Jump to content


Photo

ht vs ram


  • Please log in to reply
5 replies to this topic

#1 Dasa

Dasa

    Guru

  • Super Hero
  • 14,787 posts

Posted 29 November 2015 - 07:08 AM

i was playing again lastnight and thought i would share the results

this is dragon age 3 using the frostbite 3 engine which is one of the few to see a improvement from ht

cpu is 3770k

dragonage_zpsok1sgkeu.jpg


Silverstone TJ08-EW 6700k@4.7 1.375v - Z170-GENE - 2x8g 3866 16-16-16 - 1070@ 2100\9600MHz  -Samsung 830 64G, Sandisk Ultra II 960G, WD Green 3tb - Seasonic XP1050 - Dell U2713 - Pioneer Todoroki 5.1 Apogee Drive II - EK VGA-HF Supreme  - Phobia 200mm Rad - Silverstone AP181 Project Darkling
3770k vs 6700k RAM Scaling, HT vs RAM, Arma III CPU vs RAM, Thief CPU vs RAM
Parts Guide


#2 TheFrunj

TheFrunj

    Guru

  • Atomican
  • 14,277 posts
  • Location:Atomic Labs

Posted 29 November 2015 - 08:52 AM

That's a 6% difference at stock and a 5% difference at 4.2 with hyperthreading on. Interesting that it's that threaded, I wonder if they had developed it with the PS3 Cell chip in mind and later ported it to x86.

 

Memory bandwidth is making an impact too, going from OC 1600 HT to OC 2400 HT is a 13% improvement.


"TheFrunj doesn't guess, when he forms words with his lips, the world will Warp and change to do his bidding." - Shikimaru No longer writing for Atomic :( http://forums.atomic...showtopic=49899

#3 Rybags

Rybags

    Immortal

  • Super Hero
  • 34,378 posts

Posted 29 November 2015 - 09:23 AM

I've been playing with wPrime checking improvements of normal vs HT.

As you'd expect there's a bit of a penalty if you run more threads than you have cores (or virtual cores) especially on a non-HT CPU (e.g. Core2).

On my later Xeon, an 8T vs 4T run gives 263 vs 381 seconds for a 1024K calculation.  Pretty huge difference.

Having the Ram in triple channel helping a bit there as well.

 

As Frunj said - to get full benefit, the program needs to have been written with it in mind.



#4 Dasa

Dasa

    Guru

  • Super Hero
  • 14,787 posts

Posted 29 November 2015 - 09:28 AM

redid the tests trying to be more consistent with when i started fraps recording fps it so that it wasnt still preloading files

 

Dragonage%203770k%20ht%20vs%20ram_zpscgw

 

to benchmark it i loaded a save and ran without moving view from one side of the valley to the other which takes about 14sec as you can see once you get to the end of the valley fps are much higher than at the start where it has more to render

Dragonage%203770k%20ht%20vs%20ram%20char

 

my system is totally gpu bottlenecked with just the 290 in this game at 1080p\2560 i was only getting ~40fps at the start of the test instead of ~60 so i had to drop the res right down to see any difference from the extra cpu speed

detail settings were ultra dx11

unfortunately i cant test mantle with fraps but i can tell you that it makes no difference once the res is up and im gpu bottlenecked at least in this part of the game

 

this review is why i chose to test dragonage

http://www.dsogaming...mance-analysis/

im guessing they mixed up the 3 core 4 core ht on\off in there chart


Edited by Dasa, 29 November 2015 - 09:40 AM.

Silverstone TJ08-EW 6700k@4.7 1.375v - Z170-GENE - 2x8g 3866 16-16-16 - 1070@ 2100\9600MHz  -Samsung 830 64G, Sandisk Ultra II 960G, WD Green 3tb - Seasonic XP1050 - Dell U2713 - Pioneer Todoroki 5.1 Apogee Drive II - EK VGA-HF Supreme  - Phobia 200mm Rad - Silverstone AP181 Project Darkling
3770k vs 6700k RAM Scaling, HT vs RAM, Arma III CPU vs RAM, Thief CPU vs RAM
Parts Guide


#5 Dasa

Dasa

    Guru

  • Super Hero
  • 14,787 posts

Posted 01 December 2015 - 10:51 PM

I've been playing with wPrime checking improvements of normal vs HT.

As you'd expect there's a bit of a penalty if you run more threads than you have cores (or virtual cores) especially on a non-HT CPU (e.g. Core2).

On my later Xeon, an 8T vs 4T run gives 263 vs 381 seconds for a 1024K calculation.  Pretty huge difference.

Having the Ram in triple channel helping a bit there as well.

 

As Frunj said - to get full benefit, the program needs to have been written with it in mind.

That explains why some reviews see such big gains when they disable threads to simulate a quad core if the game is still thinking it has a 8 thread cpu its getting jammed up sending 8 threads to a quad core and is in line with why dual core cpu with ht see such massive gains from ht in games that are made to work with no less than a quad core


Silverstone TJ08-EW 6700k@4.7 1.375v - Z170-GENE - 2x8g 3866 16-16-16 - 1070@ 2100\9600MHz  -Samsung 830 64G, Sandisk Ultra II 960G, WD Green 3tb - Seasonic XP1050 - Dell U2713 - Pioneer Todoroki 5.1 Apogee Drive II - EK VGA-HF Supreme  - Phobia 200mm Rad - Silverstone AP181 Project Darkling
3770k vs 6700k RAM Scaling, HT vs RAM, Arma III CPU vs RAM, Thief CPU vs RAM
Parts Guide


#6 Rybags

Rybags

    Immortal

  • Super Hero
  • 34,378 posts

Posted 01 December 2015 - 11:01 PM

Yep - it's often pointless running more threads than cores.  The scheduling overhead creates a performance hit.  Also you could have potential cache issues, more chance of needed instructions/data getting flushed by running more threads.






0 user(s) are reading this topic

0 members, 0 guests, 0 anonymous users