Does the CPU matter in gaming anymore?

[Termie]

Why yes, it does!

http://techreport.com/articles.x/23246/1

TL;DR version: Bulldozer is slower than Nehalem

Thanks for the link. TechReport does awesome work. This is the article I've wanted to read for a long, long time...I agree that BF3 multiplayer is the ultimate test of CPU effectiveness, but it's also the ultimate waste of time for reviewers looking to bench a lot of CPUs consistently.

BTW, Nehalem isn't in this review.

 

[AtenRa]

Good luck getting a repeatable test with that!

Thats why you bench for more than 60 secs, you dont get exactly the same run each time but you can see how each system performs.

http://atenra.blog.com/2012/06/08/amd-fx8150-vs-intel-2500k-1080p-dx-11-gaming-evaluation/

 

[Mallibu]

Well, it seems like not only FX loses bad in gaming scenarios, but also while encoding and gaming the same time (so much for the supposed improved multitasking from more cores ... :thumbsdown: ).
The current FX lineup is such a joke, since even Phenom II 980 stomps all over them.

 

[Concillian]

Comparisons aside --

That is an excellent way to analyze CPU contribution to gaming performance and focusing on what really matters. I'm impressed with the way they engineered their analysis.

Thumbs up to Tech Report.

I think we've always known that single threaded performance is critically important to gaming performance, this just re-emphasizes that importance. That more cores is only a good thing if you can maintain single threaded performance shouldn't really surprise anyone, should it?

 

[soccerballtux]

Battlefield 3- Bulldozer approved

+1
FWIW that game is the future. Once games go sufficiently multithreaded AMD will be back in the game. Or, they're already back in the game, because games in the future will be multithreaded so you can safely buy AMD now.

 

[poohbear]

i noticed NO difference going from a PhenomII x6-1100T @ 3.8ghz to a 2500k @ 4,4ghz with my GTX 670 @ 1080 with AA & AF on. It was the most pointless upgrade ever. Don't listen to anyone that tells u otherwise.

Once u turn on AA & maximize all the quality settings @ 1080+ resolution, ANY quad core will suffice as the workload is now entirely on the GPU. The exception is with some RTS games, but with a GTX 670 its already way beyond 60 FPS so it doesnt matter anyways. ITS JUST THAT SIMPLE.

 

[frozentundra123456]

+1
FWIW that game is the future. Once games go sufficiently multithreaded AMD will be back in the game. Or, they're already back in the game, because games in the future will be multithreaded so you can safely buy AMD now.

Or you can buy intel and get superior performance now, and not rely on some future scenario possibly coming true.

Even is this one game 8 core BD was not better than 4 core intel. It just lost by less.

 

[Hypertag]

i noticed NO difference going from a PhenomII x6-1100T @ 3.8ghz to a 2500k @ 4,4ghz with my GTX 670 @ 1080 with AA & AF on. It was the most pointless upgrade ever. Don't listen to anyone that tells u otherwise.

Once u turn on AA & maximize all the quality settings @ 1080+ resolution, ANY quad core will suffice as the workload is now entirely on the GPU. The exception is with some RTS games, but with a GTX 670 its already way beyond 60 FPS so it doesnt matter anyways. ITS JUST THAT SIMPLE.

Okay. I am happy that you have proven every single benchmark and review on the planet incorrect. You have proven that AMD's processors are awesome, and clearly superior to Intel's. Anyone that disagrees is an Intel shareholder, shrill, employee, and fanboy. The Intel-CPU cartel will no longer be able to manipulate the public. Instead, the sleeping dragon will awaken, and everyone will be rushing to ebay to buy these awesome quad core processors on ebay since they are so cheap. http://www.newegg.com/Product/Produc...82E16819103244

 

[2is]

i noticed NO difference going from a PhenomII x6-1100T @ 3.8ghz to a 2500k @ 4,4ghz with my GTX 670 @ 1080 with AA & AF on. It was the most pointless upgrade ever. Don't listen to anyone that tells u otherwise.

Once u turn on AA & maximize all the quality settings @ 1080+ resolution, ANY quad core will suffice as the workload is now entirely on the GPU. The exception is with some RTS games, but with a GTX 670 its already way beyond 60 FPS so it doesnt matter anyways. ITS JUST THAT SIMPLE.

Except that it isn't. Not everyone is going to completely saturate their GPU to make it the end all be all limiting factor like you are. I noticed a huge jump from my Q6600 to 3770k in BF3 MP.

Anyone making a blanket statement like this is who shouldn't be listened to.

Then there are the people with multi-GPUs. But I guess since that's not you, it shouldn't be considered right?

 

[Fallengod]

i noticed NO difference going from a PhenomII x6-1100T @ 3.8ghz to a 2500k @ 4,4ghz with my GTX 670 @ 1080 with AA & AF on. It was the most pointless upgrade ever. Don't listen to anyone that tells u otherwise.

Once u turn on AA & maximize all the quality settings @ 1080+ resolution, ANY quad core will suffice as the workload is now entirely on the GPU. The exception is with some RTS games, but with a GTX 670 its already way beyond 60 FPS so it doesnt matter anyways. ITS JUST THAT SIMPLE.

As others stated, that is not completely accurate. What you missed during your so called "testing" is what previous posters pointed out. Not all games are the same nor do all put the same stress upon the cpu or the gpu. Its very possible that whatever you were testing in simply wasnt able to show the differences in cpu performance.

As I stated in an earlier post and what is plastered all over the internet in accurate benchmarks, I noticed a fairly big/decent fps difference going from a E6400 @ 3ghz(even though this is only a dual-core) to an i5-2500k @ stock.

What conclusions should we draw in these discrepancies? Well, you just have to do your research and figure out where you get the benefits. Sometimes its not worth the upgrade and inexperienced people who buy upgrades without doing research deserve what they get. It all comes down to what games you play, resolution played at, monitor, video card, CPU, etc, it all factors in.

 

[bl00tdi]

+1 to the last two replies. In my case, I upgraded from a 1090T @ 4.1 to a 3570k @ 4.5 and saw a VERY significant boost in performance in my "game" of choice, flight simulator x. It just so happens to be heavily dependent on single thread performance where IB destroys Thuban, especially when overclocked. Absolutely worth the upgrade. I did my due diligence before I pulled the trigger so my expectations were consistent with the results.

 

[AtenRa]

It depends on the Game and Gameplay settings, but most games will be GPU limited once you enable AA filters at 1080p and above.

 

[boxleitnerb]

Hmm, they run Batman AC in DX-9 mode and they haven't used any AA filters on Crysis 2.
Also, it seams to me they just run the Single Player mode in BF3. That was the only game of the four used that they should tested in MP mode to really check if the CPU plays a significant part in Gaming.

Crysis 2 DX11 doesn't support MSAA. Batman AC DX11 runs like garbage even after two patches. Texture streaming issues and stuttering. But I agree, BF3 should have been looked at in MP - but it is difficult to get repeatable results there.

Thats why you bench for more than 60 secs, you dont get exactly the same run each time but you can see how each system performs.

http://atenra.blog.com/2012/06/08/amd-fx8150-vs-intel-2500k-1080p-dx-11-gaming-evaluation/

No. It is only logical that the longer the bench runs, the more likely it is that the results will be different. When doing benchmarks of dynamic scenes, shortness is key.

 

[2is]

But not all that difficult to get a good average, which would still be more telling than SP benchmarks.

 

[AtenRa]

Crysis 2 DX11 doesn't support MSAA. Batman AC DX11 runs like garbage even after two patches. Texture streaming issues and stuttering. But I agree, BF3 should have been looked at in MP - but it is difficult to get repeatable results there.

Even so, benching only 4 games and coming to a conclusion is not the best thing to do. I have benched 9 games in my DX-11 gaming evaluation review and 5 of them show no performance gains from an Overclocked CPU or from different CPU models because they were GPU limited.

No. It is only logical that the longer the bench runs, the more likely it is that the results will be different. When doing benchmarks of dynamic scenes, shortness is key.

If you die 2-3 times in 60 seconds it will have a big impact in your fps chart, if you die 2-3 times in 200-300 seconds it will not effect your average picture of the gameplay benchmark run that much.
Also you may not see a heavy fight situation in 60 seconds of gameplay simple because you will be in the wrong place on the map than the place where the heavy fighting is happening.

In BF3 MP in open large maps you should bench in 200-300 seconds or more and try to do the same things possible in order to have the best average picture of the gameplay. It is not easy but it is not impossible, it only takes more time.

 

[boxleitnerb]

Well as I told you, your test methods and/or conclusions are questionable.

If you bench MP where you have little to no influence over the action, a longer benchmark might be not at bad idea. For SP however, even 30 seconds is quite long.

 

[AtenRa]

Well as I told you, your test methods and/or conclusions are questionable.

And why is that ?

 

[KingFatty]

Anyone know what is about the CPU that helps here?

It's nice to know we have the proxy of using brand to know which will be better, but what technically is it about a CPU that will give you the performance you want?

I guess I'm trying to understand the mechanics a bit more, is it a good memory controller, or a fast cache, or what? What tangible specs do you look for in a CPU to help performance (e.g., better minimum frame rate etc), other than brand/speed?

 

[soccerballtux]

Or you can buy intel and get superior performance now, and not rely on some future scenario possibly coming true.

Even is this one game 8 core BD was not better than 4 core intel. It just lost by less.

that's true but you can't get an 8-thread Intel for $170.

 

[soccerballtux]

And why is that ?

because AMD was shown to perform well in BF3

 

[gmaster456]

Okay. I am happy that you have proven every single benchmark and review on the planet incorrect. You have proven that AMD's processors are awesome, and clearly superior to Intel's. Anyone that disagrees is an Intel shareholder, shrill, employee, and fanboy. The Intel-CPU cartel will no longer be able to manipulate the public. Instead, the sleeping dragon will awaken, and everyone will be rushing to ebay to buy these awesome quad core processors on ebay since they are so cheap. http://www.newegg.com/Product/Produc...82E16819103244

I don't see anywhere in his post where he said Intel was better.

 

[Phynaz]

that's true but you can't get an 8-thread Intel for $170.

You can get pretty darn close. So for the couple of percent price difference in a build you will be sacrificing large amounts of gaming performance. Why gimp an expensive GPU just to save a couple of bucks on the CPU?

You get what you pay for.

 

[Phynaz]

It depends on the Game and Gameplay settings, but most games will be GPU limited once you enable AA filters at 1080p and above.

No, it doesn't. Intel is better at gaming. Unless you can prove Techreport wrong by doing your own frame latency testing?

 

[frozentundra123456]

that's true but you can't get an 8-thread Intel for $170.

Who cares how many threads you have. A 2500k with 4 threads gives better performance in all but a very few highly threaded scenarios. Especially in games it is better in almost every case. Even a 4thread i3 2120 is equal or better in some games.

 

[boxleitnerb]

And why is that ?

You've still not updated your Shogun 2 results. In a real battle when watching the action and zooming in, you will never ever have such high fps. The replay doesn't include camera movements, you have to position it yourself to where some action will be - as you would when actually playing the game. Also I found out recently that while fps may be high, animations will be choppy if the CPU has too much work. Do a quick 4 vs 4 battle and watch the units walk/run at the beginning. I have 40-50ish fps, but the guys skip some animation steps because it is simply too much with this many units. Animation and fps seem to be somehow decoupled. You don't learn that from running a graphics benchmark.

As for Civ 5, every review that uses actual savegames gets a different result than you did. HardOCP, Computerbase, look them up. Also I would think it is self-evident to test turn time in that game as well.

BF3 is just weird. That a 920@4GHz is so much faster than a 2500K and the 2500K OC giving that large a benefit doesn't make sense at all. I told you, repeatability is important. Especially to analyze the outcome if something went wrong like here. Calls into question all the BF3 results. To analyze and question results is also part of a test. Don't just run your fraps and post the numbers without thinking if they make sense.