The Intel Core i3-7350K (60W) Review: Almost a Core i7-2600K
by Ian Cutress on February 3, 2017 8:00 AM ESTOffice Performance
The dynamics of CPU Turbo modes, both Intel and AMD, can cause concern during environments with a variable threaded workload. There is also an added issue of the motherboard remaining consistent, depending on how the motherboard manufacturer wants to add in their own boosting technologies over the ones that Intel would prefer they used. In order to remain consistent, we implement an OS-level unique high-performance mode on all the CPUs we test which should override any motherboard manufacturer performance mode.
All of our benchmark results can also be found in our benchmark engine, Bench.
Dolphin Benchmark: link
Many emulators are often bound by single thread CPU performance, and general reports tended to suggest that Haswell provided a significant boost to emulator performance. This benchmark runs a Wii program that ray traces a complex 3D scene inside the Dolphin Wii emulator. Performance on this benchmark is a good proxy of the speed of Dolphin CPU emulation, which is an intensive single core task using most aspects of a CPU. Results are given in minutes, where the Wii itself scores 17.53 minutes.
The 7350K, with a high single thread frequency, easily surpasses the i5 and i7-2600K here. That being said, there's a slight difference to the Skylake i3, perhaps down to various generation specific code differences.
WinRAR 5.0.1: link
Our WinRAR test from 2013 is updated to the latest version of WinRAR at the start of 2014. We compress a set of 2867 files across 320 folders totaling 1.52 GB in size – 95% of these files are small typical website files, and the rest (90% of the size) are small 30 second 720p videos.
WinRAR is more geared towards a variable threaded environment but also memory speed. The fact that the Core i5 is above the Core i3 shows that having actual cores helps, regardless of frequency - the additional hyperthreads for the Core i7-2600K also gives it the win, despite the memory frequency difference.
3D Particle Movement v2
3DPM is a self-penned benchmark, taking basic 3D movement algorithms used in Brownian Motion simulations and testing them for speed. High floating point performance, MHz and IPC wins the day. This is the second variant of this benchmark, fixing for false sharing in the first version, and lending itself to better multithreaded performance.
3DPMv2 is still new, so we don’t have too many results for it so far - but again this is another situation where having actual cores helps. This is typically when the threads are 'heavy', i.e. spill out into various caches and require more than 1/2 the cache shared within a core each. In the case of the Kaby Lake, this means that each core has 32KB of L1 - or 32KB per thread for the i5 but only 16KB per thread in the i3.
SYSMark 2014
Engineered by BAPco (to which Intel is a consortium member), this set of tests are designed to be an office/data/media/financial range of tests using common well-known CAD, image editing, web browsing and other tools to put out a score, where a score of 1000 is attributed to an old Core i3 using a mechanical harddrive. Here we report the overall score, however the test breakdowns can be found in Bench.
Because SYSMark is a variety of tests that rely on response and throughput, here is where the Core i3 comes into play over the i5 and i7-2600K. With the i5 it's about equal, but the years of IPC increases put the i7-2600K now behind the Kaby i3.
Web Benchmarks
On the lower end processors, general usability is a big factor of experience, especially as we move into the HTML5 era of web browsing. For our web benchmarks, we take well-known tests with Chrome as installed by SYSMark as a consistent browser.
Mozilla Kraken 1.1
Kraken favors high frequency and IPC, so the i3 takes a large lead over the i7-2600K for this sort of workload.
Facebook
Twitter
RSS
186 Comments
View All Comments
JordanV - Tuesday, February 14, 2017 - link
I think the sales argument for the big Intel chips as video encoders has been for x264 where the faster NVENC, VCE, and QuickSync technologies offer lower quality at a given bitrate for higher quality x264 settings. For most people, the hardware encoders are enough but for many others, the quality is not sufficient.The quality difference between hardware and software HEVC is smaller with higher quality software h265 encodes beating the quality of your Pascal x265 encodes but with a big performance penalty. It's not worth it for most people, but if you have limited bitrate/storage and want the best quality, it might be.
HerrKaLeun - Friday, February 3, 2017 - link
Thanks for the great review, Ian.Considering one needs an expensive Z-board to OC, for most people buying an i5 makes more sense.
I don't understand why so many people complain about Intel allegedly not making enough progress. Now you get a dual-core that comes close (or even exceeds in single threaded benches) to the former flagship quadcore. If you want to have a CPU that vastly exceeds the "old" quadcore, Intel also has newer quadcores. It is not like the i3 is the end of the lineup.... For the $317 that the 2600k used to cost you can get a KabyLake non-K i7, which sure vastly exceeds in performance (and much lower TDP). I assume someone who could afford an over $300 CPU 6 years ago can afford $300 now and upgrading to an i3 may not be what that person would do anyway. the trend goes to more cores.... most people here complain about Intel not offering mainstream hexa and octa cores... not sure why the same people allegedly are so eager to get dual-cores.
zodiacfml - Friday, February 3, 2017 - link
Dual core is too weak for me. Web browsing can use more cores.Hulk - Friday, February 3, 2017 - link
Sorry to be dense.What does 2+2, 4+2, 4+3/e mean?
babysam - Saturday, February 4, 2017 - link
The first number refers to the number of CPU cores. Te second number refers to the IGP configuration (the number of shaders, which may be a little bit different across generations, e.g. Haswell GT3 has 40 shaders, while Broadwell/Skylake GT3 have 48 shaders).The extra e means there is an extra eDRAM cache (Crystalwell) on the CPU package.
Hulk - Saturday, February 4, 2017 - link
Thanks.AndrewJacksonZA - Saturday, February 4, 2017 - link
Thank you babysam.babysam - Saturday, February 4, 2017 - link
Thank you for you article (especially when many of us are waiting on the information of new CPU of both AMD and Intel). It is always good to have something to play (overclocked) with, but this is a little bit expensive.When I read the analysis of the first page, I see the lack of information on the CPU die size and transistor count disclosed by Intel recently. Also, I feel strange that the effect of the change of the 32nm to 22nm (from Haswell to Broadwell) have such a large difference on the 2C+GT2(which Intel claims there is a 37% reduction of the die, which can be seen in the table) and the 4C+GT3(which the difference are much smaller) CPU die. I feel even stranger when I see the Skylake 4C+GT3e die is a bit smaller than the Broadwell 4c+GT3e die. So I am quite curious on the sources of the die estimate.
P.S. I found the origin of the 234mm^2 of the Skylake die size estimate.
https://techreport.com/forums/viewtopic.php?t=1177...
which based on the images of the following
http://www.anandtech.com/show/10281/intel-adds-cry...
It seems that the die described is the Skylake-H(which is a 4C+GT4e configuration). This makes the 241.5mm^2 estimate of the Broadwell 4C+GT3e a little bit unrealistic (Skylake GT4e have 72 shaders, while Broadwell GT3e have 48 only)
babysam - Saturday, February 4, 2017 - link
Just find the die size of the Broadwell-H (4C+3e) in this documenthttp://www.intel.com/content/www/us/en/embedded/pr...
According to the document, the die size of Broadwell-H (4C+GT3e) should be 13.7mmx12.3mm = 168.51mm^2
(Many thanks for the hints: https://forums.anandtech.com/threads/broadwell-cor... , which the got the answer two years ago.)
WoodyBL - Saturday, February 4, 2017 - link
Am I the only one noticing that the i5-4690 was beating the i5-7600k in a lot of benchmarks? I'm having a hard time processing how that was even possible...