As you can see, the effect of Hyper-Threading is diminishing. This makes it easier to understand Intel's drive to separate Core i5 and Core i7 in terms of other specifications, too. In particular, already from the 2nd generation on, these lineups got different cache capacities. From the 3rd generation on, all Core i7 also got the best graphics cores. For Core i5, those remained more like an exception, because virtual multithreading alone is not enough to justify the difference in positioning and price. It won't be a surprise if the share of Core i7 will continue to reduce for mainstream platforms.
Why would you want to speed up office suite if those work fast enough already? Well, "fast enough" isn't "instant". Besides, there are cheaper processors, too, for which 2.4 GHz isn't always achievable (Core CPUs in higher-end netbooks, for example, usually have clock rates of 1.0 to 1.5 GHz). Moreover, Finereader is the only benchmark that supports multithreading well. The rest favor single-thread performance that grows with every CPU generation. Yet, the 3rd one shows a smaller boost than the 2nd.
As you can see, the 3rd generation provides half as much boost as the 2nd. This fits in the general picture well.
This is when Hyper-Threading is a drawback: many today's games need four threads but not more. (Or at least they can't handle more threads as efficiently.) For Nehalem, HT was useful in heavy-load modes, but a minor increase in single-thread performance provided by the 2nd and 3rd generations of Core CPUs renders HT useless for gaming rigs.
Again we turn to experimental tests, because results promise to be interesting. The idea is simple: five benchmarks are run one after another in 15-second intervals; all tasks are sent to background (none of their windows are active). The result is a geometrical mean of completion times...
...Aaand is quite unexpected. These diverse threads give four cores the same treatment no matter the platform generation. However, the efficiency of Hyper-Threading spikes in this case (as opposed to parallelism within one piece of software). This isn't going to help regular users much, because most processes they run remain in standby while in background. But such users prefer Core i5 CPUs anyway. In turn, Intel Core i7 processors lean towards professional use. This might be useful to know for buyers of Ivy Bridge-based Xeon E3s: the junior 1220V2 and 1225V2 do not support Hyper-Threading and are closer to entry-level workstation CPUs. For a server, there's more sense in buying at least 1230V2. All the more so 8 threads are cheaper in this segment than they are in the desktop one.
Overall Score and Final Thoughts
Before making any conclusions, you should note that our benchmarks aren't optimized for Ivy Bridge, being more than a year old. Many programs have been updated since that time, so when we update our test method (which is going to be soon), results may change, processor ratings, shift.
Yet we can see that there's much sense in Ivy Bridge whatever the case. The new processors are at least on par, sometimes, much better than the old ones. This is nice compared with the move from Nehalem to Westmere that sometimes resulted in performance drops. In other words, the popular opinion that Ivy Bridge owes everything to the improved Turbo Boost isn't completely true. The 3rd generation does have certain architectural benefits. Nothing major, though. For comparison, the most noticeable change in the 2nd generation was the introduction of the ring bus to mainstream CPUs and the increase in L3 cache frequency. In turn, Ivy Bridge focuses on improving graphics and reducing power consumption (aside from the move to the 22nm, of course). That minor performance boost is provided by internal optimizations which is more like a bonus. At least despite we have some 'radically multi-threaded' software in our test method that favors Hyper-Threading so much, Ivy Bridge with HT disabled catches up with Nehalem with HT enabled, given the same clock rates, number of cores and cache capacity.
Speaking of Hyper-Threading, let's calculate how efficient it is in general. This should be indicative, considering how much diverse software we use in tests. So, to the 1st generation of quad-core Intel Core processors, HT provided about 11.3% boost. For the 2nd generation, that result reduced to 10.5%. Finally, to the 3rd generation, it provides 9.3% more performance. The software is same, test conditions are the same, the technology itself hasn't become any worse, but it's getting less and less useful nevertheless—for each single application at least. A single thread loads a physical core so well that it leaves no resources for anything else. However, Intel has seemingly changed something in the 3rd generation in this regard, judging by our experimental test results. You might say they could've done it in the 2nd generation instead, but it was a large step forward already.
That's basically what progress has looked like for these three years. After three generations, the Intel Core architecture has improved much. Well, you can see that cores themselves only gained about 10% performance, but a finer process technology and optimizations also allowed to increase clock rates directly and by means of Turbo Boost. The reduced power consumption is also a plus.
We don't know for sure what happens next, but we do know that it's the end of the line for LGA1155. Moreover, there's information that Haswell CPUs for LGA1150 won't be just 'upgraded Core' but representatives of a new architecture that is promised to be not just a step but a leap forward. Let's wait and find out!
Write a comment below. No registration needed!