Gaming performance between 4300 & 8350 and 4-threaded performance between AMD & Intel

[Faa]

Yeah we know there are 5 games atm that take advantage of 6-8 cores (next time don't copypaste the recommended requirements) where a 8350 would do much better than a 4300. We know a single module has only one FPU (>.>), so the difference is 2 vs 4 FPU's between a 4300 & 8350 so that's twice the throughput of a potential of having 100% more 4-threaded performance which could end up having twice as much FPS. Except that we are dealing with a FlexFPU that would work as a single 256bit with a single thread or split in two 128 bit FPU's with 2 threads. Now the difference between a 128 bit FPU & 256 bit FPU is almost nonexistent in most workloads and definitely games - saying there's just no difference between a 4300 & 8350 in every game out there except a few. With just 2 threads between a 8350 & 4300, there's just no difference unless you're a troll because we got 2 modules that can each take a thread. The advantage of a 8350 coming with 4 front-ends vs 2 front-ends starts from 3 threads but I'm limiting it to 4 threaded performance because anything up isn't relevant. The main reason why you're getting a slight performance advantage between 10-20% is due to that each thread can have her own dedicated front-end rather than 2 threads sharing each the same front-end that could potentially lead into the front-end bottlenecking (yes AMD front-ends are garbage).

Benchmark time.

Type of benchmark that's irrelevant here;



As you can see all CPU's here are performing equally because the GPU was the bottleneck and at that point the CPU is irrelevant and therefore it doesn't prove a 4300 performs as good as a 8350 or the 8350 is better than the 4300. What we should focus on having all CPU's being the limiting factor so we can see the difference between x & y CPU. Starting off with a game that's only taking advantage of 2 cores and no more;



We see here a large difference between a 8350 & 4790K definitely proving a 8350 was bottlenecking so we can start comparing the performance between a 8350 & 4300. No difference. Moving on to a 4-threaded game;



10% for like 150$ more. WP, amazing value. Pay more for more cores for a better price/performance.



I assume this is a game that takes advantage of 4 cores as the difference between the i3 & 4690K was large; no difference again.





No difference.



No difference.

http://www.dsogaming.com/editorial/report-despite-claims-most-pc-games-are-still-unable-to-take-advantage-of-more-than-4-cpu-cores/

Dude showed his CPU usage so posting benchmarks from the ones I could find;



All of them at the bottom.



Just to put in a conclusion; the 8350 is an overpriced 4300. That FPU theory is flawed. If you are wondering the difference between a 4670K/4300, look up at synthetic benchmarks between the two to find out what the 4-threaded performance difference is as most games only take advantage of 1-4 threads and most reviewers testing Intel with the GPU being the bottleneck not even showing the full potential. The i5 easily gets twice the performance a 4300 can provide; http://www.anandtech.com/bench/product/1261?vs=700 

Don't buy a 8350. It's nothing more or less than an old 2005 low-end CPU offering the worst gaming value (aside from the i7's). 
 

[SoftenButterCream]

Yes, but the 8300s is good since it has weaker ipc, and most people get it for video editing on a budget.

[SoftenButterCream]

fucking mobile.. Double post

[WinNut]

inb4 total blue vs red shitwar

[Imabigmac]

Well this should be interesting... 

 

 

inb4 total blue vs red shitwar

No because it is not a Intel vs AMD thread, it is a AMD vs AMD thread. The issue if any which we both think there will be is, that the 8350 was just shot down and a lot of people stand by it. 

[Fate]

A thing I thought was funny, that I mentioned in another thread before Faa made this and I'll mention it again, is that pretty much everybody on this forum would agree that an FX-4300 flat out bottlenecks a 970 in the very large majority of games, yeah? But not everybody agrees that an 8350 bottlenecks a 970. Why?

We clearly see the FX-4300 performing very similar in most games to a FX-8350, yet apparently the 8350 is much better some how. 

[SoftenButterCream]

A thing I thought was funny, that I mentioned in another thread before Faa made this and I'll mention it again, is that pretty much everybody on this forum would agree that an FX-4300 flat out bottlenecks a 970 in the very large majority of games, yeah? But not everybody agrees that an 8350 bottlenecks a 970. Why?

We clearly see the FX-4300 performing very similar in most games to a FX-8350, yet apparently the 8350 is much better some how.

Exactly, no point in pairing a old cpu with weak ipc, with high end graphics if the cpu itself can't handle it. Its good for editing though

[Faceman]

Why haven't moderators helped clean up all the garbage people spew recommending FX processors when hard evidence exists to show how bad of value it really is. If you already own one, that's fine, but don't run around recommending it like it's the best thing since sliced bread.

[Dabombinable]

Why haven't moderators helped clean up all the garbage people spew recommending FX processors when hard evidence exists to show how bad of value it really is. If you already own one, that's fine, but don't run around recommending it like it's the best thing since sliced bread.

I can say personally that if your able to get one of the later LGA775 quad cores (QX9775/Xeon X5450) and overclock it to 4GHz (about the same as an FX 8350 at stock speeds), you will notice that the 4 threads of the older intel cpu (2008) will easily beat the crap out of any AMD cpu at the same frequency. FX cpus use a longer pipeline just like the Pentium 4 that Intel ended up getting rid off, which is why they have the high clock speeds, but tiny performance gains when overclocked. Plus they are shown to be faster when opperating with 1 core per module, instead of 2. (4m/4c-4m/8c)

[Faa]

Plus they are shown to be faster when opperating with 1 core per module, instead of 2. (4m/4c-4m/8c)

Right, when you run a singlethreaded thing you won't gain a thing from disabling the 2nd ALU cluster (which people call a core) because the FPU already operated at 256 bit and the front-end only had to deal with one thread. This trick is only, it's not even a trick, is only good to show the 4-threaded performance difference between a 4300 (2M/4C) vs a 8350 (4M/4c). Whenever you have lets say 5 big threads, you'll have more performance from 4m/8c than 4m/4c. If you have 4 big threads you'll have the same performance between 4m/4c & 4m/8c if the Windows scheduler did its job correctly like this -> Each thread (4 in total), should have been in a different module regardless of 4m/4c or 4m/8c.

Thread = Software thread. Not refering to hardware threads like i7 4 cores and 8 threads.