i5 CPU and Gaming
A few weeks ago I got curious enough about the effects of a CPU’s speed on gaming that I decided to start doing some serious testing. I mean for a long time I have said that it really was not an issue, that even slower CPU speeds could still deliver great gaming but that was in an abstract from disjointed testing. So with a few i5 chips, a Gigabyte Z77 motherboard and an EVGA GTX 660Ti video card I set about to find out if the theory held up.
To begin my journey I needed to set some parameters and so I chose the most common chip comparison I normally use the i5 3450 vs the i5 3570K. I thought we should add a bit more to the mix so I enlisted the help of my i7 3820 so we could add hyper threading and I got all the speed settings for the i5 3470, 3550 and 3330. The testing was all performed on a clean Windows 8 install using the same drivers for the GTX and all games as well as benchmarks set to the same settings in 1080 resolution. The settings were all done to the highest offered in game with no extra or outside tweaking. This was done to reflect a normal user experience level. The base clock speeds were tested along with the 3570 down clocked to allow testing at the base speeds of the other processors. All systems had 8 gig dual channel memory provided by Kingston set at 1600 for speed. The only test performed outside the normal setup was the i7 3820 which was performed on a Sapphire Pure Black X79, however it was limited to dual channel on the memory.
For games I wanted to pull from a board range of games that are popular, again I was aiming for looking at what the typical gamer is going to see. With this in mind I loaded up Skyrim, Borderlands 2, Black Ops II, Planetside 2, League of Legends, Star Trek Online, World of Tanks, Witcher 2 and Hawken. All of the games were benchmarked using Fraps and everything was set to highest in game levels at 1080 resolution. To this mix I added some synthetic benchmarks with 3DMark, Heaven and PCMark, using the presets in the benchmarks for testing.
Now I could give you a,long list of numbers and frame rates here, I have a spreadsheet full of them but that is not how we roll here and everyone that has been a listener over the years knows this. Instead of looking for a bunch of individual numbers I was working through the data to look for a deeper pattern or meaning as to how the various PCU speeds effected the overall gaming and system performance.
Lets begin with the overall experience, using every chip speed I tested, ALL of the games played smooth and delivered a great gaming experience across the board. During all of the game play testing not a single game suffered from any changes to the CPU speed. This is across, in the case of the i5 chips, a pricing delta of $190 to $230, according to current Newegg pricing. Now in fairness this delta is very small and in my opinion Intel can of convoluted the market with the current selection. But that is for a bit later in this discussion. Benchmarks showed the difference but then again we expect them to, however the difference was only seen in the numbers.
Speaking of the numbers, what did the performance numbers show? Well for purposes of this article what I did was measure the performance delta with the i5 3330 at the lowest speed being set as the base. I then took the differences in performance as we worked up the testing in a percentage gain format and once I had all the the tests calculated I took the average of those scores to give an overall performance difference.
The result was within the range I expected but still not what I expected. At the extreme of the delta, the i5 3330 vs the i5 3570 the performance difference was a whapping 4.7%. Now again this is an overall average so this reflects in my opinion the best mix of the benchmarks from the games and the overall scores from the synthetics to give the total system impact of the CPU speed. This test is NOT about pure CPU horsepower, but about the impact on the system as a whole and it appears that impact is not very high.
Now I did not feel that it was fair to stop there, as enthusiasts are all about the overclock. So with that in mind I tested the i5 3450 at 3.9 GHz, the max overclock it can achieve on a Z77 board without effecting bus speeds. I then did the same type of overclock on the 3570 which thanks to it’s unlocked multiplier was able to achieve 4.1 GHz without any tweaking using the same method. Adding these into the mix the delta overall climbed to 6.1%. Again these overclocks were achieved with now real tweaking and on stock cooling. So from an i5 3330 to an i5 3570 @4.1GHz we see a gain in total system performance of about 6%.
Now again let me stress this is an OVERALL view of the performance delta. I did my best by offering a wide range of games to look at not just the CPU performance itself but the TOTAL SYSTEM impact that the various CPU speeds were making. This means that while my testing results are showing this smaller overall system impact in certain areas the impact can be higher, or lower. This is why I averaged out the gains to get the numbers I arrived at. I am not listed all the various speeds and numbers because the delta here is so small that it would not be useful, the differences are next to nothing as you step up.
Now as I mentioned earlier in this piece I did also add testing with an i7 3820, the reason I am not adding it’s numbers into the mix is because the price point on the chip, current $300 on Newegg puts it way outside the the i5’s and the numbers from testing get skewed a bit. Using 3DMark as a prime example, we see a pretty large performance boost due to the addition of the Hyperthreading. this boost however does not reflect within the gaming benchmarks and so skewed the numbers from the i7 pretty bad compared to the others. If I remove the 3DMark score or I turn off hyper threading the 3820 falls right in line with the i5’s.
Now that might seem a bit like I am throwing out test results but that is not an accurate assessment. For the nature of this test I am seeing that the i7 brings nothing to the table for the gamer at this time in any meaningful way that is reflected in real life performance. Now I am NOT saying the i7 does not deliver better performance, it does within software that leverages the Hyperthreading fully. That however does not appear to be something that current crop of games do.
What about if we push the overclock, does the performance keep ramping? Yes it does, if we push the I5 3570K with some voltage tweaks and begin to ramp it up, the performance delta does continue to climb, however my testing has shown that my friend Shannon Robb was accurate when he said the gain begins to tapper off. I was able with some take the 3570 to 4.5 GHz but I stopped there when I started looking at the results. At 4.5GHz the performance delta has only climbed to 7.4%, that is a minor gain and as the speeds climbed I watched the gain fall off. Would higher speeds results in more performance, of course it would but the percentage of gain would continue to decline as the speed climbed and this would mean more effort for less results.
What does this mean however, for me, the gamer that wants to build a new gaming rig? First do not get caught up in the unlocked multiplier craze of today. I see enthusiasts forums actually call builds that do not include an unlocked multiplier dumb and this is not cool. I stand by my position that of the i5 lineup right now the sweet spot if the i5 3450. It is an easy chip to find out in the wild and delivers outstanding gaming performance at stock speeds plus can, if you desire, get a nice little boost to 3.9GHz without fancy cooling or tweaking. In the grand scheme of things that is plenty of performance for a great gaming experience and has some headroom.
Now this does beg the question of what is Intel thinking in it’s marketing? From the i5 3330 to the i5 3570K we have 8 Models and yet only a 5% or so delta overall? What is worse is this chip group is all within that $180 to $230 price range. Now I understand the desire for a high end model that opens up the overclock option, I understand a low power option like the S chip and I understand a middle of the road or mainstream chip to fill out a lineup with options. However the convoluted mess that we see with 8 chips is nothing but confusing for the consumer.
So with this information in hand my recommendations are pretty straight forward and simple. If you want to build a new system the best value right now is the i5 3450. The stock speed is plenty for a great computing experience in everyway, as gamer you just add a good gaming video card to this and you are set. If you like to tinker and want a littler more performance then you can overclock to 3.9 GHz and push this chip to it’s limit. If however you are more than a little into tinkering and you want those bragging rights, then spend the extra money for the i5 3570. While at stock speeds it does make sense over it’s little brother, if you like to push things to the limit it is a great chip for tinkering and seeing just what you can make the system do.
As for the other chips in the Intel i5 lineup? Well I would buy if they are the only choice or the pricing is the same. For example I have seen the i5 3550 for the same price as the i5 34560, well it would be silly not to buy it as you get a little better base speed for free. However as far as paying more for the minor boost up the lineup, they are really not worth extra money that could be sued for a better video card, more RAM, an SSD or or other component changes that would have a bigger overall system impact. What about the lower power chips, the S series? Those I must say I am torn on, the reason is the i5 chip is already a very low TDP chip and the reduction might be nice on paper but I am not sure if worth it at the end of the day for an extra cost.
Next up AMD….
Thank you to the folks at EVGA, Gigabyte, Intel and Kingston for providing the components used in this testing.