Fallout 4 CPU scaling

[MajinCry]

How did you get these numbers?

Boris' Fallout 4 ENB thingy. Doesn't do any graphical shiz yet, but it allows you to see the number of draw calls being made.

 

[Face2Face]

diamond city in not brutal at all at cpu.That scene is far more demanding than diamond city.
So far only skylake was able more than 60fps in that scene.

Alright, so are we talking 1080p ultra, including god rays in this seen?

 

[Head1985]

Resolution dont matter.Just be sure you are not GPU limited.You can lower resolution to 720P but keep max details.
With my GTX 970 1500/8000 in 1920x1080 all max i am at 85% GPU usage with 6700k 4.5Ghz and DDR4 3000Mhz.

 

[Face2Face]

Resolution dont matter.Just be sure you are not GPU limited.You can lower resolution to 720P but keep max details.
With my GTX 970 1500/8000 in 1920x1080 all max i am at 85% GPU usage with 6700k 4.5Ghz and DDR4 3000Mhz.

Alright cool. I'll give it a try tonight. I have a 5820K upgrade in the planning stages, so this will be a major factor in my final decision. I plan on playing this game for a long time.

 

[Head1985]

well i dont have data from haswell-e yet.I hope someone with haswell-e can test it.AMD drivers have bad cpu overhead so they have very bad results other results are with Nv GPU.

Results so far:
NV cards
6700k 4.5Ghz 3000DDR4 cl14 70fps
6700k 3Ghz 3000DDR4 cl14 54fps
Xeon 6core sandy bridge 4.3ghz quad channel DDR3 1600cl9 52fps
4670k 4,7Ghz 2400DDR3 cl11 52fps
2600K 4.5Ghz 1600DDR3 cl9 49fps
3570k 4.2Ghz 1600DDR3 cl9 43fps

AMD cards
4930K 4.4Ghz quad channel 2400DDR3 cl11 + r9 290 33.3Fps
I7 4820K 4,2Ghz quad channel DDR3 1866 + r9 290 28fps
AMD FX 8370E 4,5GHz 1600DDR3 cl11 + r9 290 22fps

 

[Erenhardt]

well i dont have data from haswell-e yet.I hope someone with haswell-e can test it.AMD drivers have bad cpu overhead so they have very bad results other results are with Nv GPU.

Results so far:
NV cards
6700k 4.5Ghz 3000DDR4 cl14 70fps
6700k 3Ghz 3000DDR4 cl14 54fps
Xeon 6core sandy bridge 4.3ghz quad channel DDR3 1600cl9 52fps
4670k 4,7Ghz 2400DDR3 cl11 52fps
2600K 4.5Ghz 1600DDR3 cl9 49fps
3570k 4.2Ghz 1600DDR3 cl9 43fps

AMD cards
4930K 4.4Ghz quad channel 2400DDR3 cl11 + r9 290 33.3Fps
I7 4820K 4,2Ghz quad channel DDR3 1866 + r9 290 28fps
AMD FX 8370E 4,5GHz 1600DDR3 cl11 + r9 290 22fps

It is not driver overhead and your own number prove it. Quite the opposite, it seems nvidia gets better boost with faster CPU.

Your conclusions neeed to be backed up by the data you post. Doing contrary is IMHO the wrong way to do it.

 

[Head1985]

Its cpu overhead
CPU overhead=GPU drivers cant utilize full potential of cpu=low fps in cpu bottleneck scenes.

 

[ShintaiDK]

It is not driver overhead and your own number prove it. Quite the opposite, it seems nvidia gets better boost with faster CPU.

Your conclusions neeed to be backed up by the data you post. Doing contrary is IMHO the wrong way to do it.

All his data supports a CPU bottleneck.

 

[Erenhardt]

All his data supports a CPU bottleneck.

FX is a beast then, only Haswell has only 30% better IPC. Go get them FX CPU while they last.

diamond city in not brutal at all at cpu.That scene is far more demanding than diamond city.
So far only skylake was able more than 60fps in that scene.

This scene is not demanding on the CPU at all. I get far more %CPU utilization doing 180' turn. CPU temps get higher aswell.

Its cpu overhead
CPU overhead=GPU drivers cant utilize full potential of cpu=low fps in cpu bottleneck scenes.

what?

 

[Spjut]

Digitalfoundry has also tested some different CPUs in their test
http://www.eurogamer.net/articles/digitalfoundry-2015-the-best-pc-hardware-for-fallout-4-4023

 

[SPBHM]

diamond city in not brutal at all at cpu.That scene is far more demanding than diamond city.
So far only skylake was able more than 60fps in that scene.

yes, the game have worse parts than Diamond City, that's the problem you really have to play the game a lot to get some relevant data, or you get the gamegpu type of results.

 

[ShintaiDK]

FX is a beast then, only Haswell has only 30% better IPC. Go get them FX CPU while they last.

No its not, the FX is a turd in this game. As expected being such a weak CPU series.

 

[Erenhardt]

4930K 4.4Ghz quad channel 2400DDR3 cl11 + r9 290 33.3Fps
I7 4820K 4,2Ghz quad channel DDR3 1866 + r9 290 28fps
AMD FX 8370E 4,5GHz 1600DDR3 cl11 + r9 290 22fps

All his data supports a CPU bottleneck.

FX is a beast then, only Haswell has only 30% better IPC. Go get them FX CPU while they last.

No its not, the FX is a turd in this game. As expected being such a weak CPU series.

Say what?! Then so is haswell and ivy. I guess those two with dual channel 1600mhz memory would be slower than his fx.:thumbsup:

 

[Head1985]

So much chatting and no results:\ why no one put here some results from my save?
Just load save and take screenshot

Remember Default FOV(80) and max detail.Resolution dont matter(just dont be GPU bottleneck if you have slow GPU)

 

[ShintaiDK]

Say what?! Then so is haswell and ivy. I guess those two with dual channel 1600mhz memory would be slower than his fx.:thumbsup:

Yes, super FX performance.

 

[Erenhardt]

Their results are always way off from all test done. They hardly even test. They show what they THINK the results should look like.
Also, their test was made with nv gpu if you didn't noticed. "Geforce GTX980Ti" in polish translates into Geforce GTX980Ti in english.
We have user data that proves fx performs almost as good as ivy and haswell at the same clocks speeds and core counts. Most of the advantage for intel comes from faster memory.
Try harder.

 

[ShintaiDK]

Their results are always way off from all test done. They hardly even test. They show what they THINK the results should look like.
Also, their test was made with nv gpu if you didn't noticed. "Geforce GTX980Ti" in polish translates into Geforce GTX980Ti in english.
We have user data that proves fx performs almost as good as ivy and haswell at the same clocks speeds and core counts. Most of the advantage for intel comes from faster memory.
Try harder.

What does GTX980TI have to do with anything? The FX is crap and continues to be crap. You just want to move the bottleneck somewhere else.

And they dont even use a high demanding benchmark setup.

 

[Erenhardt]

What does GTX980TI have to do with anything?

I don't know, you started posting GTX980ti CPU scaling results when we were talking amd driver overhead. Desperation or what?

 

[ShintaiDK]

I don't know, you started posting GTX980ti CPU scaling results when we were talking amd driver overhead. Desperation or what?

I think you need to check the results again. Your Red Team membership clouds your judgement again. FX also tends to perform better with NVidia drivers due to the multithreading. Its not for fun AMDs own GPU division dont want to touch FX CPUs.

 

[frozentundra123456]

Say what?! Then so is haswell and ivy. I guess those two with dual channel 1600mhz memory would be slower than his fx.:thumbsup:

Those are some very "creative" interpretations. You cant compare FX with fast memory to intel with slow memory.

With the same memory, 4770k is consistently faster in the very benchmark you show, and with faster memory, the gap widens.

With 1333 memory, an i3 is even faster than the FX in average fps. Edit: actually i3 is faster than FX with either memory.

 

[Head1985]

So no one here have fallout4 but everyone discus it or why you dont input results can i ask?
like rally its best to test it and we all can see and compare results in very CPU bottleneck scene.

but instead you all just put here one or another graphs:\
Or its too hard load save game take screenshot and put here results?

 

[Erenhardt]

I think you need to check the results again. Your Red Team membership clouds your judgement again. FX also tends to perform better with NVidia drivers due to the multithreading. Its not for fun AMDs own GPU division dont want to touch FX CPUs.

The numbers are against you. Clearly faster CPU help nvidia more than amd i latest games:
AMD gpu scaling with CPU: from 2,5Ghz to 4,5Ghz is close to none:

not much of CPU bottleneck here. Otherwise 4.4 Ghz (grey) would be much faster than 2,5 Ghz (green)

On the other hand, nv:

15% difference between those two CPUs.

Same in fallout:

FuryX is faster with 2.5 GHz 4C/4T haswell than with 3.5Ghz (40% faster) 2C/4T haswell.
Contrary to nvidia, which is having problems with 2.5GHz and needs 40% more IPC on 2C/T haswell to offset the core deficit.

Looking from different perspective, Fury X scales 8% with increasing CPU frequency by 40% (2.5Ghz 4C/4T vs 3.5Ghz 4C/4T), while nvidia by 13%.

AMD is not dependent on CPU frequency as much as nvidia, ergo, nvidia has more driver overhead.

 

[Erenhardt]

So no one here have fallout4 but everyone discus it or why you dont input results can i ask?
like rally its best to test it and we all can see and compare results in very CPU bottleneck scene.

but instead you all just put here one or another graphs:\
Or its too hard load save game take screenshot and put here results?

If you move the mouse a bit to the side, fps increases by a lot. There is no difference on screen with the amount of object etc. Benchmarking a buged code is pointless.

game is running at 43-44 FPS with this SS

look a little down and it gets back up to 60

look up with a whole lot more visual stuff and back up to 60

 

[ShintaiDK]

The numbers are against you. Clearly faster CPU help nvidia more than amd i latest games:

You completely missed the performance delta in your images. As I said, you lost any objective view.

The FX8370 goes from 61.7 with Fury X to 75.2 with GTX980TI. And you use a game that scales well in cores. Not to mention looking on the FPS, they could have found much more CPU demanding locations.

 

[ShintaiDK]

So no one here have fallout4 but everyone discus it or why you dont input results can i ask?
like rally its best to test it and we all can see and compare results in very CPU bottleneck scene.

but instead you all just put here one or another graphs:\
Or its too hard load save game take screenshot and put here results?

You already got Skylake results and you asked for Haswell-E