Sign in with
Sign up | Sign in

Test System And Benchmarks

AMD FX Vs. Intel Core i3: Exploring Game Performance With Cheap GPUs
By

One of the challenges we face every time we analyze game performance is that most titles offer a large number of customizable graphics settings. Typically, we apply the same settings to every contender. The result is often that low-end GPUs hold performance back when they try to push the same options as the high-end models. Those results end up being unrealistically-low. Nobody plays a game on a Radeon HD 5570 at 12 FPS using the same settings as a Radeon HD 7970.

In order to keep our scores in tune with reality, we're going to try using different settings for each graphics card. How do we choose the right options? Some folks find that, in some games, 30 FPS is perfectly playable, while others won’t take any less than 60 FPS across all of the titles they enjoy. There’s no objective way to choose settings, so we test each game twice, each time using a different performance target.

Our first performance target is a 30 FPS minimum. The video game industry traditionally considers this to be the goal for smooth game play. So, we choose settings that keep the game's minimum frame rate around 30 FPS, but at the highest detail and resolution possible. If you value resolution and visual fidelity over all else, this is the kind of performance target you’re probably shooting for.

The second target is for gamers who want a more fluid experience. In this case, we’re looking for about 60 FPS average, and we don’t want to dip below 40 FPS. This imparts a much smoother feel than 30 FPS minimum, even though we’ll probably have to drop visual settings and resolution to achieve it with lower-end hardware. Gamers who enjoy first-person shooters often favor responsiveness over visual fidelity, especially in competitive environments.

Our FX-4100 is actually an FX-8120 with two of its Bulldozer modules turned off. We've run exhaustive testing against an actual FX-4100 to confirm that the performance is similar. The only complication is a lack of granularity in setting its Turbo Core multiplier. So, the CPU itself runs at 3.8 GHz (rather than ranging between 3.6 and 3.8 GHz). With that established, a 100 or 200 MHz difference isn't going to impact our results noticeably.

Finally, we want to point out that the $125 Core i3-2100 is only $3 less than the Core i3-2120, a CPU that is 200 MHz faster. Although the -2120 would be our recommended buy, we're using the -2100 because that's what we have on-hand.


Socket AM3+
LGA 1155
CPU

AMD FX-4100 (Zambezi), 3.6 GHz Base, 3.8 GHz Turbo Core

Intel Core i3-2100 (Sandy Bridge), 3.1 GHz, Hyper-Threading enabled
Motherboard

Biostar TA990FXE
Socket AM3+, Chipset: AMD 990FX

Asus P8P67 Pro
LGA 1155, Chipset: Intel P67 Express
Networking
On-Board Gigabit LAN controller
Memory

Corsair Vengeance LP PC3-16000, 2 x 4 GB, 1600 MT/s, CL 8-8-8-24-2T

Graphics

AMD Radeon HD 5570
650 MHz GPU, 1 GB DDR3 at 900 MHz

AMD Radeon HD 6770

850 MHz GPU, 1 GB GDDR5 at 1200 MHz

AMD Radeon HD 6850

775 MHz GPU, 1 GB GDDR5 at 1000 MHz

AMD Radeon HD 6950
800 MHz GPU, 1 GB GDDR5 at 1250 MHz

Hard Drive

Western Digital Caviar Black 750 GB
7200 RPM, 32 MB Cache, SATA 3Gb/s

Power

ePower EP-1200E10-T2 1200 W
ATX12V, EPS12V

Software and Drivers
Operating System
Microsoft Windows 7 x6, Service Pack 1, KB2645594 and KB2646060 installed
DirectX
DirectX 11
Graphics Drivers

AMD Catalyst 12.1

Benchmark Configuration
3D Games
Metro 2033
Version 1.0.0.1, Built-In Benchmark
Battlefield 3
Version 1.0.0.0, Operation Swordbreaker, FRAPS runs
Elder Scrolls V: Skyrim
Version 1.4.21.04, FRAPS runs
DiRT 3
Version 1.2.0.0, Built-In Benchmark
Just Cause 2
Version 1.0.0.2, Concrete Jungle Benchmark
StarCraft 2
Version: 1.4.2.20141, Tom's Hardware Guide Benchmark
Ask a Category Expert

Create a new thread in the Reviews comments forum about this subject

Example: Notebook, Android, SSD hard drive

Display all 167 comments.
This thread is closed for comments
Top Comments
  • 29 Hide
    lemlo , February 23, 2012 3:55 AM
    Quote:
    Whats with the line graph Don, it's hard to read especially with the choice of color on the lines. Bring back the bar graphs.:) 



    The line graph is better way to show it's behaviour over a period of time rather than a flat average, which doesn't explain frequent dips or long stretches of smooth gameplay in fps and such.

    A very informative and realistic article, nice work Tom's. Lets hope AMD has something with piledriver.
  • 28 Hide
    compton , February 23, 2012 3:28 AM
    I think if you're really budget limited, but need to build a system today, buying 1155 makes a little more sense. Get a decent motherboard, then get an 1155 Celeron G530/540 or Pentium. Then save up some money, and upgrade to an Ivy Bridge CPU later. The i3 is great, but the i5s are a much better deal $/performance.

    So if you need a new system and can't afford an i5, just buy a cheap 1155 SB as a placeholder until you can [afford an i5]. Like the conclusion states, the upgrade path is there -- I just think that if it's a temporary step, you might as well save $60 to $80 if you're upgrading in the next 4 months anyway. You'd be surprised how fast the SB budget parts are, and they're fast enough to get you through till IB.

  • 22 Hide
    jjb8675309 , February 23, 2012 3:28 AM
    This is a great article, Toms needs more budget rundowns like this and perhaps more games in the test sweet that exploit the cpu difference more...
Other Comments
  • 11 Hide
    ilysaml , February 23, 2012 3:18 AM
    Great...AMD is still capable.
  • 22 Hide
    jjb8675309 , February 23, 2012 3:28 AM
    This is a great article, Toms needs more budget rundowns like this and perhaps more games in the test sweet that exploit the cpu difference more...
  • 28 Hide
    compton , February 23, 2012 3:28 AM
    I think if you're really budget limited, but need to build a system today, buying 1155 makes a little more sense. Get a decent motherboard, then get an 1155 Celeron G530/540 or Pentium. Then save up some money, and upgrade to an Ivy Bridge CPU later. The i3 is great, but the i5s are a much better deal $/performance.

    So if you need a new system and can't afford an i5, just buy a cheap 1155 SB as a placeholder until you can [afford an i5]. Like the conclusion states, the upgrade path is there -- I just think that if it's a temporary step, you might as well save $60 to $80 if you're upgrading in the next 4 months anyway. You'd be surprised how fast the SB budget parts are, and they're fast enough to get you through till IB.

  • 16 Hide
    esrever , February 23, 2012 3:41 AM
    would be nice if more benchmarks were done, there are a lot of popular games that would be nice to test like civ 5, l4d2 or similar source game, mw3, SWTOR ect. Even if some of them aren't the most demanding games it would be nice to see them as they would be more relevant than dirt 3 or battlefield 3 single player.
  • 5 Hide
    jp182 , February 23, 2012 3:48 AM
    esreverwould be nice if more benchmarks were done, there are a lot of popular games that would be nice to test like civ 5, l4d2 or similar source game, mw3, SWTOR ect. Even if some of them aren't the most demanding games it would be nice to see them as they would be more relevant than dirt 3 or battlefield 3 single player.


    It would be nice if they through Civ 5 or MW3 in but at least on the FPS front, I think BF3 has a bigger following on the PC and the same thing goes for Skyrim. Not sure how many people are still playing Just Cause 2 though. In either case, I think this has more to do with being able to compare these results to the results from past benchmarks they've run.
  • 13 Hide
    manu 11 , February 23, 2012 3:50 AM
    Thanks for appreciating our feedback, thank you very much. Great Article As always.
  • 29 Hide
    lemlo , February 23, 2012 3:55 AM
    Quote:
    Whats with the line graph Don, it's hard to read especially with the choice of color on the lines. Bring back the bar graphs.:) 



    The line graph is better way to show it's behaviour over a period of time rather than a flat average, which doesn't explain frequent dips or long stretches of smooth gameplay in fps and such.

    A very informative and realistic article, nice work Tom's. Lets hope AMD has something with piledriver.
  • 9 Hide
    amdfangirl , February 23, 2012 4:04 AM
    Sucks that the Core i3 can't be overclocked like the legendary e4xxx series or the e2160 which you could get a 100% OC with.

    If DC Sandy Bridges could be unlocked, they would be so good for gaming.
  • 14 Hide
    erunion , February 23, 2012 4:04 AM
    Love the FPS graphed over time. Keep using them!
  • 16 Hide
    king_maliken , February 23, 2012 4:08 AM
    Brandon SI love my i5 2500k and I will never go back to amd


    This is all kinds of wrong... "NEVER" is really idiotic to say in this situation, you don't know, AMD might come out with something that will in the future be the best performer. You son have a lot to learn yet and probably have a lot of living left to do.
  • 5 Hide
    darkchazz , February 23, 2012 4:15 AM
    Hey Toms, you want good game for testing cpus ? -> GTA IV
    :p 
  • 21 Hide
    CaedenV , February 23, 2012 4:21 AM
    Brandon SI love my i5 2500k and I will never go back to amd

    That is short sighted. I love my Intel build, but in 4-5 years when I upgrade again I will jump all over AMD if they have something good... but it just is not looking good now.
    reyshanWhats with the line graph Don, it's hard to read especially with the choice of color on the lines. Bring back the bar graphs.

    I love the line graphs! They show what we need to know, and more than the overly simplistic min/ave/max.
  • 8 Hide
    williehmmm , February 23, 2012 4:24 AM
    I complained and complained on the most recent CPU recommendations list that the FX 4100 should not be 3 tiers lower than the i3 2100.

    The i3 is the better/faster chip for gaming, but not so much that you should spend a lot upgrading to it from a "somewhat parallel" performing FX 4100.

    Quote - "I don’t recommend upgrading your CPU unless the potential replacement is at least three tiers higher. Otherwise, the upgrade is somewhat parallel and you may not notice a worthwhile difference in game performance."

    This article at least shows that their will not be a noticeable difference in game performance. And I would go as fas as to say that once overclocked, there would be no difference whatsoever between the FX and an i3 (which is locked and can't be overclocked).

    Absolute respect to Toms for taking onboard these points and hopefully rearranging the table for the March CPU hierarchy chart. And it would seem the G860 & i3 should be closer together too.

  • -5 Hide
    williehmmm , February 23, 2012 4:35 AM
    Also this "Even a $200 FX-8120 won’t solve your problem; our tests show that chip acts just like the FX-4100 in gaming environments."

    and this -

    http://www.tomshardware.com/reviews/fx-8150-zambezi-bulldozer-990fx,3043-18.html

    The FX 8150 seems to be absolutely equal to an i5 2500k at extreme resolutions 2560x1600, ultra detail levels, x8 AA. Would the FX 4100 deliver the same equal performance at these resolutions?

    I've never even seen a monitor that has that kind of resolution. 4 megapixels. However eyefinity and Nvidia surround do. 2.7 megapixels for a 720p 3 screen setup and 5.2 megapixels for a 1680x1050 3 screen setup.

    Would a $110 FX4100 deliver the same FPS as a $230 i5 2500k? The FX 8150 did.
  • 4 Hide
    Stardude82 , February 23, 2012 4:54 AM
    comptonGet a decent motherboard, then get an 1155 Celeron G530/540 or Pentium.


    Totally, I bet you could feed a 6770 just fine with a $50 CPU. No need for these extravagant $120 chips.
  • -8 Hide
    ivaroeines , February 23, 2012 4:56 AM
    I will still buy AMD, not because they are better or cheaper than Intel i know they arent, but because they are good enough for me and that Intel need to be challenged. To me it seems like "Tom's hardware" is on a crusade to bring down AMD, looking back to the release of the Core i7 cpu's "Tom's Hardware" have used encryption suits in testing cpu performance and made a big point of how excellent Intel cpu's are at encryption, i dont know of anyone besides myself( i use it only for testing ) who use encryption on their computers. I have always felt that the encryption suits "Tom's hardware " employ have been used to show how bad cpu's AMD makes, AMD dont make bad cpu's( to my knowledge that is ), at any given price point up to $800 to $1000 i think most people wouldnt recognize the difference between a AMD and a Intel system in a blind test.

    The thing that made me raise my eyebrows in regards to "Tom's Hardware" wasnt any cpu test, but the tests of the HD 79** series, in a test where the results showed better results in most benchmark and trashing the competition in 1, the conclusion from "Tom's Hardware" was the this gpu wasnt any good and i almost felt like "Tom's Hardware" was warning me against buying such a card.

    This may just be the rantings of a AMD fanboy, but i think "Tom's Hardware" need to see if they are as objective as they claim to be.
  • 3 Hide
    de5_Roy , February 23, 2012 5:17 AM
    nice article! you addressed and cleared up a lot of issues concerning the 4100.
    good to know that fx 4100 is viable for using with entry level gfx cards. actually it makes more sense to couple the $110~ cpu with a similarly priced gfx card.
    as for cpu limitations in cfx/sli: imo those are more likely to come out in budget pcs than higher end pcs as budget gamers might want to upgrade their gfx card or add another for cfx/sli relatively sooner than people who build with cfx/sli in mind or start with 2x cards. intel's h61 and most h67 mobos would be useless for cfx/sli but a lot of cheap p67 and z68 mobos can support cfx/sli. i suspect that a lot of people who bought pcs with llano apus might eventually want to upgrade /add gfx cards without changing their apu and mobo.
    another issue most amd users/amd-biased people usually avoid: power consumption. in budget gaming pcs power consumption matters because higher load power consumption results in higher wattage (likely costlier) psu. the locked core i3 uses less power than fx4100, so builders can afford to add higher tdp gfx cards or save money with smaller psu yet use a high perf. gfx card.
    overall core i3 is still much more suitable for budget gaming.
  • 6 Hide
    de5_Roy , February 23, 2012 5:28 AM
    Quote:
    something about encryption (read the reply below) and tom's intel bias (totally baseless), something about being amd fanboy and something about the 7970 'preview' article (off topic....) .

    phenoms and athlons didn't have aes hardware acceleration. llano apus don't have it either. trinity might have it, but i doubt that. zambezi actually supports aes hardware acceleration. 4/6/8 core (2/3/4 modules) zambezi can outperform 2/4 core intel cpus respectively in encryption benchmarks. some review sites that favored fx8150 used encryption benchmarks (among others) and it's high scores in those benchmarks to pitch it as a 'great cpu'. they also undermined it's power consumption. :) 
    intel's sb core i5 and i7 support aes acceleration, core i3 doesn't. so in an encryption benchmark, the core i3 would lose to fx 4100. your claim is baseless.
  • 6 Hide
    haplo602 , February 23, 2012 5:36 AM
    One of the best articles I read in a while. Observing the bottlenecks to shift from CPU to GPU and back is great. Also it showed that if the CPUs are paired with adequate GPUs, there is virtualy no difference in most applications.

    The 6850 looks to be the best one to use as it becomes the bottleneck at 1080p for most tested games.

    Also can we have a similar article with Nvidia cards please ? :-)))
Display more comments