[gamegpu.ru] CPU scaling in 2015 release Games

[AtenRa]

The results doesnt portrait a good future for Dual core (With or not HT) CPUs. More and more games use higher core count and we havent even seen DX-12 games yet.

AMD FX8350 does extremely well for its price and the old Sandybridge Core i7 2600K is still one of the best for its age. The Core i5 2500K even at default clocks is more than enough for the majority of games, even today 5 years after its release.


http://gamegpu.ru/action-/-fps-/-tps/the-witcher-3-wild-hunt-test-gpu-2015.html

http://gamegpu.ru/action-/-fps-/-tps/dying-light-test-gpu-2015.html

http://gamegpu.ru/action-/-fps-/-tps/grand-theft-auto-v-test-gpu-2015.html

http://gamegpu.ru/action-/-fps-/-tps/star-wars-battlefront-test-gpu-2015.html

http://gamegpu.ru/action-/-fps-/-tps/fallout-4-test-gpu-2015.html

http://gamegpu.ru/action-/-fps-/-tps/assassin-s-creed-syndicate-test-gpu-2015.html

http://gamegpu.ru/racing-simulators-/-gonki/project-cars-test-gpu-2015.html

http://gamegpu.ru/action-/-fps-/-tps/just-cause-3-test-gpu-2015.html

http://gamegpu.ru/action-/-fps-/-tps/metal-gear-solid-v-the-phantom-pain-test-gpu-2015.html

http://gamegpu.ru/action-/-fps-/-tps/call-of-duty-black-ops-iii-test-gpu-2015.html

 

[TheELF]

witcher 3 min FPS
fx-6100 35 = 6 cores or 3 depending on who you ask
i3-4330 40 = 2 cores
fx-6300 41 = as fx-6100

Not exactly scaling and it's the same for pretty much any game,it has nothing to do with core count.

 

[wege12]

I think it was pretty clear to most that this is how the industry is/will be moving. It will be extremely interesting to see the results with some DX12 games. Glad I purchased the i7-5820k (oc'd to 4.6ghz) instead of the i7-4970k.

 

[AtenRa]

witcher 3 min FPS
fx-6100 35 = 6 cores or 3 depending on who you ask
i3-4330 40 = 2 cores
fx-6300 41 = as fx-6100

Not exactly scaling and it's the same for pretty much any game,it has nothing to do with core count.

Witcher 3 min FPS

i3 4330 = 40fps
i5 2500K = 46fps
i7 2600K = 58fps

 

[ShintaiDK]

The results doesnt portrait a good future for Dual core (With or not HT) CPUs. More and more games use higher core count and we havent even seen DX-12 games yet.

AMD FX8350 does extremely well for its price and the old Sandybridge Core i7 2600K is still one of the best for its age. The Core i5 2500K even at default clocks is more than enough for the majority of games, even today 5 years after its release.

First of all this belongs to the CPU section.
That "dual core" is constantly beating 6 to 8 of the "other" cores.

Tell me why this GTA V is different from the one you posted.

Fallout 4 when its CPU demanding looks like this.

And so on and on. You can try cherry pick all you want, but the FX is still crap anno 2015 and will also be in 2016.

 

[Innokentij]

The AMD cpu's sure are weak both compared to clock for clock and even when they have more cores. Hoping next gen will be more competitive, not that im gonna buy em but cheaper intel CPU for me.

 

[AtenRa]

Tell me why this GTA V is different from the one you posted.

The one you posted is from the release date of April 2015, the one I posted is from December 2015.

 

[desprado]

sad thing for AMD is that even 2500k is better than best best CPU in most CPU demanding games like GTA V, Dying light, fallout 4 and star wars.

 

[frozentundra123456]

Game.gpu used to by my go-to source for game benchmarks, but they seem to be falling behind. 4770k is the newest quad intel? They should have skylake with fast ram and 4790k or at least 4770k overclocked to 4790k speeds.

 

[el etro]

Thanks to the consoles. DX12 may help even more.

AMD did this.

 

[Yuriman]

I have to agree with some other posters here - it looks like the i3 4xxx is doing quite well. It's generally somewhere between an FX-6300 and an FX-9xxx in performance, averaging between the FX-81xx and FX-83xx. If you're not overclocking those AMD chips, you're just as well off with a Haswell i3, based on the benches you posted, and better off if heat/noise/size/power are a concern.

Sandy Bridge i3's are showing their age, but HT has only become better with each generation, causing i3's to close the gap with i5's somewhat; it's not a simple clockspeed/IPC increase. Also, the i3 2100 was not slotted against the FX 8 cores when it was released in 2011. It predates the FX-4100 by 6 months, was similarly priced, and still comes out ahead.

 

[Sweepr]

Game.gpu used to by my go-to source for game benchmarks, but they seem to be falling behind. 4770k is the newest quad intel? They should have skylake with fast ram and 4790k or at least 4770k overclocked to 4790k speeds.

Agreed, 2013 Haswell is old news so these results don't reflect how Intel and AMD lineups compare right now. Looking at Haswell-based Core i3-4330, I have little doubt Skylake Core i3 would outperform FX6300 and even FX8350 in these game results, so it looks like dual-cores are not that bad compared to AMD's current lineup despite OPs efforts to prove otherwise.

Also whoever believed that FX8150 would be the better choice in this generation (X1/PS4-titles) must be feeling bad right now, Core i5-2500K is still faster with only 4C/4T 5 years after launch.

 

[Mercennarius]

More cores/threads will become increasingly more important especially with DX12. I posted a few weeks ago showing Ashes of Singularity was using 12+ threads on my system.

 

[skipsneeky2]

You could nearly double your fps on that i3 if you drop enough settings in GTA V.The game maxed out chokes the hell out of a i3.Might not look as good but you can sure get away with much cheaper hardware to run the game.

I had posted some numbers in the GTA V cpu thread how bad a i3 3220 chokes on max settings vs a i7 3770,it was a 50% bottleneck with a gtx 970 at 1080p.At low the frames were perfectly playable.

 

[TheELF]

More cores/threads will become increasingly more important especially with DX12. I posted a few weeks ago showing Ashes of Singularity was using 12+ threads on my system.

Did you enter the game at all and see that it only ran on one core?
Gpu benchmarks scale very well with threads, game engines do not!

You only have to look at Crysis 3 welcome to the jungle to figure out why devs don't use multi threaded rendering in games,it sucks big time!
Very low minimums, and very frequent ones as well,that only get amplified by the high max framerates due to multithreaded rendering.
http://www.tomshardware.com/reviews/crysis-3-performance-benchmark-gaming,3451-8.html
Giving you huge jumps in framerate unless you do a fixed camera game with it like ryse did.
Look at that, 100 fps difference between min and max FPS.
https://www.youtube.com/watch?v=lEafN99Q4K4

 

[Mercennarius]

Did you enter the game at all and see that it only ran on one core?
Gpu benchmarks scale very well with threads, game engines do not!

You only have to look at Crysis 3 welcome to the jungle to figure out why devs don't use multi threaded rendering in games,it sucks big time!
Very low minimums, and very frequent ones as well,that only get amplified by the high max framerates due to multithreaded rendering.
http://www.tomshardware.com/reviews/crysis-3-performance-benchmark-gaming,3451-8.html
Giving you huge jumps in framerate unless you do a fixed camera game with it like ryse did.
Look at that, 100 fps difference between min and max FPS.
https://www.youtube.com/watch?v=lEafN99Q4K4

Yes I did. The game can use 12+ threads in game.

DX12:

DX11:

DX11 and DX12 can both utilize multiple threads in game, though DX12 is significantly more efficient when spreading the work load. DX11 effectively relies on 1 main thead while DX12 can spread the work load much more evenly.

 

[TheELF]

Yes I did. The game can use 12+ threads in game.

What you show are not cores in use but core usage OVER TIME,there is a very big difference.

Look at post #7
http://forums.anandtech.com/showthread.php?t=240423

 

[Mercennarius]

What you show are not cores in use but core usage OVER TIME,there is a very big difference.

Look at post #7
http://forums.anandtech.com/showthread.php?t=240423

Notice the workload allocation across threads between DX11 & DX12. On my system when I limit the game to only 6 or fewer threads I begin losing performance. The game has increased performance with more threads available.

 

[inf64]

Notice the workload allocation across threads between DX11 & DX12. On my system when I limit the game to only 6 or fewer cores I begin losing performance. The game has increased performance with more threads available.

Notice that your overall CPU utilization is around ~20% or so, which effectively means that only 1/5 of your 24 logical threads are doing work all the time while the rest are doing nothing. Or in other words, only ~5 threads are fully active all the time out of 24.

 

[Mercennarius]

Notice that your overall CPU utilization is around ~20% or so, which effectively means that only 1/5 of your 24 logical threads are doing work all the time while the rest are doing nothing. Or in other words, only ~5 threads are fully active all the time out of 24.

That's not how that is calculated. I could have all 24 threads being utilized at 10% each and my overall utilization would be only 10%. I could also have two threads at 100%, one at 40%, and the rest at 0% and my overall utilization would still be 10%. In the screenshot I posted i'm only using 20% of my total CPU workload potential, but it does not mean only 20% of the threads are active. If that were the case you would see spikes or drop outs on each of graphs.

 

[poofyhairguy]

Wow, I feel pretty vindicated upgrading my 2500k to a 2600k last year instead of a platform upgrade to a midrange Haswell.

 

[Jacky60]

These benchmarks just show how piss poor AMDs have become since their heyday in the early 2000s. They are OK for web surfing and watching video but I'd never buy or recommend one for gaming. I hope this changes but doubt it well with them so starved of resources.

 

[AtenRa]

That's not how that is calculated. I could have all 24 threads being utilized at 10% each and my overall utilization would be only 10%. I could also have two threads at 100%, one at 40%, and the rest at 0% and my overall utilization would still be 10%. In the screenshot I posted i'm only using 20% of my total CPU workload potential, but it does not mean only 20% of the threads are active. If that were the case you would see spikes or drop outs on each of graphs.

That is correct, the game uses more than 5 threads but each thread is not using 100% of each Core performance.

Here are a few examples, FX8350 + HD7950 on RYSE Son Of Rome in DX-11 and Dragon Age Inquisition with Mantle.

Just because the CPU monitor is not at 100% doesnt mean it only uses 2-3 threads.

 

[TheELF]

That's not how that is calculated. I could have all 24 threads being utilized at 10% each and my overall utilization would be only 10%. I could also have two threads at 100%, one at 40%, and the rest at 0% and my overall utilization would still be 10%. In the screenshot I posted i'm only using 20% of my total CPU workload potential, but it does not mean only 20% of the threads are active. If that were the case you would see spikes or drop outs on each of graphs.

Exactly, and task manager does not tell you which case is true...
The only sure thing from your pics is that Dx12 uses 21% at a throttled 2,91Ghz while Dx11 uses 18% at 3,36Ghz.

If you want to then run ashes in a smaller window with process hacker in the background,double click the games exe in hackers window and choose the "threads" tab this will show you threads in real time and not the juggling that windows does.

 

[TheELF]

That is correct, the game uses more than 5 threads but each thread is not using 100% of each Core performance.

Here are a few examples, FX8350 + HD7950 on RYSE Son Of Rome in DX-11 and Dragon Age Inquisition with Mantle.

Just because the CPU monitor is not at 100% doesnt mean it only uses 2-3 threads.

Jup and it also shows why the 8core has such low performance...
1 out of 8 cores is 12,5% and in both games you can see one thread reaching close to that,which is the maximum,keeping the rest of the threads from doing more work.
Getting ~50% out of a CPU in multithreaded games....