Many reviewers and hardware enthusiasts have either migrated to the Intel Core i7 and Core i5 platforms, or are in the process of doing so. Unfortunately, this upgrade is not without its pitfalls for the unwary. When it comes to the Intel Core i7 and Core i5 processors, the biggest pitfall you have to watch out for is the Intel Turbo Boost Technology. Most of us would have heard of this innocuous-sounding feature. Unfortunately, it is this passing familiarity that causes most of us to virtually forget all about it. Even as seasoned reviewers and hardware enthusiasts, we were once caught with our pants down and we have no doubt that many other reviewers and hardware enthusiasts too would have made similar mistakes. It is even possible that many are still unaware of the problems that Intel Turbo Mode can cause. Link : ED#114 : Intel Turbo Boost Technology - Trapping The Unwary
Turbo boost is NOT overclocking By definition, overclocking is running the processor at a clock above the manufacturer's specifications. And so, the clock speed increase due to a manufacturer-supported and fully warranty-covered feature can't be called overclocking. If the processors beat another processor by increasing their clock speed, the processors are definitely better than that other processor. It's not an unfair advantage. The processors have a feature that the other processor doesn't have, and it's only fair that they win. You also say that it's difficult to get accurate benchmark results. Actually it's very easy. Just make sure that you're running the benchmarks with a constant ambient temperature with the same cooler and same workload. You'll find that the Nehalems turbo up by the same amount each time. The very purpose of benchmarks is to let the end user know which processor to buy in a given price range. Suppose when you turn on turbo, Processor 1 beats Processor 2 by 10%. And when you turn off turbo, Processor 2 beats Processor 1 by 5%. In this case, for the end user Processor 1 is obviously the better choice if both cost the same, since they aren't going to turn off turbo (there's no need). But since you'll publish benchmarks with turbo off, you'll be misleading them. Of course, turning turbo off is beneficial when overclocking because it often leads to stabler overclocks at high clocks. But overclockers probably already know that, so it's better to publish benchmarks with turbo ON.
I don't think I will agree with you one that. Turbo Boost is definitely considered as overclocking. Whether it voids warranty or not is a completely different matter. i7 920 is rated to run at 2.66GHz. Turning on the Turbo Boost will "overclock" it to 2.8GHz. Note that I use the work overclocking here? There's the meaning of overclocking. Unless you have a temperature controlled room, it's impossible for the ambient temperature to stay constant. It is very misleading because Turbo Boost will only be enabled when the processor is cooling. In our country, we are constantly getting 30+C temperature, and it is relatively easy to get the processor to overheat. And if I were to compare my results with some other websites, I may not get the same results. And worst of all, I noticed most websites do not publish their Turbo Boost settings.
Turbo Boost is not activated whole time. And judging from all OC reviews for i7 and i5 (from 2'8 to 4 Ghz with air cooling) I don't see any problem with "turbo boost" at all. Well, no problem for the normal user. As article says, " it is a dangerous feature for reviewers and hardware enthusiasts " but they already know every option that exist in BIOS. It's like "Hmm.. C-state doesnt allow us to upper speed. Remove that option, Mobo manufacturer!" _
Turbo Boost should be activated the whole time. As a feature for end user, there's definitely no issue, and worked as advertised. But as a hardware reviewer looking for consistency, making fair comparison with other processors, I would definitely prefer to know which is actually faster clock per clock and the reason for being faster. I was fairly interested in i5 processors. After knowing that the Turbo Boost actually skew results, and clock per clock, it is not really that much faster than C2D or C2Q. Personally, I overclock all my processors, so I don't really need Turbo Boost. Currently I own E8400 and overclocked to 3.8GHz using stock voltage. i5 is not that appealing to me anymore.
Turbo Boost is activated whenever there's a heavy load on the CPU. Intel defines that as "when the Operating System (OS) requests the highest processor performance state (P0)." In addition, Intel clearly states that the effect of Turbo Boost on CPU frequency will vary according to a variety of factors including load : I agree that enabling Turbo Boost can be necessary if we want to show the real world effect of a Core i7/i5 processor versus another processor. However, we must do so with the full understanding that the Core i7/i5 processor isn't actually running at the official clock speed. Not only will we have to report this in any review / article that enables Turbo Boost, we also have to point out that the results of a Turbo Boosted processor cannot be used to determine a particular CPU's processing efficiency. For example, let's say we look at the x264 HD Benchmark 3.0 results (Pass 1) and compare the results of the Core i5-750 and the Core 2 Quad Q9450, both of which run at 2.66 GHz. Without Turbo Boost, the Core i5-750 is just 4.6% faster than the Core 2 Quad Q9450. With Turbo Boost disabled, we can accurately determine the efficiency of the Core i5 processor when compared to the Core 2 Quad processor of the same clock speed. We can see that the Core i5 processor isn't really that much faster than a Core 2 Quad processor on a clock-to-clock basis. However, if we intentionally or accidentally ignore the effect of Turbo Boost on the CPU's performance, then Turbo Boost will skew our perception of the processor as the Core i5-750 would be running at 2.8 GHz or 3.2 GHz. The Core i5 will now appear MUCH faster than the Core 2 Quad processor on a clock-to-clock basis. This is merely one example of how Turbo Boost can skew our understanding of how fast are the Core i7/i5 processors compared to the older Core 2 processors or even AMD processors. What I'm trying to point out with this article is that we should all be aware of what Turbo Boost does. For reviewers and those who wish to accurately record a CPU's performance at a particular clock speed, the proper understanding and judicious use of Turbo Boost is an absolute necessity. PS. Turbo Boost is, by definition, a dynamic overclocking feature because it goes beyond the official clock speed of the processor, and it does so dynamically based on a variety of factors.
I don't quite agree with that. The only part of Turbo Boost that can be called a "feature" would be its dynamic nature which adjusts the clock speed according to a few factors. Other than that, all it does is overclock the processor. That itself isn't a feature per se, and is something that all processors can readily do. I agree that you can get repeatable results with Turbo Boost on. However, it is possible that you may NOT get accurate results because Intel does not specify the exact conditions in which Turbo Boost is enabled. These conditions may even differ for different models. For example, we may benchmark the Core i7-965 EE and the Core i7-940 with Turbo Boost on. Both processors officially have the same Turbo Boost profile. But what if the controlling factors (voltages, temperature, load factor, etc.) are different? Intel could set Turbo Boost to run in the Core i7-965 at a higher temp and voltage than the Core i7-940. This may skew the results because while we may assume that Turbo Boost is running for both CPUs, the difference in operating conditions may actually cause Turbo Boost to perform intermittently or not at all in one of the processors. I agree... if that is what we want to show. For example, in a review of a Core i7 processor, we may want to show the Core i7 processor with Turbo Boost enabled to show its real world performance. However, we must also mention that the processor is NOT running at XX MHz but rather within a range of frequencies from XX MHz to YY MHz. In our review of the Intel Core i7-975 EE, I intentionally turned off Turbo Boost because the processors we tested all had the SAME Turbo Boost profile - they were all overclocked by a maximum of 2 speed bins. Because they would all be overclocked by the same amount, Turbo Boost would have no net effect on the performance difference between the Core i7 processors we tested. For example, the Core i7-975 would still be 133 MHz faster than the Core i7-965 with or without Turbo Boost.
You are correct in saying that leaving turbo boost on unintentionally will lead to overestimation of a processor's per-clock efficiency. Nehalems are not more efficient than Core 2's on a per-clock basis in some workloads. But to the end user, per-clock efficiency means nothing (enthusiasts are an exception, of course). Instead what matters is per-dollar efficiency. I myself overclock all my processors. I was just saying that you can mislead general public by turning off turbo. So it's probably better to include both turbo-off and turbo-on results in the benchmarks. The turbo-off numbers will be the worst-case performance of the processor (for example if the ambient temperature is too high or appropriate cooling is not available).
Oh, I don't intend to turn off Turbo Boost for all reviews or articles. In fact, I'm not advocating that reviewers or hardware enthusiasts should disable it whenever benchmarking a Core i7/i5/i3 processor. All I'm saying is that we should be aware of what Turbo Boost does, and decide if we should turn it on or off. In some conditions, we will want Turbo Boost enabled, while in other conditions, it should be turned off. For example, in our x264 HD Benchmark 3.0 results, we are asking everyone to turn off Turbo Boost so that we can list the results at the exact frequency. For this purpose, we are recommending that should you want to overclock, you should do it manually so we can directly compare it to other processors on a clock-to-clock basis.
The sad thing is, most reviews, even the high profile ones, are ignoring the Turbo Boost feature and say i7 is much faster than C2Q at equivalent clock speed. This is clearly very misleading, I was also fooled by the results, thinking that i7 is much faster clock-per-clock. Most sites failed to mention the Turbo Boost, and almost none of them compare the results with and without Turbo Boost. When I see the results of the x264 benchmark, I noticed something is not right. The C2Q are surprisingly competitive, then only I started noticing that Turbo Boost is skewing the results for most websites, since Adrian mentioned that Turbo Boost must be disabled to make a fairer comparison. I almost bought an i5 system, but now knowing that my C2D is not really that far off i5, and I'm already running at 3.8GHz, there's no point in upgrading. Almost none of my applications supports more than 2 core multithreading.
Turning off Turbo boost seems fair for comparison Besides, turbo boost seems to be affected by ambient temperature i suppose. So the colder pace you live the better boost it will have?
That depends. Intel set maximum limits that Turbo Boost can achieve... but whether you reach those limits depends on a variety of factors, temperature being one of them. But generally, using Turbo Boost in cooler climates (or with better coolers) is likely to generate better results than doing so in warmer climates... or with poorer coolers.
