CPU VS RAM Performance & CPU Threading Benchmarked

[Leon86 13]

nvm armamark, good.

rarely that much difference between memory, especially in games.

Edited April 21, 2013 by Leon86

[jumpinghubert 49]

we have a sticky thead about it. Thats better than nothing (the years before).

[fabio_chavez 103]

i dont get white, while he has some point, that is maybe not baseles, hes trying to come up with logic and stuff while beeing completely ignorant to the fact that he has no justification to trying to blame one particular developer for the issues, (the performance issues are there, and they have been around for some time but he has not the slightest insight what the real causes are in the context of the developer side, instead he boasts about widely known information about the issues as if he personally did like research to exclusively reveal it to the public. Also if you are not willing to live up to academic standards in debating, stop claiming you are capable of logic reasoning and schooling people about ad hominem attacks while attack people ad hominem is also not making you look credible because it makes you exactly and precisly what is called a troll.

you are an agressive selfrighteous troll that is trying to take this alledged flaw of BI development as an excuse for beeing a dick.

if you had any brains in you then you would have to come to the conclusion that BI would allready have fixed this significant problems if it was within their limit of resources, and its persistence is not due to irrational or malevolence and definately not because of their lack of know how.

edit: one could argue that BIs information policy about this is not nice, that is correct, but - given the circumstances that a company will align its information policy with its economical necessities - to think that to hassle a developer in the forum will fix the hardware utilization, that is not correct.

And its not like BI is trying to sell a broken product due to deception...

Edited April 21, 2013 by Fabio_Chavez

[MavericK96 0]

we have a sticky thead about it. Thats better than nothing (the years before).

Agreed. At least it seems to have some more importance placed on it, which is good since the game is still Alpha and hopefully it can be fixed.

Dwarden...sorry you have to deal with trolls, but I guess until the issue gets fixed they're probably not going to go away. So you should...you know...fix it quickly. :D

(Kidding :p...though I do really want to see this fixed)

[Insanatrix 0]

So asking a developer to be more direct is trolling now? I'm sure someone has a spare pair of balls somewhere....

[fabio_chavez 103]

by what standard can this be considered "asking", i ask you?

really, you are implying that more than 2 cores makes a difference? god, thanks for ignoring every post proving otherwise, good to know all the effort to try to clear the subject from every angle was completely in vain, showing how you guys are indeed not going to do absolutely nothing about it. another half a decade with a dual core game here we go.

or to make sure everyone gets it straight: "really, you are implying that more than 2 cores makes a difference?" <- no he wasnt and there was nothing indicating him doing so, the interpretation was based on anger rather than on reason.

why would someone respect that?

Edited April 21, 2013 by Fabio_Chavez

[MavericK96 0]

So asking a developer to be more direct is trolling now? I'm sure someone has a spare pair of balls somewhere....

I'm pretty sure in the large stickied thread a dev said something like "we're looking into it" and it's labelled as "assigned" on the issue tracker. Not sure what else you guys want...a status update might be nice but maybe they're focused on other issues right now (hopefully the performance one is a top priority, though) or there's just not much to say.

[white 1]

i dont get white, while he has some point, that is maybe not baseles, hes trying to come up with logic and stuff while beeing completely ignorant to the fact that he has no justification to trying to blame one particular developer for the issues, (the performance issues are there, and they have been around for some time but he has not the slightest insight what the real causes are in the context of the developer side, instead he boasts about widely known information about the issues as if he personally did like research to exclusively reveal it to the public. Also if you are not willing to live up to academic standards in debating, stop claiming you are capable of logic reasoning and schooling people about ad hominem attacks while attack people ad hominem is also not making you look credible because it makes you exactly and precisly what is called a troll.

you are an agressive selfrighteous troll that is trying to take this alledged flaw of BI development as an excuse for beeing a dick.

if you had any brains in you then you would have to come to the conclusion that BI would allready have fixed this significant problems if it was within their limit of resources, and its persistence is not due to irrational or malevolence and definately not because of their lack of know how.

I´m not trying to prove it on this topic, the sum of all that has being discussed shows the same thing over and over, i could pick other users posts, or benchmarks like the ones i linked to show the same point. because a lot of people have replicated this, even on a discussion on steam that had a ton of information about this that was closed by Dwarden on steam. I asked the author to share them here and he say he wouldnt bother because that wouldnt change BIS approach to it. SO i respected him by not reposting. but The person i was arguing with on this topic knows and have read all that was there. But like you when someone on a forum attacks me directly, but not my argument, its pretty obvious this person cant argue against what ive stated so then tries to discredit me. this is a mediocre approach and will be pointed out, exatcly like you are trying to do now with your "if you had any brains". can you comprehend how pathetic that approach to arguing sounds? maybe not after all you are using it.

Another simple and logic conclusion is that not enough people have complained about this in the past for it to be taken as a priority, and fanboys arent helping at all working directly against what they call "whining". I don´t blame Dwarden as an individual, he has taken upon himself the role of the company´s voice, and i take his statements seriously as if they were made by the company itself, like i would with any other employee of any company out there. when i direct my posts to him im trying to get an counter argument from BIS, not the individual alone, that would be fruitless.

its been said by him that the game supports multicore for a long time, but then he says that they are aware of the issue and are going to fix it. well what issue? the game not handling multicore properly? because thats the low usage issue. which is it? he also posted and has insisted that no games have high cpu usage or are very efficient with multithreading, even posting an BIS article about it. but when anyone can verify that the game barely uses more than 2 cores (yes there are a ton of threads that spill into other cores, like the new physics and clouds) because the main threads are bottlenecked by the big thread on the first core and anything else usefull fit fine on the second core (shown by not having significant performance gain with more than 2 cores), and that we can produce games with high usage that scales performancewise with 4 cores or more, then those statements are false, or only apply to certain games, but are not a rule of how multithreading performance in games work.

what i take from those statements is whats implied on them, that the engine already is multicore (with proper multithread limited by how multithreading works while already supporting 4 cores or more) and that not much can be done about that, since we shouldnt expect "unrealistic gains" with their "optimizations". they will fix something, but what? multicore support that they say they already have? that unknown misterious issue that wasnt specified?

blaming amd drivers also dont help much, since other games handle 4 cores, and sometimes more, just fine, and gets me curious about how would amd respond to that serious statement.

edit: one could argue that BIs information policy about this is not nice, that is correct, but - given the circumstances that a company will align its information policy with its economical necessities - to think that to hassle a developer in the forum will fix the hardware utilization, that is not correct.

And its not like BI is trying to sell a broken product due to deception...

Then theres that. right now the alpha doesnt play "fine" with the recommended specs shown on steam, allegedly a rare few have no problem with it though, but like ive said before, i dont use exceptions as the rule. but i have some issues with that for 2 reasons, its no secret the game has a performance problem, same limitation that always existed on ArmA 2, its here on a more demanding game now, its on the most voted issue, usually is on the first page on the steam discussion with a new topic, its on other foruns, reddit aswell, and so, why it isnt on the known issues/sitrep? there are plenty of irrelevant things there though. that reminds me of the "dont expect unrealistic gains" thing. which to me sounds like "it wont change much if at all", and since beta will be in about 2 months according to them (end of Q2)and the rule of thumb is that this kind of main design issue, that i agree it could be considered acceptable in Alpha, shouldnt even get to beta, because i dont expect a game in beta to have its fundamental inner workings (AI threads, AI simulation syncing) reworked. Makes no sense for them to not be clear about it all while letting the userbase shitstorm go on.

But still, i hope i´m wrong, i really do. Otherwise i wouldnt be posting, i would love for them to shut me up with clear official statements about it.

I could make a small exercise here, the questions i would like simple yes/no to:

1 - Do they know what causes the low cpu usage on cpus with more than 2 cores and are they going to fix it before launch? not try, but fix it (i would guess Y and N)

2 - Will the game share its heaviest thread load across all cores? (or at least 4)? (Kind of the same question but more specific. I would guess N)

3 - Will AI syncing be reworked/fixed so multiplayer doesnt have such a huge impact with offscreen simulations away from the view? (That reminds me of the Suma statement from 2 years ago, which isnt very favorable this close to beta, but that could also solve the performance issue even with the game being "dualcore", i would guess N again.)

4 - Will there be GPU accelerated (particles/physix/clouds) support at launch? (not as important, but i guess those could offload somethings off the cpu at least for nvidia users, thats always usefull and what ususally games with physix do)

there you go, i replied truthfully while not deliberatly attacking what would be an failed attempt to imagine who you are and your brain capabilities, you know, like you did.

Probably ill get another warn or have everything erased for replying another completely offtopic post directed at me, but even Dwarden encouraged this offtopic discussion by keeping replying to them (my first posts where directly about CPU performance and the topic). and you shouldnt have posted to begin with, your problem is that you lack context.

Thats all.

Edited April 21, 2013 by white

[Insanatrix 0]

by what standard can this be considered "asking", i ask you?

or to make sure everyone gets it straight: "really, you are implying that more than 2 cores makes a difference?" <- no he wasnt and there was nothing indicating him doing so, the interpretation was based on anger rather than on reason.

why would someone respect that?

You mind actually quoting me instead of someone else if you're going to argue something I said?

Also he simply said that the developer was ignoring the facts that users have placed in front of him countering what he is stating. I'm sorry but to be blunt, he is.

I'm pretty sure in the large stickied thread a dev said something like "we're looking into it" and it's labelled as "assigned" on the issue tracker. Not sure what else you guys want...a status update might be nice but maybe they're focused on other issues right now (hopefully the performance one is a top priority, though) or there's just not much to say.

After 6+ years though? Anyone with some logic can understand that it's because they don't want to rewrite their engine, they just want to reuse it over and over again with the inherent flaws that come with it. The problem is that those flaws are becoming more and more serious with the way technology is moving and this engine being unable to cope with that. They've pretty much confirmed that by saying that they don't wanna go 64 bit and the engine is already multi-threaded, which you can see by running any API monitor or hardware graph that it's very limited multi-threading and it's more thread swapping than it is true multi-threaded or parallel processing.

This was said years ago, now we are hitting a point where there are going to be no more 32 bit OS's, There haven't been 32 bit processor's in years and quad/hexa/octo-core processor's are becoming the norm. Yet that development ethos is still apparently upheld today.

[white 1]

by what standard can this be considered "asking", i ask you?

or to make sure everyone gets it straight: "really, you are implying that more than 2 cores makes a difference?" <- no he wasnt and there was nothing indicating him doing so, the interpretation was based on anger rather than on reason.

why would someone respect that?

exthread is about sharing those specific threads into more cores, it does imply that by using it or not would alter the performance with 2 or more cores, thus the reason he asked, but it doesnt and has been shown a few times before when showing that 2 or more cores dont make much of a difference with or without launch parameters, if at all. i can believe that it was a naive question but that would also imply that he ignored every post about the exact same benchmark done before. even the one i posted in which i used affinity aswell, exatcly like in this topic, and like i mentioned, is frustrating to have a huge topic about it with dozens of ppl posting and trying things out only to be ignored.

and why would i be angry at him? makes no sense. i´m frustrated indeed, somewhat disappointed, but it doesnt change or affect my arguments. you keep trying to attack me as a person instead of trying to understand and counter my arguments. not knowing enough and being eager to post is your problem, should stick more to facts than your imagination. and now i also know that the low cpu/gpu usage topic is being ignored/dismissed, and that tells me a lot.

Edited April 22, 2013 by white

[dasa 20]

Did you turn "core parking" off?

no but i tried it then and it made no difference at 4.7ghz with ht enabled it was still getting around 42fps

After 6+ years though? Anyone with some logic can understand that it's because they don't want to rewrite their engine, they just want to reuse it over and over again with the inherent flaws that come with it. The problem is that those flaws are becoming more and more serious with the way technology is moving and this engine being unable to cope with that. They've pretty much confirmed that by saying that they don't wanna go 64 bit and the engine is already multi-threaded, which you can see by running any API monitor or hardware graph that it's very limited multi-threading and it's more thread swapping than it is true multi-threaded or parallel processing.

This was said years ago, now we are hitting a point where there are going to be no more 32 bit OS's, There haven't been 32 bit processor's in years and quad/hexa/octo-core processor's are becoming the norm. Yet that development ethos is still apparently upheld today.

that is what i was thinking

it would take a lot of time and money to do such a massive engine overhall and there is no way it will be done so late in the arma 3 development the fact is they probably cant afford it

i wouldnt be surprised if none of the current devs even had the skills required to do it as from what i can gather it is no easy task

[mobile_medic 43]

... spend some moment on trying find my answers about the performance issues in multiple threads

then you become aware we know about it ...

just go on thread list (not search, not new posts) and there is this little icon saying "Developer posts in thread") click it

and then next to thread # there is icon which takes you to another developer post and so on ...

hard to try it explain to the very same people who ignored the posts in first place ....

With all due respect, the thing we are still not aware about is if you actually intend to do anything significant about it, or give it the attention that it deserves.

I am sure you know about it. It would be shocking if you did not after this many years of this issue being reported, and talked about.

Perhaps you could alleviate much concern by simply adding the longest-standing, most reported, most up-voted issue with your game(s) by adding it to the so-called list of "Top known issues" of the "most obvious known issues", since it is clearly both of those things, and has been going back several years now.

I hope you can understand that 'of course we know about it' is kind of stating the obvious, and only adds to the concern of your customer base as it does not even begin to rise to the level of conversation and reporting that has taken place over the last years on this very fundamental issue with the Arma series that makes what could be an absolutely epic game entirely unplayable for a good number of us.

It is very clear when i watch my performance nosedive as my hardware utilization also nosedives as a result of something to do with (at least) the presence of AI. When I am on an empty map, gpu usage is 90-100% and fps performance is great. If I play small user-made issues, I get largely good performance and good gpu utilization. Otherwise, I watch my gpu being utilized at 30% while I get 15-25fps and the game is rendered unplayable. Same like Arma 2 (which I spent nearly as much time troubleshooting as I did actually getting to play it before finally giving up. Well, maybe I've been fooled twice here, by actually giving you money again. And, it is the seeming lack of attention to the issue after having high hopes that this "new engine" would address the most fundamental issue of Arma 2, that is utterly frustrating. In my humble opinion, if this issue is still outstanding on release of Arma 3, BIS deserves much bad press for their response and (apparent) lack of attention. From this side, that's what it looks like.

--

I will submit that having that one long sticky thread (while obviously good for keeping the board at large more organized) does not actually offer much hope to me personally. Especially when you read through it. It tends to confuse and cloud the root of the issue. Particularly, when combined with lack of meaningful (in my view) input from the developers. This is not a small issue. This is a fundamental, and long-standing issue. Something made even more frustrating by the fact that you folks actually do regularly patch your games. To watch such a long-standing issue continue to go largely un-addressed is not exactly encouraging.

So, forgive us if we are extremely skeptical about a hopeful resolution to this long-standing issue. It is not without evidence though.

Peace.

Edited April 22, 2013 by Mobile_Medic

[dasa 20]

just ran a quick test with tighter timmings for comparison

4.7GHz DDR1333 7-7-7-24-1T 38.9FPS

its still slower that 1600 9-10-10 but not by much

[Dwarden 1125]

this thread turned into another copy of http://forums.bistudio.com/showthread.php?147533-Low-CPU-utilization-amp-Low-FPS

where i posted that low utilization (not MT related) relates to some bugs which need fix first before the MT issues

so instead to have discussion about benchmarking , RAM and threading w/e. it's turning to complain fest about low FPS or else

also, what else is there to say than we know about it and working on it?

+ hint: see difference between

https://en.wikipedia.org/wiki/Software_release_life_cycle#Alpha

and

https://en.wikipedia.org/wiki/Software_release_life_cycle#Beta

[mobile_medic 43]

Please add it to the known issues, then.

Most of us are well aware that this is alpha, and if this were a brand new issue, that would be one thing. But, b/c this issue spans multiple iterations of the Arma series, personally, I at least hoped for some more detailed acknowledgement and recognition. It is concerning, to say the least, from this side of the fence to see passive recognition only, rather than acknowledgement that places it squarely where it belongs ("Top known issues") considering the (perhaps incorrect) assumption from the get go was that this would have been a primary issue to be addressed with any new engine. For all the improvements to game play, some of us only get a glimpse of that hard work, due to this long-standing issue.

For me. After so many years, I would expect at least "we know that this is a very important, outstanding issue. Please be assured that we are doing absolutely everything we can to resolve it. We will keep you up-to-date with our progress as we move forward in the Alpha. This will get fixed. We understand some of you may be concerned seeing this issue carry over into the A3 alpha, but rest assured it is at the top of our list, and we are working on it." The responses thus far, do not seem to appreciate the importance of this issue.

I look forward to being proven wrong.

I apologize for contributing to the derailment of this thread.

[dasa 20]

im not sure that its worth getting so worked up about

is it disappointing that its the same outdated engine with a few tweaks? yes but then there is so many other games out there doing the same thing and some even have the cheek to charge $90 while arma 3 is currently only $33

from what i have seen so far the changes and improvements being made will be worth the asking price

im just glad they are still making it for pc and not trying to turn it into something like operation flashpoint 2

clearly they have done some things right to have people so passionate about this game and how it runs :)

Edited April 22, 2013 by dasa

[Insanatrix 0]

Common misconception is that because people voice a negative criticism about a problem or flaw and then seek for recognition of the problem or flaw and a statement about said flaw, they are somehow terribly worked up or upset.

My issue is that after experiencing the same flaw/problem since ArmA 2, I really just want to know if this issue will ever be fixed basically. An actual statement about it that focuses on it rather than "Yeah we're working on it" and then nothing about it ever again.

[Tonci87 163]

Good job derailing a potentialy usefull thread....

[Forty40 10]

The biggest problem and you didn't added it to the know issues page?

Please add the issue to this page:

http://alpha.arma3.com/known-issues

...so most users will know that the issue is really recognized by the dev team.

Atm it really looks like you're trying to hide the problem in the hoping that we will forget about it.

[white 1]

this thread turned into another copy of http://forums.bistudio.com/showthread.php?147533-Low-CPU-utilization-amp-Low-FPS

where i posted that low utilization (not MT related) relates to some bugs which need fix first before the MT issues

so instead to have discussion about benchmarking , RAM and threading w/e. it's turning to complain fest about low FPS or else

also, what else is there to say than we know about it and working on it?

+ hint: see difference between

https://en.wikipedia.org/wiki/Software_release_life_cycle#Alpha

and

https://en.wikipedia.org/wiki/Software_release_life_cycle#Beta

that means this will apply?

"The alpha phase usually ends with a feature freeze, indicating that no more features will be added to the software. At this time, the software is said to be feature complete."

[Leon86 13]

Ive done a few runs of armamark as well with my i5-750 1024x768 settings mostly high, 2000 object viewdistance.

done 3 runs on each setting, discarding the first (first run always stutters a bit, stuff having to come off the harddrive, slightly less consistent)

Cpu 1800 MHz, ddr3-1600, ~2100 marks

Cpu 3200 MHz, ddr3-960 , ~3080 marks

Cpu 3200 MHz, ddr3-1600, ~3550 marks

Cpu 4000 MHz, ddr3-1600, ~4400 marks

I tried to run the same cpu / memory clocks to see what the blck speed does to performance but it wouldn't really let me, no real difference between 200*9 cpu + 200*3 and 150*12 + 150*4, 120*15 +120*5 wouldn't post.

Looks like there's some benefit to running decently fast memory compared to the slowest possible, but cpu clockspeed is more important. My memory (kingston valueram 1333) was running 9-9-9-24-1T timings (on 1.5V)

Edited April 22, 2013 by Leon86

[dnk 13]

@OP

Did you adjust timings for the memory with each increase or were they set at a single CAS/etc?

I was considering doing something like this myself, but sort of got tired of all this. Anyway, thanks for it!

@Dwarden, White, Insanatrix, etc

I've certainly noticed 1 AI underutilization bug (whatever it was, devs haven't explained it) that got fixed regarding AI in cities, so it's clear that these are at least part of the issue, but are you saying that all this time (since A1) that the AI issues have been due to these bugs, or has the engine improved somehow to better handle AI AND it has some new bugs?

I think if you could explain how the engine has improved regarding AI processing and bottlenecking and threading, it would clear things up (or not).

Edited April 22, 2013 by DNK

[Leon86 13]

^^^

he was running at the same timings, but he did a final run with 1333 tighter timings.

(Edit: Just tested 4.7GHz DDR1333 7-7-7-24-1T 38.9FPS)

[dasa 20]

Ive done a few runs of armamark as well with my i5-750 1024x768 settings mostly high, 2000 object viewdistance.

done 3 runs on each setting, discarding the first (first run always stutters a bit, stuff having to come off the harddrive, slightly less consistent)

Cpu 1800 MHz, ddr3-1600, ~2100 marks

Cpu 3200 MHz, ddr3-960 , ~3080 marks

Cpu 3200 MHz, ddr3-1600, ~3550 marks

Cpu 4000 MHz, ddr3-1600, ~4400 marks

I tried to run the same cpu / memory clocks to see what the blck speed does to performance but it wouldn't really let me, no real difference between 200*9 cpu + 200*3 and 150*12 + 150*4, 120*15 +120*5 wouldn't post.

Looks like there's some benefit to running decently fast memory compared to the slowest possible, but cpu clockspeed is more important. My memory (kingston valueram 1333) was running 9-9-9-24-1T timings (on 1.5V)

the faster the cpu the more starved for memory bandwidth it becomes i suspect a 3#70k would see slightly higher benefits than my cpu if at the same clock speed

upping the fsb use to bring decent gains back in the day it will be interesting to see if it makes any difference on haswell when it arrives around june as early info suggests it will have a greater range of bclk adjustment than sandy or ivy

[Leon86 13]

My 960 vs 1600 isn't really valid as nobody would be running 960 on with timings over 7. Anyway, I think my memory is "fast enough" for lynnfield, at least where arma is concerned.