Early last year, AMD introduced its brand new “Zen” micro-architecture for desktops, with the Ryzen brand. Offering much higher performance than the previous ‘Construction’ cores ending with Excavator, Ryzen was a great success. AMD had pulled off something great with Ryzen, they beat their initial target of 40% performance per clock – reaching 52% higher IPC than Excavator. And while the Ryzen processors can’t quite match their Intel counterparts in pure IPC or single-threaded performance, they absolutely dominated their equivalently priced blue-team parts in multi-thread, all while still offering very serviceable single-thread. After all, it was the quad-core i7 7700K going up against the octa-core Ryzen 7 1700 in terms of price.
Towards the end of 2017, Intel responded with the ‘Coffee-Lake’ series of CPUs. Lead by the flagship mainstream i7 8700K processor – now featuring six cores instead of four. Across the board, Intel has now offered more cores than previous ‘7th generation’ parts. Even the lowly i3s are now native quad core parts (admittedly without Hyper Threading, so they still only have four threads). Combined with improvements to Intel’s 14nm process (now called 14nm++), Coffee Lake can reach the same high frequencies as Kaby Lake, but with two additional cores to boot.
This year, AMD has announced price-cuts to the majority of their 1st generation Ryzen CPUs, most likely in preparation for the incoming 2nd generation known as ‘Pinnacle Ridge’. These price cuts add even more value proposition to the already well priced Ryzen parts, so how exactly does Ryzen now fare against a similarly priced Intel CPU from the Coffee-Lake generation?
But wait…
Now, a lot of people will ask “why not test the i5 8400, since they are both hexa cores?”, and while this is a valid question, I feel the $169 i3 8350K is somewhat of an underrated product. It’s essentially a cheaper i5-7600K with an additional 2MB of L3 cache. Offering a native quad-core die with four threads and complete overclocking support, it seems like a processor geared towards gamers. It’s also cheaper. Quite a bit cheaper at the time of writing. Combined with the already higher platform cost of Z370, I believe this is still a valid comparison.
In addition, some might notice that the i3 8350K is in fact cheaper than the Ryzen 5 1600 for the processor alone, and that’s true. But…
The answer to that is the fact that overall platform cost for Ryzen is lower than Z370, and the money saved from the lower cost B350 Motherboard allows me to get a more expensive Processor, whilst still costing less than the i3 8350K and its cooler (which it doesn’t come with).
Now that I’ve gotten that out of the way, let’s take a look at the actual silicon powering these Processors.
The CPUs
i3 8350K
- 4 cores
- 4 threads
- 256kb L2 cache per core
- 8MB L3 cache
- Dual-channel memory interface
- 4 GHz base clock
The i3 8350K is a native quad-core die based on the ‘Kaby Lake-R’ silicon. The CPU cores feature the same underlying micro-architecture as the 2015 Skylake architecture. The chip has 256kb of L2 cache per core and an 8MB inclusive Last Level L3 cache. In addition, the CPU cores communicate via Intel’s “ringbus” topology, rather than the new “mesh” architecture introduced with Skylake-X CPUs. Unlike the current Ryzen processors, the 8350K includes an onboard GPU on the silicon, seen here occupying over a third of the die space.
A die shot of Kaby Lake-R, the chip used for the 8th generation i3 series.
Ryzen 5 1600
- 6 cores
- 12 threads
- 512kb L2 cache per core
- 16MB L3 cache
- Dual-channel memory interface
- 3.2 GHz base clock
- 3.6 GHz boost clock
AMD’s Ryzen 5 1600 is based on the same die as the entire Ryzen lineup, and even the Epyc Server processors. This die is called a “Zeppelin”. It is a eight-core die, with cores split into two clusters called Core Complexes. One of these cores per CCX are disabled on the Ryzen 5 1600, creating a hexa-core part in a 3+3 configuration. Each core features a 512KB L2 cache. The L3 cache is split into two 8MB blocks, one per CCX, and is left fully operational on the 1600. The L3 cache is a victim cache, unlike Intel’s inclusive cache. This means that the cache only accepts data evicted from L2, rather than allowing it to be loaded into the Cache directly. Simultaneous Multi Threading is left enabled so each core can handle two logical threads, giving this processor 12 logical threads in total.
The “Zeppelin” die, clearly visible are the two Core Complexes.
Test systems
For this comparison the following components were used:
- Asrock Z370 KILLER SLI / Asrock AB350M
- 16GB (2x8GB) Avexir 2666 MHz C16-18-18-36
- NVIDIA GeForce GTX 1070 Ti with driver version 390.65
- Seasonic FOCUS 850W Platinum
- Samsung 960 PRO (M.2 NVME) / Samsung 850 PRO (SATA)
In addition, on the i3 8350K, a be Quiet Slim 120W cooler was used, whereas the 1600’s stock cooler was used.