Arma 3 CPU vs RAM performance comparison 1600-2133= up to 15% FPS gain

[dezkit 28]

About to upgrade from some 2x4gb 1333 old dell sticks to some corsair 2x8gb 2400 mhz sticks. I will deliver some results soon.

[St. Jimmy 272]

I've currently G.Skill RipjawsX 2x4GB 1600MHz CL9-9-9-24 and I'm getting G.Skill RipjawsX 2x4GB 2133MHz CL9-11-10-28 this week. Need to make tests and interesting to see some results. Hopefully my motherboard just doesn't suck and I can squeeze everything and maybe bit lower timings out from the new sticks :p

[cosmic10r 2331]

I've currently G.Skill RipjawsX 2x4GB 1600MHz CL9-9-9-24 and I'm getting G.Skill RipjawsX 2x4GB 2133MHz CL9-11-10-28 this week. Need to make tests and interesting to see some results. Hopefully my motherboard just doesn't suck and I can squeeze everything and maybe bit lower timings out from the new sticks :p

I'm in a similar boat... Please post results as I may upgrade if the increase is worth it

[St. Jimmy 272]

1600MHz CL9-9-9-24:

- 53fps avg. on Altis benchmark

2133MHz CL9-11-10-28:

- 58fps avg. on Altis benchmark

Settings: Resolution is 1920x1080

Rig that the tests were done

i7 2600K @ 4,4GHz

GTX 680 2GB

ASUS P8Z68-V LX

Samsung SSD 840 250GB

WD Caviar Blue 500GB

Windows 7 Ultimate 64-bit

That's 9% increase. Next I'll try to squeeze some timings lower (aaaand I haven't been succesfull after 2 months...) If I can get some lower timings to match my 1600 sticks 10% could be achievable. Seems like I got very close results that the OP had.

/Couldn't lower timings. Maybe turning object detaills down to standard would show some increase but that 9% is already enough proof for me :)

-----------------------------------Edit 31.10.------------------------------------------

Now that I got my new cooler installed I dare to OC my CPU more. From 4,4GHz -> 4,7GHz made 5fps increase.

This means that upgrading my memory and OC more my CPU I got 10fps increase in total in the benchmark. So if you're buying a new PC you should really go for a good memory and CPU cooler. If I would've bought these parts now and not after I would've only spent 30€-40€ (pretty much the price of the CPU cooler) more because there isn't much price difference with CL9 1600 and 2133 sticks. So getting low timing and high MHz sticks is worth it in Arma unlike in other games.

Edited December 4, 2014 by St. Jimmy

[St. Jimmy 272]

/Oh never mind... There are some nice timing tests also in the OP.

So anyone who's going to upgrade their RAMs really need to watch the timings also. Because CL8 1600MHz can be as quick as CL11 2133MHz. So don't be fooled that MHz are the only thing that have effect ;)

Edited November 28, 2014 by St. Jimmy

[dasa 20]

Its nice to see more people getting results there is so many gpu bottlnecked ram benchmarks out there its easy to doubt that my results were accurate till more people have a go

St. Jimmy how did you go with tightening timings? best i managed was 2133 9-10-10 1t

Now that you have your oc up to 4.7ghz your gains from the ram have probably gone up at least 1% due to the cpu being a bit more starved for bandwidth

Its a shame this old 2600k cant post with the ram at 2400mhz

Edited December 2, 2014 by dasa

[St. Jimmy 272]

St. Jimmy how did you go with tightening timings? best i managed was 2133 9-10-10 1t

Now that you have your oc up to 4.7ghz your gains from the ram have probably gone up at least 1% due to the cpu being a bit more starved for bandwidth

I've tried to tightening timings but it always fails :( So I can only get what's rated for the RAM (9-11-10-28-2T). I don't know should I try to give it bit more voltage like 1.70V to achieve anything more. I've 1.65V currently.

[markku 0]

I get new ram 2133mhz timing 9-11-10-28 old was 7-8-7-24 1600mhz and i can see little better performance but only little. As i set 137 AI in the bornholm map fps drops 22. Now i get 24-25 fps so not very good performance increase. Seems to me that 2600k @ 4.8ghz is holding back my setup. GTX 780ti EVGA should be good, ram and ssd should be good. Also Bornholm is huge fps killer too. I try different NVIDIA drivers but only 1fps more or less.

Oh, those fps is the lowest what i get...

[St. Jimmy 272]

I get new ram 2133mhz timing 9-11-10-28 old was 7-8-7-24 1600mhz and i can see little better performance but only little. As i set 137 AI in the bornholm map fps drops 22. Now i get 24-25 fps so not very good performance increase. Seems to me that 2600k @ 4.8ghz is holding back my setup. GTX 780ti EVGA should be good, ram and ssd should be good. Also Bornholm is huge fps killer too. I try different NVIDIA drivers but only 1fps more or less.

Oh, those fps is the lowest what i get...

Well if you watch the second last picture in the first post it shows that 8-8-8-24 1600MHz isn't too far from 9-11-10-31 2133MHz so your improvement shouldn't be too big. Just about 2-3fps.

AI will likely always be biggest fps killer no matter what setup you've (unless CPU OC'd maybe at 10GHz could help to get 60 fps with 137 AI :D ).

[dasa 20]

other than for a test to see if it works i wouldnt bother with 1.7v for 24\7 use it may be worth increasing the cpu pqi\vtt\vccio though incase its the memory controller is holding you back 1.05v is default anything up to 1.2v should be ok for 24\7 use from memory some run 1.35v apparently

[ruthberg 7]

4770k @ 3.5 GHz, DDR3-1333, CL9:

Altis benchmark: 39 FPS

4790k @ 3.5 GHz, DDR3-2400, CL9:

Altis benchmark: 48 FPS

4790k @ 4.6 GHz, DDR3-2400, CL9:

Altis benchmark: 59 FPS

[dnk 13]

nvm..

[AL_ 10]

Hello guys, thank you for the tests, this looks interesting..

We play one, quite demanding mission, where my pc is running it as a server. I use the famous TPW_MODS script, which adds quite a lot of load on my pc as well I guess, and the game can sometimes run at as low as 18fps. Maybe I could gain a few more fps with faster RAM? Currently I'm running i5-750 CPU at 3,68GHz, paired with 4x2 GB RAM at 1500MHz (give or take, not sure now as it's dependant on the external cpu freq.) at 8-8-8 timings

[St. Jimmy 272]

1600MHz 7-8-8-24

1866MHz 8-8-9-24

2133MHz 9-11-10-28

2133MHz 8-10-10-27 Likely the fastest.

Those are some fastest factory timings and clocks I could find between 1600-2133MHz. Some low voltage kits coul get close to those if you OC but that's just a camble if you don't know what you're doing. I haven't watched faster MHz and DDR4. Would be interesting to see some DDR3 vs. DDR4 comparsions.

[forteh 11]

Apologies for bumping this thread but would like to contribute :)

I just upgraded my ram from 8gb crucial ddr3 1600 (10-10-10-28 I think so nothing special) to 16gb teamgroup vulcan ddr3 2133 (11-11-11-31) and have also experienced considerable increases in benchmark framerates.

Running i5 2500k@4.6ghz / gtx660 / ssd / win7 pro. Arma settings are all very high with 2500 view & 1500 object distance on 1080p. Average fps has jumped from 61 to 73 fps, almost 20% increase! GPU load is 70-80% so I should be able to up some settings should I want to.

Not had a chance to test it in MP yet but fps at the start of the Huntersix mission has gone from around 23 to around 33 which holds in line with the benchmark results for SP play; around 10fps increase which is pretty huge for arma3.

[xon2 102]

@forteh ...so did you do some more testing? The interesting part of course is whether you can raise the bottom fps. It would be awesome of one could stop arma 3 dropping below 30 with a ram upgrade. Could you provide us with more comparison data?

Did anyone here build a good ddr4 system, any great improvements there?

[I give up 152]

There will be benefits in matters of performance with RAM working at higher speeds (with modules with low (CAS) latency) IF you are using a RAMDISK, otherwise will be more like a placebo.

[fabio_chavez 103]

i read that DDR4 can as well now hit very high clockspeed (up to 4ghz?), anyone ever tested that on arma3?

bratwurste, afaik the significant fps increase from high ram clocks is pretty consistently affirmed/validated, how come you claim its placebo?

[xon2 102]

i guess i will just borrow some better ram from a friend and test it myself....

[forteh 11]

I've not got any more comparison data in afraid, this was based purely on the altis benchmark results. If I get time I'll put together a quick scenario of 100ai fighting each other and then compare the typical fps with the ram running at 1600 and 2133.

It's definitely not a placebo unless msi afterburner is susceptible to placebos?

Running a ram disk would be even faster, I tested one a good while a go on 1600mhz and it garnered a very substantial speed increase, this was with my old i5 750 not my current 2500k so the apparent bottleneck might be less now. If I get time I'll set up the ram disk again and retest the same scenario.

[I give up 152]

Exactly, the main reason for bottleneck in Arma 3 is the Hard Disk. Arma 3 due to its architecture "needs" Hard Disk.

Arma 3 is not like some other games where we only use the Hard Disk when we load the map/mission for the first time and after that is only a matter of CPU/GPU/RAM.

Arma 3, after loaded, requires a continuous Hard Disk usage and this happens practically every time we move the mouse or press a key.

And the thing is, the current technology for Hard Disk in matters of performance/speed is light years back, when compared with CPU/GPU/RAM. This with most applications in general is not a issue, but for Arma 3 it is.

About RAM speed, with current technology higher frequency means higher latency, the CAS for 2133/2400 are way high when compared with 1600 or even 1866. This means that the gains that we may have in frequency are lost in latency.

In a game like Arma 3 RAM is used continuously, several GBs are continuously loaded into RAM and continuously flushed, this is a continuous process.

In this architecture if RAM speed is important, RAM latency is also important. Higher frequencis can help with minimum Frames per Second but do not improve the performance in general because it will increase response time of the CPU and even GPU (in a game like Arma 3).

Due to these facts, in my book I call it a placebo and do not worth the investment, imo.

In my tests, the setup with I achieved better performance in matters of FPS was with 8 GB of RAM, 1866Mhz, CAS 7 using a pair of Corsair Dominator.

Now if you really want to improve the performance for Arma 3 do not waste money in CPU, GPU and RAM.

For Arma 3 any mid range CPU, any mid range GPU (ATI/Nvidia) and any mid range RAM (1600/1866) are perfectly fine.

Like I have said several times, if you really want to get better performance you have 3 options.

1. Get a PCI Express SSD.

2 Get a RAMDISK.

3. Get a couple of SSD in RAID0.

(ordered by performance)

Want to reach the heaven?

Get a couple of PCI Express SSD in RAID0.

/nuff said.

[fabio_chavez 103]

Bratwurste you give very little solid reasoning for ur claims, afaik ram timings are neglectible over bare clock speed and i doubt it will be worth the loads of extra money to switch from my ssd to fancy raid or ramdisc, id rather spend a little extra on overclockable ram if that really makes a difference. Especially fast ddr3 is cheap these days though if i build a new system its going to be ddr4, id like to hear some first hand experience with a3 and fast ddr4 in particular

[jumpinghubert 49]

Like I have said several times, if you really want to get better performance you have 3 options.

1. Get a PCI Express SSD.

2 Get a RAMDISK.

3. Get a couple of SSD in RAID0.

(ordered by performance)

highscore in nonsense

To change from slow hdd to ssd eliminates stuttering so it eliminates slow downs in fps, change from ssd to raid or ramdisk it changes nothing.

Edited July 26, 2015 by JumpingHubert

[I give up 152]

Bratwurste you give very little solid reasoning for ur claims, afaik ram timings are neglectible over bare clock speed and i doubt it will be worth the loads of extra money to switch from my ssd to fancy raid or ramdisc, id rather spend a little extra on overclockable ram if that really makes a difference. Especially fast ddr3 is cheap these days though if i build a new system its going to be ddr4, id like to hear some first hand experience with a3 and fast ddr4 in particular

Theoretically, higher frequency means more bandwidth, but does not automatically means better performance, like I said we need to consider also the timings.

For benchmark purposes generally higher frequency gives better results, for real world apps is a different subject. While some apps perform better with higher frequencies, some other perform better with lower timings.

Let's give an example with 2 cases.

(Just as note. We need to have in mind that above CL7 for 1600 is garbage, above CL9 for 2133 is garbage and above CL10 for 2400 is also garbage)

First case.

With most of (decent) 1600 CL7 modules we can easily overclock it to around 2000 CL9.

So, if I can reach basically the same with my current modules why I would drop it? For 0.0001 % performance gain? This makes sense? If this make sense to you, go for it.

If I have (decent) 16 GB 1600 CL7, worth to drop these to go for 16 GB 2133 CL9? In my opinion, No.

Second case.

If I have some 1600 CL9 (or above CL9), I would drop it already, it is garbage. In these circumstances (or if was building a new system) probably I would go for 2133 CL9. Currently is the best choice considering performance vs stability, imo.

And yes, in these circumstances you will have a performance gain. For Arma 3, around 5% increase in FPS and it wil become a bit more smooth/stable.

You can get some decent 16 GB 2133 CL9 for around 130/150 bucks.

First case again.

If I have already a decent 1600 CL7, at this point and having Arma 3 in mind, definitely I would use the money in a second SSD. It really makes a difference, at least to me does.

Also with RAID0 we do not lose anything, if we set 2 SSD of 240GB in RAID0 we will have 480GB of space available in our Hard Disk.

(For DDR4, we need to wait for the next CPU generation and see how it goes. With current and because DDR4 operates with really high timings/latency, there is no performance gain when comparing with DDR3)

Edited July 27, 2015 by Bratwurste

[jumpinghubert 49]

In the past we had people with arma2 on ramdisk. No difference in performance, only less stuttering. Every one knows latency is also important for bandwich (and bandwich is important for higher fps especially in arma)

Please bratwurste don´t highjack this very useful thread with your wrong statements. Thanks in advance.

You ignore Chavez´ hint the significant fps increase from high ram clocks is pretty consistently affirmed/validated in this thread.

Edited July 27, 2015 by JumpingHubert