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

[Guest]

I love the intent of this article, but i hate the way the data is presented.

The number one performance metric i care about by games is lowest framerate, highest latency. I dont care what the average is if it freezes or stutters.

By showing average latencies....you are defeating the whole purpose of this article... What would be MUCH better is a chart that shows how many times a high latency happened.

How much lag is noticeable is a matter of debate. I dont think anyone would argue that you can notice 10 milliseconds. But 100 would be easily noticeable.

Id like to see a chart like this:

# of frames 0-10ms : xxx (who cares, you wont notice)
# of frames 11-15ms: xx
# of frames 16-20ms: xx (probably the first spot that some people would notice, sub 60 fps)
# of frames 21-25ms: xx
# of frames 26-30ms: xx
# of frames 31-35ms: xx (starting to get noticable to most, micro pauses, ~30 fps)
....
# of frames over 50ms: xxxx (not good, this would be less than 20fps, anyone should be able to notice)
# of frames over 100ms: xxx (massive fail, this would be less then 10fps)

Those last 2 lines are really what i care about, they are the fail lines, they are the DO NOT BUY THIS HARDWARE lines.

Above applies to standard displays, a 3d display, where it alternates left eye right eye, might display noticeable issues at lower latencies, i'm not sure i don't own one.

 

[cleeve]

I love the intent of this article, but i hate the way the data is presented...

...By showing average latencies....you are defeating the whole purpose of this article... What would be MUCH better is a chart that shows how many times a high latency happened...

We don't show averages only, we show percentiles also.

What you propose is what the percentile results cover, which we have already done!

 

[blazorthon]

[citation][nom]SuperVeloce[/nom]are you sure about that? As far as I know, AMD did integrate both controllers on chip even before Intel...[/citation]

Memory controller, yes, but not the PCIe controller AFAIK. I think that is still on the motherboard. It's been a whiel since I've read up on it, so maybe I'm mistaken, but I'm pretty sure about it and that it's part of why AMD's differing chipsets can have different numbers of PCIe lanes without a chip such as a PLX chip in the top models.

 

[guerrero]

So my i3 2120 does better then the newer i3 3220? hhhmmmm

 

[Guest]

That's precisely what the new chart does!

As we mention, we compare CONSECUTIVE (side-by-side) frames in the stream. That means we're not looking at each frame on their own, we're looking at how each frame compares to the frames next to it.

sorry but the article even states the chart lacks data:

As you can see, this chart does not reflect raw frame rates. That's not its job. And that's fine with us because we're still going to continue capturing average frame rate for the foreseeable future. It may not tell the entire performance story, but it remains an important metric. We're simply adding the new data to help fill in the blanks.

to be blunt, it isn't hard to record the frame times using FRAPS and export the CSV file to excel and make a graph.

oh, but you didn't discuss how you measured latency nor qualified it. so i guess all the data and how it is measured is suspect . . . nevermind.

 

[cleeve]

to be blunt, it isn't hard to record the frame times using FRAPS and export the CSV file to excel and make a graph.

Yeah, you kind of need to actually read the article to understand.

Give page 2 more than a cursory look and then try again.

 

[Wendigo]

[citation][nom]cleeve[/nom] There's no universal law that states that 30 FPS is the minimum for smooth gaming. It's a sort of common denominator that's somehow made its way as a de-facto...[/citation]

I may be mistaken, but as far as I know, the 30fps rule comes from the fact that it's the lowest "smooth" fps that you can get if you're using vsync in your game. The next step is 20fps, which has obvious stuttering for anyone.

 

[Guest]

Yeah, you kind of need to actually read the article to understand.

Give page 2 more than a cursory look and then try again.

excuse me, i have read and re-read the article several times. i have also dissected your graphs and how it is presented. now my comprehension may have fallen due to age and abuse i have subjected my brain cells to when i was much younger.

to say, this graph represents that graph doesn't work. you may know how the data was collected and you may know what you are representing but, no matter how many times i reread 5 paragraphs; its not explained how an average of ~40ms latency equates to a baseline of 0ms nor how one instance of a 15ms difference is represented by a 29.9ms 95th percentile.

 

[blazorthon]

[citation][nom]sarinaide[/nom]Blaz I am not doubting results, results are just not universal hence why different unless you universalize the test methods.Sure you get skimped Z77's which you can spot a mile away and enjoy the RMA's that go with them. When Sandy was out got to test Gigabyte boards, anything below UD5 was stay away territory despite the costs involved. This has carried over to Ivy and basically every manufacturers low end boards have the higher RMA rates, I know this because I do this as a hobby job. Also $100 Z77 going on the above is very skimped on less SATA, Less USB, Slower DIMM's, low quality VRMS, limited PCI-e bus interface etc etc. Consider the lower end AM3+ boards still support more than the above and a $100 FM2A85 board features almost everything the UP7 used has, I don't think you really get good value below $200 with Intel chipsets. I certainly don't see a i3 as better out the box than AMD across the line, it is certain games and mostly single player in general its either just in front or behind the FX6300. As for overclocking AMD parts to get performance, I have the 8350 ES chip which is clocked at 3.3ghz, it runs every game well and significantly faster than the FX8150 in general feel less twitchy and beats my Thuban in gaming and sizeably in general computing. In general with socket 1155 EOL soon and the rate at which intel chips disappear after its generation going on how quickly 1366 went away, if stuck on a i3 that is as good as it will ever get going on the fact a buyer of said part will be a budget builder and have no money to upgrade to a i5 before they become non existent. As we know with AMD socket stability allows flexibility in upgrading. Going on this and experience the only feasible intel SKU's are i5k's, i7k's and the 3930k and 3960X. I did mention in the intel thread that Intel also skimps its low end by putting top end igpu's on high end chips which are never going to use them and the i3's that do support it are often like $150 which is ludicrous at every level when a A8 and A10 completely run it into the ground, even dual core APU's are capable of running intel HD into the ground considering the benches ran i7's against a6's and a4's, while haswell will close the gap the policy remains when it seems stupid to have the low end bereft of the better iGPU.[/citation]

Many low end boards can have high RMA rates, but the same is true for many high end boards too, even from companies such as Gigabyte and Asus that are known for making some of the best boards. You simply have to find specific models that you like and hope for good luck no matter how expensive your board is. Also, from what I've seen, there are a lot more good, cheap Z77 boards than there were good, cheap Z68 boards.

I don't know many people who needs more than four to eight USB 2.0 ports and I don't know anyone who needs more than two USB 3.0 ports, so not having a ton of USB ports is rarely an issue. I also don't know many people who need more than two to four SATA 6Gb/s ports and four to six SATA3Gb/s ports, so they're rarely an issue too. With Ivy, VRM is usually not a big deal. Ivy simply doesn't suck a whole lot of power even with common overclocks such as around 4GHz to 4.6GHz.

I also don't know many people who need ridiculous memory performance. Even cheap boards can run reasonable amounts of memory (IE 8GB) at DDR3-2133 with good timings just fine and that's well into the upper end of the bell curve for memory performance in consumer systems. Most people don't need more than one PCIe x16 slot and for the few who want more, two is not a difficult number to find on some cheap boards. More than two is probably nonexistent on any cheap Intel board, but more than two is not often used in consumer systems anyway.

Yes, I too think that AMD boards tend to have better connectivity at a given price. I agree that AMD's socket changes are done in a more convenient way for upgrading. I agree that AMD's stances on overclocking are better than Intel's. I agree that Intel's IGP allotment in CPUs is "ludicrous" in some ways. I even agree to a point that someone with an i3 probably wouldn't bother upgrading to an i5 or i7 of the same platform later on because they'll probably not be worth the money by the time that customer could use the performance improvement if the trend of CPUs on dead platforms not going down in price as newer platforms with better price/performance ratios emerge keeps going.

I was not saying that the LGA 1155 i3s necessarily beat AMD across the line. AMD simply doesn't really top them out in average gaming performance. In some games, AMD's four to eight core models can beat the LGA 1155 i3s to varying degrees, but overall, the i3s and many of AMD's models are fairly similar in average gaming performance.

 

[cleeve]

excuse me, i have read and re-read the article several times. i have also dissected your graphs and how it is presented. now my comprehension may have fallen due to age and abuse i have subjected my brain cells to when i was much younger.

to say, this graph represents that graph doesn't work. you may know how the data was collected and you may know what you are representing but, no matter how many times i reread 5 paragraphs; its not explained how an average of ~40ms latency equates to a baseline of 0ms.

You still aren't getting the point, based on this and your previous responses.

That's OK. Take from it what you will.

 

[cleeve]

I may be mistaken, but as far as I know, the 30fps rule comes from the fact that it's the lowest "smooth" fps that you can get if you're using vsync in your game. The next step is 20fps, which has obvious stuttering for anyone.

Vsync isn't married to 30 FPS, it's married to half the monitors refresh rate (in many instances this is 60 Hz).

 

[Guest]

You still aren't getting the point, based on this and your previous responses.

That's OK. Take from it what you will.

i had hoped that if i was missing something glaring than another member would have jumped in.

so would there be even o hope of it being explained in the future to the functional brain dead such as myself?

 

[littleleo]

I know there was a price range involved but it seems all the top AMD CPUs popular in gaming were covered but the top Intel model for gamers wasn't the i5-3570K. That would have been interesting to see how it would have compared. I've seen them at MicroCenter for $189.99 so they would have qualified...

 

[Fulgurant]

No De5_Roy, I just don't know how people can be spoon fed that a i3 is better than every AMD CPU to date. I started with the Motherboard issue, that Z77 board will inflate the i3's results over anything a B75 or H61 (the chips realistic chipset) can produce, we have done this before and have found that the general difference is 2-5FPS from mid to high end can be around 10FPS from a low end to high end board, that is the reason why you pay so much. If performance was the same then they wouldn't cost it. Also what possible reason is there not to use the Crosshair V formula Z update which every reviewer will tell you is the best undisputed AM3+ board available.

(Emphasis mine.)

In other words, "High-end motherboards must increase performance because otherwise they wouldn't cost more." Boy, that's a vendor-friendly outlook. You're the customer every commercial enterprise dreams about, congrats!

High-end motherboards cost more because they have more features. For most users, those features are superfluous. Hell, these days, even low-end motherboards can have features most users will never realistically use. The one motherboard feature that does have a significant effect in performance comparisons is overclockability, but this isn't an overclocking article, and since it isn't an overclocking article, you can't kill Tom's for not running out to find the most expensive AMD mobo available.

When looking at intel B75 and H61 is a disaster, Z77 is expensive (like Z68 was) so why not a full spectrum review on build off's to price point and relative performance.

and the kicker its a locked multiplier and really not fitting for anything more than a H61/B75 platform which I wouldn't even give my mother.

You're so right, dude. Clearly, a chipset designed for small-business use will be a total disaster relative to an ultra-Xtreme Fatal1ty g4m3r board. It's not like businesses need reliable products or anything. And we know that everyone needs overclocking options!

High-end motherboards might be slightly more reliable, on average, than cheaper ones, all else being equal. Unfortunately, all else usually isn't equal; it's possible that feature-creep makes high-end parts more prone to quirky behavior. In any case, here are the return rates for major motherboard manufacturers, as of mid-November; notice that that there aren't a plethora of low-end boards on the hall-of-shame list at the bottom (return rates over 5%): http://www.behardware.com/articles/881-2/components-returns-rates-7.html

On the whole, in fact, motherboard return rates are pretty low. The bottom line, as usual in these discussions, is that occasionally every company sends out a dud, no matter the product. You have a really good chance of not receiving one, though.

 

[grich96]

So are the i3 3220 and i3 2120 better than the fx 8350 and the fx 8320? That seems hard to believe.

 

[cleeve]

so would there be even a hope of it being explained...?

Ok, here goes:

The chart you propose is only a representation of raw frame times. It does not impart any knowledge of frame-to-frame latency.

Two graphics cards, both with identical results on your proposed chart, could represent vastly different scenarios because of varying frame rates during the benchmark. One might have a static frame rate and a ton of latency; the other might have varying frame rates and high frame-to-frame latency. There's no way to tell!

I believe frame rates and latency are most helpful if you consider them separate entities. They are parts of the same problem, but they describe very different phenomenon.

 

[cleeve]

So are the i3 3220 and i3 2120 better than the fx 8350 and the fx 8320? That seems hard to believe.

Not really. The FX series has trailed the i3 in gaming since its inception.

This has been in our, and other sites', benchmarks for a long time.

Keep in mind, we're focused on gaming here. When applications are used, particularly multi-threaded applications, things can change in a big way.

 

[cleeve]

I know there was a price range involved but it seems all the top AMD CPUs popular in gaming were covered but the top Intel model for gamers wasn't the i5-3570K. That would have been interesting to see how it would have compared. I've seen them at MicroCenter for $189.99 so they would have qualified...

The 3870K is a mere 100 MHz faster than the 3550K. It makes no noticable difference unless you're overclocking, and we didn't do that.

 

[Fulgurant]

The 3870K is a mere 100 MHz faster than the 3550K. It makes no noticable difference unless you're overclocking, and we didn't do that.

Plus, Microcenter deals tend to be in-store-pickup exclusives. For whatever reason, Microcenter's been selling Intel CPUs at absurdly low prices (3770k for ~$220!) for months now, but since you can't order those items online, citing Microcenter's prices in an article like this one wouldn't be fair.

 

[Outlander_04]

Bottom line

Intel i5 3570K + Radeon 7850 1 gig = $380

AMD FX 6300 + 7870 GHz edition = $380

One of those two games WAY better .

In case casual readers cant work out the answer ...Its not the one with the"must have" intel processor

 

[Outlander_04]

[citation][nom]cleeve[/nom]Not really. The FX series has trailed the i3 in gaming since its inception. This has been in our, and other sites', benchmarks for a long time. Keep in mind, we're focused on gaming here. When applications are used, particularly multi-threaded applications, things can change in a big way.[/citation]


Several things need to be said to qualify that statement , I believe .

The first would be that the intel architecture is clearly better in some game engines because of the potentially stronger individual threads . This shows most clearly when you test older game engines . Today if you were picking a suite of games for testing would you include skyrim or starcraft? Probably not . They are not that current and are only used because they have a history in the Tomshardware benchmarks .
As newer games are released the pendulum will swing towards cpu's with a higher thread count giving some weight to the FX being more future proof .

You also have to consider that your benchmarks are run on clean installs on machines with little or no other software , and do not give a clear indication of the situation for an online gamer . More threads, more cores you are always better off in those situations . The intel dual cores lose a lot of gloss

 

[Guest]

Ok, here goes:

The chart you propose is only a representation of raw frame times. It does not impart any knowledge of frame-to-frame latency.

Two graphics cards, both with identical results on your proposed chart, could represent vastly different scenarios because of varying frame rates during the benchmark. One might have a static frame rate and a ton of latency; the other might have varying frame rates and high frame-to-frame latency. There's no way to tell!

I believe frame rates and latency are most helpful if you consider them separate entities. They are parts of the same problem, but they describe very different phenomenon.

*sigh*

thanks for the explanation i have known since the first Tech Reports "inside the second" article explaining the latency metrics back in september. and i am telling you again, what you are doing is misrepresenting the data which you haven't explained how you collected.

good luck with that- i am out.

 

[Fulgurant]

You also have to consider that your benchmarks are run on clean installs on machines with little or no other software , and do not give a clear indication of the situation for an online gamer . More threads, more cores you are always better off in those situations . The intel dual cores lose a lot of gloss

As an avowed Intel devotee, I have to admit that the above-quoted point legitimately falls in AMD's favor at the low-end of the CPU market. It's a point that isn't emphasized enough, IMO.

All of the other stuff that's typically dredged up -- complaints about motherboard inequalities, complaints about game selection, complaints about how the games are coded (as if any review site or any gamer, for that matter, has any control over how developers program games -- isn't anywhere near as compelling as the practical argument you summarize above.

What AMD gives you at sub-$200 price points is a quality-of-life advantage. And if you're looking for an extremely small, cheap, and power-efficient rig capable of playing games at reasonable frame rates, you can't beat AMD's integrated graphics solutions. I, for one, can't wait to see what Richland can do. Should be fun.

 

[cleeve]

*sigh*

thanks for the explanation i have known since the first Tech Reports "inside the second" article explaining the latency metrics back in september. and i am telling you again, what you are doing is misrepresenting the data which you haven't explained how you collected.

good luck with that- i am out.

And I'm telling you again, you don't appear to grasp the concepts. The chart you propose has no merit. There's no way to discern microstuttering from it.

I'm not out at all, though. In fact, I'm still here if you're honestly trying to get some clarification.

I'd even be willing to have a Skype discussion with you if you'd like to talk about it further.

 

[cleeve]

Today if you were picking a suite of games for testing would you include skyrim or starcraft? Probably not .

Totally disagree with you there. Skyrim and StarCraft II remain some of the most significant games of our day.