Intel vs AMD "Best CPU for Gaming" benchmarks.

http://udteam.tistory.com/647

Plenty of charts. Has the Core i7 4960X and the FX-9590 for the high end and plenty of other price points in between. Tests include low resolutions and higher resolutions. Nvidia 780Ti was used for the testing to eliminate bottlenecks.

So what do you think of the author's work and results? Discuss.

Honestly, it's not that bad, and I like the fact he was thorough in testing a wide range of processors from both AMD and Intel.

In the end, if you're not that nitpicky with losing a few frames per second versus an Intel processor, but still have 60 FPS or higher in a game with the same settings, AMD is still not a bad processor for gaming. People just have to realize that.

Hopefully, in time, AMD's next processor-- Excavator and the unnamed x86-64 CPU architecture in 2016-- will show more improvements.

Nice comparison but it would be even better if they include Opterons. From the results CPU doesn't make that much of a difference so save money on an AMD FX-6300 or better and put the savings toward a better GPU with the advantage of more CPU cores for virtualization.

Lost planet 2 really showed a difference in frame rate. The Intel chips were twice as fast in that game! Sleeping Dogs and a few others at 1680x1050 favored Intel, but the AMD was over 100 fps so big deal. I also look forward to seeing what AMD is working on for the future.

Kinda shoddily done. No explanation of the the game settings used nor what levels or methods used for doing the in-game runs. I also would have like to see the Intel and AMD CPU performance charts for each game combined into one chart for a clearer picture where Intel and AMD CPUs stood in comparison with each other.

Also, practically worthless benchmarks for the BF4 results IMO since more than likely they used the single-player game. It's BF4 FFS. Of course you're going to be playing multiplayer more than single player. Considering that BF4 MP is more stressful than BF4 SP, I believe more helpful/accurate performance results would have been gained had the tests been done in BF4 MP.

I mean, kudos to the author for gathering up all those CPUs and doing the testing. But still not quite as useful as it could have been.

Things I noticed:

In synthetic benchmarks, AMD seems to have a nice, even footing with Intel.
In actual games, AMD's top chips are either between the i3 and i5, or between the i5 and i7 (game-dependent).
AMD's 4.7GhZ CPU is not much faster than its 4GhZ CPU in most benchmarks.

So we can assume:
Around 4GhZ, an AMD chip does not seem to bottleneck a single GPU so there is probably very little reason to buy the 4.7GhZ chip when you can overclock the 4GhZ chip. Even if you can't get to 4.7GhZ, it won't matter much.
For games that are CPU-heavy, the better choice is Intel (I'm not sure anybody really thinks otherwise anyway these days, it's been pretty clear).
Synthetic benchmarks are still synthetic, but at least the devs appear to try to take advantage of any hardware you can throw at it.
It takes roughly 500-700MhZ more for an AMD CPU to do the same work as an Intel one (which we've known since Bulldozer came out, so no surprise there).

cageymaru said:

Lost planet 2 really showed a difference in frame rate. The Intel chips were twice as fast in that game! Sleeping Dogs and a few others at 1680x1050 favored Intel, but the AMD was over 100 fps so big deal. I also look forward to seeing what AMD is working on for the future.

Click to expand...

Hyper-Threading may have accounted for it?

LAiN_OfTheWired said:

It takes roughly 500-700MhZ more for an AMD CPU to do the same work as an Intel one (which we've known since Bulldozer came out, so no surprise there).

Click to expand...

I thought the range was more like 2 times that. Although I would have to closely look at benchmarks to see..

I think these charts show that for gaming, most users will not notice much of a difference between amd or intel. Only in price. Oh ye AMD!

Nice comparison... but what about the overclockers? The Intel CPUs run away with the show when you consider O/C potential.

Dangman said:

Kinda shoddily done. No explanation of the the game settings used nor what levels or methods used for doing the in-game runs. I also would have like to see the Intel and AMD CPU performance charts for each game combined into one chart for a clearer picture where Intel and AMD CPUs stood in comparison with each other.

Also, practically worthless benchmarks for the BF4 results IMO since more than likely they used the single-player game. It's BF4 FFS. Of course you're going to be playing multiplayer more than single player. Considering that BF4 MP is more stressful than BF4 SP, I believe more helpful/accurate performance results would have been gained had the tests been done in BF4 MP.

I mean, kudos to the author for gathering up all those CPUs and doing the testing. But still not quite as useful as it could have been.

Click to expand...

"Note that for every game title, all graphics quality options in in-game control panel are set to the highest possible settings. For example, for Battlefield 4 I used an "Ultra" preset as well as "DX11+DDOF" for Bioshock : Infinite benchmark."

Despite not being perfect, it is still an excellent job of data collection. I too am glad to see that AMD may not be the performance king, but certainly offers some competition in the price/performance area. We need someone to keep Intel and nVidia from complete market domination.

LAiN_OfTheWired said:

Things I noticed:

In synthetic benchmarks, AMD seems to have a nice, even footing with Intel.
In actual games, AMD's top chips are either between the i3 and i5, or between the i5 and i7 (game-dependent).
AMD's 4.7GhZ CPU is not much faster than its 4GhZ CPU in most benchmarks.

So we can assume:
Around 4GhZ, an AMD chip does not seem to bottleneck a single GPU so there is probably very little reason to buy the 4.7GhZ chip when you can overclock the 4GhZ chip. Even if you can't get to 4.7GhZ, it won't matter much.
For games that are CPU-heavy, the better choice is Intel (I'm not sure anybody really thinks otherwise anyway these days, it's been pretty clear).
Synthetic benchmarks are still synthetic, but at least the devs appear to try to take advantage of any hardware you can throw at it.
It takes roughly 500-700MhZ more for an AMD CPU to do the same work as an Intel one (which we've known since Bulldozer came out, so no surprise there).

Click to expand...

r-486 said:

Hyper-Threading may have accounted for it?

Click to expand...

Lost Planet has always been super sensitive to both ipc and core count. It was proven that the game engine could actually leverage 8 cores about the time the Intel skull trail platform came out with dual quad core cpu's. It remains as one of the few gaming benchmarks that can. It has been a staple of CPU benchmarking ever since. Hardocp.com has used it for years. Although an older game, this game engine isn't doing anything special, its not like it was using x87 instructions that Intel dominated for years in SuperPi with.

What is really telling in these benches is how badly a quad core hyperthreaded Intel CPU crushes the best performing octo core AMD chip. You want to know just how far AMD is behind? Look at the 1680x1050 Lost Planet 2 bench which is obviously CPU bound.

Think about it for a second...
Intel i7 4790K Haswell (3.6ghz base, 4.0ghz boost): 101fps
Intel i7 4690K Haswell (3.5ghz base, 3.9ghz boost): 91fps
versus AMD FX9590 (8 cores, 4.7ghz base, 5.0ghz boost): 56fps
AMD's chip is 55% as much work as the Intel i7 chip and 62% of the amount of work of the i5 chip all the while the AMD chip has a full ghz advantage (25% clock speed advantage) and 4 more physical cores on a game engine that runs 8 threads. Hyperthreading only gave a 10% boost in frames compared to the similarly clocked i5 4690K.

Lets use the i5's for comparison sake since hyperthreading muddies things a little. The i5's each increase in speed by 200mhz and 2fps per model. That's 1fps per 100mhz increase. Therefore we can guess that a hypothetical i5 from the same family at 5ghz would run at 102fps. Lets make a hypothetical octocore intel chip from 2 of these i5's running at 5ghz. The game engine has shown an increase of 24% in FPS going from 4 threads to 8. Please note the similar numbers here to the link I provided are simply concidence due to math shown previously. That's now 135fps for 8 intel cores at 5ghz versus 56fps for 8 AMD cores at 5ghz.

Now you have a direct comparison of core to core performance of an Intel Haswell versus an AMD Excavator. Intel's chips do 2.4 times as much work per clock cycle as AMD.
Yikes. This is on a game engine that loves multithreading and uses commonly used CPU instructions, which are 2 comparison points that AMD supporters have emphasized for years to make the the outlook appear rosier.

Fortunately for AMD, software really hasn't caught up yet, so many of these weaknesses aren't shown in real world. I also gave a worst case scenario, so many of you will never see this kind of difference. However, it does show how bad things *could* get for AMD if they don't catch up soon. What is really bad is the AMD FX9590 is currently going for $329.99, and Intel's i7 4770K is currently going for exact same price. Half decent AMD and Intel overclocking boards are going for relatively the same price, with only the super high end Intel boards showing any dramatic price differences. Where's the price vs performance ratio now? It barely favors AMD anymore, and that's only if you go higher than a 4770 and super high end board.

Sure, AMD works for 90% of the gamers out there. That's wonderful. But AMD needs to catch up quickly before software does. AMD can't even keep from bottlenecking their own GPU's in crossfire.

It truly is sad. I want AMD to catch up. I really do.

I think they will. In two years. After intel has gone 10nm (assuming no delays) and probably will out-IPC haswell in the interim steps to that.


As to the case of Lost Planet, eh.

Every game from here on out is going to be optimized for octo-jaguar as a baseline. even a quad richland will be able to keep up with ~actual gameplay~, even if intel continues to stomp on benchmarks in 8-thread-land.

Intel will probably favor PC exclusives and bottlenecked situations in the extreme high end, but for the most part that end of the market is shrinking fast in favor of tablet chips and phones..and Intel is moving into that fairly ~okay~ without resorting to ARM.

I feel like we've hit a plateau of general performance computing that's lending itself nicely to miniaturization and process optimization rather than a continuation of moore's law.

CaptNumbNutz said:

Lost Planet has always been super sensitive to both ipc and core count. It was proven that the game engine could actually leverage 8 cores about the time the Intel skull trail platform came out with dual quad core cpu's. It remains as one of the few gaming benchmarks that can. It has been a staple of CPU benchmarking ever since. Hardocp.com has used it for years. Although an older game, this game engine isn't doing anything special, its not like it was using x87 instructions that Intel dominated for years in SuperPi with.

Click to expand...

Pardon the ignorance from a non gamer, that has never even heard the name "Lost Planet", before, but i have an objection to the phrase "this game engine isn't doing anything special".

To my humble opinion, this engine is doing in deed something special. It is blatantly programmed to favour Intel...

Because, how you otherwise have:

FX-8350> 54 fps
FX-8300> 53 fps
FX-6300> 52 fps

I mean, really? 2 fps difference for 2 more cores at 500Μhz higher clock per core? If you want the best "real world" test for CPU strength, with "honest" code (non strong bias against AMD), grab a x264 test. Even there, if there is a small optimization in favour of one, i 'd say it's Intel, because Dark Shikari (the dev) some years ago, i think in a discussion, preferred Intels over AMD.

But, saying that this game is the perfect index of CPU stength, where 2 more AMD cores account for just 2 fps, is a bit odd, isn't it?

If such a game is the most accurate representation, then what happens when you run x264 and you see a perfect scaling with the FX-8320 showing 2 more cores comparing to the FX-6300? Α miracle?

That's because he is wrong- Lost Planet 1 can use 6 cores. Even that was huge overkill when it released.

Also using 780ti is best case scenario for AMD cpus as NVIDIA drivers scale much better with multiple cores.

But apart from few wrongly chosen test places in benchmarks where there's barely any diffrence from celeron to i7 they are quite correct in showing that 5Ghz 8 core FX falls somewhere beetween i3 and low end stock i5.

Does he have a definitive one for 2013? Just curious. The site is mixed language, so I'm finding it hard to find things there.

Edit: Also why is this specifically in the AMD section...

Michaelius said:

That's because he is wrong- Lost Planet 1 can use 6 cores. Even that was huge overkill when it released.

Click to expand...

Interesting. That would explain the imperfect scaling. Going from 4 to 8 cores. I can take that spitball math and simply remove that last 24% speed increase from the scaling and it still doesn't make the picture any rosier.

Michaelius said:

But apart from few wrongly chosen test places in benchmarks where there's barely any diffrence from celeron to i7 they are quite correct in showing that 5Ghz 8 core FX falls somewhere beetween i3 and low end stock i5.

Click to expand...

That was what I was shooting for. The high end FX chips really are supposed to compete with high end Intel LGA1150, but they don't come close in performance, or even price per performance.

I think this is yet another fantastic win for Intel's 50 dollar monster. Amd can't outperform the Celeron with anything less than a quad, and that's dangerously close to the 100 mark!

:/ I may use Intel but I've always been a fan of AMD as the underdog back in the AthlonXP and Duron days. I keep hoping they will, one of these days, come out with something that gives Intel a run for it's money.

I have an FX-8320 in one rig and an i7 4930k in another. I don't see much of a difference between the two other than the i7 being a bit better at handling high part counts in Kerbal Space Program. That and the FX-8320 didn't rape my wallet.

I think this is the 1000th discussion on the same subject and unless one lives for benchmarks, there is little to it.

Current CPUs, even the weaker AMDs, are capable to run any game fluidly with a good GPU.
New gen consoles sport very weak AMD octacores. This means, that as long as they are the top consoles, FX octacores will most likely be able to run anything without problem.

This guy here, proved that a 1100T can run QUAD 7970, without the 1100T bottlenecking the cards and with all GPUs hitting 100%. Oh and the 1100T at stock speed.

http://www.overclock.net/t/1489949/amd-cpus-vs-intel-cpus-for-gaming/40#post_22279648

What more do you need? Even i, who do a lot of video encoding at home, find it hard to justify to myself buying an FX-8320, when i have one 1090T and one FX-6300. Because i start thinking of heat, stress on my cheap motherboards and frankly, the hexacores are quite fast on the job.

Relax, when i read these discussions, sometimes it seems as if games have brought the current AMD CPUs to their knees and there is some urgent need to upgrade. There isn't. Here in Italy, many gamers still game happily on Phenom x4 without complains. They simply get faster GPUs.

DeathFromBelow said:

That and the FX-8320 didn't rape my wallet.

Click to expand...

That's just a red herring.

It's been well known for quite some time that the i7 desktop processors are overkill for games. This article once again confirms that by showing i5 with 95% the performance of the i7, showing that only the small clock speed difference matters.

You could get the same performance from an i5, or possibly i3 or Pentium. Kerbal, like many games STILL being released, only uses one core, so single-threaded speed is king.

It's really just your fault you paid too much for an i7, but didn't actually need it

AMDs vs Intel debate is shifting as more devs get onboard with:

a) Better thread count
b) Integration of various processes

the only place AMD still is a failure is Single Core Performance. But again, thats old games and programs. Kind of a moot point as the future is shifting to better written apps.

Another point is chipset/mb feature set with AM3+, but i guess with their hard push towards APU development as the future that will probably go away.

Plenty of those CPUs look fine for gaming, but the biggest surprise for me was how well the Celeron G1840 did for its price segment, especially considering how it's a low clocked dual core CPU.

Faethon said:

I think this is the 1000th discussion on the same subject and unless one lives for benchmarks, there is little to it.

Current CPUs, even the weaker AMDs, are capable to run any game fluidly with a good GPU.
New gen consoles sport very weak AMD octacores. This means, that as long as they are the top consoles, FX octacores will most likely be able to run anything without problem.

This guy here, proved that a 1100T can run QUAD 7970, without the 1100T bottlenecking the cards and with all GPUs hitting 100%. Oh and the 1100T at stock speed.

http://www.overclock.net/t/1489949/amd-cpus-vs-intel-cpus-for-gaming/40#post_22279648

What more do you need? Even i, who do a lot of video encoding at home, find it hard to justify to myself buying an FX-8320, when i have one 1090T and one FX-6300. Because i start thinking of heat, stress on my cheap motherboards and frankly, the hexacores are quite fast on the job.

Relax, when i read these discussions, sometimes it seems as if games have brought the current AMD CPUs to their knees and there is some urgent need to upgrade. There isn't. Here in Italy, many gamers still game happily on Phenom x4 without complains. They simply get faster GPUs.

Click to expand...

Funny thing about benchmarks is you can prove almost everything.

Look how this great bargain 60 EUR cpu kicks ass of those expensive i5/i7



or does it ?



True miracle isn't it ?

http://pclab.pl/art51525-2.html

But maybe we don't need to spend money on GPUs too ?




Wow why are we buying those $300 gpu ?


PS. CPU X was Pentium G860

Michaelius, i get the point with tests being able to prove several things, but this doesn't mean that "Red's" test doesn't prove a thing too. His point isn't whether 60euro CPU can run Quad 9790 nor that you need not buy 300$ GPU. His point is that Dirt3 can be run with quad 7970 on a stock 1100T. I think this is impressive enough. Can a Pengium G860 do that? I think not... So, why mix tests...

samm said:

Another point is chipset/mb feature set with AM3+, but i guess with their hard push towards APU development as the future that will probably go away.

Click to expand...

APUs aren't going to disappear. If they do, they'll return. History of CPUs and PCs in general shows components getting integrated, because it's faster and cheaper that way. Northbridge is no longer on the motherboard. It's in the CPU. MCC is no longer in the motherboard. It's in the CPU. Cache no longer on the motherboard. It's in the CPU.

r-486 said:

APUs aren't going to disappear. If they do, they'll return. History of CPUs and PCs in general shows components getting integrated, because it's faster and cheaper that way. Northbridge is no longer on the motherboard. It's in the CPU. MCC is no longer in the motherboard. It's in the CPU. Cache no longer on the motherboard. It's in the CPU.

Click to expand...

Sorry for my wording, it may have confused you. In my original post, I was stating that the deficit in chipset features that current AMD has will go away as they delve deeper into the apu platform, which is here to stay for a very very long future

Amd's high end cpus pretty much come overclocked, Intels don't. In most tests from reputable sites there also seems to be a much larger gap between the two companies.

I have an amd apu in my second pc but that's only because I live close to a microcenter and got it for stupid cheap. I'm glad people still believe the bullshit about Amd being close in performance because competition is good, but there is no competition currently, and I hope that changes. I also think buying haswell is completely pointless compared to Intels other offerings.

My AMD FX-9370 runs at 5GHz with a H100 cooling it. I could have just gotten a FX-8350 and done the same thing. 1GHz overclock on what is nothing more than a FX-8350 seems like a legit overclock to me. I have no idea how far the little APU can overclock.

I suppose that the i7-4770K can hit 5K also. But to say that the AMD chips are already overclocked from the factory, which implies that they can't be overclocked more, is stretching it a bit.

cageymaru said:

My AMD FX-9370 runs at 5GHz with a H100 cooling it. I could have just gotten a FX-8350 and done the same thing. 1GHz overclock on what is nothing more than a FX-8350 seems like a legit overclock to me. I have no idea how far the little APU can overclock.

I suppose that the i7-4770K can hit 5K also. But to say that the AMD chips are already overclocked from the factory, which implies that they can't be overclocked more, is stretching it a bit.

Click to expand...

but comparing the overclock percentages there is no comparison. Yes the amd chips can still be overclocked but not nearly as much.

defaultluser said:

That's just a red herring.

It's been well known for quite some time that the i7 desktop processors are overkill for games. This article once again confirms that by showing i5 with 95% the performance of the i7, showing that only the small clock speed difference matters.

You could get the same performance from an i5, or possibly i3 or Pentium. Kerbal, like many games STILL being released, only uses one core, so single-threaded speed is king.

It's really just your fault you paid too much for an i7, but didn't actually need it

Click to expand...

People point out these 6 core i5s? All I see are 6 core 12 thread i7s.

I personally don't care if games only use a few cores. That means I get those other cores to do other things with. If we're moving towards multi-monitor setups, which I think we are, I don't want a single game eating all 8 cores of my CPU and all my ram. I'd like to game on "Windowed Fullscreen" mode, use the other monitors for other media, while maybe also encoding a gameplay stream or doing other things. This wouldn't be possible if 8 cores are sitting at 100% because of a single game. I don't want a multicore CPU that can't do anything else while a game runs because that game is using all of the cores "efficiently". CPU's are general purpose and I should have extra cores for other things.