Jump to content
Sign in to follow this  
Folterknecht

WoT performance scaling with RAM speed - benchmarked (now with iGPU)

Recommended Posts

Hi!

 

To answer that question I spend several hours yesterday testing that out. The sytem used was as follows:

  • Win7 Pro 64bit SP1 running from an SSD (Samsung 830 Pro 512GB)
  • WoT 0.9.10 was also installed on the same SSD. Using the SD client
  • Nvidia 355.98 Driver

 

  • i5 3570K @4GHz (40x 100MHz)
  • 2x 4GB DDR3 1866 MHz  9-10-9-28 1T G.SKILL SNIPER
  • ASRock Z77 Professional
  • GTX960 2GB running at constant 1455 MHz core and 1753 MHz memory speed. VRAM usage between 1 - 1.2 GB

 

ingame graphic settings

ramperfromanceu8aj7.jpg

 

methodology:

RAM and CPU settings were dialed in via BIOS, the system was started running MSI Afterburner (for ingame monitoring) and CPU-Z to make sure all BIOS settings are applied correctly. After that the replay (attached) is run and XVM was outputting a dataset for each benchmark run containing around 3430 data points (one every 0.1sec) listing the current FPS. After the replay was over the system was restarted and the process repeated 3 times for every RAM setting except 2000MHz, there I only made one run. The 3 results for avg, min and max FPS for each speed where averaged.

 

2133 MHz isn't possible with the combination of my RAM and motherboard, they top out around 2080 MHz .For 2080 MHz RAM speed I would have to use a differenct BCKL and Multi, making the resulting numbers nearly irrelevant anyway for this comparison, because I can't hit excatly 4 GHz CPU speed that way.

 

The result? No real suprise there, WoT does not scale with RAM speed:

  2x 4GB DDR3 1333 CL9-9-9-24   2x 4GB DDR3 1600 CL9-9-9-24 1T   2x 4GB DDR3 1866 CL9-10-9-28 1T   DDR3 2000 CL10-11-10-32 1T
Avg 1. run 67,5   67,5   67,7   67,7
Avg 2. run 67,4   67,7   67,7    
Avg 3. run 67,6   67,7   67,9    
               
Min 1. run 47,5   46,9   47,3   47,2
Min 2. run 46,8   47,3   46,8    
Min 3. run 47,3   46,9   46,8    
               
Max 1. run 93,3   94,3   94,3   93,5
Max 2. run 92,8   92,9   94,7    
Max 3. run 93,5   92,3   96,0    
               
AVG 67,5   67,6   67,8    
MIN 47,2   47,0   47,0    
MAX 93,2   94,3   95    

 

Are there other things to learn here? Well ofc there is still the question to be answered, if that also applies to playing WoT on an integrated GPU, which I'm planning to do in the near future. I 'll also post my findings here. 

 

Furthermore I highly recommend using more than 4GB of RAM on a 64bit OS for playing WoT. You can get by with 4GB, if you havn't polluted your OS with all kind of nonsense running in the background and close all background tasks. But just keeping a browser with several tabs open can in combination with WoT exceed 4GB, which would lead to all kinds of OS-background activity involving your storage. This mainly effects HDD users, but we all know that WoT is sensitive in that regard.

Does more than 8 GB RAM help? Nope, not with playing WoT as long as you dont plan on creating a RAM Disk!

 

So when you go out and buy RAM it doesn't matter if it's 1333 or 2133 MHz - for WoT using a dedicated graphic card with at least 2GB VRAM!  But the prices in the US/CA and the EU are more or less the same from 1333 MHz to 1866 MHz (sometimes even 2133 MHz on sale), so I 'd go for the faster RAM (1st) with the lowest timings (2nd - CL numbers), as there are applications and games where RAM makes a difference, though "relative small". Just make sure you buy enough capacity and use kits with 2 or 4 modules (3 for trippel channel board like X58 and X79), for the benefits of dual/trippel channel (0-5% performance advantage over single channel).

 

http://wotreplays.com/site/2246165#mountain_pass-f0lterknecht-is-3

 

Update - Integrated Graphics (iGPU - HD4000)

 

Ok, now we 'll be looking at the iGPU performance of the HD4000 in my i3570K. Hardware, software and methodology were the same as with the GTX960 installed. The HD4000 wasn't overclocked and the BIOS settings were also the same as with my dGPU. The only additional BIOS setting worth mentioning was 1024MB of shared memory.

 

Ofc the ingame settings had to be adjusted - I switched back to the old render (standard graphic)

wotigpuingamesettingsh7zad.jpg

 

 

That had the following optical results

wotramperformanceigput8qsi.jpg

 

 

  2x 4GB DDR3 1333 CL9-9-9-24   2x 4GB DDR3 1600 CL9-9-9-24 1T   2x 4GB DDR3 1866 CL9-10-9-28 1T   DDR3 2000 CL10-11-10-32 1T
Avg 1 45,9   48,7   50,1   50,0
Avg 2 45,9   48,6   50,0    
Avg 3 46,1   48,6   49,8   50,0
               
Min 1 28,3   30,2   31,2   32,3
Min 2 29,3   31,7   30,8    
Min 3 30,1   30,9   32,2   32,0
               
Max 1 73,7   79,0   82,0   81,2
Max 2 72,2   79,2   81,2    
Max 3 73,4   78,5   81,0   81,6
               
AVG 46,0   48,6   50,0    
MIN 29,2   30,9   31,4    
MAX 73,1   78,9   81,4    

 

Again I made a few runs with 2000MHz RAM speed. This time though the results show something thats worth remembering - higher timings counteract the higher speed to a degree. The increase in performance isn't that great*.

Puting the 2000 MHz results to the side, we see a performance gain of slightly above 5.5% for avg and min FPS stepping up from 1333 to 1600 MHz. The jump from 1333 to 1866 MHz brings a 8.7% plus for avg FPS and +7.5% for min FPS.

 

So for little to no price increase we get a decent increase in avg and min FPS in case of an iGPU - not bad. The performance plus for AMDs APUs is even bigger in many games. See AMDs marketing slide below:

 

Spoiler

http://www.techpowerup.com/reviews/AMD/A10-7860K/images/mem_gains.jpg

 

*You get a rough idea of how fast a RAM kit is compared to offers from different vendors or even the same manufacturer by calculating the so called performance index number. I stumbled above this handy little trick in an article by Dr. Ian Cutress from Anandtech. This works as follows:

Divide the speed number of your memory Kit for example 1333 MHz by the timings. In our case its CL 9.

1333/9 ~ 148

Now do the same for the memory kit you want to compare it to (stay within DDR3)  - for example

2133/12 ~ 178

 

As you can clearly see 178 > 148 and therefor in most cases the kit DDR3 kit with 2133 MHz and CL12 will be faster than the 1333 MHz CL9 kit.

 

20151023_2151_ussr-R19_IS-3_37_caucasus.wotreplay

Share this post


Link to post
Share on other sites

Thank you for this information!

 

I've been meaning to run some different benchmark tests of my own but have been too lazy to work when I'm not at work.  Aside from the game, It's really hard for me to think about anything computer related when I'm not cloistered in my cube 50-60 hours a week...

Share this post


Link to post
Share on other sites

thats interesting, good to know.

 

how would RAM speed affect different games and applications though? games like BF4/GTA 5 for example, along with general usage?

The difference can be really noticable, depending on the game or programm. Packers like WinRar for example are generally more sernsitive to RAM speed than many games. On the other hand RTS games with dozens or hundrets of units also tend to be RAM sensitive - see below (Witcher 3 vs Total War Attila):

 

ramscalingwitcher3b9uqr.jpg

ramscalingtotalwaratty6u53.jpg

 

 

Share this post


Link to post
Share on other sites

Very interesting. As you know from my other threads I bought an SSD (it will go in a computer that has an i5 in it) for one computer and a whole new computer (finally retired an imac from running WoT in boot camp under windows 7). The new computer has an i7 6600 Skylake running at 4.0ghz on an ASUS Z170, with 2400mhz RAM with 16 gb and an nvidia GTX 980 graphics card. It is setup with 500gb SSD on C drive and 2 terabyte HDD as a D drive, and only WoT, Windows 10, and Team Speak so far on the C drive (have yet to load all my wife's work software - sshh don't tell her yet).

Presently WoT 9.10 has all graphics settings maxed out and FPS is running at 117-122. Clearly the RAM is not really helping make WoT perform. Would you guess that it is the speed of the cores or the graphics card pushing WoT to that FPS?

Share this post


Link to post
Share on other sites

Both

WoT needs a strong single thread performance (IPC) and some decent clockspeed on the CPU side of things. And a strong GPU be it from ATI or Nvidia to move the pixels around.

Share this post


Link to post
Share on other sites

Update - Integrated Graphics (iGPU - HD4000)

 

Ok, now we 'll be looking at the iGPU performance of the HD4000 in my i3570K. Hardware, software and methodology were the same as with the GTX960 installed. The HD4000 wasn't overclocked and the BIOS seetings were also the same as with my dGPU. The only additional BIOS setting worth mentioning was 1024MB of shared memory.

 

Ofc the ingame settings had to be adjusted - I switched back to the old render (standard graphic). Side note - enabling 2x MSAA nearly cut FPS in half with this iGPU.

wotigpuingamesettingsh7zad.jpg

 

 

That had the following optical results

wotramperformanceigput8qsi.jpg

 

 

 2x 4GB DDR3 1333 CL9-9-9-24 2x 4GB DDR3 1600 CL9-9-9-24 1T 2x 4GB DDR3 1866 CL9-10-9-28 1T DDR3 2000 CL10-11-10-32 1T
Avg 145,9 48,7 50,1 50,0
Avg 245,9 48,6 50,0  
Avg 346,1 48,6 49,8 50,0
        
Min 128,3 30,2 31,2 32,3
Min 229,3 31,7 30,8  
Min 330,1 30,9 32,2 32,0
        
Max 173,7 79,0 82,0 81,2
Max 272,2 79,2 81,2  
Max 373,4 78,5 81,0 81,6
        
AVG46,0 48,6 50,0  
MIN29,2 30,9 31,4  
MAX73,1 78,9 81,4  

 

Again I made a few runs with 2000MHz RAM speed. This time though the results show something thats worth remembering - higher timings counteract the higher speed to a degree. The increase in performance is slowing down there*.

Puting the 2000 MHz results to the side, we see a performance gain of slightly above 5.5% for avg and min FPS stepping up from 1333 to 1600 MHz and the jump from 1333 to 1866 MHz brings a 8.7% plus for avg FPS and +7.5% for min FPS.

 

So for little to no price increase we get a decent increase in avg and min FPS in case of an iGPU - not bad. The performance plus for AMDs APUs is even bigger in many games.

 

*You get get a rough idea of how fast a RAM kit is compared to offers from different vendors or even the same manufacturer by calculating the so called performance index number. I stumbled above this handy little trick in an article by Dr. Ian Cutress from Anandtech. This works as follows:

Divide the speed number of your memory Kit for example 1333 MHz by the timings. In our case its CL 9.

1333/9 ~ 148

Now do the same for the memory kit you want to compare it to (stay within DDR3)  - for example

2133/12 ~ 178

 

As you can clearly see 178 > 148 and therefor in most cases the kit DDR3 kit with 2133 MHz and CL12 will be faster than the 1333 MHz CL9 kit.

 

Edited by Folterknecht

Share this post


Link to post
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

Sign in to follow this  

  • Recently Browsing   0 members

    No registered users viewing this page.

×
×
  • Create New...