Will-my-pc-run-Arma3? What cpu/gpu to get? What settings? What system specifications?

[Groove_C 267]

Wrong. But well, don't want to insist.

[Reezo 45]

18 minutes ago, Groove_C said:

Wrong. But well, don't want to insist.

 

Sorry if I seemed to be arguing, I actually wanted to know more about this. I am actually live streaming this: I set my i7 9700K at stock CPU + XMP Profile and run YAAB. Then I am going to overclock the memory and see what kind results this yields.

 

You are welcome to join (as is everybody else) on my channel, I am live, there!

 

 

[Groove_C 267]

Need to also increase SA (system agent / memory controller) and I/O analog/digital voltages when very low timings and/or very high frequency.

Core voltage might need to be increased as well.

 

ASRock Timing Configurator 4.0.4 for Z390

 

Secondary and tertiary timings are very important as well.

If PC is not crashing, it doesn't mean, that everything is fine with RAM settings. Can result in lower FPS.

[Reezo 45]

Thanks @Groove_C - very useful.

 

So far I am benching exactly the same, no benefit:

 

(CPU ON STOCK)

RAM 3200 CL14 1.35v = LOW: 53.5 // HIGH: 80.5

RAM 3600 CL14 (same primary timings but all auto on the others) = LOW: 79.9 // HIGH: 52.4

 

I am clearly not getting anything out of this so far and that's because the timings might be adapting themselves (given the many "Auto" I have).

 

Experimenting more, now! Live! Hot! Amazing 🙂

 

UPDATE: Finding out lots of things about this! So good! Thanks, I am doing this live in the stream and there's a lot to be learnt!

[Groove_C 267]

Download and install this and launch it after reboot:

Asus TurboV Core 1.03.02

[Reezo 45]

17 minutes ago, Groove_C said:

Download and install this and launch it after reboot:

Asus TurboV Core 1.03.02

 

Can't install that because I have a Gigabyte otherboard..

 

--

 

Anyway, thanks! Here's what I found out: I had not noticed that during training the BIOS was lowering my OC settings for RAM when not working.. and I though it was training fine as on the AMD I either got to POST or to not POST.. interesting!

 

I just got 85fps instead of 80fps at low settings by going from 3200 14-14-14-34 to 3733 16-16-16-36 Command Rate: 1 .. and I did not even try a serious RAM overclock..

 

All secondary and tertiary are still on auto.. wow!

This worth a second episode! Thanks,

[Groove_C 267]

You can set SA and I/O analog/digital to 1.25 V both. So 3x 1.25 V.

[Reezo 45]

31 minutes ago, Groove_C said:

You can set SA and I/O analog/digital to 1.25 V both. So 3x 1.25 V. 

 

Definitely worth talking about it tomorrow in an addendum episode, I was clearly able to get +5fps from 80 to 85 at low settings, just by randomly going from 3200 CL14 to 3733 CL16 without even being serious at it. At ULTRA settings I did not get any improvement but I guess because the CPU/GPU are more under stress (I kept the stock CPU values..) the faster RAM does not make that much of a difference. Once I bring it back to 5GHz I am sure it will be able to squeeze some more FPS out of the system, or possibly even raise the lowest minimum because of the lower latency (I'd rather have lower 1% minimum than higher FPS as I am a G-Sync player at 75Hz lol :))

 

I had not noticed that the motherboard was doing what it wanted after unsuccessful training.. in my Asus x470 I would either die not posting or post successfully (and maybe crash in Windows.. but still..) here the motherboard is assessing and lowering values in case of initial inability to post.. that was deceiving 😉

[Groove_C 267]

RAM OC helps a lot, once in CPU limit. Proven! Sure, at higher res and video settings there is less of a difference, but still should be. If it's for free, why not use it.

[Reezo 45]

7 minutes ago, Groove_C said:

RAM OC helps a lot, once in CPU limit. Proven! Sure, at higher res and video settings there is less of a difference, but still should be. If it's for free, why not use it.

 

Couldn't agree more. Stay tuned, tomorrow I'll add a chapter related to pushing the RAM in ArmA 3 from my 5GHz profile I have. I played a simple mission today 100% in urban area in Altis and I was like "uuuuuuh babeeeee" <--yes.

[Groove_C 267]

new Intel CPUs vulnerability -> CacheOut

[Reezo 45]

2 hours ago, Groove_C said:

new Intel CPUs vulnerability -> CacheOut

 

Quote

"I don't care if all I get is more fps in ArmA 3"

 

teasing ;))

[Groove_C 267]

Well, when BIOS and/or Windows updates to fix (partially) vulnerabilities, it costs performance, so...

[Horus 83]

6 hours ago, Groove_C said:

new Intel CPUs vulnerability -> CacheOut

Skylake/Cascade Lake and up are affected with this https://software.intel.com/security-software-guidance/insights/processors-affected-l1d-eviction-sampling

 

[Groove_C 267]

And not even released 10th gen Comet Lake CPUs are affected as well, since same architecture )))

So they will have a lot of vulnerabilities right on the release )))

And new vulnerabilities on the way. Just not publicly disclosed yet, since no solution for them, yet.

 

Cristiano Giuffrida, Vrije Universiteit, Amsterdam

Security engineering at Intel (or rather lack thereof) is still business as usual. These issues aren't trivial to fix.

But after eighteen months, they're still waiting for researchers to put together proofs-of-concept of every small variation of the attack for them? It’s amazing.

We don’t know the inner workings of Intel's team. But it’s not a good look from the outside.

If Intel isn’t looking and is making no effort, who knows how many vulnerabilities are left?

[Reezo 45]

I was able to finish the conclusions video before I get attacked on my brand new, unfixed i7!

 

Less than 10 minutes, people, here it is, the final answer for me:

 

[Groove_C 267]

Well, as I could see, you benched only 1x per setting, which is really wrong, because of great discrepancy there can be between each run, thus resulting in wrong results for AMD and Intel and you also couldn't manage to OC your RAM on Intel.

So for me, the results can't be concidered at all.

One needs to run at least 3x per quality setting and per CPU and/or RAM setting, to make and average of 3 runs, for min, avg and max.

And if the discrepancy between runs is high (2 FPS or more), continue to bench (more than 3x) and completely drop anomal runs. If min is too low or max is too high.

Also closing all known/unnecessary background tasks, like antivirus, firefox, discord, monitoring software (MSI Afterburner/EVGA Precision X/ HWiNFO64), nVidia Shadowplay, Discord, Skype... that can influence the results.

 

@ standard I can have like 36, 43, 39 FPS min, 68-70 FPS avg and 95, 108, 103 FPS max.

So I have to understand what is anomal, too low 36 FPS or too high 43 FPS, too low 95 FPS or too high 108 FPS.

To clarify this, I have to run more at same settings, to see what is closer to the reality, in the way, what comes as result more often, so I can exclude, 36 or 43, 95 or 108.

To have a clear picture.

[Reezo 45]

Hi @Groove_C I am sorry to hear I did not give you anything back after all the help you gave me. I thought somehow this could be added to the pool of tests all people out there did, with the same degree of real-world bench. I thought actually benching two identical builds that differ only in motherboard and CPU ("only" being still a lot, obviously) was something else.

 

1) I did test 5x each setting and took the best result of 5. I have to say anyway that all results were equally far off of no more than 2FPS so the delta is maintained whether I use the best, the worst or the average of 5. I had no anomalies in the results in AMD or Intel. Just a normal 2 fps delta at most. When I say Intel was % better better than AMD I could pretty much keep that difference in all five tests (worst of, best of, average)

 

2) As I said I did not manage to OC the RAM on Intel but that was not the point of the video, I did overclock the two systems at the best that I could using their strong points. I guess if I had overclocked the memory on Intel I would have even had better results, thus making the Intel build even more ahead of the AMD (that's a natural guess)

 

3) I did close all unnecessary tasks and tests were done in the most equal conditions I could.

 

 

Is there any video out there that you could point out that could give a clearer picture than mine? I am sorry I haven't found one before making my own, it could have been a good inspiration.

 

I believe saying "it cannot be considered at all" is a bit of a stretch.. I mean I find it hard to believe I did something completely useless? To me (and I am just entitled as much as anybody else in here) it was a great hard proof to see that in both conditions (even my own two hands building the rigs) I was able to see the numbers of how things felt.

 

I already said that both rigs were perfectly playable and enjoyable, I don't think I am even bashing any of the two? I've read that you are cable of unbiased thought and also added you to the credits in the video because of all the help you gave me, I cannot possibly fathom that I made something totally unacceptable 🙂

 

And then again, it's just some guy (me) making some test, come on, cheer up brother, come over here and gimme a hug 🙂

 

[Groove_C 267]

You say you tested each change in video quality and CPU/RAM settings 5x, but in your live videos I could see you run 1x and then directly upload to imgur with titles for screens, so...

And also making average of all 5 runs is not the way I would do, if in some runs min or/and high are more than 2 FPS off.

 

I'm not a fanboy, and if Intel or AMD sucks -> they suck. I can bash on Intel for known for years vulnerabilities, even if I myself have more than 1 Intel. I don't care. I don't "protect" something, because I'm the owner of. Would have be stupid if I did so.

 

I know other people that have benched Intel vs. AMD and Intel was not behind AMD @ min FPS. They did it with RAM at +- same settings and with CMA AVX2 x64.

[Reezo 45]

1 minute ago, Groove_C said:

You say you tested each change in video quality and CPU/RAM settings 5x, but in your live videos I could see you run 1x and then directly upload to imgur with titles for screens, so...

And also making average of all 5 runs is not the way I would do, if in some runs min or/and high are more than 2 FPS off.

 

Makes sense but trust me, I did ..and - if it was crashing in one of the five tests - I proceeded to trash all tests and find a BIOS setting that was stable.

 

The 5 runs I had all benched within 2 fps delta. The worst bench from AMD and the worst bench from Intel at the same settings are still the same % apart. Even if I had taken the best of AMD and the worst of Intel, the Intel would still be over +10% better average fps. Again, I can't stress enough that this is what happened here in my beloved room 🙂

[froggyluv 2135]

*Grabs more popcorn for the infamous 1-3 FPS nerd wars

[Groove_C 267]

And you say that 9700K is better than 3700X, because average and max FPS are better, which is also completely wrong, since Intel dips to 33 and AMD to only 38, which is more important, obviously.

What brings me 110 FPS peak if I have just little above 30 FPS in heavy situations, whereas the other system delivers almost 40 FPS in same heavy situation.

[Groove_C 267]

And well, it's a already known fact, that in A3, 5.0 GHz Intel 9xxx is same as 4.2 GHz Ryzen 3xxx, with same RAM settings and CMA memory allocator.

[Groove_C 267]

Max FPS is not important at all, since it can peak to a value maybe for 1 ms during several hours gameplay.

I use max FPS just as indicator of anomaly run vs. run, to see if I need to run the benchmark again, but I don't pay attention to it at all.

Average is what you will experience the most of the time, but then, when you're in a large village with a lot of shooting, explosions, destructions, players and AI, your FPS will be closer to minimum.

So average FPS is to consider, but minimum FPS is not less important, if not more.

[Groove_C 267]

A 3700X 4.2 GHz with RAM OC is better than 9700K 5.0 GHz with RAM OC, since same performance in A3, but less $$$, less heat and lower power consumption + you can stream, Blender etc.

And yes, I say it despite myself having morethan 1 Intel system.