Begone Bulldozer. AMD’s Zen architecture is here to lay to rest those long, nightmarish days of lacklustre AMD processors. Genuine competition in the CPU space is here again with the top-end Ryzen 7 1800X offering serious computational grunt for half the price of their Intel rivals.
The CPU is only one half of the PC picture, you’ll also need the best graphics card to support your shiny new processor.
It’s hard to believe it’s been quite so long in the CPU wilderness for AMD, but since they gambled early on pushing thread-count over single-core performance with the Bulldozer architecture Intel have ruled the roost.
Admittedly the Zen architecture is arguably doing the same thing, pushing the industry forwards towards a cheaper multi-core future, but the timing for that is now arguably better and AMD have also made moves to shore up the single-core performance of their latest chips.
The big question for us though is how much does the $499 (£499) Ryzen 7 1800X close the gap with their Intel competition when it comes to primarily single-threaded gaming performance? The initial noises at launch weren’t great, so have things levelled out now a few months later?
Click on the jump links below to quickly skip sections.
- AMD Ryzen 7 1800X specs
- AMD Ryzen 7 1800X benchmarks
- AMD Ryzen 7 1800X performance
- AMD Ryzen 7 1800X verdict
Before we get carried away digging around the performance metrics of AMD’s chip we need to have a look at how it’s been put together. The Zen architecture represents a brand new, ground-up x86 CPU design, and AMD is right when they say that’s a real rarity in today’s mature PC component market.
The basic make-up of the 14nm FinFET Zen design surrounds the core complex (CCX). This is a modular chunk of silicon sporting four discrete CPU cores and allows the architecture to scale the core count upwards from there. The octa-core chips then feature a pair of these CCX modules at their hearts to provide this full eight-core design. This modular layout is what allows AMD’s upcoming Zen-based Naples server chips to pack a full 32 physical cores inside them and why there is speculation about a new AMD high-end desktop range too.
The major departure from the old Bulldozer architecture is the use of simultaneous multi-threading (SMT) to allow the Zen-based chips to better utilise the available compute pipelines and spread the load from a single core. Essentially it nominally delivers two concurrent processing threads from one core, doubling the thread count of processors which take advantage of the SMT tech.
Interestingly though AMD lists the SMT support of the Zen architecture as optional, paving the way for them to release lower-spec desktop Ryzen chips without the extra thread count, or potentially mobile CPUs without multi-threading support.
Partly it’s this increased parallelism which provides the Zen architecture with its boosted single-core performance. AMD have also specifically improved other parts of the design to target a higher instructions per clock (IPC) rating for Zen too. The architecture is now better able to predict what work is going to be required next, thanks to its neural network-based branch predictor, and is also able to chuck more work into the individual execution units too.
AMD have also made changes to the cache structure of the Zen architecture, providing each core with 512KB of dedicated level 2 cache memory and a full 8MB of level 3 cache shared between each four-core CCX module. AMD estimates this provides around five times the cache bandwidth per core compared with the final Excavator generation of the Bulldozer CPU family.
In between the individual CCX modules AMD are using an interface they’re calling Infinity Fabric. It’s this high-speed interconnect which allows the different parts of the chip, from the cores to the memory to the system controllers, to communicate and feedback to each other. The Infinity Fabric interconnect is intrinsic to the SenseMI tech AMD have also jammed into the Zen architecture.
It’s the SenseMI features which allow a Zen-based CPU to dynamically adjust its voltage and frequency on-the-fly as energy and temperature demands require. The Precision Boost and Pure Power features mean the chip can almost instantly push up the frequency and dial down the power on different parts of the CPU in milliseconds. When a part of the processor is not needed it gets gated off and the logic then doesn’t draw any power, which can then be distributed elsewhere.
Finally the Extended Frequency Range (XFR) is a feature which apes what Nvidia have been doing with their graphics cards - offering a clockspeed boost if there is enough thermal headroom available to the chip. XFR is primarily available to the ‘X’ models of Ryzen CPUs, allowing those SKUs to push beyond the clockspeed limits of Precision Boost, essentially given them an automated overclock of around 100MHz beyond their rated Boost clock.
The R7 1800X is the flagship CPU in AMD’s Ryzen range of processors and represents the current pinnacle of their new Zen architecture. Inside the R7 1800X are a pair of CCX modules giving it the full eight-core beans with 16 threads of processing grunt to back that up.
That layout means it’s filled with 4.8 billion transistors, with the smallest being based on the 14nm FinFET design from GlobalFoundaries. Those are all packed inside a die that’s a little over 195mm2. And, because it’s rocking two CCX modules, the R7 1800X has a total of 16MB of L3 cache inside it. It’s also the highest-clocked of all the Ryzen chips, with a base clockspeed of 3.6GHz and a maximum boost speed (on a single core) of 4GHz.
As an ‘X’ rated Ryzen processor the R7 1800X comes with a TDP of 95W, like the R7 1700X, to help it get its XFR on when there’s the thermal support available. In practice we didn’t really see any evidence of XFR pushing up the clockspeeds of the 1800X, though with this early platform our measuring tools aren’t necessarily 100% accurate. CPU-Z, AMD’s own overclocking tool and Windows all seem to give different results in terms of actual second-by-second clockspeed.
In terms of platform the R7 1800X drops into any AM4 motherboard, as all good AMD chips from here on out should. Unless this HEDT madness actually comes to pass… We’ve been testing in an X370 board as that is the chipset which offers the best chance of high-performance overclocking. The B350 chipset will also offer lower level overclocking while the low-end A320 doesn’t at all.
This top-end Ryzen CPU is quite remarkable. In pretty much any computationally intensive benchmark you care to throw its way the R7 1800X will greedily chomp through it before rapidly coming back for more. It’s one of the finest performers in the Cinebench test I’ve ever seen, even the $1,700 deca-core 6950X only just has the edge in the multithreading benchmark.
And that’s with our sample seemingly refusing to top 3.64GHz when all its cores were occupied. My first taste of XFR hasn’t been the most exciting...
The similarly-specced eight-core 5960X, itself a $1,000 CPU, can’t keep up with the 1800X and the similarly-priced Core i7 6800K is well off the pace too. As you might expect the quad-core, eight-thread 7700K brings up the rear… except in the single-threaded stakes. That’s where the cheaper, lower-spec Core i7 has the edge, with a 22% lead in the single-core Cinebench test.
The AMD R7 1800X does though struggle on the memory side too. At launch support for different DDR4 memory can be a little flaky and our regular 3,200MHz Crucial Ballistix kit was too rich for Ryzen’s blood. AMD suggested a 2,667MHz speed for the number of DIMMs we were jamming into the Asus Crosshair VI Hero AM4 board. And because the AM4 platform is a resolutely dual-channel one the quad-channel Intel X99 setups are able to dominate it on the memory bandwidth side.
But it’s really the gaming performance that we care about, and that’s where a lot of the initial concerns for Ryzen have… er… arisen. Before I get into that though a quick note about testing - I’ve heard a lot of people getting stick for testing at 1080p when they themselves play at 1440p or 4K and want to see results for that. But here I’m testing the R7 1800X, not the GTX 1070 we’ve used in our test rig since we started benching CPUs here at PCGamesN.
At 1440p or 4K all you’re really testing is the GPU performance, even at 1080p you can run into GPU limitations - as with Wildlands and Rise of the Tomb Raider - but you will still get a better idea of the performance delta between different chips when you’re not overly taxing the graphics silicon.
The performance difference though can almost disappear at the higher resolutions, so you may ask what does it matter? But as GPU power increases over time, as it inevitably does, the hidden performance differences will become more apparent as the platform matures. You may not care about it now, but you might in a year or two.
The good news though is that the performance delta between the top-end AMD chip and the Intel competition isn’t that great. There are though a couple of cases which display a severe performance difference, most noticeably in the older Total War and Grand Theft Auto DX11 tests where the R7 1800X is around 30% slower on average than the cheaper i7 7700K.
You might point to the incredibly low minimum frame rate in GTA V with the Intel chips, but if you take the 99th percentile frame time into account Intel is still able to post smoother overall frame rates. The 99th percentile frame rate for the 1800X is 71fps, while the Intel chips are 91fps, 77fps and 111fps for the 6800K, 5960X and 7700K respectively.
There’s also double-digit percentage performance differences between the two chips when it comes to Hitman and Civilisation VI in DirectX 12 modes. But with the 3DMark, Doom, Ghost Recon Wildlands and Rise of the Tomb Raider benchmarks there was essentially nothing between them.
We knew that Intel’s high-performing cores would mean they could retain their lead as the go-to purveyors of gaming CPUs, but the fact AMD have closed the gap this much with the limited R&D resources they have at their disposal should be applauded.
The most interesting comparison chip though probably isn’t the Kaby Lake 7700K. For my money it’s the Core i7 6800K which provides arguably the sternest test for the flagship Ryzen CPU. The 6800K is a six-core, 12-thread processor which costs around $100 (£100) less than the octa-core 1800X. In the CPU benchmarks the Ryzen chip has it beat, though it still delivers great multi-threaded performance for the cash. But where it stands out against the 1800X is in that gaming performance. It’s quicker and a good bit cheaper - that’s where my money would still go.
Ryzen is also a hot and thirsty architecture by comparison too. Even if AMD are to believed that their own Ryzen monitoring software is reporting the temperatures incorrectly (or correctly so “all AMD Ryzen processors have a consistent fan policy”) it’s still pretty toasty. The ‘X’ chips are supposedly reporting temperatures 20°C higher than the real CPU temps, but even then the 1800X is idling - under liquid chilling - above the 40°C mark.
And that plays into possibly why the Ryzen 7 1800X isn’t a particularly impressive overclocker. The 1800X is designed to boost up to 4GHz on a single core and it’s simple to switch that around so all cores are running at that speed, but pushing any further is a lot tougher. We managed an almost completely stable 4.1GHz all-core overclock by upping the multiplier, but had better luck (i.e. a more stable chip) pushing up the BCLK and dropping the multiplier. Even then the actual performance improvement was negligible and with AMD themselves recommending you don’t run with a core voltage of 1.45v, or else risk the longevity of your silicon, it doesn’t seem to make a lot of sense running Ryzen above stock clocks.
The Ryzen platform as a whole does also seem a little slow out of the blocks. It's nothing to do with the storage performance of the setup, but it seems to take an age for our system, with a high-end Asus X370 board, to even get to the POST screen.
I don’t love the Ryzen 7 1800X. It’s certainly not the processor I would instantly recommend any PC gamer out there buy, yet it’s tough to argue against what an important processor it is. With Ryzen AMD haven’t taken the CPU market by storm, they haven’t ousted Intel and they’re not suddenly going to become top dog. But what they have done though is become competitive and relevant once again.
You wouldn’t buy a Ryzen chip just for its gaming performance - you’d be leaving too much GPU power on the factory floor if you did - but there are people for whom a computationally-advanced $500 (£500) octa-core processor will be the stuff of their silicon dreams. Content creators, streamers and YouTubers will love the sheer CPU horsepower on offer even if it does limit their graphics card’s performance.
The fact there is even a question as to whether you should consider Intel or AMD for your next purchase, and that it’s not just an automatic decision, is huge. Of course it helps AMD have gone so aggressively into the market with the pricing of their new chips. Undercutting Intel’s eight-core range by so much was always going to be disruptive, but it was something AMD had to do to get noticed.
It does though mean AMD are reinforcing the old-school status quo, still occupying the more price-conscious end of the CPU spectrum. Money-no-object you’d go for the Intel platform whether you’re a gamer or a serious content producer. They still have the absolute best processor technology and motherboard platforms around. But AMD have produced a great-value eight-core processor for those who can’t, or don’t want to, spend over a grand on one.
But there’s still the fact most of us will be weighting our next component purchase on graphics and gaming performance over compute power or maximum core/thread count. With the 1800X costing more than either the 6800K or 7700K - even taking the platform costs into account - and unable to perform as well in games with the same GPU, it’s clear Intel will still hold the gamers’ hearts.
That said, we’ve long been suggesting the Ryzen 7 range aren’t the chips gamers should be excited about. The Ryzen 5 range, with the stunning R5 1600X, is able to deliver almost Intel i5 levels of gaming performance while offering three times the thread-count for the same price. That's where the smart gaming money goes.