Framework CPU: Choosing Intel i5-1135G7 vs i7-1165G7

As you know, the initial Framework Laptop offering has a choice between three CPUs, one i5 and two i7’s. The top-ranked i7 has vPro, which I do not need (and do not trust given Intel’s record), so I’m only looking at the Intel Core i5-1135G7 and the i7-1165G7. Intel CPUs are ubiquitous so there are lots of sites that compare “benchmark” numbers for these two processors. What’s interesting is the disparity amongst the reviews. The majority I’ve found states that there is about a 10-20% difference in common workloads. But one site says the i7 is about twice the speed of the i5.

Here are some of the majoritarian-view sites:

Here is the outlier that claims the i7 “wiped the floor” with the i5. Note the first comment after the review (but read “rethink” where the reviewer somehow said “overthink”).

Benchmarking is notoriously subjective, that is, all benchmarks should be taken with the proverbial grain or two of sodium chloride. There are too many variables that affect “performance” to exactly quantify one component. For example, in the FW Laptop case, choosing the SN850 version of the flash disk over the SN750 might make as much difference for less money. Or using paired memory DIMMs (eg., two x 16Gb instead of one 32GB, even though that might make upgrading later more expensive). I remind you: might.

Here, as reference, are the official Intel specs page for the two processors.

i5-1135G7 i7-1165g7

Note that Intel uses the term “Configurable TDP-up Frequency” for what we call “normal CPU speed.”

As always, you have to make your own judgement call. In my case, the performance increase for the i7 over the i5 does not seem to justify the $300 premium that Framework is charging (as was noted earlier, the MSRP difference on the two chips at retail is $117).
FW has not offered an explanation (nor are they required to), but perhaps there are other support chips that need changing, or perhaps they just need the money from i7 sales to fund the i5 sales, like Tesla sold Model S to fund development of Model X, and the same for Model 3).

At any rate, as always, YMMV: your mileage may vary.


The testing looks really bad. The Swift 5 Pro has an Nvidia MX350, so the gpu numbers probably aren’t really i5 vs i7 at all.

I also agree the $300 seems steep, but I went for it anyway. 10-20% is not an insignificant difference, and the bigger gpu is big plus for me. Plus, if you shop around for storage and ram, you might be able to save $100+ depending on the config you’re going for

1 Like

Excellent post.

The people running demanding workloads, like compiling or rendering, should know whether the extra cache or frequency will make a difference. Or they might just run serious workstations :wink:

1 Like

Please for the love of god, never use Userbenchmark, they show up first in google results and not enough people know how they weight their results

Take this comparison for example Ryzen straight up wipes the floor with Intel in every scenario but Userbenchmark gives a slight edge to Intel

I’m not a fanboi for either side but I detest the well-documented bias that site has

Avoid them and use reputable review sites…please


Framework could show a few benchmark scores aswell as the fps in a few games in their store or somewhere in their forums. It would make it easier for consumers to see the difference between the different processors option and later it could also shows the difference between multiple laptops from framework. They had to do some internal testings to see how their laptop perfroms anyway that they could show

With benchmarks you kinda need to know what they are even trying to measure. There are a lot of usecases for a processor and it will perform differently in every one. You can’t really summaries the performance in one score in a good way. Good reviews will show you a great variety of benchmarks for this reason for all kinds of applications. In an cpu comparison a good reviewer should make sure that the laptops used are identical in almost everything but the cpu. That also goes for the TDP and cooling of the cpu.

About usebenchmark (click to expand)

Userbenchark is favouring intel cpus over amd cpu to the point where it gets hilarious. When amd ryzen cpu started to dominate over intel cpus in their benchmarks because of their high multicore perfromance they reduced the impact of it from 10% to 2% of their “effective speed score” despite multicore becoming more and more important. It also seems to put a huge weight on memory latency where intel is better than amd. All in all you can get some hilarious results.
A test from notebookcheck where amd outperforms intel in al scores but intel is still rated higher.

Or a 64 core threadripper TR being beaten by a intel 4790K from 2014 despite being faster in single-, quad-, octa- and multicore perfromance but a bit slower in memory latency

Things also get hilarious when you look at the comments from userbenchmark
About the 10900K:

Intel’s Comet Lake flagship, the i9-10900K, is the fastest gaming and desktop CPU currently available. This ten-core hyperthreaded processor can easily be overclocked so that all twenty threads run at an eye-watering 5.2 GHz. […] It’s stellar performance is second to none […].

About ryzen 5800X:

Our benchmarks show that the 5800X is comparable to Intel’s $250 USD cheaper i5-9600K. Gamers that do not wish to pay “marketing fees” can invest those savings in a better GPU which will produce an unquestionably better gaming PC.


For myself I went with the i5 because it should be better on battery performance (I could just be assuming this in my head since they all have the same TDP range). But I also choose the cheaper one because I’m sure one day (might be a couple years) they’ll offer an AMD mainboard and I’d probably upgrade to that. i5 should be more than strong enough to run anything I need… Which honestly, all a laptop cpu needs to do for me is be able to run the OS effectively*. I’ll do more CPU intensive stuff on it than just the OS but having to wait a little bit longer while compiling is fine.

*this just stems from the last laptop I owned over a decade ago that low end models… were laughable at how well they run an OS. Luckily we’re no longer in that era though.

For very light gaming, the 96 cores of xe igpu on the 1165g7 vs 80 on the 1135g7 were worth the upgrade cost.

I expect future mainboards to move on to ddr5. I want my first mainboard to have tolerable performance for a few years, since the total cost of upgrading it will be significant.

Probably the biggest difference in terms of the 1135 vs. 1165 CPU (not the GPU) is the increased cache. The increased clock speed helps too, but the cache more. The difference between the 1165 and 1185 is not as pronounced but they have the same cache and just a small clock speed difference. I could justify small cost increase going from the 1135 to the 1165 but not the large cost difference going to the 1185.

There’s a HUGE review of the 1165 here:

Too bad they didn’t compare all Tiger Lake UP3 CPUs. It would be interesting to see such a detailed comparison between the 1135, the 1165 and the 1185.

All I went by for such a comparison is PassMark, which doesn’t have the best reputation.


@Fraoch Techspot is a very reputable reviewer, they also recently demonstrated the effect that cache can have on performance here


On the topic of cache and perfromance. Amd plans to bring out a refresh of their ryzen 5000 desktop CPUs with an 3D stacked cache which increases the amount of cache from 32MB to 96MB. Amd also showed a few gaming benchmarks with an average of 12-15% perfromance increase thanks the higher amounts of cache.

I would be careful with comparing a desktop CPU to a low power laptop cpu. With the perfromance of a desktop CPU the cache and ram is more likely to become a bottlneck than with the perfromance of a laptop CPU

I’m definitely not trying to do that, I’m linking to an article that demonstrates the effect cache can have on performance, the fact the benchmarks were run on desktop CPUs is kinda irrelevant to that, namely because its hard to demonstrate this effect on laptop machines that have different cooling solutions and restrict RAM timings and such

moar cache is moar better all else being equal

1 Like

Yes, of course. My desktop is an AMD Ryzen 9 5900X. 12C/24T, 105W TDP, 64MB L3 cache. 3.7GHz, 4.8GHz max. It will absolutely blow the Core i7-1165G7 out of the water in any metric you’d care to measure, but:

  • it has 4 times higher TDP
  • it has 3 times the number of cores
  • it has over 5 times the cache

What’s more interesting is comparing CPUs with the same constraints against each other. Cache helps more than clock speed. The number of cores helps, but up to a point - check that review, although the Ryzen 7 4800H dominates multi-thread performance, the 1165 beats it in a fair number of single-threaded applications - and that’s despite the 4800H’s 45W TDP.

I would have preferred Ryzen on Framework, but the Tiger Lake CPUs are pretty good.

The 15W section is pretty interesting in your test. Btw the 4800H only consumes 12W on a single core Cinebench run, thats why it has the same single core performance as a 12-25W 4800u. Both have the same max turbo clock of 4.2Ghz. In comparison the i7-1165G7 in the Schenker Vision 15 consumes around 23-24W for 4.7Ghz unless you lower the TDP. Also cinebench isn’t influenced a lot by ram speed and latency. With faster ram you only see like 3% improvements in cinebench while in games would run like 15% faster. So more cache isn’t going to have much influence in this benchmark. The main contributer of faster ram and more cache are video games ironicaly.
I agree that the biggest contributer for the little cpu perfromance difference between tiger lake i5 and the i7 is most likely the cache. The rest most likely be because of slightly higher clock speeds thanks to better binning on the i7. Also the base and turbo clock speeds on the spec sheets can’t really be trusted. A higher base clock or turbo clock is no guarenty that the cpu can clock higher at the same power consumption as before

1 Like

That is the exact opposite of what benchmarks are supposed to do

Benchmarks are repeatable and verifiable and should be easy to compare across platforms and should be limited in one aspect or another to measure performance in a specific arena

Reputable reviewers take great pains to preserve their reputation, that’s the only currency they have

If you feel otherwise then that’s because you have found reputable reviewers


These are all great resources-Jarrod and Tim from HardwareUnboxed are the only ones on that list that I know do Laptop reviews/mobile part reviews

1 Like

This is also inaccurate-I assume you are basing those numbers off of the price listed on intel ARK

Those number are tray prices-not retail and are irrespective of volume discounts, promo funds or any number of other things

The price framework pays may be higher, it may be lower-we the consumers will never know since we don’t have eyes on their contract that is almost certainly under NDA

I went i5, just because I have no use case where I’ll be needing extra performance. I have my 5800x on my desktop for pretty much anything CPU intensive.

One thing I’m curious about is multicore workloads. I saw a couple reviews stating that the i5 can sometimes outperform the i7 in some all-core benchmarks. i7 obviously trounces it in single core workloads - which will matter more in the vast majority of use cases, but if you’re using it more as a desktop replacement and have some intense multithreaded workloads probably worth checking. FWIW I get about 5423 in Cinebench R23 with the i5.

1 Like

I went with the i5. I weighed the option of the 1165. But none of the options are quite to the level of a major powerhouse for heavy tasks anyway. So I decided to go with the i5 which should be fine for everyday tasks and is the cheapest option. Now maybe I won’t feel so bad upgrading when a new mainboard comes out with something more exciting. If that doesn’t happen, I still saved $300. And I can keep repeating that to myself that if I ever start to wish I’d went with the i7, lol.

1 Like