't a good reference point for that reason (oops). But to answer your question my CPU usage is 85-100% and GPU is around 25-40%. Indicating CPU throttle....right?

 

This is after I made the change you suggested.

 

My friend sits at 70% GPU, 80% CPU (the same one with the 1070 and similar CPU) in game.

 

Thanks for your time looking into this with me.