Gaming CPU

[AtenRa]

The tests was done with a HD6970. Do the math yourself if you add a HD7970.

Keep living in denial while your look at your FX case.

Whats the next argument you gonna present? That people are supposed to play in 5760*1200? Else its unfair to the dulldozer?

Do you have 1080p AA/AF enabled gaming benchmarks (except the 2-3 that we already know are running faster with Intel) to show me that FX CPUs are bottlenecking the GPU ??

 

[Black96ws6]

I believe it is the other way around. FX8120 is $50/70 cheaper than 2500K/3570K

FX8120 = $169,99
http://www.newegg.com/Product/Produc...961&Tpk=fx8120

2500K = $219,99
http://www.newegg.com/Product/Produc...82E16819115072

3570K = $239,99
http://www.newegg.com/Product/Produc...82E16819116504

also,

For $50/70 less, you get higher performance in multithreaded apps OC or not.

http://forums.anandtech.com/showthread.php?t=2241712&highlight=

As for the games,

We all know there are 2-3 games (DX-9) where Intel is faster. But in the majority of the games you will not see more than 5-10% difference between those three CPUs.

Better spend those $50-70 for a better GPU

Those prices for the 2500k are way off. NewEgg just had a sale on them for $199, and MC has a sale on them RIGHT NOW for $169:

http://www.microcenter.com/single_product_results.phtml?product_id=0354589

Hmmn....$169 for BD or a 2500k....that's not even a contest...

 

[Black96ws6]

Or if you want to give Ivy Bridge a try, 3570K for $189 right now:

http://www.microcenter.com/single_product_results.phtml?product_id=0388577

 

[AtenRa]

FX8120 $149,99
http://www.microcenter.com/single_product_results.phtml?product_id=0375765

Next time you get a price for a CPU from MC, try get the price of the second CPU from the same store aswell

 

[Makaveli]

FX8120 $149,99
http://www.microcenter.com/single_product_results.phtml?product_id=0375765

Next time you get a price for a CPU from MC, try get the price of the second CPU from the same store aswell

how dare you bring that kinda logic in here

the nerve of some people.......

 

[Black96ws6]

FX8120 $149,99
http://www.microcenter.com/single_product_results.phtml?product_id=0375765

Next time you get a price for a CPU from MC, try get the price of the second CPU from the same store aswell

I wouldn't take BD even if it was $40 cheaper.

I'd eat Cup o Noodles for a week and get the Sandy Bridge CPU

 

[Makaveli]

I wouldn't take BD even if it was $40 cheaper.

I'd eat Cup o Noodles for a week and get the Sandy Bridge CPU

lol

 

[borisvodofsky]

I wouldn't take BD even if it was $40 cheaper.

I'd eat Cup o Noodles for a week and get the Sandy Bridge CPU

Glad to see you've taken the Cup Noodle spin-off to my Top Ramen.

BTW, have you tried BOWL noodle?

 

[AtenRa]

You really have to hate yourself to buy AMD for gaming these years.

Thats a big pile of BS. You live in a world of denial.

Bulldozer is the biggest flop there is for gaming.

I wouldn't take BD even if it was $40 cheaper.

I'd eat Cup o Noodles for a week and get the Sandy Bridge CPU

Just have a look in the link of my sig bellow.

 

[boxleitnerb]

How about posting more information what exactly these benchmarks are? Patch version, timedemo, savegame, how many cars/soldiers/..., what level etc. The information is incomplete.

Also one has to wonder if the scenes in question represent CPU-intensive portions of the respective games. For example do you really expect 80fps in Shogun 2 in large battles with thousands of soldiers? In a real battle, fps will be half that if you're lucky. In almost every game you can find CPU and GPU bottlenecked parts depending on where you are and what you do. The art of benchmarking is to find those scenes that matter for the components you test and make a repeatable benchmark out of it.

Look at PCGH and how well they document their benchmarks:
http://www.pcgameshardware.de/aid,7...-Skyrim-und-Spielbarkeitsgrenzen/Spiele/Test/

CPU intensive scenes for CPU tests, detailed instructions, detailed information about the sequence, patch version etc and a video of the scene in question. This is how it should be done.

 

[Rvenger]

Just have a look in the link of my sig bellow.

Well if Piledriver is close to a 10% increase over Phenom 2 with lower power consumption than Bulldozer. I think it will do well.

 

[AtenRa]

I have updated the review with the version of each game, IQ settings and what benchmark was used.

If you would like any other info just ask

 

[boxleitnerb]

Okay, thank you. Looks like you chose timedemos quite often. They don't really represent ingame performance in many cases. What you basically did was benchmark your graphics card.

 

[AtenRa]

I wanted to have a run with the 2500K at 4.5GHz but i didnt have the time, i had to give the GPU back. Since most of the benchmarks are GPU limited you will not see any performance gains from 4GHz to 4.5GHz or even at 4.8GHz.

The FX8150 was using a watercooled kit, it was my Core i7 920@4GHz that is using the Noctua NH-D14. For the Core i5 the Noctua NH-U12P SE2 is one of the best aircoolers today. I dont see why you have any problems with the cooling in those test systems.

As for the socket 1155 motherboard, it is a Z68 motherboard that will allow you to OC the Core i5 2500K at 4.5GHz or above. Again i dont understand why you sea a problem with the hardware used.

As for the benchmarks used,
I have found them to represent most of the time a very close to in game performance. I use them in order to have a consistent run every time. Gamers with High End graphics cards will game with every IQ setting enabled available.

So, those benchmark runs and IQ settings represent what most users will use when they actually game at 1080p.

 

[boxleitnerb]

As for the benchmarks used,
I have found them to represent most of the time a very close to in game performance. I use them in order to have a consistent run every time. Gamers with High End graphics cards will game with every IQ setting enabled available.

So, those benchmark runs and IQ settings represent what most users will use when they actually game at 1080p.

I challenge that. The problem with timedemos/integrated benchmarks is, that CPU load is often lower than in the game itself. You want to show performance of a game? Then you should actually play it and benchmark it while playing. Sure, it takes work and patience to produce consistent results, but that's the way it is. And it removes any doubt about these results that may arise.

For example look here for Civ 5:
http://www.computerbase.de/artikel/prozessoren/2011/test-amd-bulldozer/45/

A vast difference between maximum and minimum zoom. Your benchmark results are weird anyway. The FX gains almost nothing from OC, the 2500K more than doubles? Makes absolutely no sense. This is why I would never trust these integrated benchmarks when it comes to CPUs. The only way to do it right is to make it yourself.

 

[bononos]

....... As for the benchmarks used,
I have found them to represent most of the time a very close to in game performance. I use them in order to have a consistent run every time. Gamers with High End graphics cards will game with every IQ setting enabled available.

So, those benchmark runs and IQ settings represent what most users will use when they actually game at 1080p.

I find that hard to believe and you got into the same debate in another thread for the same reasons that the others are pointing out- that your benchmarking methods don't push the cpu enough as much as really playing the game.

And focusing on 1080 and above is iffy since 100hz/120hz monitors are getting more affordable and fps players will appreciate the higher framerates above >60fps.

 

[AtenRa]

I challenge that. The problem with timedemos/integrated benchmarks is, that CPU load is often lower than in the game itself. You want to show performance of a game? Then you should actually play it and benchmark it while playing. Sure, it takes work and patience to produce consistent results, but that's the way it is. And it removes any doubt about these results that may arise.

For example look here for Civ 5:
http://www.computerbase.de/artikel/prozessoren/2011/test-amd-bulldozer/45/

A vast difference between maximum and minimum zoom. Your benchmark results are weird anyway. The FX gains almost nothing from OC, the 2500K more than doubles? Makes absolutely no sense. This is why I would never trust these integrated benchmarks when it comes to CPUs. The only way to do it right is to make it yourself.

I will just quote from Civ V notes.

1. Late Game View Benchmark.
This benchmark is designed to simulate a late game workload. This scenario exercises all aspects of the game engine pipeline since all simulation and renderable object types are represented at a frequency consistent with a game that has been in progress for 300+ turns

Running the LateGameView benchmark is the same as playing the game yourself.

One more difference in your link (http://www.computerbase.de/artikel/prozessoren/2011/test-amd-bulldozer/45/) is that they benched at DX-9 when i bench at DX-11.

 

[AtenRa]

I find that hard to believe and you got into the same debate in another thread for the same reasons that the others are pointing out- that your benchmarking methods don't push the cpu enough as much as really playing the game.

And focusing on 1080 and above is iffy since 100hz/120hz monitors are getting more affordable and fps players will appreciate the higher framerates above >60fps.

If you have more than 1080p resolutions you becoming more and more GPU dependent. The majority of PC users currently are using 60Hz 1080p displays.

 

[Magic Carpet]

Intel is better but it costs more. I have to agree with AtenRa. At higher resolutions, GPU becomes more of a deal, than CPU. Hence, Faildozer can still be a "recommendation". Especially when you factor in, a motherboard choice, the Intel equivalent, will cost more and sometimes will give you less options. Only a few (the more expensive ASRock), for example, come with legacy support (FDD, PATA), whereas AMD mobos are often bundled "free".

If you have the dough, get Intel. But if you don't, you can easily game with AMD as well. You will get used to it, not a big deal. Humans adapt quick.

Max Payne 3 is another example, clearly bottlenecked by GPU. Even 2.5 Ghz is enough for it. Battlefield 3 MP is the only game that really benefits from behemoth CPU power. And if you don't overclock, power consumption isn't an issue with Zambezi chips (most people don't do that in fact). Maniacs like myself are.. a minority in my place anyway.

 

[Durvelle27]

If you have more than 1080p resolutions you becoming more and more GPU dependent. The majority of PC users currently are using 60Hz 1080p displays.

does it matter between 120Hz and 60Hz

 

[boxleitnerb]

I will just quote from Civ V notes.

Running the LateGameView benchmark is the same as playing the game yourself.

One more difference in your link (http://www.computerbase.de/artikel/prozessoren/2011/test-amd-bulldozer/45/) is that they benched at DX-9 when i bench at DX-11.

Okay, but your results are still strange. Anyway, there are situations where the FX is clearly behind by about 30%:
http://www.hardocp.com/article/2011/10/11/amd_bulldozer_fx8150_gameplay_performance_review/4

This may not always be the case, but as I said - to test CPUs properly, it is important to find a typical scene with a CPU limit. You don't have to look 10 hours for it and it doesn't have to be an absolute worst case scenario if that occurs only once in the game. Still, to chose integrated benchmarks is a bad idea in most cases.

 

[AtenRa]

Okay, but your results are still strange. Anyway, there are situations where the FX is clearly behind by about 30%:
http://www.hardocp.com/article/2011/10/11/amd_bulldozer_fx8150_gameplay_performance_review/4

This may not always be the case, but as I said - to test CPUs properly, it is important to find a typical scene with a CPU limit. You don't have to look 10 hours for it and it doesn't have to be an absolute worst case scenario if that occurs only once in the game. Still, to chose integrated benchmarks is a bad idea in most cases.

From the HOCP article you linked.

for that we used an AMD Radeon HD 6970. The driver version used was AMD Catalyst 11.10 Preview driver, which contains performance improvements for Battlefield 3 Beta.

I have used an HD7970 with Catalyst 12.4. Not only that but they didn't use 4xMSAA, i did. You cannot compare the results of the two reviews

 

[boxleitnerb]

From the HOCP article you linked.

I have used an HD7970 with Catalyst 12.4. Not only that but they didn't use 4xMSAA, i did. You cannot compare the results of the two reviews

The graphics driver does not improve CPU performance. If they were held back by the 6970, the FX and the 2600K would perform the same.

For the 100th time:
It doesn't make sense to benchmark CPUs with AA on as this is calculated by the GPU. I begin to think that you just don't want to understand that. You can put 4x 7970 in there and play at 5760x1080 with SSAA - in this particular scenario the FX-rig will produce at best of 41fps avg and the 2500K-rig 58fps.

It is so easy, really:
Test the CPU in a CPU demanding scene - x fps
Test the GPU in a GPU demanding scene - y fps

Will a GPU-limited benchmark show how the system fares in CPU-intensive scenarios? No.
Will a GPU-limited benchmark show how the system will perform if you upgrade your graphics card? No.

You cannot test a 2 component system (CPU and GPU) with one single benchmark. In no field I can think of does this make sense.

 

[cmdrdredd]

does it matter between 120Hz and 60Hz

It's a topic for another thread but I'll answer anyway. It does in the fact that when you use vsync you're stuck at 60fps on a 60hz monitor. If you remove vsync you get screen tearing. With a 120hz monitor you can use vsync and go all the way to 120fps which is a lot better. 120hz monitors generally show motion at 60fps as a smoother video to your eye. Some people claim it's placebo but side by side I can tell. I'm going to upgrade to 120hz myself in the near future.

When testing CPUs and locking to 60hz you're limiting the CPU's ability to push framerate and lowering the load. Games like Battlefield 3 show extremely different characteristics when you remove the 60hz limit and turn vsync off. The CPU starts to have a hard time keeping up with the GPU as it's load goes up.

 

[AtenRa]

The graphics driver does not improve CPU performance. If they were held back by the 6970, the FX and the 2600K would perform the same.

For the 100th time:
It doesn't make sense to benchmark CPUs with AA on as this is calculated by the GPU. I begin to think that you just don't want to understand that. You can put 4x 7970 in there and play at 5760x1080 with SSAA - in this particular scenario the FX-rig will produce at best of 41fps avg and the 2500K-rig 58fps.

It is so easy, really:
Test the CPU in a CPU demanding scene - x fps
Test the GPU in a GPU demanding scene - y fps

Will a GPU-limited benchmark show how the system fares in CPU-intensive scenarios? No.
Will a GPU-limited benchmark show how the system will perform if you upgrade your graphics card? No.

You cannot test a 2 component system (CPU and GPU) with one single benchmark. In no field I can think of does this make sense.

1080p DX-11 Gaming Evaluation. That is the topic of my review.

First line in the review.

This review is for people that really would like to know how those two CPUs are performing in today’s DX-11 games with a High-End GPU like the HD7970 at 1080p with gaming IQ settings.

Im not trying to bench the CPU or the GPU. Im trying to show you what performance to expect when gaming at 1080p with a high end GPU paired with those three processors.

The scenario is simple. You have $500-550 to spend for the GPU. You buy the HD7970 and you have a 1080p 60Hz monitor. Does any of the three CPUs of the review will bottleneck your new High-End GPU in today's DX-11 games at 1080p ??? There are some Games that Intel CPUs are faster by a few fps and FX is faster in others but most of the DX-11 games are GPU limited.

Will Overclocking your CPU give you any substantial performance gains in today's DX-11 Games at 1080p ??? In some Games it will, in most of the others it will not because your GPU will dictate the performance limit.

Benching at 1024x768 will not give you any valuable data when you actually gaming at 1080p. Core i5 2500K may be 2x faster than FX at 1024x768 but at the resolution and with the IQ settings you are actually playing at (1080p) it doesn't matter because the game is GPU limited.