There was a discussion not long ago about why Will Fastie (@willf) was using an i5 for his build rather than an i7. I opined that the best way to go was to find the “sweet spot” in terms of price vs. performance rather than trying to future proof the computer if it means going outside of the “sweet spot.”
I used an example of my own choice to buy a Dell XPS 13 (model 9310) with the i5 CPU rather than one of the two i7 options. My thought was that since all three of them use the same case, the same heat sink, and the same fan setup, it would mean that the thermals would greatly constrain the more expensive CPUs.
I started ruminating on that topic, and I decided to give the ol’ benchmarks a try once again.
I used the trial version of Geekbench (which only outputs its result to their web site; if you want offline testing, you have to pay for it) to test the XPS in a bunch of different configurations (most of which were meant to save power). The out-of-box setup proved to be one of the fastest (UEFI CPU performance strategy set to balanced, CPU idle driver intel_pstate with powersave governor, and the CPU energy performance policy set to balance-performance. These were the default settings in the UEFI and in an Ubuntu-based Linux distro.
On Geekbench, my XPS 13 with the i5-1135g7 CPU scored 1446 on single-core and 5524 on the multi-core. These are rather meaningless numbers without something to compare them to, so I looked up the faster of the two i7s that are available in the 9310. I looked through all of the results and selected the very fastest of the i7-1185G7s I could find, with 1611 and 6115 for single and multicore, respectively.
Those numbers represent a boost in performance of 11.4% and 10.7%, respectively. That’s not much! The cost for this upgrade would come to a substantial $270, fully a quarter again what I paid for the XPS (which I got on Dell’s anniversary sale, 35% off). It’s doubtful you’d even notice 12% greater performance, and it surely would not be enough to make the difference when the little laptop is someday facing obsolescence.
The i7-1185G7’s integrated GPU is also a bit better than the i5-1135G7’s, but I haven’t tested that (yet?). The same thing applies here as before. It’s the same cooling system as with the i5, and that extra performance again means more heat. The CPU and integrated GPU share a TDP budget, as they’re both on the same chip. The more heat the GPU puts out, the less thermal headroom the CPU has to do its thing, and vice versa.
This is a specific case, of course. A desktop would not be as thermally constrained as a laptop, and certainly a super thin and light laptop like the XPS. Laptops are, though, the most popular form factor of PC supposedly, so it is useful to look at the various CPUs in this context.
I also checked the geekbench scores of some desktop versions of the 11th-gen CPU in my XPS… in this case, the i5-11600k and its i7-11700k counterpart. These things could be in any motherboard in any kind of desktop case with any kind of cooling, in addition to the myriad configurations that can be set up within the OS and UEFI, so I just looked for the fastest ones in each.
The single-core performance of each chip seems to be about the same (about 1800), while the i7’s multicore score (~10k) tops the i5 (~8k) by 25%, which is not surprising given the i7’s two extra cores (8, versus the 6 in the i5). Because of the vagaries of the tests in question, these relative performance figures are approximate. I would expect the i7 to show a slight increase in performance over the i5 because of the larger “Intel smart cache” and the very slightly higher clock speed (only 0.1 GHz). In real world performance, I doubt anyone would notice it, and it’s no surprise that the hypothetical performance difference disappears into the “noise” on the benchmark results.
In this case, buying the i7 would only get you that ~25% (maybe a bit more) better performance on workloads that maximize the potential of having an eight-core, sixteen-thread CPU. Some workloads do, like software compilation or video rendering (and multicore benchmarks), but most people don’t run those sorts of heavy loads that lend themselves to massive parallelization. For those people, the extra performance potential of the i7 over the i5 won’t be realized, now or in the future, most likely. The tasks that don’t lend themselves to parallel computation still won’t do so in the future. It’s not all about those darn lazy developers who don’t write a program to utilize the multiple cores effectively, as I’ve heard a lot of people say!
With the i5-11600k, you’ve got six cores with twelve threads, and that’s plenty for most workloads. That matches the i7-8750H in my G3 gaming laptop, and is 50% more core-y than the i7s available in the XPS 13/9310 and the ones available for my desktop PC, the i7-2600k and the i7-2700k. Intel is all over the map with their i-series nomenclature.
On any of the many tasks that are still limited by single thread performance, the i5-11600k is just as good as the i7-11700k, at least from these results. If you average out the ~0% gain in single-core and the 25% gain in multiple core, you get 12.5%, not that different from the results in the two mobile CPUs for the XPS.
In this case, though, the difference in CPU prices is about $100, with the i5 at ~$300 and the i7 at ~$400 (based on current prices on Newegg as I write this). The difference in price is the same as the difference in cores (six or eight, respectively), so if you have workloads that can actually use sixteen logical cores consistently, it would be a prudent move to go for the i7, as it’s just as much in the sweet spot as the i5– if the i7’s extra cores are needed.
In the case of the XPS and the mobile CPUs, I’d definitely take the i5 and the $270 (which is what I did). In another laptop model where there is more cooling available, the performance difference may be better.
Dell XPS 13/9310, i5-1135G7/16GB, KDE Neon 6.2
XPG Xenia 15, i7-9750H/32GB & GTX1660ti, Kubuntu 24.04
Acer Swift Go 14, i5-1335U/16GB, Kubuntu 24.04 (and Win 11)
