DDR3 1600 non k i7 bottleneck

[skyrimsucks]

i'm wondering how bad (if any) is bottleneck of 1600 mhz cl10 ddr3 ram (single channel 8 gb) for i7 4770 (3.7 ghz on turbo). i have seen some fallout 4 benchmarks where non k i3 with 2400 mhz ram had like 20% ADVANTAGE over 4790k on 1333 mhz memory. i just cant belive the ram bottleneck is so bad for intel cpus considering intel recommended 1600 mhz ram for haswell line...

 

if any1 has 4th gen i7 with 2400 mhz ddr3 i'd like to know how much performance is lost when ram is set to 1600 (when cpu is at stock speed).

[Guest]

I can do such test for you later today.

 

I7 4790K 4.0G

CORSAIR 1866MHz

 

I can slow down the memory at 1600MHz and tell you the result.

 

[skyrimsucks]

I can do such test for you later today.

 

I7 4790K 4.0G

CORSAIR 1866MHz

 

I can slow down the memory at 1600MHz and tell you the result.

thx, i'd greatly appreciate. i know bethesda coding is garbage but such a humongous ram bottleneck would mean that intel fucked up and their cpus cannot be utilised properly without 2400+ mhz ram under certain loads.

[Guest]

OK, starting the test.

i7 4970 4GHz. Ram clocked at 1866MHz. GTX 980 TI 1.1GHz. Full screen 1960x1080.

I will load a save game, doing nothing for one minute, then spawning 10 deatclaws using the console, waiting another minute.

I will watch the CPU and the GPU charges (I can track them for minutes), and the FPS (but only Min/Max/Average.)

 

No mods will be used, all DLCs will be included.

 

Results of the first test (Mem at 1866MHz):

* CPU Min 11% average 38%, max 67% (only on Fallout4.exe, excluding all the other tasks)

* GPU Min 6%, average 11% peak 61% (sadly I cannot exclude other tasks here, but there was pretty much nothing running except the tracking tool)

* FPS steady 60:  min 59 max 60 average 60

 

 

Resetting my PC, and limiting the Ram speed in the EFI screen.

See you in a short while for the second test.

 

[Guest]

OK, results of the second test:

 

i7 4970 4GHz. Ram clocked at 1599MHz. GTX 980 TI 1.1GHz. Full screen 1960x1080.

Loaded a save game (same as previous test, was in a heavy built Sanctuary), nothing for one minute, spawn 10 deathclaws using the console, wait another minute.

No mods will be used, all DLCs will be included.

 

Results of the second test (RAM at 1599MHz):

* CPU Min 8% average 39%, max 65% (only on Fallout4.exe, excluding all the other tasks)

* GPU Min 7%, average 11%, peak 59%

* FPS Min 56, Max 60, Average 59. The FPS drop was for a fraction of second just after I spawned the deathclaws and all settles started bombarding them.

 

 

At the end, I don't think RAM at 1600MHz is a bottleneck, at least not on my architecture.

 

[skyrimsucks]

thx, its quite interesting.

i guess considering fallout engine do not work properly above 60 its irrelvant to make more use of cpu anyway. the large differences may have to do with drivers and other factors. or they simply fudged data to sell high end ram...

 

http://www.gamespot.com/forums/pc-mac-linux-society-1000004/fallout-4-ridiculusly-high-performance-gained-by-f-32754964/

https://www.reddit.com/r/pcmasterrace/comments/3subz6/ram_speeds_becoming_increasingly_more_important/

 

i will try to disable 2 cores of my i7 to emulate i3 and compare ram 1600 to 1333 from curiosity.

[skyrimsucks]

i couldnt get any visible results (crowded sanctuary with exploding sentry bot) even when i disabled 2 cores and reduced ram speed to 800. i guess whatever impact ram speed have its even more diminished in weaker gpu rigs(r7 250x). good to know since my mobo do not support anything above 1600 and i dont plan on getting top tier gpu.

 

i've read that fallout tend to stutter if not used from ssd (or very fast hdd). constant hard drive access would put extra strain on ram which could explain why ram speed would be more relevant than in most games/programs.

 

http://www.eurogamer.net/articles/digitalfoundry-2015-faster-hard-drives-boost-xbox-one-fallout-4-performance

[Lendova]

With a 250X, you have to have massive GPU limit all the time. 

[skyrimsucks]

With a 250X, you have to have massive GPU limit all the time. 

and still shitty gamebryo cant even use my gpu at 100%. max gpu load i had is 97%. unless i'm looking on sentry bot on fire or something like that (when i dip to 20 fps) i got around 40-45 fps most of the time on 1920x1200 res.

[Lendova]

 

With a 250X, you have to have massive GPU limit all the time. 

and still shitty gamebryo cant even use my gpu at 100%. max gpu load i had is 97%. unless i'm looking on sentry bot on fire or something like that (when i dip to 20 fps) i got around 40-45 fps most of the time on 1920x1200 res.

 

What is your GPU load when the system is not busy loading textures from HDD/memory? 

I ask this, because my much stronger GPU with much weaker CPU is maxed (98-99%) almost all the time. 

[skyrimsucks]

 

 

With a 250X, you have to have massive GPU limit all the time. 

and still shitty gamebryo cant even use my gpu at 100%. max gpu load i had is 97%. unless i'm looking on sentry bot on fire or something like that (when i dip to 20 fps) i got around 40-45 fps most of the time on 1920x1200 res.

 

What is your GPU load when the system is not busy loading textures from HDD/memory? 

I ask this, because my much stronger GPU with much weaker CPU is maxed (98-99%) almost all the time. 

 

 

as far as i know, shitbryo "background streaming technology" is suppose to make busy hdd/ram and presumably also vram all the time. so this game seems not as much as computing as bandwith depending. during loading screen my gpu is around 47% then it fluctuates around 90-98 but most of the time its around 97%. my gpu has weak floating point performance (geometry processing bottleneck) but perhaps i could push shading or some other things up. i will increase shadow distance and see if there is any difference.

[skyrimsucks]

yea i got stable 99% after i turned god rays, antialiasing, maxed shadows and other bullshit. i will revert to earlier config to keep max framerate my 1.216 teraflop card can do. i guess it would be nice if afterburner included software that measures which card components are stressed...