Following AMD’s Ryzen press event and the public release of a few benchmark scores for their new chips, I found myself pondering the comparison Lisa Su made between the new Ryzen 1700 and Intel’s 7700K. It would appear that Kaby Lake has a frequency advantage as well as slightly better IPC than Ryzen, which compensates by packing more cores into a smaller TDP, and I started wondering what kind of performance differential this will lead to in video games.
Along the way, I spotted a claim that Ryzen’s single thread performance won’t be high enough to make it a good gaming chip. Later on, I saw someone claim that the 6900K was never designed for gaming and would suck in games. While I can understand why someone would make the first of these claims, a last attempt to diss on AMD before the chips hit retail, the later claim piqued my interest, so I decided to look into.
Given that Ryzen has been shown to have a similar IPC to Broadwell-E, and that comparisons between Ryzen and Intel’s flagship i7, the 7700K, are inevitable, I decided to do my own comparison between the 7700K and the 6900K to get some idea of what kind of gaming performance lead the i7 should expect to hold over AMD’s head. In my head, this didn’t seem like it was unreasonable to expect in the slightest. After all, improved core count is one thing, but the 7700K holds an IPC advantage along with 1GHz base and 800MHz turbo frequency leads. As multi-threaded as games have become, the i7 is still a quad-core CPU with multi-threaded, the widely acknowledged “sweet-spot” for gamers right now. I didn’t expect the 6900K to be able to overcome that performance improvement trifecta to outspeed the 7700K.
I started by looking for direct comparisons between the two in reviews and benchmarking sessions, but they were harder to find than I expected. After a little digging around, I managed to find a couple of sites, Anandtech and Guru3D, that had reviewed both chips, tested them in games and kept the same testing methodology. You can find the four reviews I’ll be referencing here, here, here and here.
Let’s look at Anandtech’s results first. These are average frame rate figures for controlled benchmark runs performed at 1080p with Ultra graphics settings, rendered on an Asus Strix GTX 1080 4GB graphics card:
Moving on to Guru3D’s results, again average frame rates at 1080p on a GTX 1080:
The DirectX versions used by the games in these tests is as follows:
- DX9 – GRID: Autosport
- DX10 – Total War, GTA V
- DX11 – Alien, Far Cry, The Division
- DX12 – Tomb Raider, Hitman
Finally, it’s worth noting that Guru3D’s Firestrike scores for the 6900K and 7700K were almost identical at 17487 and 17854 respectively.
The Anandtech results are interesting in a very old and boring way that’s remained relevant for coming up two decades now. All these games are GPU bottlenecked. Neither the 7700K nor the 6900K can be said to have any advantage here at all, although it does raise the question of why Anandtech uses only these games to test CPUs on.
Guru3D’s results are more interesting – hard not to be when there’s actually some differential worth mentioning. The 7700K beats the 6900K by a worthwhile margin in 3 of the 4 tests, while the 6900K narrowly edges out it’s only win in a DX12 title. While an 11% lead in The Division to the 7700K seems worthwhile when taken at face value, it’s worth considering that every result here from both processors is in excess of 100fps. While higher FPS is ‘A Good Thing™’, the diminishing returns of perceived frame rate increases over 100fps (outside of VR) is also ‘A Thing™‘, especially when technology like Nvidia’s G-sync variable refresh rates are employed. I would like to find more game tests to analyze, but my searches haven’t shown up any more direct comparisons I can use.
It’s not a surprise to me that people can accuse the 6900K of “not being a gaming CPU” in Internet forums and article/video comments sections. Given it’s huge price premium and the benchmarks above showing no corresponding performance increase, it’s easy to start at “the 7700K is the better value and performance choice for gamers who don’t stream,” mentally adjust that to “the 6900K performs worse in games than a cheaper CPU,” and then boil that down to “the 6900K sucks!” Even though it’s far from true, I can understand the proliferation of that opinion through the Chinese whispers of modern social media.
There are effectively two ways of lining up CPUs across manufacturers for comparison. If you do it by performance, the Ryzen SKU’s AMD is launching are showing early indications of having achieved performance parity with Intel’s enthusiast chips at a fraction of the cost. If you do it by price, folks are likely to look at benchmarks results showing Intel’s mainstream i7’s outperforming Ryzen in games and immediately write Ryzen (and AMD) off as irrelevant. While AMD may have a whole raft of released aimed to match Intel’s performance while undercutting them on price at every product level, it’s likely Ryzen’s first impression may be underwhelming for a lot of folks.
I do wonder why AMD aren’t starting at the bottom of the product stack, leveraging Ryzen’s universal unlocked multipliers against Intel’s anti-overclocking stance to attract prospective overclockers away from Intel’s small, expensive range of K series chips. If you have to choose between spending your money on a quad-core Ryzen, very possibly with eight threads, or an i3-7350K, it’ll come down to the clock speeds you can reliably hit with each chip to decide which way your money goes.
And this is where things get really interesting. YouTuber Austin Evans got himself in a little bit of hot water yesterday when his Ryzen event coverage had to be pulled and re-edited after he violated his NDA with AMD. Yours truly saw the original upload before it was yanked and thought it worthwhile tipping off a few publications about the material that was removed from the version that emerged later in the day. Firstly, a reference to AMD’s Vega graphics card was overdubbed. Badly. Watch the video, you’ll know exactly where I mean but don’t worry, there was nothing substantial removed about AMD’s upcoming graphics card. A later segment showing the a liquid nitrogen overclocking session was also removed. I won’t go into details, but let’s just say a new Cinebench R15 world record score was set…