Gaming Shoot-Out: 18 CPUs And APUs Under $200, Benchmarked

[army_ant7]

Though some of us may know what the outcome would be, it would've been nice seeing a 2nd/3rd gen i7 in here, just to show what effect 8 HTT threads have on some games. Other than that, I like the spectrum of CPU's used.

Only the FX-4300, A10-5800K, and FX-4170 have a 95th percentile lag of more than eight milliseconds.

You may have meant "...are the only quad-cores (or dual-modules) that have a 95th...of 8ms or more"

Also, about the explanation of how 75th and 95th percentile work, IMO the wording needs a little work, just like someone before me suggested. (no offense) It's just that saying that:

the 75th percentile result shows us the longest lag between consecutive frames that we see 75 percent of the time

leaves room for misunderstanding (unless you're already familiar with what percentiles are exactly, which I wasn't).
For instance, one may wonder which 75% of the time? From the beginning to 75% of the time? For 75% of shortest latencies?
It took me quite a while to figure out. (It bothered me since I wasn't able to figure out what the data actually meant, aside from the shorter the better.) I only figured it out after going through the (long list of) comments and also the Wikipedia article on percentiles. Maybe I'm just stupid... Hehe...

 

[Guest]

Look how they threw in a non sub $200 processor just so they could have Intel at the top of the charts in most tests. Reminds me of when the Athlon 3000 was blowing away everything except the p4extreme ( a $1000 processor vs a $200 one).
Anyone with a brain knows that the quad core A8 series is a massive win for AMD in the budget space. For low end gaming, htpc etc, intel isn't even in the ballpark. These processors integrate a perfectly respectable gpu into the processor itself. Note the lack of comparisons with , say hd4000 graphics, which also occupy the budget gaming space ( supposedly ! hd4000 is a joke ).

 

[idecris]

Look how they threw in a non sub $200 processor just so they could have Intel at the top of the charts in most tests. Reminds me of when the Athlon 3000 was blowing away everything except the p4extreme ( a $1000 processor vs a $200 one).
Anyone with a brain knows that the quad core A8 series is a massive win for AMD in the budget space. For low end gaming, htpc etc, intel isn't even in the ballpark. These processors integrate a perfectly respectable gpu into the processor itself. Note the lack of comparisons with , say hd4000 graphics, which also occupy the budget gaming space ( supposedly ! hd4000 is a joke ).

no gamer with common sense will game with an A8 and it's integrated graphics.

even the A10-5800k integrated 7660D is just like an x1950 which is idiotically not good enough for modern gaming.

gaming builds is synonymous with discrete graphics.

this is not your forum.

 

[idecris]

FOR THE AMD FANBOYS (to shut you up already)

I'm betting you that

AMD A10-5800K 3.8GHz Quad-Core Processor at $127.29 Amazon with
PowerColor Radeon HD 6670 1GB Video Card at $50.97 Newegg for total = $178.26

will not even come close to:

Intel Celeron G555 2.7GHz Dual-Core Processor at $53.99 Amazon with
EVGA GeForce GTX 650 Ti 1GB Video Card at $124.99 Newegg for total = $178.98

A10 will never be the best budget option. Ever. For idiots probably.

'Nuff said.

 

[idecris]

I'm also betting that the:

Intel Celeron G555 2.7GHz Dual-Core Processor at $53.99 Amazon

will beat any AMD processor

at any processor + video card price point. What do you say Cleeve? Will I win?

Architecture, true value and priceerformance, the real reason to choose Intel.

I'm not an AMD hater, I had an Athlon XP 2500+ Barton build way back and it was the processor with the value and priceerformance then.

But Intel's really just doing very well now especially in terms of value.

 

[SuperVeloce]

lol, try crysis 3 on that low clocked celeron... crysis 3 on dual cores without HT is virtually unplayable.

 

[idecris]

[citation][nom]SuperVeloce[/nom]lol, try crysis 3 on that low clocked celeron... crysis 3 on dual cores without HT is virtually unplayable.[/citation]

LOL! just saw the Crysis 3 benchies... I now want my Vishera CPU hahaha

That APU Crossfire budget rig still sucks though.

 

[Guest]

Don't know about u guys, but sounds like more of a bunch of INTEL fanboys to me. I dont know if this is even on topic, but I do a lot of gaming, and am frequently amazed that the freakin dude with the dm1z ( amd a8 I believe), can actually sometimes beat me at battlefield, on my i7 ultrabook. Can also say I'm pretty impressed too.

 

[palladin9479]

LOL! just saw the Crysis 3 benchies... I now want my Vishera CPU hahaha

That APU Crossfire budget rig still sucks though.

You shouldn't be CFing APU's in general, unless their in some sort of budget laptop build. APU's are for low end small box's without dGPU's, having a dGPU present kind of goes against the entire concept of an APU. The i3 got such high scored for so long because the tests were all focused on one to two cores actually doing work. This ultimately penalized any CPU with more cores. Right ~now~ we're starting to see games come out that need 3~4 cores.

Recently friend of mine built the following for his wife,

Asrock FM2 board
A10-5800K
8GB DDR3-2133 memory
Case + PSU + HDD

Ended up being a really cheap small system that she uses to play online games with him.

 

[SuperVeloce]

sandy/ivy i3 with HT is still going strong, but Pentiums are now getting a beating

 

[rojodogg]

I was very disappointed with the choice of GPU, this is low budget ($200) CPU's yet you used a High end GPU (READ EXPENSIVE) that most of us would never be able to put in a budget system. Why not use a GPU that reflects what most people would use combined with a budget CPU (a GPU under the $200). That would give us a better understanding on this test. Noted that the APU's also crossfire with radeon 6000 & 7000 GPU's this should have been considered in the test also. I am sure that those would have shown a Huge performance boost compaired to the non APU's giving us a better judge on the bang for the buck theory. I did like the test but wondered why you used such high end MOBO's when these CPU's would be used by most with less expensive MOBO's in the $130 and under range. Once again the reason we buy and use CPU's $200 and under is due to cost and bang for the buck so the other hardware should reflect this in your testing. Just my opinion this would help more people out in making educated choices when purchacing a CPU.

 

[SuperVeloce]

what? you need to eliminate GPU bottleneck to see CPU scaling. it's a CPU gaming benchmark, for system builds look for an article about that...

 

[rojodogg]

$399.99 for the Intel MOBO Newegg
$164.99 AMD 990fx MOBO Newegg
$94.99 FM2 MOBO Newegg
$114.99 FM1 MOBO Tiger
$504.99 for the Video Card Tiger

Price always effects our choices and performace choices. I understand wanting to eliminate bottlenecks but this test seem's a little skewed to me and not real world friendly. I like Intel CPU's and Have AMD CPU's. I own a Internet and Computer store plus I build systems for customers. I want the best bang for the buck and this compairison is not a equal test of the market CPU's under $200. I bet if we use a Gigabyte GA-H77-DS3H MOBO LGA1155 Intel CPU's would score lower ($96 Mobo).

 

[palladin9479]

sandy/ivy i3 with HT is still going strong, but Pentiums are now getting a beating

i3 won't survive anything released over the next few years, take a long look at the recent crysis 3 benchmarks. If someone has an I3 they shouldn't immediately need to replace it, but it's something they'll have to plan on. For anyone buying / building right now I'd recommend staying away from anything with only two cores, plan on four from here on out. Honestly anything build in the last year or so should of been 4+ cores. When you build new you want to plan for next years games not last years.

 

[Fulgurant]

i3 won't survive anything released over the next few years, take a long look at the recent crysis 3 benchmarks. If someone has an I3 they shouldn't immediately need to replace it, but it's something they'll have to plan on. For anyone buying / building right now I'd recommend staying away from anything with only two cores, plan on four from here on out. Honestly anything build in the last year or so should of been 4+ cores. When you build new you want to plan for next years games not last years.

Eh, I understand where you're coming from, but I think your tone is (perhaps unintentionally) a little alarmist.

The Core i3 will have about as much future-proofing value as most any low-end CPU ever has -- which is to say, less than mid-range CPUs, and probably a little more (on a performance per dollar basis) than high-end CPUs. But ultimately the question is subjective: how long a given piece of hardware will "last" depends on your tolerance for its performance running the software that's important to you.

I submit that the average i3 buyer will be satisfied with middling performance in games over the next two or three years. By no means will the i3 become obsolete in the next couple of years, unless the user's expectations are too high.

All of this is just a long way of saying that games aren't suddenly going to require 4+ cores overnight. Dual-core processors will die a slow death in much the same way that single-core processors took years and years to die. It is not in game developers' best interests to limit their target audience.

 

[idecris]

It is not in game developers' best interests to limit their target audience.

true. not to mention that the largest market for games globally are the budget gamers. think about that.

 

[palladin9479]

Eh, I understand where you're coming from, but I think your tone is (perhaps unintentionally) a little alarmist.

The Core i3 will have about as much future-proofing value as most any low-end CPU ever has -- which is to say, less than mid-range CPUs, and probably a little more (on a performance per dollar basis) than high-end CPUs. But ultimately the question is subjective: how long a given piece of hardware will "last" depends on your tolerance for its performance running the software that's important to you.

I submit that the average i3 buyer will be satisfied with middling performance in games over the next two or three years. By no means will the i3 become obsolete in the next couple of years, unless the user's expectations are too high.

All of this is just a long way of saying that games aren't suddenly going to require 4+ cores overnight. Dual-core processors will die a slow death in much the same way that single-core processors took years and years to die. It is not in game developers' best interests to limit their target audience.

Not alarmist just stating that anyone building new right now should avoid anything with less then four real cores.

Games requiring 4+ "cores" didn't happen last night, it happened last year and has been slowly gaining momentum. It's not a switch on a wall, instead it's a gradual occurrence with each new iteration requiring more then the previous ones. BF3 in large multiplayer maps required 3~4 cores to run really well, something that wasn't really demonstrated by the single player looped demo benchmarking everyone was doing. Crysis 3 does more physics calculations on the CPU then many previous games, those calculations are whats being passed to the additional cores. DirectX 11 allows for parallel rendering of single threaded operations. As more games utilize DX11 features you'll see a sharp uptick in the amount of parallel work done and as a result the number of simultaneous threads needed.

So anyone with an i3 "now" is going to be ok for another year or so, no need for immediate replacement. Anyone building new is going to need to think about running four or more simultaneous threads, this means low end i5's and some of the four "core" PDs. Essentially the days of an i3 beating a low end i5 and scoring around the same as a high end i5 are about to end.

 

[Fulgurant]

Not alarmist just stating that anyone building new right now should avoid anything with less then four real cores.

So anyone with an i3 "now" is going to be ok for another year or so, no need for immediate replacement. Anyone building new is going to need to think about running four or more simultaneous threads, this means low end i5's and some of the four "core" PDs. Essentially the days of an i3 beating a low end i5 and scoring around the same as a high end i5 are about to end.

Right. And my point is that the i3 is a budget CPU. It shouldn't be on par with an i5. Anyone in the market for a budget CPU (or who bought one recently) should be ok with less-than-i5 performance.

If anything, the i3 has been too good a value thus far (gaming-wise). That's a good thing. The fact that that situation won't last much longer shouldn't be held against the product, which will continue to perform just fine for the market for which it was intended.

If the user has the money to spend on an i5, then sure, I agree with your premise. But if you insist that anything less than an i5 will become inadequate for a budget user in the near future, I have to disagree. The question, as always, comes down to the user's expectations. You're right to point out that some users might have grown spoiled with respect to the i3's current status (in games), but that's not the i3's fault.

 

[palladin9479]

If anything, the i3 has been too good a value thus far (gaming-wise). That's a good thing.

Actually it's was a very bad thing. The i3 wasn't that good but that the modern software we use for metrics was that bad. An i3 has exactly 50% of the processing capability of an i5 (6 ALUs vs 12) and both have less then an fx8xxx (16). The problem was efficiently utilizing that processing power which games weren't able to do. Someone who purchased an i5 for example would only be utilizing 50% or less of that processor's capability, the rest would be dormant and underutilized.

That's changing right now as we speak, started last year and will continue onward. Where previously something like an i3 had good performance / price vs other processors, nowish / soon that will no longer be the case. Next year is when you can start to expect titles to appear that require more oomph then an i3 will be capable of delivering. We experienced something similar to this when we went from single core CPU's do duals.

 

[army_ant7]

@palladin9479
Though I don't disagree with the i3 having 50% of the processing capability of the i5, we still have to remember how that may not be the case for actual performance as the i3 has Hyper-Threading and is thus able to more efficiently utilize that 50% of processing capability compared to how the i5 uses its 100%, at least in some situations, I think.

Aside from that, seeing as how the 2nd/3rd Gen. i3 still outperformed the quad-AMD CPU's in the benchmarks used for this article, I would say that it does provide good actual performance for a 2 core (4 thread) CPU. I don't doubt the possibility of games becoming more and more performance-intensive in the future, but maybe the i3 would then become what the Athlon/Phenom II X4's are now--"mostly" sufficient gaming CPU's which I think a lot of people are still satisfied with. No doubt the difference in actual application performance between the i3 and i5 would widen as applications get better threaded, but that's not to say that the i3 wouldn't still be sufficient for people.

 

[Fulgurant]

If anything, the i3 has been too good a value thus far (gaming-wise). That's a good thing.

Actually it's was a very bad thing. The i3 wasn't that good but that the modern software we use for metrics was that bad. An i3 has exactly 50% of the processing capability of an i5 (6 ALUs vs 12) and both have less then an fx8xxx (16). The problem was efficiently utilizing that processing power which games weren't able to do. Someone who purchased an i5 for example would only be utilizing 50% or less of that processor's capability, the rest would be dormant and underutilized.

Heh, you know what I meant. The situation (in which the i3 was competitive with the i5 in games) was a good thing as far as the i3 was concerned, a point in the product's favor.

In any case, there are diminished returns on parallel processing power. It's not a flaw in the software industry, per se, that software doesn't rush to exploit the latest multi-core configuration. It's a matter of cost efficiency; coding for parallel processing takes extra time and effort (and sometimes, beyond a certain point it's just not feasible). And if the market isn't saturated with users who possess the latest multi-core configuration, then that effort isn't worthwhile.

That situation changes slowly over time. I still have a dual-core Opteron 165 rig in my living room that I use for casual 2d stuff and network storage. I bought it in 2005, back when dual-core CPUs were basically state of the art. Given that I still use that rig, it'd be easy for me to say that buying a dual core back then was the right choice; a single-core proc would probably have a hard time booting Windows 7. (And frankly, the Opteron 165 was an insanely good deal back then.)

But I can't honestly say with any degree of certainty that I couldn't have netted similar or even superior value out of a cheaper single-core proc at the time, followed by another cheaper CPU later, and so on and so forth.

All of that said, and by the way, an i3 doesn't have exactly 50% of the processing power of an equal clocked i5. At worst, the i3 has (2 * 1.3 Hyperthreading) / 4 = 65% of an equal-clocked i5's processing power -- and that's just the abstract figure. In fact, extra cores don't scale linearly across all tasks; you can blame the software all you like, but it is what it is. Granted, the i5 can clock significantly higher than an i3, but that's beside the point.

That's changing right now as we speak, started last year and will continue onward. Where previously something like an i3 had good performance / price vs other processors, nowish / soon that will no longer be the case. Next year is when you can start to expect titles to appear that require more oomph then an i3 will be capable of delivering. We experienced something similar to this when we went from single core CPU's do duals.

(Emphasis mine.)

Agree to disagree. I still think you're over-stating your case. Games will remain playable on an i3 for a long time to come. If a given user expects to run the latest first-person-shooter games at max settings on an i3 over the next two or three years, that's the user's problem, not the i3's.

 

[palladin9479]

Stop

Hyperthreading adds absolutely zero capacity to a CPU, it doesn't generate 30% more resources. What HT does is allow software to more efficiently utilize existing CPU resources.

Case in Point

One SB core has exactly three integer units. This means it can process up to three integer micro-ops per cycle. Now x86 is a very serial code, it does things one instruction at a time so the possibility of doing more then one integer micro-op is very slim. Branch prediction and instruction prefetch allow the CPU to attempt to do additional integer options ahead of time though you rarely get more then 1 extra instruction.

HT brings this further by allowing the OS to schedule another set of code that the CPU will then attempt to run on those underutilized resources. This is a very important distinction to make because some programs are better optimized for prefetch and branch prediction then others, the better optimized ones won't get much from HT as their already utilizing the additional resources. So instead of having 3 ALU's dedicated to each thread you can share those three ALU's between two threads. You still have 3 ALU's per core, your just using them more efficiently.

This is also the problem that BD and to a lessor degree PD ran into. AMD's 8xxx chips have 16 ALU's divided up amongst eight cores at 2 ALU's per. In pure theoretical numbers that chip can push more micro-ops then the i7's. Where AMD failed was the additional latency of their caching and the poor performance of their branch predictor. Those problems severely limit the speed at which you can shove instructions through the pipelines and that's who you got SB crushing PD in IPC. The FX CPU's had more resources but had a poor utilization rate.

 

[army_ant7]

@palladin9479
Good explanation of Hyper-Threading and the BD/PD situation. (Really, I for one appreciate it.)
Though both still do seem to support how the i3 isn't necessarily a bad choice.

Though it won't happen all the time (based on your explanation), HT can substantially help the i3's actual performance (which is what really matters as shown by your BD/PD explanation) through efficiently utilizing the possibly underutilized resources compared to a core i5.

In most games as it seems, BD/PD < i5. Just as your said, BD/PD may have more muscle, but it inefficiently uses it.
Now, this isn't an analogy pointing to the i3 being equal to or greater than the i5 (that would of course be preposterous for me to say ), but that the i3 doesn't always just perform 50% of what the i5 does. It could perform more than 50% due to HT.

Like you were saying, as games become more threaded, the performance gap between the i3 and the i5 will widen, but that's not to say the i3 won't be sufficient anymore for some (or most). It has been seen outpacing the Phenom/Athlon II X4 in games where the the latter two have been able to spread their...threads. (Though I'm not saying that you're saying those AMD quad-cores will be better than the i3 just because they're quad-cores.)

(Another thought is that current octa-core BD/PD may have better experiences with up-coming games. )

 

[de5_Roy]

it's worse to build a new gaming pc with a core i3 (sb/ivb) right now than it was in 2011. lga 1155 is dead end and core i3 cpus are still overpriced. in a few months haswell cpus will launch and existing lga1155 platform will be phased out. so new core i3 users will have to buy overpriced i5 or switch to haswell/amd next year.
when you upgrade to a quadcore (you should) you end up paying more. core i3 buyers tend to buy h and b series motherboards, so even after upgrading to any core i5 you can't oc the cpu. most of the h and b series don't support cfx/sli. some b series mobos support evenly splitting pcie lanes but that's it. pentium to i5 is a good upgrade but core i3 right now is of poor value as more games start to use more real cores let alone multiplayer games.
i dunno how haswell core i3s are gonna perform, so those are out of the equation right now.

 

[Fulgurant]

Stop

Hyperthreading adds absolutely zero capacity to a CPU, it doesn't generate 30% more resources. What HT does is allow software to more efficiently utilize existing CPU resources.

No, 30% is an estimate of HT's practical performance benefit. Although I -- like army_ant -- appreciate your detailed analysis of the hardware's mechanics, I also can't accept the premise that we should draw a hard and unpassable line in the sand between the hardware's theoretical capabilities and the current bottom-line performance of that hardware in various applications. For the purpose of this discussion, you're pointing to a distinction without a difference.

If it's reasonable to assume that game developers will optimize their software, more and more, for 3+ core CPUs, then it's also reasonable to assume that they will optimize their software for one of Intel's flagship features.

Clearly, Hyperthreading adds something -- and it's something that i3s have, but i5s don't.

This is also the problem that BD and to a lessor degree PD ran into. AMD's 8xxx chips have 16 ALU's divided up amongst eight cores at 2 ALU's per. In pure theoretical numbers that chip can push more micro-ops then the i7's. Where AMD failed was the additional latency of their caching and the poor performance of their branch predictor. Those problems severely limit the speed at which you can shove instructions through the pipelines and that's who you got SB crushing PD in IPC.

Yes, exactly. And until such time as AMD fixes those problems, their theoretical advantage is functionally irrelevant. As a consumer, am I supposed to care why AMD is getting crushed by Intel in IPC?