Single Thread or Multi Thread?

AdrianJaden

Distinguished
May 25, 2015
132
0
18,680
What do games mostly use?

Ex: GTA V, BF4, Minecraft, CSGO, etc.

I am planning to see which games use those threads. I am stuck between an i3 and an FX 6350. 6350 has a better multi thread score while the i3 has a better Single thread score. I can't decide which works perfect with these games.
 


Oh. Most games use more than one, but the a lot of load is often on one core more than another. Some games (rarely) still use one. Sorry for that misunderstanding 😉

However, not many games use more than 4, 4 is usually the most but some have reached 6.
 
It would help to look up cpu benchmarks for the games you're interested in. Most of the time there are fairly thorough (several cpus) tested against one another. Even if a program is 'multithreaded' it doesn't necessarily mean the cpu with more cores/threads will be the better performer. That only works if comparing processors with the same architecture, aka a 6350 vs an 8350 since they're both fx. If you knew for sure a program was heavily multithreaded (used more than 6 threads), the 8 core/8 thread would likely give more performance. Comparing amd and intel doesn't always work, not just for clock speeds but thread/core count. Intel cpu cores get more work done faster, more efficiently which is how even in heavily multithreaded scenarios a 4c/4t i5 easily keeps up with (if not surpasses) an 8c/8t fx.

The same holds true for the i3 vs 6350.

For instance, here's bf4 multiplayer benchmarked. The i3 even gets higher fps at 1920x1080 than the fx 8350, much less the 6350.
http://www.hardwarepal.com/battlefield-4-benchmark-mp-cpu-gpu-w7-vs-w8-1/

Yes with a proper motherboard and aftermarket cooler the 6350 can be overclocked. Factor in the price of an aftermarket cooler and you've easily got $140-150 sunk into the 6350 just to try and meet or slightly beat the i3. Toss another $20 onto that cost, get a locked core i5 and get significant performance gains.

I figured this was a good example of a multithreaded game in a multiplayer scenario to show how the cpus really stack up. Minecraft is another title where intel cpus really excel since it loves strong fast cores. It's a very cpu heavy game, something amd isn't equipped for.

Cs:go isn't all that intensive, many people play with high frame rates on mobile processors.
http://www.notebookcheck.net/Counter-Strike-Global-Offensive-Benchmarked.81183.0.html
 


For the games I listed, What thread do they use?
 
As far as I am aware most games are single threaded. Some new games do take advantage of multiple cores but I don't think that means they are multi threaded.

My best suggestion is to look at the specs of the game and then see if the cpu's you have listed can play the recommended specs. If it can, then you should get smooth gameplay without changing the default settings in the game. Otherwise it means that they are too weak and you will need to start changing the settings to get better framerates.
 


So I would mostly play demanding games like BF4 and GTA V. Since they are multithreaded, Should I get the FX 6350? I will mostly play those games.
 


But what's important is the distribution of those threads on various cores.
 
That's why I posted the links to the benchmarks. It clearly shows the i3 performing better than the fx even in 'demanding multi threaded' games. So looking at the benchmarks, if the i3 in many of the games you listed outperforms the fx 6350, should you get the fx? It's up to you. That's what benchmarks are for, to get around the 'theory' and show what does what. Someone could argue all day long that a game is multithreaded so therefor an fx chip is better since it has more threads. That contradicts the outcome though when a simple dual core 4 threaded cpu like the i3 outperforms it.

Without getting bogged down by details and theory, benchmarks help show real data - how many fps is each one producing? Worrying about how many threads a game does or doesn't use can become irrelevant. This is precisely why synthetic benchmarks like passmark don't mean much. Passmark isn't a real program, bf4 is a real program. According to passmark, the fx 6350 has a higher score than the i3 4160 - so why does the i3 outperform the fx in bf4? If someone wants to produce great synthetic benchmark scores all day long then the fx 6350 is perfect. It doesn't really amount to much though in real world use.
 


As I've been explaining here for over 5 years now, this is EXPECTED.

Using more threads doesn't magically give you a free performance benefit; this only occurs if you have removed a CPU bottleneck that is otherwise lowering application throughput. If no bottleneck exists, you get next to no benefit by threading in applications not dominated by CPU performance. Games, for example, are driven by GPU performance, and as long as the CPU is working fast enough to keep the GPU fed, there is no CPU bottleneck, and the CPU has very little performance impact on the final result, with the only performance differences being driven by single-core throughput, which favors Intel over AMD. That's why the Core i3 performance equally to the FX-8000 series in games most of the time.

Now, core count would have a significant impact in latency based statistics, since, using the Core i3 as an example again, you'd have more threads waiting before they get executed. And if you delay the GPU render thread, it's going to result in some wild latency swings due to the GPU stalling out. So while the Core i3 may give better pure FPS by virtue of superior single-core performance, it will often have significantly worse latency due to the GPU thread having to wait, causing the game to miss its 16ms refresh window, delaying the frame output until the next refresh.

This is the main reason why I believe lower level API's like DX12 are going to heavily bias performance toward Intel, since the Core i3 and Pentium lineups in particular is going to benefit the most from lower CPU performance requirements.