Puget Systems print logo
Read this article at https://www.pugetsystems.com/guides/617
Article Thumbnail

Introduction to Intel S-series Processors

Written on January 29, 2015 by Matt Bach


If you are in the market for a new PC, you're probably all too aware of the huge variety of different CPU models available. We've covered the differences between the desktop Intel Core i3, i5, and i7 CPUs in our Haswell Core i3 vs. i5 vs. i7 - Which is right for you? article, but it turns out that there are actually two or three different product lines within each of those brands. You might may be familiar with the "K" or "X" lines, but did you know there are also "S", and "T" lines as well? According to Intel, these letter designators mean:

  • K/X - "Unlocked" - allows for CPU overclocking. "X" is simply the letter denoting the highest-end desktop CPU available from Intel in that generation.
  • S - "Performance-optimized lifestyle" - a little lower performance in exchange for lower power draw.
  • T - "Power-optimized lifestyle" - much lower performance, but also much lower power draw.

In this article, we want to cover the S-series product line in particular because the specifications for S-series processors are actually very similar to that of standard Intel CPUs. To determine the advantage of S-series CPUs we will first look at the official specifications differences between a couple of S-series processors and their standard counterparts then move on to our own testing and benchmarking to determine the actual performance, power draw, and thermal differences.

Specification Differences

To compare the official specifications between S-series CPUs and their standard counterparts we took four of the latest generation Haswell CPUs. There are, of course, more models available from Intel, but we've found that the differences between S-series and their standard counterparts to actually be very consistent, so this is a good representation of the S-series as a whole.

Basic CPU Specifications i7 4790 i7 4790S   i5 4690 i4 4690S
Base Freq. 3.6 GHz 3.2 GHz   3.5 GHz 3.2 GHz
Max Turbo Freq. 4 GHz 4 GHz   3.9 GHz 3.9 GHz
Cores/Threads 4/8 4/8   4/4 4/4
TDP (wattage) 84 W 65 W   84 W 65 W
Smart Cache 8 MB 8 MB   6 MB 6 MB
T case 72.72 °C 71.35 °C   72.72 °C 71.35 °C

This isn't a complete list of all the official specifications (you can view a complete list at Intel's Ark page) but it covers most of the major specifications and all of the ways in which the CPUs officially differ. According to Intel's official specifications, there are really only three differences between the two different model types.

First, the S-series CPUs have a much lower base frequency than the standard CPUs. You would think this means that they will always be slower, but we've shown in the past that the base frequency is somewhat of a meaningless specification due to Intel's Turbo Boost and Speedstep technologies. These technologies mean that an Intel CPU will only rarely operate at the advertised base frequency which brings into question whether this spec actually has any bearing on real-world performance.

The second difference is the lower TDP wattage on the S-series processors. Again, this is a somewhat misleading specification since according to the Intel Ark specification pages, TDP is defined as:

... the average power, in watts, the processor dissipates when operating at Base Frequency with all cores active under an Intel-defined, high-complexity workload

In other words, TDP is only what the wattage of the CPU is when running at the base frequency, not the maximum frequency as determined by Turbo Boost. And since we have found that the base frequency isn't a great indicator of performance, this means that TDP is also a questionable specification. It certainly indicates that the S-series processors should run at a lower wattage, but unless you disable Turbo Boost it does not in any way guarantee it.

Finally, the last difference is the lower T case temperature on the S-series CPUs. This is related to Intel's overheating protection and basically is the thermal cutoff point where an Intel CPU will automatically lower its operating frequency to try to reduce the amount of heat the CPU is generating. More information on this is available in our Impact of Temperature on Intel CPU Performance article. Realistically, as long as a CPU has adequate cooling this specification should never have a bearing on the performance of the CPU.

To sum up, all the official specifications really tell us is that S-series processors may be a bit slower and may run a bit cooler than their standard equivalent but it is far from conclusive. For that, we will need to perform our own testing. This testing gets into some technical details, so if you are only concerned about our conclusions, feel free to jump ahead to the Conclusion section.

Test Setup

Since Intel's specifications are not very great at showing the actual difference between the S-series and standard CPUs, we decided to compare a Core i7 4790 and a Core i7 4790S to determine the actual power draw, thermal, and performance differences between the two CPUs.

For this testing, we used the following hardware:

All driver and Windows updates were applied prior to testing. Coretemp was used to record both the CPU temperature and operating frequency along with a "watts up? PRO" power meter with USB logging to record the system power draw.

Real world difference between 4790 and 4790S

Starting with the overall system's power draw, we found that while the system was idle or in sleep mode, the two CPUs had almost identical power draw. While running our various benchmarks, however, we started to see a mix of results. In both Unigine Heaven 4.0 Pro and Cinebench R15 (single threaded mode) both CPUs had almost identical system power draw. Only Prime95, Linpack, and Cinebench R15 (multi threaded mode) resulted in the Core i7 4790S using less power. Even then, the difference was only 5-10 watts (or 4-7%) which is far less than the 19 watt difference that Intel advertises.

Higher wattage should directly correlate to higher temperatures so it is no surprise that this graph is very similar to the power draw graph. The Core i7 4790S runs at approximately the same temperature as the Core i7 4790 at idle, during Unigine Heaven, and during the single threaded portion of Cinebench R15. Really, the only significant difference in temperature occurred during Linpack and Cinebench R15 (multi threaded mode) where the 4790S ran 4.5 ºC and 3.75 ºC (or 5% and 8%) cooler respectively.

With the actual system performance results we start to see a very clear pattern. Just like the power draw and temperature results, Unigine Heaven and Cinebench R15 (single threaded mode) perform almost exactly the same with both CPUs. However, Linpack and Cinebench R15 (multi threaded mode) ran about 5% slower with the 4790S compared to the 4790.

So, why the big variance between the different benchmarks? Obviously, a pattern exists, since the 4790S and 4790 really only showed a difference in power draw, temperature, and performance during Prime95, Linpack, and Cinebench R15 (multi threaded mode). What we believe to be happening is in many ways related to the base frequency specification we discounted during the Specification Differences section. It turns out even if a CPU never runs at its advertised base frequency, this specification is still an indicator of how effective Turbo Boost may be depending on the number of threads an application can use.

Performance based on number of active threads

As we pointed out earlier in this article, due to Intel's Turbo Boost technology an Intel CPU will almost always operate above its rated base frequency when put under load. What is often overlooked, however, is that the Turbo Boost frequency at which a CPU runs depends on the number of cores actively being used. In fact, the maximum Turbo Boost frequency Intel lists on their specification sheets only applies if the CPU is using just a single core. As increasingly more cores are used simultaneously, the maximum Turbo Boost frequency is reduced.

Likely, the inconsistencies we found in the previous section are related to how many threads each benchmark is able to utilize and how that affected the amount of Turbo Boost each CPU received. To determine the actual amount of Turbo Boost available to our test CPUs based on the number of active threads we ran Prime95 on each CPU with the bechmark limited to between one and eight active threads.

As you can see in the above chart, it is clear that as more CPU threads are being used, the 4790S runs at an increasingly lower frequency than the standard 4790. This means that even though the 4790S and 4790 technically have the exact same maximum Turbo Boost frequency, they are very different as more threads are being used. Depending on the number of threads, the 4790S is anywhere from .1 to .2 GHz slower than the 4790. This isn't much - only about 2.5 to 5.3% slower - but it lines up almost exactly with what we say in our real world benchmarks.

S-series processors are supposed to be lower wattage (that is how they are primarily advertised by Intel) so the slightly lower performance is to be expected and in some ways is secondary to the power draw of the S-series CPUs. What we found was that the 4790S did indeed draw less power when five to eight threads were being used, but with fewer than five threads the 4790S actually pulled more power than the standard 4790 by an average of 6%.

We have no explanation for why this is, although it could simply be that we had an unusually hot running 4790S and an unusually cool running 4790. Either way, our testing with these sample CPUs indicates that the 4790S is actually only lower power when more than half of the total CPU threads are under load.


To summarize both our research and testing results, there are three major ways in which Intel S-series processors differ from the standard models:

  • Power Draw: Our testing between a Core i7 4790S and a Core i7 4790 indicates that S-series processors draw less power when more than half the threads on the CPU are being used. On the other hand, our testing also showed that the 4790S actually draw slightly more power when less than half threads are being used.
  • Temperature: Even though the 4790S used more power in some applications than the 4790, we didn't really see the 4790S running any hotter than the 4790 in those instances. When the 4790S does use less power, however, we saw a drop in temperature of up to about 5 °C. This isn't insignificant, but at the same time is not going to be a game changer in most situations.
  • Performance: On applications that can only utilize a few threads, the S-series CPUs are essentially identical in performance to their standard counterparts. If an application can use a higher number of CPU threads (or if many applications are running at once), however, the S-series CPUs can be as much as 5% slower than the standard models. 

Overall, we would say that S-series CPUs can simply be thought of as being slightly slower, slightly cooler, and a little more power efficient than their standard counterparts. None of it is by very much - and in many situations the two models are really identical - but if you want a system that uses slightly less power in exchange for slightly lower performance than the S-series processors may be worth considering. One thing to keep in mind is that S-series CPUs are usually priced exactly the same as the standard CPUs so unless the small power savings is important you will get better performance for your dollar from the standard models.

Tags: Intel, S-series
Tai The Guy

Plot twist: every S, U, T, Y series processor is really just a standard model CPU, with a microscopic manufacturing flaw, that ALMOST passed CQ. And so they're re-branded with the appropriate suffix, and sold along side their superior flawless brethren - saving Intel millions

Posted on 2019-10-06 07:54:52
Jorge Luiz Rodrigues Vilella

Actually, U and Y chips are entire different chip designs in both configuration and die size. They're definitely NOT a result of binning.

S and T are, indeed, binned chips that didn't make the cut to be mainstream/K processors. In some generations, i5s are defective i7s that just couldn't keep themselves stable with Hyperthreading too.

But hey, that's the norm in silicon! You could get ten different i5-2500Ks with different overclocking capabilities depending on luck alone. The GTX970s used an 1GB VRAM chip that couldn't keep itself stable when loaded >512MB, combined with 3 good ones to get 4GB VRAM but really slow speeds when you get to the last half gig...

Posted on 2020-06-28 09:19:56