Which games do take advantage of many CPU cores?

Q

Quartz-1

Supreme [H]ardness
Joined
May 20, 2011
Messages
4,257
Many games make use of only one or two cores. Some make use of four. DX 12 uses four. I think I've read that Ashes of the Singularity will use twelve. But which games make use of very many cores? IOW which games will take advantage of Threadripper, Epyc, Xeon, i9, etc?
 
When i play PUBG i see 10 of my 16 threads active at all times. Thats the only game ive played on Ryzen though. Came from a i5-3570k lol
 
Make_more_cores.jpg
 
Adding more cores is like adding more RAM. Essentially useless for gaming. No, a workstation CPU for encoding, heavy server uses and whatever else professionals may require won't work better for gaming. If AMD wants to beat Intel for gaming they need higher IPC and or higher GHZ. Ryzen works fine in most games, but the performance crown goes to the i5, and i7s can be helpful in a handful of games.

Going from a 4 core i5 to a 6 core 12 thread Ryzen gained me zero frame rates in a number of games.
 
How about Java-based games like Minecraft? Any reports Minecraft uses more than 4 cores?
 
drescherjm said:
Are they all over 50% most of the time??
Click to expand...
around that, Once the game loads fully it idles down to like 25%

Flogger23m said:
Adding more cores is like adding more RAM. Essentially useless for gaming. No, a workstation CPU for encoding, heavy server uses and whatever else professionals may require won't work better for gaming. If AMD wants to beat Intel for gaming they need higher IPC and or higher GHZ. Ryzen works fine in most games, but the performance crown goes to the i5, and i7s can be helpful in a handful of games.

Going from a 4 core i5 to a 6 core 12 thread Ryzen gained me zero frame rates in a number of games.
Click to expand...

while I do believe my fps has increased coming from a 4.5ghz 3570k, I dont have benchmarks to back up my "butt dyno" feelings. What I can say is that the smoothness of the game has increased by A LOT, Not to mention being able to alt tab and handle way more tasks while gaming was worth the purchase alone. While ryzen isnt the IPC king, I think everyone here knew that was going to happen. But now MOARRR CORESSSS is becoming much more accessible. I just got done building 5 new computers for myself and friends for Destiny 2. While i built them, they aren't uneducated. AMD got a high end enthusiast like me to switch and made my more casual friends tell me they wanted ryzen 5. While we loose 5 fps top end, Its a better experience overall.
 
Quartz-1 said:
Not if the game makes use of them.
Click to expand...

It has been explained to me a number of times that games are making very good use of multi core as is. Playing a game isn't like encoding a video, so more cores will never scale nearly as well on it. It will also very heavily from game to game. It would be like upgrading a GPU to make Excel run faster. The application type won't benefit much if at all from it.

Dan said:
while I do believe my fps has increased coming from a 4.5ghz 3570k, I dont have benchmarks to back up my "butt dyno" feelings. What I can say is that the smoothness of the game has increased by A LOT, Not to mention being able to alt tab and handle way more tasks while gaming was worth the purchase alone. While ryzen isnt the IPC king, I think everyone here knew that was going to happen. But now MOARRR CORESSSS is becoming much more accessible. I just got done building 5 new computers for myself and friends for Destiny 2. While i built them, they aren't uneducated. AMD got a high end enthusiast like me to switch and made my more casual friends tell me they wanted ryzen 5. While we loose 5 fps top end, Its a better experience overall.
Click to expand...

Nothing changed on my end running the same settings, although my RAM is limited to 2666 MHZ on my board as is. I doubt it will change much though with faster RAM. I did benchmarks and ran each test three times Across the board frame rates were essentially the same. 50 frame rates on my i5 looked just as smooth as 50 frame rates on Ryzen.

As for doing other things while gaming, I don't see the point. Outside of a browser I never leave things running in the background. If you're encoding a video (which can use all cores) your frame rates will likely be impacted. And if you're playing a game, you'll be distracted when you're minimizing. And don't do it during online games, because we all hate the guy who goes AFK and makes us loose the match. :p I see it the same way as those slobs that used to hog computers at the library years ago before I had a laptop. They'd sit there chomping away at food doing absolutely nothing for 1-2 minutes at a time, and they juggle their nasty crap in one hand and peck away at the keyboard with the other. They aren't getting anything done quicker, all they're doing is being nasty slobs. Mini rant thread diversion is officially over.

Back on topic, what we need from AMD now is big jumps in IPC performance and if possible higher MHZ numbers. Otherwise Intel will bury them in the coming months as they will be better in games and multi tasking for the same price ranges. With luck, AMD can bring Ryzen 2 fairly soon with such improvements so I can upgrade and hopefully keep the same AM4 motherboard. Otherwise, I am going back to Intel.

I only spend $100 in any case after selling my i5 platform. :)
 
That may be somewhat the case now, though it has been changing. However, with all of the major consoles having X86-based 8 core CPUs now, I think you're going to see much more adoption of multi-threaded techniques in use in PC games. Where it's been a very gentle upward slope until just recently, I think we'll see a steeper curve very soon. Just the simple fact that these games can be pretty easily ported now, is one factor. Also, the new APIs have been developed with this in mind. (DX12 / Vulkan) IMO you can count on the trend rising sooner than later.
 
In response to the PUBG comment, I don't experience much CPU utilization in PUBG. I have a 6-core and it typically uses about 20% or less of that.

I've experienced good utilization in Gears of War 4 and Civilization 6. Dx12 and should scale up to 6-cores better according to info released by Microsoft, but it doesn't look like it'll properly utilize much more than that. Although some games (Dx11 titles and OpenGL) touch a lot of cores and threads, they don't actually strongly utilize more than 3 cores as far as I've seen.
 
Rise of the tomb raider in DX12 will use all your cores and I think Watch Dogs eats up the CPU cores too IIRC.
 
Personally, I'd rather see lower % utilization over more cores than higher in many cases. Sure, it means you CPU isn't being pushed fully, but that could mean a lot of things. What it does mean (unless you're running a heavily threaded workload in the background) is that the game is using multiple threads, and is spreading them out as it should. This means that if you get a spike in activity from the game in a CPU heavy situation, you have plenty of "head room" and also if you get a spike of OS or other background activity, it's not going to impact the game. If the game is using 95% on all cores, then if something spikes, you'll notice it in the game. (whether it came from the game of another task) I guess for me the ideal situation would be to see about 50-60% usage on all cores in a demanding game. That way you have room for more, but the game is taking advantage of available resources as well.

If a skilled dev can push things and make a noticeable improvement in a game, then that's great. However, I think we're just seeing what current games actually need of a CPU. As long as we're moving away from pure IPC on a single or maybe two cores, and seeing things spread out over more (without even necessarily taxing all of them) we're moving in the correct direction. Game AI, physics, etc. will eventually advance to where we'll see overall utilization per core go up, but that may take longer. (most devs don't want to leave the consoles in the dust) Maybe the new XBox will help push that line up a little bit, but who knows.

Right now, I'm typically happy with better GPU utilization (assuming the game is visual-heavy I suppose), and moderate multi-core usage.
 
Dan said:
When i play PUBG i see 10 of my 16 threads active at all times. Thats the only game ive played on Ryzen though. Came from a i5-3570k lol
Click to expand...

I assume you are meaning logical cores and not threads here as it seems like a common misconception.
Looking at how many of your (logical) cores are being utilized foes not mean it actually scaling across that many cores.
This is due to the simple fact that you are measuring over a time interval and that a thread ( real threads) can easily jump between (logical) cores in that time interval
Aka 1 thread running at 100% load can easily looks like 4 (logical) cores being used 25%, it does not mean it actually benefits form all 4 (logical) cores
This is exactly the same mistake Kyle did when testing Doom when it first came out for scaleabilty

To this check these kinds of things uses process explorer and check the actual threads of the game.



@op
Doom2016 Vulkan mode easily uses 8 main threads with pretty close scaling. and probably more if i had more (logical) cores to test with.
 
Last edited:
SvenBent said:
I assume you are meaning logical cores and not threads here as it seems like a common misconception.
Looking at how many your (logical) cores are being utilizied odes not mean it acttally scaling across that many cores.
this is the smipe lfacts that you measure over a time intervall and that a thread ( real threads) can eassily jump between (logical) cores in that time interval
Saa aka 1 threads runnig at 100% load vcan eassily looks like 4 (logical) cores beeing used 25% it does not mean it actually benefits form all 4 (logical) cores
This is exactly the same mistake Kyle did when testing Doom when it first came out for scaleabilty

To this check these kinds of things uses process explorer and check the actual threads of the game.



@op
Doom2016 Vulkan mode easily uses 8 main threads with pretty close scaling. and probably more if i had more (logical) cores to test with.
Click to expand...

I'm guessing that any Tech 6 game would do the same, so we can expect the same from Quake Champions too.

Edit: Just realized that QC doesn't run on a full featured Tech 6 engine, so it may actually behave a bit differently than Doom. The new Wolfenstein game is Tech 6 though...
 
Last edited by a moderator:
Dan said:
When i play PUBG i see 10 of my 16 threads active at all times. Thats the only game ive played on Ryzen though. Came from a i5-3570k lol
Click to expand...

Here is the count of heavy threads in games i've test so far

BF1 : 5
CS:GO : 2+2+2 4+
Doom 2016 (vulkan): >8
Killing Floor: 1-2
League of Legends: 2
Lords Of The Fallen: 4+
Path of Exiled : 2
Skyrim SE: 1+
WarFrame: 3
World of Warcraft: 1+
UT2004: 2
XCOM 2008 demo: 3
Wolfensteain new order: 4+


cs:go is a bit weird. it has 2 main threads then up to 2 that has abit lesse utilization and then sometimes up to another 2 that has a bit lesse again. however at no points do it have more than 4 heavy threads hence why 4+ cores are recommende ( 4 for the ehavy onw and maybe HT or another cores for the lesser ones)

+ simple means ther is a lot of extra tiny threads that could maybe used an extra core
 
Looks like Tech 5 is pretty threaded too then. (Wolfenstein) That's actually what I'd expect, but hadn't actually checked it before. Actually curious about Tech 4 now too. I know they were doing some multi-threading as early as Q3A, but it didn't take off until later. Think I'll give Doom 3 and Quake 4 a quick try when I get a minute. You know... for science!
 
J3RK said:
Looks like Tech 5 is pretty threaded too then. (Wolfenstein) That's actually what I'd expect, but hadn't actually checked it before. Actually curious about Tech 4 now too. I know they were doing some multi-threading as early as Q3A, but it didn't take off until later. Think I'll give Doom 3 and Quake 4 a quick try when I get a minute. You know... for science!
Click to expand...
I remember messing with R_SMP bakc on my abit BP6 system but i beleive it was unstable and actually got removede in later patches (as well as the hidden software rendering)

If i recall correctly when i teste ?Quake3 arena it was simple never more than one heavy thread even in R_SMP 1 mode but that is running a updated version.

i'll look into it tommorow

-- update --
Image1.png


Quakearen in 1.27g3 is very close to be 100% single threade
teste with lightmpa or vertext lignings
full screen vs windows mode
as well as arena and team arena mode


wht we see he is a single main threads that is using 98.4% of a cores (remember the states are based of 8 logical cores in this instance so we multiply with 8 to get i for one core)
and a secodanry threads at only 12.48% so there is a small gain using a dual cores but hardly an important one

in short: this game needs 1+ cores
 
Last edited:
Dan said:
around that, Once the game loads fully it idles down to like 25%



while I do believe my fps has increased coming from a 4.5ghz 3570k, I dont have benchmarks to back up my "butt dyno" feelings. What I can say is that the smoothness of the game has increased by A LOT, Not to mention being able to alt tab and handle way more tasks while gaming was worth the purchase alone. While ryzen isnt the IPC king, I think everyone here knew that was going to happen. But now MOARRR CORESSSS is becoming much more accessible. I just got done building 5 new computers for myself and friends for Destiny 2. While i built them, they aren't uneducated. AMD got a high end enthusiast like me to switch and made my more casual friends tell me they wanted ryzen 5. While we loose 5 fps top end, Its a better experience overall.
Click to expand...

When I redid my desktop, I actually dropped cores - going from Q6600 to G3258 (four cores to just two), and I only took it in the neck in maybe TWO games in my rota at the time. I've changed the rota - however, rather surprisingly, the lack of cores is not a factor in it. That was, however, two years ago. Since that core-halving, I've done only one other hardware upgrade - and that wasn't the CPU, either - instead, it was a GPU upgrade (from GTX550Ti to GTX 1050Ti) and the on-GPU memory (from 1 GB to 4 GB). That impact (the GPU upgrade) IS a shocker - performance, visually and otherwise - is up across the board. In fact, it's up compared to the Q6600, despite having two fewer cores. That goes so far against the supposed grain, that I have to wonder what the grain really is.
Last night, I continued my revisit of games that I had dropped when I did my core-account halving with Crysis 2 Maximum Edition (which dates back TO said Q6600 - in other words, it sat on a shelf since). Where would the '58/Pascal tag-team find itself? (Yikes - I need new socks.) Firewalled - Ultra everything. On a dead-stock '58. Because I have been unable to find the last version of the Adrenaline Crysis 2 benchmark tool, I've been unable to properly benchmark the game, I can't explain how that happened - however, according to "conventional wisdom" it certainly wasn't supposed to. Basically, so much for conventional wisdom. It points to what I have been seeing in every DX11 title I have revisited this week - the GPU in DX11 is not merely a significant factor, but may actually be the MOST significant factor in terms of performance. In other words, unless you are foreswearing DX11 for DX12 for most of your gaming, if you have a GPU will less than 4 GB of RAM, upgrade that before doing any other non-SDD upgrades - your games WILL thank you, and so will your eyeballs.
 
J3RK said:
That may be somewhat the case now, though it has been changing. However, with all of the major consoles having X86-based 8 core CPUs now, I think you're going to see much more adoption of multi-threaded techniques in use in PC games. Where it's been a very gentle upward slope until just recently, I think we'll see a steeper curve very soon. Just the simple fact that these games can be pretty easily ported now, is one factor. Also, the new APIs have been developed with this in mind. (DX12 / Vulkan) IMO you can count on the trend rising sooner than later.
Click to expand...

DX12-based games all have at least an i5 as the stated minimum (the sole exception being Forza 6: Apex) - the likely impetus of THAT means that developers leveraging either API are likely developing using 6-core or better coding hardware - which IS more plentiful today than it was when DX11 and the engines that leveraged that API existed. It's more common in consumer hands as well - what is the versus number in terms of core counts not in terms of AMD vs. Intel, but within each company's CPU lineup (intra-Intel and intra-AMD, in other words)? Back in the tailchase of DX11, there WAS AMD's Phenom and Phenom II - the original six-core consumer-facing beastie - however, what was prevalent in terms of developer hardware? Segue to now - developer hardware and target hardware are more closely matched than ever (except at the very floor of the consumer space). Unusually, the floor of the consumer space (by core count) is not Android (not with ARM's big.LITTLE having been 4+-core for a decade plus), but Windows (dual-cores and APUs). Exactly what games are targeting that floor, for the most part? They aren't considered "mainstream" games - not by any stretch.. The mainstream games are definitely multi-core (four cores or better) and this is across DX12, Android, consoles, etc. (Injustice 2 is easily THE example of the current generation - it's out on consoles, and on Android right now, with iOS and a Windows port dead-last. I've been playing it on Android (no console in the house - and my desktop is dual-core); you WILL need an i5 or Ryzen 5 or better to play it on Windows. Still, except for screen-size, the game - even on Android (Samsung Galaxy S7), is a core-feaster - I wouldn't DARE think of throwing in on my Galaxy Nexus (which is my default phone). Basically, Ryzen 5 targeted DX12.
 
I seem to remember BF4 utilizing all cores at 70% utilization or higher on my old fx8320 4.5GHz @1440p, maxed settings on a 64m large maps game.
 
alxlwson said:
I seem to remember BF4 utilizing all cores at 70% utilization or higher on my old fx8320 4.5GHz @1440p, maxed settings on a 64m large maps game.
Click to expand...
Please understand form above explanation why that does not really give any exact information about how many cores it can utilize though.

if it did use 70% on all cores over the same measuring interval it just means it has at least 6 threads that uses the CPU. (including all background load) but i have a hard time picturing you would have exactly 70% on all cores cause it very rarely line up that perfectly and nice.
 
SvenBent said:
Please understand form above explanation why that does not really give any exact information about how many cores it can utilize though.

if it did use 70% on all cores over the same measuring interval it just means it has at least 6 threads that uses the CPU. (including all background load) but i have a hard time picturing you would have exactly 70% on all cores cause it very rarely line up that perfectly and nice.
Click to expand...


Who said anything about "exact"? I know I didn't.
If I'm in my desktop doing nothing, and my cpu time is essentially idle, and then I turn on bf4 in the configuration that I mentioned above, and task manager shows bf4.exe to be 70%+ and that the logical cpu graph indicates 70%+ on each of the 8 cores... Is this difficult to understand?
Knowledge and reading comprehension is key.

Someone correct me if I'm wrong, but if a program has eight threads, it can utilize eight cores. Windows handles the core/thread scheduling, not the game engine.
 
Last edited:
Dan said:
around that, Once the game loads fully it idles down to like 25%



while I do believe my fps has increased coming from a 4.5ghz 3570k, I dont have benchmarks to back up my "butt dyno" feelings. What I can say is that the smoothness of the game has increased by A LOT, Not to mention being able to alt tab and handle way more tasks while gaming was worth the purchase alone. While ryzen isnt the IPC king, I think everyone here knew that was going to happen. But now MOARRR CORESSSS is becoming much more accessible. I just got done building 5 new computers for myself and friends for Destiny 2. While i built them, they aren't uneducated. AMD got a high end enthusiast like me to switch and made my more casual friends tell me they wanted ryzen 5. While we loose 5 fps top end, Its a better experience overall.
Click to expand...

I think some of that smoothness comes from reinstalling a pc to be honest. My 5960x is on a two year old install right now, and I just put a 980ti in the old 4770 and reinstalled, and the 4770 feels buttery in comparison.
 
alxlwson said:
Who said anything about "exact"? I know I didn't.
If I'm in my desktop doing nothing, and my cpu time is essentially idle, and then I turn on bf4 in the configuration that I mentioned above, and task manager shows bf4.exe to be 70%+ and that the logical cpu graph indicates 70%+ on each of the 8 cores... Is this difficult to understand?
Knowledge and reading comprehension is key.

Someone correct me if I'm wrong, but if a program has eight threads, it can utilize eight cores. Windows handles the core/thread scheduling, not the game engine.
Click to expand...

Sure someone has an attitude...
But yes it difficult to understand "70%+" when you say "70%" because they are not mathematically the same.
About comprehension is key... You are sure right on that one, including mathematics and what the different signs means.


Also about correcting you. if you read above as is suggest you would know what that is not entirely true in the way you represented your data. So you where already corrected but that only got you pissed for some weird reason.
It does NOT need 8 threads to utilize 8 cores at 70% each it only takes a minimum of 6.
 
Last edited:
SvenBent said:
But yes it difficult to understand "70%+" when you say "70%" because they are not mathematically the same.
About comprehension is key... You are sure right on that one, including mathematics and what the different signs
Click to expand...

alxlwson said:
I seem to remember BF4 utilizing all cores at 70% utilization or higher on my old fx8320 4.5GHz @1440p, maxed settings on a 64m large maps game.
Click to expand...
 
You are absolute correct I did missed that

Conclussion is still that it only prove a minimum of 6 heavy CPU threads so only 6 cores gives extra performance. it might be more but its not proven.
 
SvenBent said:
You are absolute correct I did missed that

Conclussion is still that it only prove a minimum of 6 heavy CPU threads so only 6 cores gives extra performance. it might be more but its not proven.
Click to expand...


How you recon only six threads?
 
alxlwson said:
How you recon only six threads?
Click to expand...
I already explained that earlier in this thread But you got so upset when i directed you towards the information. So I don't really want to.
and I'm not saying only 6 threads. I'm saying it only proves 6 threads, It can be more but then number doesn't prove anymore. Why this way is not in anyway a good way to measure scaleabilty and you should use process explore to see threads loads instead.

-- edit --

I gave a pretty long explanation but then it f#cked up. so her is the short version

1 threads does not maintaig usage on only one core over time. and you measure CPU usage over time

so here is an example if 1 threades looking liek 7 threads in your way of measuring stuff

Core_STuff.png


As you can see we have utilization on 7 cores ( 6 with 12.5% 1 with 25%)
but its only one thread running at 100% of whatever core is handed to it by the CPU scheduler in windows

We know one thread can only go to 100% of a core over time. so by looking at you 8cores of 70% you have 560% aka you need at least 6 threads to hit a bove 500%... it can be more it could be 16 threads using 35% each or 56 threads using 10% each or any combination whatsoever.

But the only thing we know for sure is at least 6 threads


Your way of thinking is a common misconception of how threads,multicore quanta and the CPU scheduler works. Heck even Kyle got it wrong when checking Doom 2016 for CPU scaleabilty.


now you can try it yourself bu running any multithread software that is cpu heavy and you can control the threads.
7-zip is very nice for this. set up the benchmark for 6 threads and you will around 70%+ on all core. not just 6 cores running at 100%

Cinebench can not be used for this since cinebench set affinity when you start benchmarking


--- edit ---
link to another thread where i show the difference
https://hardforum.com/threads/which...wares-to-lag-frozing.1941822/#post-1043172314
her again you can se the 7zip bencmhkar with only 2 threads putting load on all cores acocrding to taskmanager. and why you need to use process explorer to really determine core scaleability
 
Last edited:
You must log in or register to reply here.
Back
Top