Can someone help in trying to figure out if I am severely bottlenecking or if something else is going on that causes me to not get expected results?

[superfantastic]

I bought a rtx 3070, replaced my gtx 1070. My cpu is an i7 9800x. I ran a bunch of benchmarks to compare. 3dmark shows that the rtx 3070 is more than twice as good which got me very excited. The Heaven benchmark also showed expected results, though a bit lower than the average.

I then undervolted my 3070 a bit, I chose 950 and 1920mhz. Ran the heaven benchmark again and actually got a higher score, higher average fps and much lower temps. So again I was really excited.

But then I tried the Horizon Zero Dawn build-in benchmark.

In the images you can see my gtx 1070 benchmark on original preset, rtx 3070 original preset and rtx 3070 ultimate preset.

Apparently I only gained about 20fps on the same quality settings. There aren't a lot of benchmarks to find on youtube with these presets but what I do find shows much higher fps consistently, above 140fps in most cases.

So... what's going on? Is my cpu too weak?

[dizmo]

Probably not the greatest benchmark to run. Looks like there's probably some scenes that don't take much power to render, so both cards perform well if you're just looking at the highs. You're seeing a 50% improvement in the 95%. Not sure having VSync on would change anything, but that's a variable that isn't the same between the two tests you've run.

[superfantastic]

26 minutes ago, dizmo said:

Probably not the greatest benchmark to run. Looks like there's probably some scenes that don't take much power to render, so both cards perform well if you're just looking at the highs. You're seeing a 50% improvement in the 95%. Not sure having VSync on would change anything, but that's a variable that isn't the same between the two tests you've run.

Maybe you are right. I suppose that in any case the game will be running more smoothly especially if I cap the framerate. But it doesn't explain why the benchmarks I find online to compare with get so much higher average fps with the same card. That's why I am suspecting a cpu bottleneck but perhaps I don't have enough knowledge to deduce such a thing

[dizmo]

5 minutes ago, superfantastic said:

Maybe you are right. I suppose that in any case the game will be running more smoothly especially if I cap the framerate. But it doesn't explain why the benchmarks I find online to compare with get so much higher average fps with the same card. That's why I am suspecting a cpu bottleneck but perhaps I don't have enough knowledge to deduce such a thing

Could be, but you'll always have a bottleneck somewhere in your system.

Have you OC'd your CPU?

Did they OC their card or CPU?

RAM speeds different?

Different drivers?
Lots of things can change performance.

[Guest]

What would be the IPC difference from a 9800X to a 12400F?

 

I'd hate to say a 12400F is significantly faster and I'm using this low end cpu for comparison because, well you can't overclock it. 

 

So single core boost from your current chip to a modern chip is probably around 20% performance difference.

 

How would that effect this 1 single in game benchmark? Have no idea honestly. Would probably want to see a few different benchmark results for best comparisons.

[jaslion]

Check your cpu per core usage whilst gaming. Horizon zero dawn is seriously cpu hungry.

[jaslion]

1 minute ago, Guest 5150 said:

What would be the IPC difference from a 9800X to a 12400F?

 

I'd hate to say a 12400F is significantly faster and I'm using this low end cpu for comparison because, well you can't overclock it. 

 

So single core boost from your current chip to a modern chip is probably around 20% performance difference.

 

How would that effect this 1 single in game benchmark? Have no idea honestly. Would probably want to see a few different benchmark results for best comparisons.

A 12400f has about a 25% ipc lead over a 9800x. Horizon zero dawn is a very cpu hungry game we noticed it when playing it on the living room pc before it was switched to a 3600x coming from a 2600k at 4ghz.. The 3600x was a really noticeable boost even on a rx590 at the time.

[merco]

Well it's interesting that it shows you CPU fps and GPU fps separately, I would imagine if the CPU could keep up with the GPU you would have the 140+ you were seeing in other reviews or benchmarks.

I don't quite understand how it works, but if I understand it correctly the numbers seem to show you what bottleneck exists.

With the 1070 both CPU and GPU have similar performance, with the 3070 you get CPU 111 and GPU 149, but the total goes down to about 5 less than CPU fps in every case.

 

Here is a vague reddit thread that is loosely related https://www.reddit.com/r/horizon/comments/i7ckud/i_might_found_the_solution_to_fix_performance_of/

It sounds weird, I don't know why running it through 2 launchers would fix it.

 

In any case I wouldn't think too much about it, since it's a game benchmark and the game could have a number of reason why it runs slower than usual. If you find this behavior in other games as well I would maybe update all drivers and such.

 

 

[superfantastic]

40 minutes ago, Guest 5150 said:

What would be the IPC difference from a 9800X to a 12400F?

 

I'd hate to say a 12400F is significantly faster and I'm using this low end cpu for comparison because, well you can't overclock it. 

 

So single core boost from your current chip to a modern chip is probably around 20% performance difference.

 

How would that effect this 1 single in game benchmark? Have no idea honestly. Would probably want to see a few different benchmark results for best comparisons.

It seems you are right. While playing Days Gone I kept an eye on my fps and then looked at some footage online to compare. And I actually get better fps than some 3070 videos with the same in-game settings. However my cpu is getting way too hot so I had to limit the fps... But that's a whole 'nother problem.

[jones177]

I use HZD as a bench.

It really does not bottleneck at 1080p on a balanced system.

Here is what it looks like with some of my GPUs.

First with an i9 9900k.

 

HZD at Ultra                                                  1080p      1440p        4k.

ASUS Strix 3080                                           134fps      118fps       80fps

EVGA XC3 Ultra 3080 ti                                147fps      128fps      79fps*     

EVGA FTW3 Ultra 3080 ti                             155fps      138fps      88fps
 

The 1080p scores would be closer if there was a bottle neck.

 

It is the same with a 5900x with 2 different GPUs

The game does not use the extra cores or higher IPC of the 5900x.

 

HZD at Ultra                                                  1080p      1440p        4k.

EVGA FTW3 Ultra 3080 ti                             156fps      138fps      88fps

EVGA FTW3 Ultra 3090 ti                             168fps      148fps      98fps

 

With your setup I  would do 1440p or 3440 X 1440 and stay away from 1080p. 

  

 * The power limit of the card stops it from being good at 4k.

[superfantastic]

1 hour ago, jones177 said:

I use HZD as a bench.

It really does not bottleneck at 1080p on a balanced system.

Here is what it looks like with some of my GPUs.

First with an i9 9900k.

 

HZD at Ultra                                                  1080p      1440p        4k.

ASUS Strix 3080                                           134fps      118fps       80fps

EVGA XC3 Ultra 3080 ti                                147fps      128fps      79fps*     

EVGA FTW3 Ultra 3080 ti                             155fps      138fps      88fps
 

The 1080p scores would be closer if there was a bottle neck.

 

It is the same with a 5900x with 2 different GPUs

The game does not use the extra cores or higher IPC of the 5900x.

 

HZD at Ultra                                                  1080p      1440p        4k.

EVGA FTW3 Ultra 3080 ti                             156fps      138fps      88fps

EVGA FTW3 Ultra 3090 ti                             168fps      148fps      98fps

 

With your setup I  would do 1440p or 3440 X 1440 and stay away from 1080p. 

  

 * The power limit of the card stops it from being good at 4k.

it's just strange to me that someone else with a 3070 gets 15fps more on average than me on the same settings. It doesn't say if he overclocked it or not, but he does have a much better cpu. That's why I was thinking of a bottleneck

[Guest]

1 hour ago, superfantastic said:

It seems you are right. While playing Days Gone I kept an eye on my fps and then looked at some footage online to compare. And I actually get better fps than some 3070 videos with the same in-game settings. However my cpu is getting way too hot so I had to limit the fps... But that's a whole 'nother problem.

Your 9800X is about as fast as my 8700K but with more cores/threads. 

To compare these 2 processors to a 5600X, the 5600X is a bit faster than both of our processors. 

 

So it boils down to, you have good numbers for your setup. Looks ok to me.

choices, keep what you have and accept the FPS you do get....

Or upgrade your system. 

 

My 2060 can't shake a stick at your 3070. So you've got me bested and there's nothing I can directly compare other than straight cpu benchmarks only.

 

 

 

[jones177]

1 hour ago, superfantastic said:

it's just strange to me that someone else with a 3070 gets 15fps more on average than me on the same settings. It doesn't say if he overclocked it or not, but he does have a much better cpu. That's why I was thinking of a bottleneck

 

Run Time Spy to see if your CPU is feeding your GPU well.

 

I do not have a 3070 but I do have a 2080 ti that produces about the same frames. I have used it with Intel 6,8 and 10 core CPUs. 

My i9 9900k should be close to your i7.

 

This is the i9 9900k at stock. https://www.3dmark.com/3dm/63092267?

This is a 5ghz Intel 6 core with the same GPU. https://www.3dmark.com/3dm/51588161?

This is a stock 5800x(8 cores) with the same GPU. https://www.3dmark.com/3dm/62386371?

 

If your CPU score it closer to the i7 6 core you will under perform compared to other modern Intel 8 cores.

If your CPU score is close to the i9 8 core then your CPU is fine and it can handle 3080 ti power without issue.

 

Note that the CPUs really did not get more out of the GPU. Not even my i9 10900k could.

 

Also there can be a vast difference between GPUs in the same class. My EVGA XC3 Ultra 3080 ti produces about 10 frames less than my EVGA FTW3 Ultra 3080 ti. Its performance is closer to a Strix 3080.   

I did not get that sort of gap with the 10 and 20 series.